Department of Medical Imaging - University of Toronto

University of Toronto Department of Medical Imaging Research Day 2006DEPARTMENT OF MEDICAL IMAGINGANNUAL RESEARCH DAY 2006Date: Tuesday, April 25, 2006Location: J.J.R. MacLeod AuditoriumStarting Time: 12:00 PM with welcome by Walter KucharczykAbdominal, Pelvis, GI, and GU ImagingSession Chair:12:10 PM Ants Toi Do additional lateral cores increase prostate cancer detection compared to the standard sextantbiopsy?12:17 PM Masoom Haider Diffusion Weighted Imaging for Localization of Prostate Cancer12:24 PM Hyun-Jung Jang Enhancement Patterns of Hepatocellular Carcinoma on Contrast-enhanced Ultrasound –Correlation with Pathologic Differentiation12:31 PM Kartik Jhaveri Comparison of CT Histogram analysis to adrenal washout CT in diagnosis of lipid pooradenomas12:38 PM Emma Robinson MRI imaging and clinical response of adenomyosis to uterine artery embolization12:45 PM M. Aeja Syed Hepatic and Portal Venous Thrombosis Associated with Hepatic Abscess12:52 PM Brian Yeung Does Radiology Consultation Pre-Imaging Affect the Outcomes of CT Renal Colic Scans?12:59 PM Tae Kyoung Kim Focal Nodular Hyperplasia and Hepatic Adenoma: Differentiation with Low-MI ContrastenhancedUltrasound – Work in Progress1:06 PM Neil Rosta Emergency CT Imaging for Acute Abdominal Aortic Disease - Are Utilization PatternsChanging?1:13 PM An Tang Hepatic vein transit times using a microbubble agent to predict severity of hepatic disease noninvasively1:20 PM Bina Lanka Impact of Contrast Enhanced Ultrasound in a Tertiary Clinical Practice1:27 PM Richard Bessell-BrownePheochromocytoma and paraganglioma: Risk of adverse events with IV nonionic contrastmaterial for CT1:34 PM Martin O'Malley Testicular Cancer Surveillance: Standard versus Low Dose CT1:41 PM Marie Staunton Can CT reliably distinguish moderately from poorly differentiated hepatocellular carcinoma?Breast, Chest, and CardiacSession Chair:1:50 PM Roberta Jong The Results of the Digital Mammographic Imaging Screening Trial - DMIST1:57 PM Oana Moscovici Characteristics of Breast MRI Screen-detected lesions that should be sent for targetedUltrasound2:04 PM Raafat Abou Saif The Value of Brest Imaging in Diagnosis of Breast Cancer in Patients with Nipple Discharge2:11 PM Mousumi Bhaduri Outcome of Breast MRI for screening breast cancer in mixed population of women2:18 PM Andre Pereira Assessing the performance of CAD for lung nodule detection: effect of radiation dose2:25 PM Hamid Bayanati Low-dose Computed Tomography in Prior Asbestos-exposed Workers: Assessment of lungnodules and pleural plaques2:32 PM Igor Sitartchouk Dynamic CT perfusion for Lung Nodules Characterization2:39 PM Hany Mehdizadeh-KashaniEffect of slice thickness and algorithm reconstruction on the performance of CAD for lungnodule detection2:46 PM Katherine Zukotynski Contrast Echocardiography Grading Predicts Pulmonary Arteriovenous Malformations at CT2:53 PM Patrick Cervini Out of Hours CT Pulmonary Angiograms: Accuracy of Resident Reporting3:00 PM Philip Buckler Pulmonary High Resolution CT Findings in Patients with Congenital Bilateral Absence of theVas Deferens3:07 PM Afsaneh Amirabadi T2 MRI for Early Diagnosis of Myocardial Iron Overload in b-Thalassemia3:14 PM Hong Huang Cardiac MRI Assessment of Cardiomyopathy in Patients with Cirrhotic Ascites3:21 PM Errol Colak The Utility of a First-Generation Computer-aided Detection (CAD) Tool for the Diagnosis ofPulmonary Arterial Filling Defects

University of Toronto Department of Medical Imaging Research Day 2006Musculoskeletal, Physics, and PaediatricSession Chair:3:30 PM Nadja Saupe Diffusion Tensor MR Imaging and Fibertractography of the Human Calf: Results at 1.5T and3.0T3:37 PM Jim Li Dynamic joint imaging using the similarity-based navigator (SIMNAV) motion compensationtechnique3:44 PM Katherine Zukotynski Ultrasonography for Children with Hemophilic Arthropathy: How we do it.3:51 PM Sean O'Connor Neonatal Findings of Fusion of the Forniceal Columns and Absent Cavum Septum Pellucidum3:58 PM Shantel Minnis The Value of Ultrasound in Assessing the Need for Full vs Focused CT in Children Evaluatedfor Acute AppendicitisNeuroimagingSession Char:4:07 PM David J. Mikulis Gadolinium Enhancement Predicts Hemorrhagic Transformation in Acute Ischemic Stroke4:14 PM Noel Fanning Cochlear implantation in patients with cochlear otosclerosis4:21 PM Santanu Chakraborty Value of CTA-Source Images in Acute Stroke using ASPECT Scoring4:28 PM Tabassum Ahmad Comparison of CTA to DSA in evaluating etiology of non-traumatic, non-subarachnoidintracranial hemorrhage4:35 PM Adrian Crawley Differences in cerebrovascular reactivity in males versus females obtained using BOLD MRIand alternating states of end-tidal pCO24:42 PM Julien Poublanc Investigation of short TE inflow-based contrast for fMRI4:49 PM Fang Liu Gender Differences in Water Diffusion of the Corpus Callosum : A Diffusion Tensor ImagingStudy4:56 PM Louis-Martin Boucher Reviewer discrepencies in analyzing CT Heads and Shunt Series in children in which VP shuntobstruction is suspected5:03 PM Sandeep Bhuta 3T Intracranial arterial wall imaging:Clinical Impact5:10 PM Jeff Mandelcorn ASPECT Scoring of CT Perfusion in Early Stroke Visualization and Assessment5:17 PM Seon Kyu Lee The Evaluation of Intracranial Atherosclerosis Using the BOLD MRI technique, Part1: BOLDMRI Mapping of Cerebro-Vascular Reserve (Cerebro-Vascular Reserve) with Carotid ArteryOcclusive Disease (CAOD): Value in Identifying Patients for Carotid Artery Revascularizations5:24 PM Peter Howard Optimizing Gadolinium Enhanced MRA for Demonstration of the Artery of Adamkiewicz5:31 PM Ryan Wada CTA can predict hematoma progression in spontaneous intracerebral hemorrhage5:38 PM Veera Bharatwal Correlation between nasal dermoid cysts and foramen cecum sizeVascular and Interventional RadiologySession Chair:5:47 PM John Kachura Radiofrequency ablation of renal cell carcinoma using a multitined electrode: preliminaryexperience5:54 PM Ronjon Raha Use of Prophylactic Antibiotics for Implanted Chest Port Insertions: Is There Reduced Risk ofInfection?6:01 PM Dheeraj Rajan Ultrahigh vs. high-pressure PTA of venous anastomotic stenosis in HD grafts: Is there adifference in patency?6:08 PM Marshall Sussman Image-Guided Navigation in the Presence of Motion6:15 PM Richard Bitar Distribution of Intraplaque Hemorrhage in Carotid Complicated Plaques Defined by High-Resolution Magnetic Resonance Direct Thrombus Imaging (hiresMRDTI)6:22 PM Murthy S.ChennapragadaDisproportionate increase in cardiac output with exercise in patients with high flow peripheralarterio-venous malformations6:29 PM Manish Taneja Iatrogenic renal trauma: Patterns of arterial injury and endovascular management6:36 PM Walter Kucharczyk Closing Comments2 of 58

University of Toronto Department of Medical Imaging Research Day 2006Abdominal, Pelvis, GI,and GU Imaging

University of Toronto Department of Medical Imaging Research Day 2006Do additional lateral cores increase prostate cancer detection compared to the standardsextant biopsy?Toi A 1 , Lockwood GA 2 , Fleshner NE 3 , Evans A 4 , Tammsalu L 1 .Departments of: 1. Medical imaging, 2. Clinical Study Coordination and Biostatistics, 3.Urology, 4. Pathology.Princess Margaret Hospital, University of TorontoPURPOSE:The literature suggests cancer yield is increased if additional biopsy cores are taken from thelateral aspect of the prostate. This has increased work and costs at a time of fiscal restrictions.We set out to determine: 1) what is the increase in cancer detection of 4 extra laterally directedcores, 2) if there is any subgroup of men that this might especially benefit, and 3) to estimateadditional costs.METHODS:A REB approved chart review was performed of all men who had sextant and laterally directedprostate biopsies. Only men presenting for initial biopsy were included. Cancer detection fromsextant and lateral cores was compared. Prognostic parameters were reviewed including age,DRE, PSA, free/total PSA ratio, TRUS and prostate volume. Costs were estimated.RESULTS:Data was available for 1010 men from March 2003 to September 2005. Patient characteristicswere: median age 63 years, median PSA 5.65, median f/t PSA 14%, DRE positive 34%; medianvolume 44cc, TRUS positive 44%. Cancer was found in 494 (48.9%). It was only medial in 107(22%); only lateral in 83 (17%), and both medial and lateral in 304 (62%). Cancer yield withpositive TRUS was 58% and with negative TRUS 30%. When medial biopsy was negative andno lateral nodule was visible, then cancer was found at lateral biopsy in 55/1010 (5.4%) of men.However it was clinically significant in only 30/1010 (3%) [Gleason > 7, or > 5% of any onecore]. Positive lateral biopsy was highly predicted by TRUS findings, slightly by PSA levels andnot by any other factors.In the pathology department, technologist time increased by 80%, and MD time by 30%. Overallcosts were estimated to have increased by about 40-50%.CONCLUSIONS:Sextant + targeted biopsy found 97% of significant cancers. Four additional laterally placedsamples in the presence of a negative TRUS resulted in 5.5% increased cancer detection but only3.0% was clinically significant. Apart from TRUS, no other factor predicted lateral cancer. Costsincreased 40-50%.4 of 58

University of Toronto Department of Medical Imaging Research Day 2006Diffusion Weighted Imaging for Localization of Prostate CancerM A Haider, MD, Toronto, ON; T van der Kwast; J Tanguay, MD; A Evans, MD; A T Hashmi;GLockwood; J Trachtenberg, MDPURPOSETo determine the performance of diffusion weighted imaging (DWI) in the localization of prostatecancer.METHOD AND MATERIALST2 weighted and DWI (b-val 600 s/mm 2 , TR/TE 4000/73ms, BW 167kHz; FOV 14cm, slicethickness 3mm; NEX 1, matrix 256x128) was preformed in 33 patients prior to radicalprostatectomy using an endorectal coil (Medrad, PA) and a 1.5T MRI system, (Excite HD, GEMedical Systems, WI). The peripheral zone (PZ) of the prostate was divided into six regions. Thetransition zone (TZ) was divided into left and right. Cancer was considered to be present in aregion if a tumor larger than 1ml in volume was seen on step section whole mount pathologyreconstructions. Apparent diffusion coefficient (ADC) maps were calculated from the DWI andreviewed at a fixed window width and level. T2 images and ADC maps were reviewedindependently. On imaging cancerous regions were defined as areas of focal low signal on T2weighted imaging or ADC maps. Each region was rated on a 4 point scale from definitely notcancer to definite cancer.RESU LTSThe area (Az) under the receiver operator characteristic (ROC) curve was significantly higher(p

University of Toronto Department of Medical Imaging Research Day 2006Enhancement Patterns of Hepatocellular Carcinoma on Contrast-enhanced Ultrasound –Correlation with Pathologic DifferentiationHyun-Jung Jang, Tae Kyoung Kim, Peter N Burns, Stephanie R WilsonDepartment of Medical Imaging, University of TorontoPURPOSE:To describe the enhancement patterns of hepatocellular carcinoma (HCC) on real-time contrastenhancedultrasound (CEUS).MATERIALS & METHODS:This study was Research Ethics Board approved. Informed consent was obtained. The studypopulation includes 112 patients (91 men and 21 women; 25 - 86 years old) with 120histologically proven HCCs - 23 well-differentiated, 84 moderately-differentiated and 13 poorlydifferentiated.All had continuous real-time low-mechanical index (MI) CEUS from wash-in ofcontrast to 300 seconds. Prospective analysis included: arterial enhancement, intratumoraldysmorphic arteries, heterogeneity, and the presence and time of negative enhancement(washout). Enhancement patterns were correlated with pathologic differentiation using Fisher’sExact test.RESULTS:In the arterial phase, 87% (104/120) showed hypervascularity, with significantly higherfrequency in moderately- (95%, 80/84) compared with well- (61%, 14/23) and poorlydifferentiatedHCC (77%, 10/13). Nine (9/120, 8%) were isovascular and seven werehypovascular (7/120, 6%). 88/120 HCCs (73%) demonstrated dysmorphic arteries. Of 104hypervascular tumors, only 42% (44/104) showed typical washout by 90 seconds. Continuedevaluation showed late washout in 27/104 HCC (26%) between 91-180 seconds and in 24/104(23%) between 181-300 seconds. The remaining nine (9%, 9/104), mostly well-differentiatedHCCs (7/9, 78%), showed sustained enhancement up to 300 seconds.CONCLUSIONS:While most HCCs with moderately-differentiated pathology show typical arterialhypervascularity, there is a subgroup without hypervascularity. Extended observation in theportal phase is important as late washout occurs slightly more frequently than washout in theconventionally defined portal venous phase. Further, sustained enhancement should not beconsidered diagnostic of a benign lesion, since it may occur, especially in well-differentiatedHCC.6 of 58

University of Toronto Department of Medical Imaging Research Day 2006Comparison Of CT Histogram Analysis To Adrenal Washout CT In The Diagnosis OfLipid Poor AdenomasJhaveri KS, Lad SV, Haider MAObjective:CT histogram analysis has been proposed as a simple post processing technique to diagnose anadrenal adenoma by demonstration of significant intralesional negative pixels (typically greaterthan 5% or 10%). The purpose of this retrospective pilot study was to assess the potential of CThistogram analysis in diagnosing an adrenal adenoma within the lipid poor category incomparison to adrenal washout CT.Materials & Methods:22 patients (age range: 51-74 years, 10 females and 12males) with 24 lipid poor adrenaladenomas (size range: 13-42mm) were identified from the radiology reporting database. Lipidpoor adenoma was defined as a nodule having unenhanced CT attenuation more than 10 HU butshowing 60% or greater absolute enhancement washout on adrenal washout CT. Adrenalwashout CT had been performed in all cases by obtaining 2.5mm CT slices on a multislice CTscanner during unehanced, 60-second post contrast and 15 minute delayed scans. Histogramanalysis was performed as post processing on a PACS workstation on the unenhanced CT phase.A region of interest (ROI) was drawn within the nodule recording mean attenuation, number ofpixels and percentage of negative pixels. A diagnosis of adenoma was made if there was agreater than 5% negative pixel threshold on the histogram analysis.Results:The mean attenuation in all nodules was > 10 HU (range 12-30HU) and absolute enhancement %washout was 60 or greater (60-79%) indicating a diagnosis of lipid poor adenoma by adrenalwashout CT.CT Histogram analysis revealed negative pixels in all 23/24 adenomas. The negative pixel %ranged from 3% to 20%. Of the 24 adenomas, 22(91.6%) showed > 5 % negative pixels and 17of 24(58.3%) showed >10% negative pixels.There was good correlation between mean attenuation and % negative pixels (Pearson's productmomentcorrelation = -0.85, 95 % confidence interval: -0. 96, -0.52, p-value = 0.0008) and therewas no significant correlation between the absolute enhancement % washout and the % negativepixel values. (Pearson’s product-moment correlation = -0.39,95 % confidence interval: -0.8,0.28,p-value = 0.2406)Conclusion:The preliminary results are encouraging for the application of CT histogram analysis incharacterization of lipid poor adrenal adenomas. If this can be confirmed in larger sample sizestudies there is a potential for reducing the actual number of adrenal washout CT examinationsfor diagnosis of lipid poor adenomas.7 of 58

University of Toronto Department of Medical Imaging Research Day 2006MRI imaging response of adenomyosis to uterine artery embolizationRobinson E, Wu LPurpose:Adenomyosis is a benign entity characterized by the presence of ectopic endometrial tissue within themyometrium. Symptoms include abnormal bleeding, pelvic pain, and pelvic pressure. Definitivetreatment is hysterectomy, but less invasive methods such as uterine artery embolization are beingconsidered, given its effectiveness in treating fibroids. Currently, UAE appears to offer encouraging shortterm results with regards to symptom control, however the long term results remain to be elucidated. Wesought to assess our experience with patients undergoing UAE at St. Michael’s Hospital, to determine theimaging and clinical response of adenomyosis to embolization.Materials and Methods:151 consecutive patients undergoing UAE since July 2004 underwent pre-UAE MRI (ax T1 FFE, axT12DFS SENSE, AX T2 TSE, sag T2 TSE, sag dynamic post gad. Adenomyosis was defined as junctionalzone thickness greater than 12mm. These patients were subdivided into diffuse symmetric, diffuseasymmetric, and focal adenomyosis. All patients completed a symptom chart (scale 1-10) documentingthe extent of their pelvic pain, bleeding, and pelvic pressure. The patients underwent UAE according tostandard protocol. They were re-imaged with MRI at 6 months post procedure, and completed a followupsymptom chart. The subgroup with adenomyosis (13 adenomyosis plus fibroids, 2 adenomyosis only)were analyzed for post- embolization change in uterus volume, fibroid volume, junctional zone thickness,junctional zone to myometrial ratio, enhancement pattern of adenomyosis prior to and followingembolization, and development of devascularization post-UAE. To assess the significance of thechanges, paired Student’s T tests were performed. To evaluate the changes in enhancement andsymptoms, the Mann Whitney U test was used.Results:15 patients (mean age 43 years) were found to have adenomyosis. They underwent pre-procedure MRIon average 14.7 days before the embolization, and 6 month follow up MRI on average 170 days postUAE. Of the adenomyosis only group , 1 patient had diffuse symmetric and one patient diffuseasymmetric adenomyosis. Of the group with combined adenomyosis and fibroids, 2 had diffusesymmetric, 8 diffuse asymmetric, and 3 focal adenomyosis. Pre-UAE uterine volume decreased 34%from 570 cm 3 to 376 cm 3 , a 34% decrease (P

University of Toronto Department of Medical Imaging Research Day 2006Hepatic and Portal Venous Thrombosis Associated with Hepatic AbscessSyed MA, Kim TK, Jang HJAbdominal imaging – University health network, Toronto, Ontario CanadaOBJECTIVE:Association of venous thrombosis and hepatic malignancy, such as hepatocellular carcinoma, iswell established. However, liver abscesses sometimes mimic malignant liver tumors on CT, bydemonstrating a solid-looking appearance with venous thrombosis. Our aim was to determine theincidence of hepatic and portal venous thrombosis associated with hepatic abscess and todescribe the CT findings.MATERIALS AND METHODS:Over a 5-year period, computerized search identified 67 patients (44 men and 23 women; meanage, 55 years) with liver abscess who underwent single-phase (n = 30) or triphasic (n = 37)contrast-enhanced CT. Images were retrospectively reviewed for the presence of venousthrombosis, any associated regional parenchymal attenuation changes, and outcome on follow-upCT.RESULTS:Twenty-eight patients (42%) had venous thrombosis, involving portal vein (PV) in 16 (57 %)and hepatic vein (HV) in 15 (53%); three (11%) had both PV and HV thrombosis. Thrombosiswas identified as non-enhancing linear or branching structures within the veins in all cases. InPV thrombosis, regional parenchymal attenuation during the portal venous phase washyperattenuating in 10 (63%), or isoattenuating in 6 (38%). In HV thrombosis, regionalparenchymal attenuation during the portal venous phase was hypoattenuating in 10 (87 %),hyperattenuating in 1 (7%), or isoattenuating in 1 (7%) (P < .001). Regional parenchymalattenuation changes during the arterial phase were variable. Follow-up contrast-enhanced CTscan (1-30 months; mean 15 months) was available in 27 (96%); venous thrombosis partiallyresolved in 17 (63%), including 13 with PV thrombosis and 4 with HV thrombosis, andcompletely resolved in 10 (37%). There was interval hepatic parenchymal atrophy in 4 withpersistent PV thrombosis.CONCLUSION:Hepatic abscesses are frequently associated with both hepatic and portal venous thrombosis,which is seen as non-enhancing linear or branching structures on CT. Regional parenchymalhypoattenuation during the portal venous phase is only seen in HV thrombosis. PV thrombosis isfrequently associated with regional parenchymal hyperattenuation and more commonly persistson follow-up CT.9 of 58

University of Toronto Department of Medical Imaging Research Day 2006Does Radiology Consultation Pre-Imaging Affect the Outcomes of CT Renal Colic Scans?Yeung B., Wu L., Hamilton P.Department of Medical Imaging, St. Michaels’ Hospital and Sunnybrook Women’s Hospital,University of TorontoPurpose:Since the mid 1990s, CT scan for renal colic has been used in the emergency department. Inonly 33-55% of patients who have these scans do we see evidence of renal colic. In the currentinvestigation of renal colic overnight at St Michael's Hospital (SMH), the emergency physicianis able to order a CT scan without radiology on call consultation. At the Sunnybrook Hospital(SH), the emergency physician must consult with radiology on call to determine if CT is themost appropiate imaging test at that time. We sought to determine if there is any significantdifference between the number of postive CT scans for renal colic between the two hospitals.Materials and Methods:The reports of CT scans for renal colic on call over a six month period were reviewed for bothhospitals. The results of the CT scans were examined and divided into three categories; positiverenal colic scan, no renal colic but an alternate cause for pain, and no cause for pain found.Statistical analysis was performed to determine if there were any significant differences betweenthe two hospitals.Results:A total of 149 CT scans from SMH and 73 CT scans from SH were performed on call for renalcolic over a six month period. CT scans were positive for renal colic in 67 (45%) patients atSMH and in 46 (63%) patients at SH reflecting a significant difference (p

University of Toronto Department of Medical Imaging Research Day 2006Focal Nodular Hyperplasia and Hepatic Adenoma: Differentiation with Low-MI ContrastenhancedUltrasound – Work in ProgressTae Kyoung Kim, MD 1 , Hyun-Jung Jang, MD 1 , Peter N. Burns, PhD 2 , Jessica Murphy-Lavallee,MD 1 , Stephanie R. Wilson, MD 11 Department of Medical Imaging, Toronto General Hospital, University of Toronto, Toronto,Ontario, Canada2 Medical Imaging research, Sunnybrook and Women’s Health Sciences CenterPurpose:To differentiate focal nodular hyperplasia (FNH) and adenoma on contrast-enhanced ultrasound(CEUS).Materials and Methods:Over a 6-year period, 188 patients with a hypervascular mass on CEUS without features ofhepatoma or hemangioma were identified. Differential diagnoses included FNH or adenoma,confirmed in 62 patients (43 FNH, 19 adenomas). CEUS was performed with low-MI continuouspulse-inversion imaging after multiple bolus injections (mean, 5) of Definity (Bristol-MyersSquibb). In 20/62 patients (32%) who were imaged with an Aplio (Toshiba) scanner, maximumintensity projection (MIP) processing with replenishment technique was used. Video clips andstatic images of these patients were retrospectively reviewed by two blind reviewers. Arterialphasefilling patterns (centrifugal/centripetal/diffuse), transient non-enhancing halo, stellatearteries, central non-enhancing scar/necrosis, and enhancement relative to liver in the portalvenous phase were documented.Results:Centrifugal filling was more common in FNH (91% and 74% for the two reviewers,respectively) than adenoma (16% and 16%) (P < .001 for both reviewers) (κ = .763). Stellatearteries were also more common in FNH (67% and 60%) than adenomas (16% and 11%) (P

University of Toronto Department of Medical Imaging Research Day 2006Emergency CT Imaging for Acute Abdominal Aortic Disease - Are Utilization PatternsChanging?Rosta, N.I.N. MD, CCFP, Chawla, P.T. MD, FRCP(C); April 2006Introduction: In the autumn of 2004, the emergency physicians at the University HealthNetwork (UHN) were given permission to order CT studies to exclude abdominal aorticaneurysms without discussing the indication for the study with a radiology resident first.Previous utilization studies conducted by residents in our programme (Kirpalani, A, 2004, Lee, T2005) which analyzed the emergency department (ER) use of CT (in University of Torontoteaching hospital ERs) to rule out renal colic and pulmonary embolism respectively, determinedthat the number of these examinations had increased, in the case of CT pulmonary angiogramsby as much as 30% over 3 years.This study analyzes the utilization of CT to rule out acute abdominal aortic disease byemergency physicians at the Toronto General Hospital (TGH) before and after the policy waschanged.Methods: A systematic review of all reports of CT abdomen and/or pelvis studies, performed inthe first six calendar months of the years 2001 through 2005, ordered by staff emergencyphysicians at the TGH, and documented in the Radiology Information System (RIS) at theUHN, was conducted by the lead investigator. All such CT studies, for which the clinical reasonindicated any concern about, or desire to exclude abdominal aortic pathology, were considered tobe a CT examination to exclude AAA (CT AAA). Positive CT AAA studies were defined asthose in which the report mentioned definite or suspected (1) abdominal aortic aneurysm(diameter ≥ 5.0 cm), (2) aortic dissection which involved the abdominal aorta, (3) aortic fistulasto the GI tract or to the venous vascular system, or, (4) aortic rupture (either contained or freerupture).Results: 4801 abdominal/pelvic CT study reports were reviewed. The absolute percentage ofTGH ER patients being sent for CT abdomen/pelvis for any indication was 3.6% in 2001, 3.3%in 2002, 4.0% in 2003, 4.5% in 2004, and 4.4% in 2005. This represents a statistically significantrelative increase year over year of 7.5% (odds ratio 1.075 (1.046 – 1.105), p

University of Toronto Department of Medical Imaging Research Day 2006Hepatic vein transit times using a microbubble agentto predict severity of hepatic disease non-invasivelyAn Tang, MDTae Kyoung Kim, MDHyun-Jung Jang, MDJenny Heathcote, MDMaha Guindi, MDRaffi Karshafian, MScPeter N Burns, PhD 4Stephanie R. Wilson, MDPurposeCirrhosis is associated with several intrahepatic and extrahepatic hemodynamic changes.Previous studies proposed the measurement of hepatic vein transit time (HVTT) as a noninvasiveindex of diffuse liver disease severity. We sought to validate the use of this index in acohort of patients with contemporaneous biopsy correlation in a variety of chronic liver diseases.Materials and MethodsThirty-four consecutive patients (mean age 49, 17 M) with clinical suspicion of fibrosis werestudied prospectively (17 HBV, 7 HCV, 5 PBC, 2 autoimmune hepatitis, 3 others). The 10 firstpatients were scanned twice by two different radiologists in a randomized sequence to evaluatethe reproducibility of the technique. After an overnight fast, the patients were scanned in pulseinversionmode (iU22, Philips, Bothell, WA) during repeated bolus injections of microbubbles(Definity, BMS Imaging, Boston MA). The time-intensity curves in an hepatic artery and veinwere analyzed with quantification software (Q-Lab, Philips, Bothell, WA). HVTT was defined asthe time from antecubital vein injection to sustained rise in Doppler signal >10% above baseline;hepatic artery transit time (HATT) as the similar arrival time in the hepatic artery. The transittimes were correlated with fibrosis stage from liver biopsies performed within a short timeinterval (mean 7.6 days).ResultsTransit time was measured successfully in 33/34 patients, with good reproducibility (IntraobserverICC on HVTT = 0.86; inter-observer ICC = 0.62).Mean (SEM) HVTT for absence/minimal fibrosis (n=14) ), moderate fibrosis (n=15) andcirrhosis (n=5) groups were 21.4 (2.4), 20.9 (1.7), 18.1 (2.0) seconds respectively.Mean (SEM) HATT for the absence/minimal fibrosis, moderate fibrosis and cirrhosis groupswere 10.5 (1.6), 9.6 (0.8), 9.4 (1.5) seconds respectively.The very low correlation between HVTT and fibrosis (r = -0.05) and HATT and fibrosis (r = +0.09) prevented us from validating previously published threshold values to distinguish mildfrom advanced fibrosis.ConclusionThis study does not confirm earlier appearance of contrast in cirrhosis and does not validateHVTT as a reliable non-invasive surrogate marker of liver fibrosis.13 of 58

University of Toronto Department of Medical Imaging Research Day 2006Impact of Contrast Enhanced Ultrasound in a Tertiary Clinical Practice.Lanka B, Jang HJ, Kim TK, Wilson SDepartment of Abdominal Imaging, Toronto General Hospital, UHNPURPOSETo assess the impact of introduction of contrast enhanced ultrasound (CEUS) in a tertiary clinicalpractice.METHODSFrom July 2003 to December 2005, 900 consecutive CEUS performed as part of routine carewere reviewed to determine the source of examination - incidental mass on baseline or clinicianreferral. The clinical impact was evaluated as positive impact (shortened time to diagnosis,reduced hospital visits, decreased referrals for further imaging, diagnosis of indeterminatemasses on other imaging, and guidance for therapy), concordant diagnosis, and negative impact.RESULTS255/900(28.3%) examinations were motivated by incidental detection of a mass at baselineultrasound, 104/255(40.8%) in high risk patients and 151/255(59.2%) in no risk patients.645/900(71.6%) patients were referred from clinicians for characterization of a mass fromprevious imaging, 445/645(69%) inside our institution and 200/645(31%) from outside scans,and for monitoring radiofrequency ablation in 45/645(6.9%). The number of patients referredfrom clinicians increased to 191 in the last six months of the study compared from 56 in the firstsix months. A significant positive impact from CEUS was found in 391/900(43.4%) patientsincluding reduced referral for other imaging in 205/900(22.8%) patients; 72.1% had no furtherimaging and 27.8% had interval follow-up. Decreased time to diagnosis occurred in329/900(36.5%) patients when CEUS was performed on the first visit and a specific diagnosismade; 136/329(41.3%) had no further imaging for confirmation. CEUS achieved the correctdiagnosis in 827/874(94.6%) known final diagnoses and significantly contributed to correctdiagnosis and alteration of patient management in 169/900(17.8%) patients. Concordant imagingwith either CT or MR occurred in 509/900(56.5%) patients. In 3(0.3%) patients a negativeimpact was present (delayed/wrong diagnosis with incorrect management).CONCLUSIONReferrals for CEUS have increased progressively. CEUS has a positive impact on patientmanagement providing rapid accurate diagnosis of incidentally detected masses and resolvingnodules found on surveillance scans and indeterminate on other imaging14 of 58

University of Toronto Department of Medical Imaging Research Day 2006Pheochromocytoma and paraganglioma: Risk of adverse events with IV nonionic contrastmaterial for CT.Bessell-Browne RJ, O’Malley MEDivision of Abdominal Imaging, Department of Medical ImagingUniversity Health Network and Mount Sinai HospitalPurpose:CT is widely used for assessing abdominal masses and localizing suspected pheochromocytomasand paragangliomas. Many institutions only perform unenhanced CT if a pheochromocytoma orparaganglioma is suspected due to a perceived risk of inducing a hypertensive crisis with IVcontrast. If IV contrast is administered in these patients, alpha-blockers may be given to preventa hypertensive crisis. The purpose of this study was to review our experience with theadministration of IV nonionic contrast for CT scanning in patients with pheochromocytomas orparagangliomas, specifically to determine if any adverse events occurred.Materials and methods:A retrospective review of radiology records, clinical notes, biochemistry and pathology resultsidentified 25 patients with 40 pathologically proven pheochromocytomas or paragangliomas whohad received IV nonionic contrast for 28 CTs at our institution. There were 15 adrenalpheochromocytomas, 12 abdominal paragangliomas, 5 neck paragangliomas, 7 metastases andone primary liver paraganglioma. Medical records and incident reports were reviewed foradverse events, particularly hypertensive crisis. Imaging features of the lesions and clinical andbiochemical evidence of catecholamine production was recorded. A PubMed literature searchwas performed for reports of hypertensive crisis related to IV nonionic contrast administration.Results:Twenty-two of the 25 patients were considered to have functional tumors. Elevated urinarycatecholamines and/or their metabolites were found in 19 of 22 patients tested. Hypertension wasdocumented in 16 patients and adrenergic symptoms were present in 10 patients.Pheochromocytoma or paraganglioma were clinically unsuspected in 12 patients at time ofimaging. No patients were on alpha blocking medications that would have preventedcatecholamine induced effects at the time of IV contrast administration. No adverse eventsrelated to IV contrast enhanced CT were documented. No reports were found in the literature ofa hypertensive crisis induced by IV nonionic contrast for CT scanning.Conclusion:There is no evidence that there is an increased risk of an adverse event, specifically ahypertensive crisis, in patients with pheochromocytomas or paragangliomas related to theadministration of IV nonionic contrast for CT even in patients who are not taking alphablockingmedication.15 of 58

University of Toronto Department of Medical Imaging Research Day 2006Testicular Cancer Surveillance: Standard versus Low Dose CTO’Malley ME, Jang HJ, Jhaveri K, Khalili K, Haider MA, Panzarella A, Anson-Cartwright L,Warde P, Chung PPurpose:To compare image quality and acceptability of a low dose CT protocol with a standard dose CTprotocol in patients on surveillance for recurrent testicular cancer.Materials and Methods:This was a prospective study approved by our Research Ethics Board. Informed consent wasobtained. All patients were scanned on an Aquilion 64 scanner. No IV contrast was given. Thestandard dose CT protocol included: acquisition 64 x 0.5mm; tube rotation 0.5s; table speed26.5mm/rotation; kV 120; automatic exposure control with minimum/maximum mA 40/440;noise SD 15; reconstruction algorithm FC04; and slice thickness 5/2.5mm. The low doseprotocol was the same except: minimum/maximum mA 20/220 and noise SD 25.5. For each CTscan, the Dose Length Product (DLP) was recorded. The annotation was removed from each setof CT scan images, and each CT was randomly assigned a case number. Three readersindependently evaluated the images for: image noise and diagnostic quality using a 5-point scale;and acceptability. The presence of enlarged lymph nodes (>10mm short axis) was recorded.Results:Twenty-five patients (21-61 years, mean 40 years) had standard and low dose CT scans. Twentypatients had seminoma and 5 patients had non-seminomatous germ cell tumor of the testis. Forstandard dose images, the mean image noise and diagnostic quality was scored as 3.16, 3.28,3.64 and 3.68, 3.72, 3.60 for each reader. For low dose images, the mean image noise anddiagnostic quality was graded as 2.04, 2.44, 2.36 and 3.28, 3.08 and 2.76 for each reader. Thedifference between image noise and diagnostic quality was statistically significant for eachreader when comparing standard and low dose images (P < 0.006). All standard and low dose CTscans were considered acceptable. Two patients had lymphadenopathy indicating recurrentdisease, identified by all 3 readers on both standard and low dose CT scans. The mean DLPs forstandard and low dose CT were 1201.3 and 532.0 mGy.cm. The mean dose reduction with lowdose CT was 56%.Conclusions:Although image noise and diagnostic quality were significantly better with standard dose CT, alllow dose CT scans were considered diagnostically acceptable and recurrent disease wasidentified by all reviewers on the low dose CT scans.16 of 58

University of Toronto Department of Medical Imaging Research Day 2006Can CT reliably distinguish moderately from poorly differentiated hepatocellularcarcinoma?Staunton M, Azadeh M, O'Malley ME, Jang H, Guindi M, Khalili K.Department of Medical Imaging, UHN and Mount Sinai Hospital, University of Toronto.PURPOSETo determine if clinical parameters or CT appearances of hepatocellular carcinoma (HCC) canpredict differentiation or recurrence.METHOD AND MATERIALSA retrospective study of adult patients with HCC who had surgery or transplantation in 2004 wasperformed. Fifty tumors in 43 patients fit the study inclusion criteria: (a) pre-operative triphasicCT with visible tumor, (b) tumor specific correlation on pathology. Exclusion criterion: Priorlocal therapy of HCC. Clinical data extracted: age (mean 58.3, range 38-73), sex (10 female, 33male), ethnicity (24 Caucasian, 19 Asian), etiology (Hep B 19, Hep C 12, Etoh 10, other 8),alphafetoprotein (Afp, median 35, range 3 cm (8/18) vs.tumor 3cm or smaller (4/32), p=0.017, OR 5.6 (p=0.028). No other significant correlation wasnoted between clinical parameters and tumor recurrence or differentiation. Tumor recurrencewas seen in 7/35 well-mod diff. tumors and 5/15 poorly diff tumors, the difference not beingsignificant (p=0.47). Imaging parameters: Interobserver agreement determined by weightedkappa for precontrast, arterial, and venous phase were 0.87, 0.48, and 0.72 respectively. Nosignificant correlation was noted between attenuation and tumor recurrence or differentiation.Vascular invasion, definition of margins, heterogeneity, and shunting also did not correlate withoutcome variables. Mean growth rate of tumor:0.45 cm/90 days (range 0-1.3); recurrence seenin 3/12 patients above and 1/11 below median rate (p=0.59).CONCLUSIONCT features of HCC do not appear to reliably predict HCC differentiation or recurrence. Size >3cm and hepatitis B infection increase risk of HCC recurrence.17 of 58

University of Toronto Department of Medical Imaging Research Day 2006Breast, Chest,and Cardiac18 of 58

University of Toronto Department of Medical Imaging Research Day 2006Digital Mammographic Imaging Screening TrialRoberta A Jong MD, FRCPCPurpose: To determine the accuracy of digital versus film/screen mammography in detectingbreast cancer in asymptomatic women.Method: There were 33 sites including academic and community hospitals and clinics in theUnited States with sites at both campuses of the Sunnybrook & Women’s College HealthSciences Centre. Asymptomatic women presenting for screening mammography were recruited.To be eligible they could not have breast implants or a prior history of breast cancer treated bylumpectomy. The manufacturers of 4 different digital systems participated. Women wererandomized to have either the film or digital examination first. These were performed by thesame technologist at the one visit. Two different radiologists independently reported theexaminations with the clinical report including both interpretations. The report included a 7-point scale of malignancy that allowed the Reader Operating Characteristic (ROC) analysis to bedone. Levels of suspicion of 4, 5, 6 or 7 were positive. The Breast Imaging Reporting and DataSystem (BIRADS) was used with 0, 4 or 5 considered positive. The age, breast density,menopausal status, race, risk of breast cancer and machine type were recorded. A positivediagnosis was confirmed by pathology within 455 days of entry. For a negative result, there wasno cancer diagnosed by biopsy or a negative follow-up mammogram at 1 year.Results: 49, 528 women were recruited over 2 years. Of these there were 42,760 women forwhom all data was available. Analysis could not be done on those who were ineligible,withdrew, had technically inadequate examinations or where the same radiologist read bothexaminations, follow-up was lacking or the patient died without a diagnosis of breast cancer. Onthe 7 point scale, 0.5% were positive on both film (F) and digital (D). 2.2% were positive onlyon D and 1.9% were positive only on F. The vast majority 94.8% were negative. Using theBIRADS scale 2.9% were positive on both D & F. 5.6% were positive only on D and 5.7% werepositive only on F. 85.8% were negative on both. 335 cancers were diagnosed within 455 dayswith 254 cancers diagnosed within 365 days. The cancer stage and histology were the same forboth technologies. Using the ROC analysis area under the curve (AUC) there was no significantdifference in diagnostic accuracy between digital (0.78 ± 0.02) and film (0.74 ±0.02) over thewhole screened population. There was no difference related to race, breast cancer risk ormachine type.There was a significant improvement in performance with digital for women < 50 years of age,with heterogeneously or extremely dense breast tissue or who were pre- or peri- menopausal.The overlap of these groups is currently being analyzed. Neither modality detected all cancers.Conclusion: There is no significant difference in diagnostic accuracy between film and digitalmammography for the whole screened population. However for women under 50 years of age,with heterogeneously or extremely dense breasts or who are pre- or peri-menopausal digitalmammography performed significantly better.19 of 58

University of Toronto Department of Medical Imaging Research Day 2006Characteristics of Breast MRI Screen-detected lesions that should be sent for targetedUltrasoundMoscovici O. 1* ; Causer P.A. 1 ; Jong R.A. 1 ; Hill K. 2 ; Warner E. 2 ; Plewes D.B. 3 ; 1. MedicalImaging, Sunnybrook and Women's College Health Sciences Centre, Toronto, Ont, Canada; 2.Medical Oncology, Sunnybrook and Women's College Health Sciences Centre, Toronto, Ont,Canada; 3. Medical Biophysics, Sunnybrook and Women's College Health Sciences Centre,Toronto, Ont, Canada.Object or Purpose of Study:To determine which MRI lesions detected in a high risk screening population should be referredfor second look ultrasound.Materials, Methods and Procedures:During the period from July 2001 to May 2005, breast MRI screening detected 159 BI-RADScategories 0, 4 and 5 lesions in 108 women at high risk for breast cancer. Targeted ultrasoundwas recommended as a part of the diagnostic workup. Included were 116/159 (72%) BI-RADS 0,4 or 5 lesions that underwent targeted ultrasound. Diagnostic follow-up MRI was performed onthe same date for 104/116 (90%) of these lesions. Additional inclusion criteria included: tissuesampling of the lesion or a minimum 1 year follow-up MRI documenting stability or resolutionof the presumed benign lesion. The sample size yielded 99/116 (85 %) lesions. The screeningbreast MRIs and targeted sonograms were retrospectively reviewed and the lesions classifiedusing MRI BI-RADS descriptors and size. The number of MRI lesions which weresonographically visible and the number of cases with a histopathologic diagnosis weredetermined. MRI lesion characteristics, size, benign and malignant pathology and benignitybased on follow-up were correlated to the presence or absence of a sonographically visible lesionusing logistic regression analysis.Results:The MRI morphologic characteristics of the 99 lesions were: 45 (45.45%) mass, 41 (41.41%)focal area and 13(13.13%) linear/ductal. Mean lesion size of the sonographic visible finding was8.1 mm [ the mean value of the MRI visible finding was11.65 mm ]. 34 (34.34%) lesions weresonographically visible. 40(40.40%) were benign based on follow-up studies, 32 (32.32%)benign and 27 (27.27%) malignant based on histopathology. Lesions either characterized as afocal area of enhancement or due to benign disease based on follow-up were significantly lesslikely to be sonographically visible (p=0.002). There is increased likelihood of mass lesions to bedepicted by ultrasound [p =0.0018], disregarding the lesion’s etiology [malignant / benign].Significance of the Conclusions:Targeted ultrasound can be omitted from the diagnostic algorithm of:1) breast MRI screen-detected lesions characterized as focal2) most linear/ductal lesions3) lesions which either resolve or are classified as likely benign after evalution with diagnostic.20 of 58

University of Toronto Department of Medical Imaging Research Day 2006Outcome of Breast MRI for screening breast cancer in mixed population of womenBhaduri, MousumiDepartment of Medical Imaging, UHN and Mt. Sinai Hospital, University of TorontoPurpose:Breast MRI has emerged as an important tool in the screening of high-risk women.The purpose of this retrospective study was to determine the factors that affect the outcome ofscreening breast MRI examinations.Materials and Methods:The review of MRI database and clinical charts revealed 534 screening breast MRI examinationsin 440 asymptomatic women (age range 22-78, mean age 47 years) with normal mammographicfindings. The predictive value of the following independent variables was analyzed: individualbreast cancer risk, mammographic density, age and number of screening round. The outcomevariables were call back rate, biopsy rate and cancer detection rate.ResultsCall-back rate after screening MRI was 22% (120 of 534 studies). Recall rate was significantlyhigher among women at first screening MRI round (p

University of Toronto Department of Medical Imaging Research Day 2006Assessing the performance of CAD for lung nodule detection: effect of radiation dosePereira AM, Paul NS, Sitartchouk I, Bayanati H, Kashani H, Roberts HDepartment of Medical Imaging, University of TorontoPURPOSETo assess the effect of radiation dose on the performance of a computer-aided detection (CAD)software for lung nodule detection.METHOD AND MATERIALSWe retrospectively analyzed the CTs of 60 sarcoma patients, which after IRB approval andconsenting consisted of three different data acquisitions. All CTs had been performed on a 16row MDCT (Somatom Sensation 16, Siemens), with 120 kV and 5 mm collimation, withstandard dose 240 mA (SD), low-dose 80 mA (LD), and minimum dose 40 mA (MD). Both softtissue and lung reconstruction of each of those 3 data sets were analyzed with a commerciallyavailable CAD system (LungCAR version 1.2 build 3.1, Medicsight, London, UK), using defaultsoftware settings. All CAD marks were reassessed and assigned as true positive TP (solid nodulefound by CAD and confirmed by the radiologist) or false positive FP (any CAD mark that wasnot a parenchymal nodule). Any nodule not marked by CAD was regarded as false negative FN.McNemar's test has been applied to assess differences between the series.RESULTSThe number of nodules present varied slightly for the different data acquisitions and ranged from274 to 278. For nodules of all sizes, CAD sensitivity was 30% in the SD soft tissue series, 31%in the lung series; CAD sensitivity in the LD soft tissue series was 35%, in the lung series 31%;and CAD sensitivity in the MD soft tissue series was 32%, in the lung series 32%. For nodules 5mm -15 mm, CAD sensitivity in the SD was 51% for soft tissue and 56% for lung reconstruction,in the LD was 56% for soft tissue and 54% for lung reconstruction, and in the MD was 50% forsoft tissue and 51% for lung reconstruction. These results were not statistically significantlydifferent. The average number of FP/case in soft tissue and lung reconstruction was 3.07 and3.18 in SD, 3.03 and 3.43 in LD and 3.33 and 3.93 in MD. Almost all FPs are easily dismissible(45% vessels, 19% bone, 16% normal pleura).CONCLUSIONWhen compared to standard dose CT, the use of lower doses (LD or MD) does not significantlyimpair CAD performance: there is the tendency towards lower sensitivity and more FP in MD. Inseries obtained with identical radiation dose, there was no statistically significant differencebetween sensitivity in soft tissue and lung algorithms; the number of false positives is slightlyhigher in lung reconstructions.CAD performance is not significantly impaired by a decrease in radiation dose, allowing lowdoseregimens for nodule detection.22 of 58

University of Toronto Department of Medical Imaging Research Day 2006Low-dose Computed Tomography in Prior Asbestos-exposed Workers:Assessment of pleural plaquesHamid Bayanati, Heidi Roberts, Andre Pereira, Igor Sitartchouk, Ashwini Kale, Marc dePerrot,Karl Uy, Narinder Paul, Michael JohnstonPurpose:Asbestos exposure is a known risk factor for lung cancer and malignant pleural mesothelioma(MPM). We use low–dose computed tomography (LDCT) for screening, and report on thepleural abnormalities.Methods:Since May 2005, we have enrolled 253 individuals with asbestos exposure at least 20 years agoand/or documented pleural plaques. A helical LDCT is performed (1.25 mm, 60 mA, 120 kV).We note the appearance of nodules (not reported here) as well as the extent, location, shape,thickness and calcification of pleural plaques. Nodules

University of Toronto Department of Medical Imaging Research Day 2006Dynamic CT perfusion for Lung Nodules CharacterizationI Sitartchouk, H Roberts, A M Pereira, H Bayanati, T Waddell, T P RobertsDepartment of Medical Imaging, University of TorontoPURPOSETo evaluate the feasibility of perfusion-sensitive dynamic contrast-enhanced computedtomography (DCE-CT) in distinguishing malignant and benign pulmonary nodules.METHOD AND MATERIALSAfter IRB approval, 48 patients were included in the study. 19 were scheduled for a CT guidedbiopsy, 27 were scheduled for surgical resection, and 2 had a solitary pulmonary nodule underconservative surveillance. A cine CT study, without table movement, is acquired during theadministration of 50 mL contrast agent (flow rate 3-5 mL/sec), for approximately 25 sec, with~1sec temporal resolution. Data analysis is performed with the CT Perfusion 3 software(Advantage Windows, General Electric Medical Systems, Milwaukee, WI), yielding maps andROI estimates of fractional blood volume (BV, ml/100g), blood flow (BF, ml/100g/min), meantransit time (MTT, 1/sec) and microvascular permeability (permeability surface area product, PS,ml/min/100g). Results from the DCE-CT are correlated with the histological/cytological resultsfrom the nodule.RESULTSHistology revealed 4 benign (1 cyst, 2 hamartomas, 1 sclerosing hemangioma) and 42 malignantlesions (21 adenocarcinomas, 7 non–small-cell lung cancers, 5 squamous carcinomas, 4metastases, 3 sarcomas, 2 typical carcinoids). Two more patients were counted as benign on astudy based on > 2 years of nodule stability. Comparing malignant lesions against benign ones,MTT and PS were not significantly different: MTT 11.1+/-1.0 vs. 12.3+/-3.7, PS 24.0+/-2.4 vs.22.3+/-12.6; BF and BV were statistically significant higher in malignant lesions: BF 102.9+/-8.5 vs. 40.5+/-7.3, p

University of Toronto Department of Medical Imaging Research Day 2006Contrast Echocardiography Grading Predicts Pulmonary ArteriovenousMalformations at CTK. A. Zukotynski 1 , R. P. Chan 1 , C.-M. Chow 2 , J.H. Cohen 2 and M. E. Faughnan 21 Department of Medical Imaging and 2 Department of Medicine, St. Michael’s Hospital,University of Toronto, Toronto, Canada.Objectives:Undiagnosed pulmonary arteriovenous malformations (PAVMs) may have devastating clinicalconsequences. Agitated saline transthoracic contrast echocardiography (TTCE) is the screeningtest of choice for PAVMs. A grading system for TTCE has been proposed but not validated.Our aim was to determine positive predictive values (PPV) of TTCE shunt grades (intensity ofopacification) for presence of PAVMs on CT.Methods:We included all patients screened for PAVMs at the Toronto HHT Centre (06/02-09/04) who hadpositive TTCE. TTCE was scored for delay (# cardiac cycles) before appearance ofmicrocavitations in the left atrium and grade of opacification. Opacification was graded: 1(minimal left ventricular opacification), 2 (moderate), 3 (extensive without outliningendocardium) or 4 (extensive with endocardial definition). CT thorax was performed in allpatients and scored as positive or negative for PAVMs.Results:104 patients had positive TTCE. Complete data was available in 88 (85%). There was asignificant association between TTCE grade and presence of PAVMs on CT (p-value=3 cardiac cycles in all cases with PAVMs on CT.Conclusions:Grading of intrapulmonary shunt on TTCE is informative in PAVM screening, with increasingshunt grade predicting increasing probability of PAVMs. TTCE grading may be used to guidedecisions in the PAVM screening algorithm.25 of 58

University of Toronto Department of Medical Imaging Research Day 2006Out of Hours CT Pulmonary Angiograms: Accuracy of Resident ReportingP Cervini, MD; N S Paul, MD; C M Bell, MD, PhDPURPOSETo evaluate the interpretation of CT pulmonary angiograms performed outside of regularreporting hours, comparing radiology resident and attending radiologist interpretations.METHOD AND MATERIALSA retrospective study of 840 consecutive CT pulmonary angiograms (CTPA) performed outsideof regular reporting hours at two large tertiary referral centers over a 2-year period (January 1,2004 - December 31, 2005). The preliminary interpretation by the on-call radiology resident wascompared to the final report issued by a chest radiologist, which was taken as the gold standard.Results of any subsequent follow-up CTPA, extremity ultrasounds and V/Q scans were recorded.The electronic patients' chart was reviewed for cases with discordant interpretations in order todetermine impact on clinical outcome. Of 840 studies, 329 (39%) were inpatients and 511 (61%)from the ER. Studies were stratified as positive, negative or equivocal.RESULTSOverall, 16% (131/840) of CTPA were reported positive by the staff radiologist. There wasagreement in 90% (752/840) of studies (κ=0.76, 95% CI, .71 - .81) with 86% (114/133)agreement for studies interpreted as positive by residents, 95% (582/612) for studies interpretedas negative by residents, and 63% (60/95) for studies interpreted as equivocal by residents. Ofresident false negative interpretations, the thrombus was located centrally in 18% (2/11), in asegmental artery in 73% (8/11) and in a subsegmental location in 9% (1/11) of cases. Studies ofgood quality had higher interobserver agreement than studies of suboptimal quality (odds ratio,4.2; 95% CI, 2.3-7.3; ρ

University of Toronto Department of Medical Imaging Research Day 2006Pulmonary High Resolution CT Findings in Patients with Congenital Bilateral Absence ofthe Vas DeferensPhilip Buckler BSc MDWilliam Weiser MD FRCP(C)Elizabeth Tullis MD FRCP(C)Elizabeth Wilcox MSc MDDanny Ng MSIV, University of TorontoPurpose:We report on the first cohort of patients with congenital bilateral absence of the vas deferens(CBAVD), a primary genital form of cystic fibrosis, to be evaluated with high resolution CT(HRCT) of the thorax.Methods:Twenty-four patients underwent HRCT between 2000 and 2005. Scans were scored according tothe method described by Bhalla for presence and extent of bronchiectasis, bronchial wallthickening, mucus plugging, abscesses, collapse/consolidation, emphysema and bullous disease.Correlation was made with clinical data including genotype, pulmonary and gastrointestinalsymptomatology, sweat chloride and spirometry. Categorical variables were analyzed with theFisher exact test.Results:25% of CBAVD patients demonstrated abnormalities on HRCT. The most commonabnormalities were mild to moderate cylindrical bronchiectasis (21%) and mild bronchial wallthickening (13%). Pulmonary function testing was normal in all patients. There was nosignificant correlation between HRCT findings and genotype, clinical symptoms or sweatchloride values.Conclusion:A substantial minority of patients with CBAVD demonstrate HRCT findings associated with CF,most commonly bronchiectasis and bronchial wall thickening. Such patients warrant ongoingclinical and radiographic surveillance for pulmonary disease.27 of 58

University of Toronto Department of Medical Imaging Research Day 2006Cardiac MRI Assessment of Cardiomyopathy in Patients with Cirrhotic AscitesHong Huang, George Therapondos, Naeem Merchant, Florence Wong,University of Toronto, Toronto, University Health Network ON, Canada.Introduction: Bicycle ergometry is used to detect systolic incompetence under stress in patients withcirrhotic cardiomyopathy. This requires strenuous exercise which is frequently difficult to carry out inpatients with advanced cirrhosis.Purpose: To investigate the use of cardiac MRI and isotonic exercise in detecting cirrhoticcardiomyopathy in cirrhotic patients with ascites and to estimate arterial compliance (COMPart).Methods: 23 patients (Range 33-67 years) and 11 healthy volunteers (32-67years) were studied. Theetiology of cirrhosis was alcohol (10), viral (8) and cryptogenic (1). EKG and cardiac MRI wereperformed at rest and during exercise (leg raised with fixed tension bands).MRI technique on a 1.5 T GEMR-Signa (R) EXCITE TM HD sistem was used. MRI parameters included short axis Fiesta images( TR3.01,TE 1.1, Flip 45, matrix 128 x 256, NEX 1, 10–mm slice thickness and FOV34 x 34).Cardiac index(CI) was calculated from stroke volume (SV) and heart rate (HR), adjusted for body surface area (BSA).COMPart was calculated as a ratio of SV/pulse pressure.Results:RESTΔ WITHEXERCISEControls Cirrhotics Controls CirrhoticsSignificancePCI(l/min/m 2 ) 1287±281 1656±212 514±81 269±58 0.03HR(bpm) 64±6 76±4 14±2 6.5±2 0.01LV measurementsLV mass(g) 68.6±8 73.3±6 NSSeptal wall(mm) 7.4±0.7 7.3±0.6 NSPost wall(mm) 7.4±1.2 5.3±0.4 NSSystolic LV functionEF % 57±2 56±3 4.0±1 3.5±2 NSDiastolic LV functionTPFR(ms) 152 ±8 214±24 0.03PFR(EDV/s) 2.6±0.1 2.8±0.1 NSEKGQTc(ms) 406±6 440±8 0.02COMPart(ml/mmHg) 0.88±0.08 0.85±0.09 1.04±0.07 0.98±0.08 NSEF: ejection fraction, LV=Left ventricle, PFR=peak filling rate, TPFR=time from end systole to PFRSummary: Baseline CI was raised in cirrhotics thus confirming a hyperdynamic circulation. Afterexercise, the rises in HR (10 vs 22%, p

University of Toronto Department of Medical Imaging Research Day 2006The Utility of a First-Generation Computer-aided Detection (CAD) Tool for the Diagnosisof Pulmonary Arterial Filling DefectsErrol Colak, Heidi Roberts, Michael Kucharczyk, Peter Kucharczyk, Katherine Zukotynski,Demetris Patsios, Igor Sitartchouk, Narinder PaulPurpose:The purpose of this study is to evaluate a computer-aided detection (CAD) tool for the automateddetection of pulmonary embolism, and to assess whether this CAD can help a resident tointerpret CT angiograms.Material and Methods:103 unselected multi-detector row MDCT scans performed to rule out pulmonary embolism wereretrospectively evaluated with a commercially available CAD system (ImageChecker CT V2.0,R2 Technology, Sunnyvale, CA). The system automatically marks potential filling defects inlobar, segmental and sub-segmental pulmonary arterial branches for further review andcharacterization. Cases were collected without knowledge of the clinical outcome or any otherunderlying diseases. The interpretation by a chest radiologist served as the gold standard forcase sensitivity. Studies were classified as positive or negative for pulmonary thromboembolismand no indeterminate or equivocal category was employed. The 103 studies were read by aradiology resident, with assessment of (1) technical quality of each scan, (2) the time required forreading, (3) the confidence, and (4) the number of correctly diagnosed cases, each without andwith CAD.Results:Of the 103 CT pulmonary angiograms, 20 were positive for acute pulmonary embolism. Themean time for reading each study was 4:29 ± 2:35 minutes without CAD and 4:55 ± 1:46 withCAD (p= 0.09). Without versus with CAD, the case sensitivity, specificity, PPV, and NPV were:70% vs. 80%, 96% vs. 99%, 82% vs. 94%, and 93% vs. 95%. Without CAD, the residentindicated little confidence in his result in 10, moderate confidence in 16, and high confidence in77 cases. With CAD, the confidence numbers were 5 (little), 21 (moderate) and 77 (high).Conclusion:Utilizing CAD software for the automated detection of pulmonary embolism by a radiologyresident resulted in a slight increase in average reading time and improvements in bothconfidence and accuracy of readings. This suggests that CAD software may play a future role asa second read for residents, particularly during on-call shifts.29 of 58

University of Toronto Department of Medical Imaging Research Day 2006Musculoskeletal, Physics,and Paediatric30 of 58

University of Toronto Department of Medical Imaging Research Day 2006Diffusion Tensor Imaging and Fiber Tractography of the Human Calf- Comparisonbetween 1.5T and 3T: Preliminary ResultsNadja Saupe 1 , Lawrence M. White 1 , Andrea Kassner 1 , Marshall S. Sussman 1 , Michael D.Noseworthy 21 Department of Medical Imaging Mount Sinai Hospital and University HealthNetwork University of Toronto, Toronto, Ontario, Canada2 Brain-Body Institute, Imaging Research Center, St. Joseph`s Healthcare, HamiltonOntario, CanadaPurpose:To compare diffusion tensor (DT) MRI of human calf muscles at 1.5T and 3T in healthyvolunteers, to determine a number of quantitative parameters to characterize diffusion anisotropyin organized muscle tissue, and to assess fiber tractography at both field strenghts using similarimaging parameters.Methods and Materials:5 healthy volunteers (4M, 1F, mean age 35 yrs) were studied. All volunteers underwent MRI at3T and 1.5T within one week. Imaging on both MR systems was performed using a standardquadrature knee coil and the same imaging protocol. Three SE-EPI pulse sequences (b-values300, 500 and 700 s⋅mm -2 ) with 6 non-collinear diffusion encoding gradients were applied tocollect 2D images through the calf of each volunteer. Image analysis, tensor calculations, andfiber tractography was performed offline using DTI studio (Johns Hopkins University,Baltimore). The eigenvalues (λ 1 , λ 2 , λ 3 ), Trace (D), fractional anisotropy (FA), relativeanisotropy (RA), and volume ratio (VR) were calculated in three different calf muscles (medialand lateral gastrocnemius, and soleus). A regression analysis looking at within-subject effectswas performed.Results:No significant differences were found in Trace (D) (p=0.19), and RA (p=0.22) values at bothfield strengths. Significantly lower values of FA (p=0.029) and VR (p=0.032) are seen at 3T. Nosignificant differences in Trace (D) were observed between muscles, irrespective of b-value.Fiber tractography was successfully performed on 1.5T and 3T datasets, defining tracts ofdiffusion anisotropy paralleling the expected anatomic course of muscle fibers in the midcalf.Conclusion:The study demonstrates useful parameters to perform DT MRI and fiber tractography at 1.5T and3T. DT MRI at 1.5T and 3T provides in vivo validation of quantitative structural analysis ofhuman skleletal muscle. Fiber tracking at 3T MR systems provides better quantitativeinformation about muscle fiber orientation and architecture of the diffusion pathway, howeverfiber tractography is still feasible at 1.5T.Clinical Relevance/Applications:DT MRI is a useful method to study muscle architecture noninvasively and provides thepossibility of in vivo quantitative structural analysis of human skeletal muscle tissue.31 of 58

University of Toronto Department of Medical Imaging Research Day 2006A Novel MRI Technique for Visualizing Dynamic Joint MotionJim Li, Lawrence White, Jeffrey Stainsby, Nadja Saupe, Marshall SussmanMusculoskeletal trauma and disease affect the structure and function of joints. Diagnosticimaging must assess both of these properties for a complete characterization of pathology. Toassess structure, imaging must evaluate joint morphology. The preferred modality for this task isMRI, due to its excellent 3D depiction of soft tissue, cartilage, and bone. To assess function,imaging must evaluate the kinematics of the joint while in motion and under biomechanicalstress. Due to its lengthy image acquisition time, MRI is typically not used for this task. Instead,rapid imaging techniques such as fluoroscopy or ultrasound are preferred. However, relative toMRI, these modalities have a much narrower range of tissue contrast, limited 3D capability, and(for ultrasound) difficulty penetrating deeper joint structures. Moreover, since separate imagingexams are used for morphology and function, often only one of these assessments is performedin the interests of cost and expediency; thus potentially missing aspects of the pathology. If aneffective MR technique for visualizing joint motion existed, then kinematics and morphologycould be assessed with superior tissue contrast, in 3D, and in a single exam.In this presentation, a novel MRI method for visualizing joint kinematics will be discussed thatovercomes the limitations of previous MR techniques. It will be shown that this method allowsone to obtain MR images of joint kinematics with high spatial and temporal resolution. Figure 1illustrates an example of this technique applied to the wrist undergoing radial-ulnar deviation.The basis of these methods will be the “navigator echo”; a type of data acquisition used tofacilitate imaging in the presence of motion.Fig.1. Images of the wrist undergoing radial-ulnar deviation.32 of 58

University of Toronto Department of Medical Imaging Research Day 2006Ultrasonography for Children with Hemophilic Arthropathy:How we do it.Zukotynski K., Doria A.S., Babyn P.S., Jarrin J., and Carcao M.Approximately 80% of hemorrhage in patients with hemophilia involves the joints, commonlythe ankles, knees and elbows. Breakdown of extravasated blood from synovial vesselsultimately leads to synovitis, cartilage loss and subchondral bone irregularity. Ultrasound isuseful in evaluating musculoskeletal involvement of joints in children. It is easily accessible anddoes not require sedation. We have developed a systematic protocol for ultrasound assessmentof the knee and ankle in children with hemophilic arthropathy. Findings on radiography,ultrasound and MRI will be shown and compared for the ankle and knee. Imaging scoringmethods currently used will be reviewed. The ultrasound protocol developed for imaging theknee and ankle in children with hemophilic arthropathy will be illustrated.33 of 58

University of Toronto Department of Medical Imaging Research Day 2006THE VALUE OF ULTRASOUND IN ASSESSING THE NEED FOR FULL VSFOCUSED CT IN CHILDREN EVALUATED FOR ACUTE APPENDICITISMinnis SObjective:Sonography (US) followed by computed tomography (CT) is often used in the evaluation ofacute appendicitis in children. CT uses ionizing radiation and carries the risk of radiationinducedmalignancy. The aim of this study was to determine whether ultrasound is reliable inexcluding significant pathology above the lower pole (LP) of the kidneys to enable the use of afocused CT, from the LP of the kidneys to the bladder base, thus reducing radiation exposure.Materials and Methods:Retrospective study between November 1999 and October 2005. Children (n=394) whopresented with acute abdominal pain and had imaging to exclude acute appendicitis with both USand CT were included. Patients with negative US above the LP of the kidneys were sub selectedfor analysis. The CT results of these patients were reviewed to determine if US had missedsignificant pathology above the LP of the kidneys.Results:340 patients (86%) with negative US above LP kidneys. CT revealed missed pathology aboveLP in 22 (7%). There was 1 case of missed appendicitis above the LP. There were 9 lower lobepneumonias, 6 pyelonephritis, 1 lung nodule, 1 esophagitis, 1 hilar adenopathy, 2 thicktransverse colon and 1 thick duodenum. NPV of a negative US above LP was 0.93 (95%CI=0.89-0.96). There were 142 patients with negative US above LP but positive findings belowLP. CT showed pathology above LP in 6 (4.2%); odds =4.4%. There were 104 patients withnegative US above and below the LP. CT showed pathology above LP in 16 (15.4%);odds=15%.Conclusion:US reliably excludes pathology above the LP of the kidneys allowing use of a focused CT, andconsequent reduction of radiation dose, only for children in whom there are findings below theLP of the kidneys to account for the abdominal pain.34 of 58

University of Toronto Department of Medical Imaging Research Day 2006Neuroimaging35 of 58

University of Toronto Department of Medical Imaging Research Day 2006Gadolinium Enhancement Predicts Hemorrhagic Transformation inAcute Ischemic StrokeObjective:To investigate the relationship between each of two high dose gadolinium MRI findings, 1)parenchymal enhancement, and 2) a hyperintense middle cerebral artery (MCA), withsubsequent hemorrhagic transformation (HT) in patients with hyperacute ischemic stroke.Methods:Twenty-four consecutive patients with ischemic stroke who underwent magnetic resonanceimaging (MRI) with high dose (45 cc) gadolinium-DTPA within 4.3 hours (±1.4) of symptomonset were retrospectively reviewed. All of these patients underwent at least one follow-up MRIor non-enhanced computed tomography (CT) study at 2-7 days post ictus. Post-gadolinium T1-weighted (T1W) images were analyzed for parenchymal enhancement and a hyperintense MCA.Gradient echo MRI and CT were used for assessment of HT.Results:Ten (41.67%) patients developed HT on follow-up imaging (hemorrhagic group). Earlyparenchymal contrast enhancement was found in 6 patients with HT (p = 0.002). A hyperintenseMCA was seen in 5 patients with HT (p = 0.006). In the fourteen patients without HT(nonhemorrhagic group), none showed early parenchymal enhancement and/or hyperintenseMCA. The sensitivity of parenchymal enhancement and the HMCA sign were 60% and 50%respectively. Parenchymal enhancement and the hyperintense MCA had specificities of 78% and74% respectively. The positive predictive values were 100% for each sign.Conclusions:Early parenchymal enhancement and a hyperintense MCA on T1WI following Gd-DTPAadministration are independent predictors of subsequent HT.Authors: Mikulis, Go, Wu, John, Silver36 of 58

University of Toronto Department of Medical Imaging Research Day 2006Cochlear Implantation in patients with Cochlear OtosclerosisFanning NF 1 , Symons S 1 , Marshall AH 2 , Shipp D 2 , Chen JM 2 , Nedzelski JM 2Department of Medical Imaging 1 , Department of Otolaryngology 2 , Sunnybrook and Women’sCollege Health Sciences Centre, University of TorontoPurpose:The objective was to correlate implant performance in cochlear otosclerosis to 1) matchedcontrol samples, 2) severity of otic capsule involvement, 3) prior ipsilateral surgery, and 4)programming issues.Materials and Methods:The study design was a retrospective case controlled study. The study cohort comprised 30individuals. Diagnosis was based on prior ear surgery (stapedectomy [n = 18] or fenestration [n =2]) and/or pathognomonic radiological findings. High-resolution computed tomography imagesof the temporal bones were assessed by two neuroradiologists and graded (range, 0-3) for theextent of otosclerosis. Operative records were reviewed. Performance, programming visits, andthe number of electrode deactivations at 6 months and at 1 year after implantation weredetermined for the individuals with otosclerosis and compared with a group of matched controlsubjects. A within-group comparison correlating severity of otosclerosis to the above was carriedout.Results:Implant performance in individuals with cochlear otosclerosis was not significantly differentfrom those without. Previous surgery on the side of implantation did not alter performance.Programming difficulty as reflected in the number of visits and electrode deactivation for soundquality reasons were comparable. Deactivation for facial nerve stimulation occurred exclusivelyin otosclerotics with the most severe radiological disease (grade 3) and was only with nonmodiolarhugging electrodes (n = 5). There was no observed difference between the radiologicalextent of otosclerosis and implant performance.Conclusions:Individuals with severe otosclerosis considering cochlear implantation can be counseled toexpect similar benefit to those without, regardless of whether prior surgery occurred on the sideof implantation or of severity of otic capsule involvement. There is a significant risk of facialnerve stimulation in otosclerotics with grade 3 disease.37 of 58

University of Toronto Department of Medical Imaging Research Day 2006Value of CTA-Source Images in Acute Stroke using ASPECT ScoringChakraborty S, Aviv RI, Shelef I, Malam S, Sahlas DJ, Symons S, Fox AJDepartment of Medical Imaging, Sunnybrook & Womens College Health Sciences Center,University of TorontoPurpose:CTA-source images (CTA-SI) increase the accuracy and reliability of detection of stroke overnon-contrast CT (NCCT). ASPECT scoring is a standardized quantification used on NCCT andimproves assessment of stroke extent. We test the added value of applying this scoring ininterpretation of CTA-SI and examine the effect of radiologists’ experience and clinical historyon these findings.Methods:23 consecutive patients presenting with stroke-like symptoms within 6 hours between Januaryand June 2004 were analyzed by 3 Neuroradiologists of varying experience. Each reviewer,blinded to clinical information reviewed a random order of NCCT and CTA-SI, documented sideof infarct and ASPECT score. The readings were repeated with and without clinical information.Sensitivity, specificity, positive and negative predictive values, accuracy and inter and intraobserveragreement was calculated.Result:For stroke detection the impact of CTA-SI over NCCT was small (3 – 5%), though the traineehad the greatest increase in accuracy and sensitivity of 9% and 20% respectively. For strokeextent CTA-SI improved accuracy and/or sensitivity of each reader by 9-17% and 0-33%compared to NCCT respectively. Inter-observer agreement was high (K 0.57-0.7) for infarctdetection and extent (K 0.64-0.91). Knowledge of clinical history demonstrated no significantchange for stroke detection or extent.Conclusion:CTA-SI is superior to NCCT in assessment of stroke extent using ASPECTS. Less experiencedreaders may benefit from CTA-SI for stroke detection but the overall improvement is small andexperience does not appear to influence accuracy significantly. Clinical history contributes littleto NCCT and CTA-SI ASPECTS.38 of 58

University of Toronto Department of Medical Imaging Research Day 2006Comparison of CTA to DSA in evaluating etiology of non-traumatic, non-subarachnoidintracranial hemorrhage.Ahmad T, Aviv R, Noel de Tilly L, Fox A, Symons SDivision of Neuroradiology, Department of Medical Imaging, Sunnybrook and Women’sCollege Health Sciences Centre, University of TorontoSt. Michaels Hospital, Department of Medical Imaging, University of TorontoPURPOSEDigital subtraction angiography (DSA) has been traditionally used to evaluate non-subarachnoidintracerebral hemorrhages in order to determine whether there is an underlying vascularmalformation. This procedure is expensive, time-consuming, and uncomfortable for the patient,and has well documented risks. Computed tomography angiography (CTA) is now commonlyused and has been show to be accurate in determining the cause of subarachnoid hemorrhage. Itsvalue in assessing other intracranial hemorrhage is not well known. We compared the accuracyof CTA to DSA in determining the cause of non-subarachnoid hemorrhages in 44 patients.MATERIALS & METHODSA total of 239 patients had both cerebral CTA and DSA between 2001-2005 at our center. Outof these a total of 44 patients fulfilled inclusion criteria.All of the CTAs were performed on a 4 or 64 detector CT (General Electric, Lightspeed,Milwaukee, WI). Multiplanar and 3D reformations were performed. All hemorrhages were alsoassessed by traditional DSA (Philips). Four blinded reviewers independently evaluated the CTAsand two of them also evaluated DSAs in separate sessions. The reviewers categorized the CTAand DSA results as: 1) AVM, 2) AVF, 3) cavernous malformation, 4) tumor, 5) aneurysm, 6)DVA, 7) negative/unknown.RESULTSCTA had a mean sensitivity of 90.75% and a mean specificity of 85.25% in diagnosis ofunderlying etiology. The mean PPV and NPV were 86.75 and 89.75 respectively.Final diagnosis was based on DSA. Disagreements were resolved based on follow up imaging,surgery or consensus. The mean kappa value was 0.71 demonstrating good inter-readerconcordance.There were 11 AVM, 3 AVF, 7 Aneurysms, 1 Tumor, 1 Moya Moya, 1 Cavernoma and 20negative cases.CONCLUSIONCTA performed on a MDCT is nearly as accurate as DSA in determining the cause of nonsubarachnoidhemorrhages.39 of 58

University of Toronto Department of Medical Imaging Research Day 2006Differences in cerebrovascular reactivity in males versus females obtained using BOLD MRI andalternating states of end-tidal pCO2Andrea Kassner, Julien Poublanc, George Tomlinson, Adrian CrawleyUniversity of Toronto &UHN, Dept of Medical Imaging and Neurology*, Toronto, ON, CanadaIntroduction: Combining inhaled CO 2 manipulation with BOLD MRI is a promising method for assessing regionaldifferences in cerebrovascular reactivity (CVR) which is a measurement of the brain’s autoregulatory capacity. Thisbecomes important for the assessment of vascular disorders in which autoregulation is compromised or exhaustedsuch as in moyamoya disease or carotid stenosis/occlusions. The characterization of these pathophysiologicalconditions in the cerebral circulation, however, requires the knowledge of potential physiological gender-dependentdifferences in CVR. The objective of this study was to evaluate gender differences in CVR in normal subjects usingquantitative CVR measurements by correlating BOLD MR signal intensity with square wave changes in end-tidalpressure of CO2 (p ET CO 2 ).Materials and Methods: Ten healthy male (age range 25 - 42 years) and 9 healthy female (age range 21-38 years)volunteers were imaged on a 1.5T GE Signa MR system 4 times (2 sequential runs on day 1 and 2 sequential runs ona separate day 1-2 weeks later) using a rebreathing circuit as previously described by Vesely et al. [1]. Each subjectwas placed inside the scanner and the rebreathing device was applied. Scanning was performed using a standardsingle-shot BOLD protocol with a spiral read out (TE=40ms, TR=2240ms, FA=85°, FOV=20mm). Scanningduration for each run was 12 minutes for an acquisition of 320 volumes. Each volume contained 28 slices and spatialresolution of the BOLD data was approx 3 x 3 mm with a slice thickness of 4.5mm. In addition, high resolution T1weighted images were acquired for co-registration purposes. Changes in p ET CO 2 were achieved by controlling thesubjects inspired gases with the aid of a nose clip, a mouthpiece, the rebreathing circuit, and a gas sequencer. Toensure that end-tidal gases are representative of lung gas concentrations, subjects were instructed to breathe deeplyduring the test. The test itself consisted of eight cycles of hypercapnia (45sec at ~ 50mmHg) interspersed with eightcycles of hypocapnia (45sec at ~ 30mmHg) all of which was regulated by an automated sequencer. Hypercapnia wasinduced by administering a gas mixture of 8% CO 2 /92% O 2 at 14L/min for 15 sec and maintained at plateau for asubsequent 30 sec by reducing gas flow to 1.5-2 L/min of O 2 . During the plateau phase, the decreased inflow offresh gases resulted in rebreathing of previously exhaled gases contained in the expiratory reservoir tube. Intervalsof low CO 2 were achieved by supplying subjects with 15sec of high flow (16-18 L/min) of 100% O 2 and maintainedby O 2 flow at a rate of 12-14 L/min. Partial pressures of end-tidal CO 2 (p ET CO 2 ) and O 2 (p ET O 2 ) were monitoredcontinuously using a commercially available capnograph and recorded digitally at a sampling rate of 60Hz/channel.After completion of the measurement, the collected p ET CO 2 data was reduced to one measure of p ET CO 2 per breathand correlation analysis with the BOLD data was performed. Prior to this, the BOLD data was co-registered tocompensate for motion artifacts. Signal of the whole brain was used as a reference to determine the shift needed tobring the CO 2 and the MR data sets in phase. Once in phase, CVR maps were calculated on a pixel by pixel basisfrom the slope of the regression of the percentage change of MR signal on the p ET CO 2 . This provides a measure ofreactivity expressed in units of % Δ MR signal/mmHg p ET CO 2 . A variance components analysis was used toestimate variability in reactivity between runs, different days and subjects. Reproducibility was quantified using theinterclass correlation coefficient (ICC). Differences between mean reactivity of males and females were assessedusing the Wilcoxon rank sum test.Results: Averages of the 4 BOLD signal changes for males ranged from 0.10 to 0.21 %Δ MR signal/mmHg p ET CO 2(mean=0.157, SD=0.029). For females, these values ranged from 0.098 to 0.161 %Δ MR signal/mmHg p ET CO 2(mean=0.123, SD=0.022). The mean value for males is statistically higher than that for females (p=0.01). Results forthe day to day variation of runs for both genders show excellent reproducibility with an ICC=0.89 (95%CI:0.92-0.97) for males and an ICC=0.96 (95%CI:0.91-0.99) for females. These vales are for the whole brain. For correlatedvalues above r=0.3, the ICC=0.95 (95%CI:0.86-0.99) for males and 0.92 (95%CI:0.81-0.98) for females . Within asession, there is approximately a 95% chance that repeated measures for entire brain values on a subject would liewithin ±0.017 %Δ MR signal/mmHg p ET CO 2 of each other. For values above r=0.3, the measurement error isslightly higher in the absolute sense. Here, 95% of repeated reads should lie within ±0.029 %Δ MR signal/mmHgp ET CO 2 of each other.Discussion: This study shows excellent reproducibility and demonstrates differences in cerebral vascular reactivitybetween males and females in the order of 27.6%. These differences may have important implications for theassessment of therapeutic interventions using CVR and need to be further investigated.References: 1. Vesely et al. MRM 2002, 2. BF article40 of 58

University of Toronto Department of Medical Imaging Research Day 2006INVESTIGATION OF SHORT TE INFLOW-BASED CONTRAST FOR FMRIJ. Poublanc, D. Mikulis, C. Wei, A. CrawleyMedical Imaging, The Toronto Western Hospital of the UHN, Toronto, Ontario, CanadaIntroductionThe effective longitudinal relaxation time T1* varies with blood inflow according to a well-knownrelation [1], 1/T1* = 1/T1 + f/λ , ie Δ(1/T1*) = Δf/λ. As T1* varies with flow, it is possible to obtainactivation maps using a simple T1 weighted sequence. This inflow-based method may possess someadvantages over BOLD and ASL. BOLD suffers from signal drop-out at the skull base because of thesusceptibility-induced field gradients (SFGs) intrinsic to a T2* weighted image. This will be reducedwith short TE T1-weighted imaging. ASL is an alternative inflow-based method that can quantitate flowchanges in fMRI but only over a relatively small number of slices. We investigate the sensitivity toneuronal activation of a simple T1-weighted sequence compared to BOLD and ASL imaging.MethodsAn anatomical scan followed by a series of functional scans was acquired on a 3T MRI system (GEMedical Systems, Milwaukee, WI). All functional experiments used a spiral acquisition, a scan time of5min, FOV=240 mm, matrix size = 64 * 64, two slices, thickness = 5 mm, acquired in superior to inferiororder. Specific parameters for each sequence were the following: BOLD (TE=35ms,TR=2000ms, θ~80°),FAIR (TE=22.5ms, TR=2000ms, TI=1000ms), two T1-inflow sequences (TE 4.7ms, θ~90°) withdifferent TRs ( TR = 650ms, TR=4000ms). The task was a bilateral finger opposition in 15 sec OFF / 15sec ON blocks except for FAIR, for which 30 sec OFF / 30 sec ON blocks were used. After standardpreprocessing steps, t-score (t) and percentage signal change (%SC) maps were calculated for eachexperiment. Peak t-score voxel was chosen in each ROIs (left and right M1 and SMA). t and %SC ofthose voxels were then averaged together.The optimal TE =T2*~35ms and TR=T1/2~650ms (for maximum contrast-to-noise (CNR) at fixed scantime) for BOLD and T1-inflow respectively.Results and DiscussionLeft to right, top to bottom: BOLD, Inflow-TR 0.65s,Inflow-TR 4s, FAIR The two first images arethreshold with t=9, the two next ones with t=3. Thecolors correspond to the %SC intensity.BOLD Inflow TR .65s Inflow TR 4s FAIR%SC 4.12 2.0 1.46 4.3t 13.5 18.7 6.3 5.8The results look consistent with our expectations (see tableabove). Although BOLD and FAIR have the highest %SC, theinflow method gives a better contrast-to-noise efficiency,which can be seen in the t-scores. The inflow technique atTR=4s and TE of 4.7ms demonstrates some residual BOLDeffect, but at a TR=650ms, the inflow effect clearly accountsfor most of the t-score. The T1 inflow effect has beeninvestigated before [2,3] and often seen as an undesirablecomponent, in terms of localization accuracy. However, in thisstudy (see images), the BOLD and inflow activations are wellco-localized. With a TE of 0 ms, which will be investigated ina future study, a simple inflow technique promises to provide arobust measurement of neuronal activation in areas where theBOLD method fails due to SFGs.References1- Kwong et al Proc Natl Acad Sci U S A. 1992 Jun 15;89(12):5675-9.2- Righini et al. Magn Reson Imaging. 1995;13(3):369-78.3- Glover et al. Magn Reson Med. 1996 Mar;35(3):299-308.

University of Toronto Department of Medical Imaging Research Day 2006Gender Differences in Water Diffusion of the Corpus Callosum:A Diffusion Tensor Imaging StudyFang Liu, PhD, Andrea Kassner, PhDDept of Medical Imaging, University of TorontoIntroductionThe corpus callosum (CC) plays an integral role in relaying sensory, motor and cognitive information between homlogousregions in the two cerebral hemispheres. Women seem to employ a greater degree of bilateral hemispheric activity than menand also seem to have greater verbal ability (1). Additionally, women have a larger callosal area proportional to cerebralvolume (2) which suggests that a larger number of fibers are crossing through and hence capacity for interhemispherictransfer is enhanced. We therefore hypothesize that we can document these differences using diffusion tensor imaging andalso quantify the differences between healthy male and female white matter of the corpus callosum using a quantity, FDi(fiber density index).Materials and Methods10 healthy volunteers (five males, five females, aged 22-34 years) underwent MR diffusion tensor imaging. All imagingwas performed on a 1.5T MRI system (GE TwinSpeed Excite, Release 11.0, GE HealthCare, Milwaukee, WI) equipped withan 8-channel head-coil. Images were acquired parallel to the anterior-posterior commissure line (AC-PC), and the midlinewas aligned to the vertical line of the images at the time of data acquisition. Diffusion tensor imaging was performed twiceusing 15 non-collinear directions of diffusion sensitization with an echo planar readout (b-value=1000s/mm 2 , matrix=128x128, FOV of 24cm; slice thickness 2mm, number of slices 60~64, TE=74ms, 4NEX).DTI data sets were spatially registered using AIR (Automated Image Registration version 3.0) (3) and post-processedusing DTI Studio (Processing Tools and Environment for Diffusion Tensor Imaging, version 2.4, H Jiang and Susumu, Deptof Radiology, Johns Hopkins University). Fractional anisotropy was calculated on a pixel by pixel basis (4), and fibertracking using the FACT algorithm (5) was then performed reconstructing all possible fibers through the image set from allpixels above an FA threshold of 0.2. Since all pixels with sufficient FA (typically several thousand pixels) are considered aspossible sources of fiber paths, it is thus quite conceivable for multiple fiber paths to pass through any given pixel. Regionsof interest (ROIs 18pixels and 8pixels, respectively) were placed in six subdivisions of the corpus callosum according to theWitelson scheme (6): CC2-Genu, CC3-Rostral body, CC4-Anterior midbody, CC5-Posterior midbody, CC6-Isthmus, CC7-Splenium. These regions were interrogated firstly to obtain the mean value of fractional anisotropy and also to determine thenumber of white matter fiber paths passing through them. From this latter measure an index of fiber density, the FDi, wasconstructed by determining the mean number of fiber paths passing through each pixel in the ROI.ROIs were placed independently by 2 readers and the two quantities, FA and FDi, were independently determined.Concordance between readers for each quantity was assessed using the intraclass correlation coefficient (ICC). Mean valuesof FA and FDi for six subdivision of corpus callosum ROI were determined by averaging the corresponding measure fromthe two separate raters. Statistical significance was determined using a paired students’ t-test.ResultsBoth readers were in good agreement (ICC=0.93). The mean values of FA of each of the callosal subregions showed thatmales had greater mean fractional anisotropy except the splenium; the mean values of FDi of each of the callosal subregionsshowed that females had a greater mean fiber density than males except the splenium. FA differences between males andfemales were non-significant for all ROIs; FDi differences between males and females were significant for CC2 and CC6(p=0.038 and 0.05, respectively).DiscussionThe new index, FDi, adds a quantitative property to the fiber tracking methodology and supports our hypothesis. FDicalculated from Diffusion Tensor Imaging in living brain corpus callosum corresponds with results obtained by J. RobinHighley et al. (7) who calculated fiber density of the corpus callosum (CC) on post-mortem brains. FDi and FA obtainedfrom DTI may offer superior description of a range of white matter pathologies e.g. in schizophrenia or developmentaldisorders. This needs to be further investigated.References1. Crow et al., Neuropsychologica 1998 2. Mitchell et al., AJNR 2003;24:410-4183. Woods RP, et al. JCAT 1993;17:536-546 4. Pierpaoli C, Basser PJ. Magn Reson Med 1996;36:893-9065. Mori S, et al. Ann. Neurol. 1999;45:265-269 6. Sandra F. Witelson, Brain 1989;112:799-8357. J. Robin Highley et al. Brain 1999;122:99-110

University of Toronto Department of Medical Imaging Research Day 2006Inter-reviewer variability in interpretation of CT heads and X-ray shunt series in childrenwith suspected ventricular peritoneal shunt obstructionBoucher LM, Shroff MDepartment of Medical Imaging, HSC, University of TorontoGOAL:CT head studies and X-ray shunt series are jointly used to assess ventriculoperitoneal(VP)shunt function. However, early signs of ventricular shunt malfunction can besubtle leading to the risk of inter-reviewer variability. Because of possible repercussions onpatient treatment, we assessed inter-reviewer variability in reading such studies.MATERIALS AND METHODS:287 shunt X-ray series and 271 CT heads previously performed on children admitted tothe Emergency with suspected VP shunt obstruction were blindly reviewed a second time. Theradiological reports were categorized into: normal study, shunt kink, or shunt disconnection forthe X-ray series, and into normal study, new increase, no change, further increase, or decrease inventricular size for the CT studies. Inter-reviewer variability was evaluated with kappa statistics.Presenting complaint, CT, and X-ray findings were correlated with shunt revision using Fisherprobability testing.RESULTS:Although none of the common clinical symptoms correlated well with shunt surgicalrevision outcome, both abnormal CT heads and X-ray series did correlate extremely well(p

University of Toronto Department of Medical Imaging Research Day 20063T INTRACRANIAL ARTERIAL WALL IMAGING: CLINICAL IMPACT.S.Bhuta, R.Farb, K.ter Brugge, J.Butani, D.Mikulis.Dept. of Neuroradiology, University of Toronto, Toronto Western Hospital, Toronto, Canada.Imaging and diagnosis of arterial pathology has been and continues to be based onmorphological changes imposed on the arterial lumen by pathological processes. In mostinstances a combination of clinical information and luminal defects seen on imaging is sufficientfor accurate diagnosis, however different vessel wall pathologies can produce similar luminaldefects limiting diagnostic specificity. Direct Imaging of the pathological processes affecting thevessel wall would be desirable in those instances where the etiology of luminal narrowing isuncertain. The ability to image the arterial wall directly at 1.5 T is hampered by limited spatialresolution and low signal to noise. Improvements in image quality are needed to accomplish this.3T MRI promises to fulfill this requirement.We present our experience in several patients in whom direct imaging of the arterial wallpermitted definitive diagnosis not provided by high quality luminal imaging includingconventional angiography. Patients were scanned on a high resolution GE 3T Signa Excite MRIsystem using an 8-channel head coil. Key sequences of the 3T protocol were high resolution T1SE pre and post gadolinium and 3D FIESTA sequences. Definitive diagnoses including giant cellarteritis of the carotid siphons and mid basilar dissection were established with high field wallimaging including histopathological confirmation wherever possible. We believe the addeddiagnostic specificity provided by wall imaging is significant and should be considered forimplementation in Neuroimaging protocols at those sites with available high field MRI systems(3T or greater).44 of 58

University of Toronto Department of Medical Imaging Research Day 2006ASPECT Scoring of CT Perfusion in Early Stroke Visualization and AssessmentMandelcorn JBackground and Purpose:The Alberta Stroke Program Early CT score (ASPECTS) has been shown to be of prognosticbenefit when applied to non contrast CT (NC-CT) and CTA-source images (CTA-SI). The scorehas the advantage of discriminating outcome of MCA territory infarct in a quick and reliablefashion with emphasis on the clinically important basal ganglia, subinsula and internal capsule.The role of CT Perfusion (CT-P) ASPECTS is uncertain. Extent of decreased cerebral bloodvolume (CBV) is known to correlate with final infarct size and with DWI MR findings. Meantransit time (MTT) and cerebral blood flow (CBF) represent the ischemic tissue at risk. Weexamine whether ASPECT assessment on CT perfusion (CT-P) is superior to NC-CT inpredicting clinical and radiological outcomeMethods:A prospective study of 39 consecutive patients presenting between January and June 2005 wasconducted. Patients were included if they presented with stroke-like symptoms within 3 hours ofonset, underwent our CT stroke protocol, were treated with chemical thrombolysis and followedup at 5-7 days. Three neuroradiologists blinded to clinical information randomly reviewed thepresentation NC-CT and CT-P studies a week apart to prevent recall bias. Presence, side ofinfarct and ASPECTS score was calculated for each study. For CT-P, abnormality was scoredwherever a qualitative abnormality was present. ASPECT scores of follow up NC-CT studies at5-7 days and modified Rankin score (mRS) / National Institutes of Health Stroke Scale (NIHS)score at 90 days were the radiological and clinical reference standards. Final infarct size wasdetermined by consensus of readers without knowledge of previous imaging. Data wastransferred to and analyzed in SPSS. The correlation between ASPECTS for NC-CT, CT-Pparameters and both radiological and clinical outcome was determined. Significance wasdetermined by Fisher’s exact test. Receiver operating characteristic (ROC) curves were drawn tocalculate optimal thresholds for determining outcome. Intraobserver agreement was alsomeasured.Results:CBV and NC-CT had equivalent correlation with final NC-CT (0.77), compared with MTT orCBF (0.63 and 0.62 respectively). CBV accuracy was higher than NC-CT. All CT-P parametersdiscriminated radiological and clinical outcome significantly. ROC thresholds for CBV andCBF/MTT differed from NC-CT. CBF and MTT accuracy equalled final infarct size butsensitivity was lower. Change in NIH score had greatest predictive value for clinical outcome.Intraobserver agreement for CBV was slightly higher then for NC-CT.Conclusion:ASPECT thresholds applied to CT-P increase accuracy for radiological and clinical outcomeover NC-CT.45 of 58

University of Toronto Department of Medical Imaging Research Day 2006The Evaluation of Intracranial Atherosclerosis Using the BOLD MRI technique, Part1:BOLD MRI Mapping of Cerebro-Vascular Reserve (Cerebro-Vascular Reserve) withCarotid Artery Occlusive Disease (CAOD): Value in Identifying Patients for CarotidArtery Revascularizations.SK Lee, J Poublanc, K terBrugge, D MikulisDivision of Neuroradiology, Department of Medical Imaging, Toronto Western Hospital/UHN,University of TorontoPurpose:The purpose of this study is to evaluate the value of the BOLD MRI CVR mapping technique foridentifying hemodynamically compromised group from the CAOD patients.Materials and methods:Thirteen patients (12 male;1 female, mean age :70.615) with BOLD MRI CVR study who werediagnosed as Carotid Artery Occlusive Disease (CAOD) based on gadolinium enhance MRangiography were retrospectively reviewed. Eleven patients were symptomatic (either TIA orminor stroke) and 2 patients were asymptomatic. Degree of symptomatic carotid artery stenosison the Gadolinium enhance MR angiography was 97.3% (range of stenosis: 85 to 100%) whichincludes 4 complete occlusions of ipsilateral carotid artery of the symptomatic side. One of twoasymptomatic patients had bilateral complete occlusion of extracranial carotid arteries andanother showed bilateral 99% stenosis of extracranial carotid arteries. BOLD MRI CVR studieswere performed using a CO2 re-breathing method during MRI. End tidal CO2 was toggledbetween 35 to 45 mmHg over 4 cycles during 12 minutes acquisitions for BOLD. Studies wereperformed on a 1.5T Signa Echospeed MRI system (GE Medical System, Milwaukee, WI)equipped with 23mT/m gradients.Results:All of asymptomatic patients (n=2, 100%) and 4 symptomatic patients (4/11, 36.3%) showedbilateral symmetric Cerebro-Vascular reserve on CVR maps. Seven out of 11 symptomaticCAOD patients (7/11, 63.6%) demonstrated ipsilateral decreased CVR on the symptomatic side.All four cases with decreased CVR who had carotid artery revascularization procedures [carotidendarterectomy(n=3), carotid stent assisted angioplasty (n=1)] showed improved CVR on followup studies.Conclusion:The CVR mapping of the brain is reflecting tissue level hemodynamic status. Non-infarctedbrain tissue with decreased CVR due to CAOD may be reversible after carotid revascularizationprocedures. The BOLD MRI CVR mapping can be a reliable and practical imaging method forthe triage of CAOD patient management.Acknowledgment: This project was supported by the Department of Medical Imaging Researchand Development Awards, University of Toronto.46 of 58

University of Toronto Department of Medical Imaging Research Day 2006Optimizing Gadolinium Enhanced MRA for Demonstration of the Artery of AdamkiewiczHoward P, Farb RDepartment of Medical Imaging, Toronto Western Hospital, University of TorontoPurpose:The largest radiculomedullary artery supplying the lumbar enlargement of the spinal cord istermed the artery of Adamkiewicz (AKA). The level of origin of the AKA is variable, but isusually between T8 and L2. Preoperative identification of the level of origin of the AKA isimportant in planning orthopedic, neurosurgical, or vascular procedures that might compromisethis vessel. Catheter angiography is currently the gold standard for demonstration of the AKA,but this invasive test also carries a risk of vascular injury. Contrast enhanced MR angiography(CEMRA) demonstrates vascular anatomy with sufficient detail to replace catheter angiographyfor many applications. The aim of this study is to optimize CEMRA for demonstration of theAKA. This is challenging as the small diameter of the AKA (0.5-0.7mm) requires high spatialresolution imaging, while high temporal resolution is required to differentiate spinal arteriesfrom spinal veins.Materials and Methods:Nineteen patients booked for enhanced MRI of the spine were studied with a first pass MRAtechnique following the bolus injection of gadolinium contrast. MRA techniques selected forhigh temporal resolution were single slab 2D projection, TRICKS, and a protocol published byHyodoh et al.(2005). MRA techniques selected for high spatial resolution included ATECO anda protocol published by Nijenhuis et al. (2004).Results:The single slab 2D projection provided the best temporal resolution but did not provide sufficientcontrast to differentiate the spinal vessels from background. TRICKS and the protocol ofHyodoh et al. have sufficient temporal resolution to show different phases of enhancement of thespinal vessels, but the lower spatial resolution of these techniques impedes visualization of smallvascular branches. Techniques with elliptic centric encoding provided the best anatomic detailof spinal vessels, at the expense of longer scan times, and with the requirement of an accuratemethod of triggering the scan.Conclusions:Gadolinium enhanced MRA can be used to demonstrate the vessels of the spinal cord. TheATECO pulse sequence in a coronal plane is a promising sequence for demonstrating the AKA.47 of 58

University of Toronto Department of Medical Imaging Research Day 2006CTA can predict hematoma progression in spontaneous intracerebral hemorrhageR. Wada, R. Aviv, A. Fox, S. SymonsSunnybrook Health Sciences CentrePurpose:The significant morbidity and mortality associated with spontaneous intracerebralhemorrhage has been shown to relate to hematoma progression. Gross contrast extravasationinto parenchymal hematomas has been previously associated with increased mortality. Wereport the presence of tiny enhancing foci within hematomas on CTA and hypothesize that theseare associated with hematoma progression.Materials and methods:CTAs performed in a 16 month period were reviewed to identify cases of intracerebralhemorrhage without a clearly defined vascular abnormality such as an arteriovenousmalformation/fistula, or major vessel aneurysm. Traumatic hemorrhages and hemorrhagictumors were excluded. Of the 44 cases, 13 patients were excluded for lack of follow-up. Allpatients underwent CTA as well as pre- and immediate post-contrast CT and had follow-upimaging 1-3 days later. A positive hematoma dot sign was defined as one or more 1-2 mm fociof enhancement within the clot on CTA.Hematoma volume was calculated using the previously validated (A x B x C)/2 methodand follow-up values were compared to those at initial presentation to evaluate for enlargement.The absolute and percentage change in hematoma volume was compared between patients withand without the focus of enhancement, and the presence of obvious contrast extravasation wasrecorded. Relevant clinical variables were also analyzed with respect to the presence of the dotsign.Results:31 patients were included in the study. 31% of patients were identified as having apositive hematoma dot sign. Initial hematoma size ranged from 4.4 to 108.1 cm 3 . Overall, 42%of patients experienced hematoma progression defined as an increase of >25% or 5 cm 3 .Hematomas with a positive dot sign were significantly more likely to progress in size (p

University of Toronto Department of Medical Imaging Research Day 2006CORRELATION BETWEEN NASAL DERMOID SINUS CYSTS ANDFORAMEN CECUM SIZEVeera BharatwalNeuroradiology FellowSupervisor: Dr. C. RaybaudBACKGROUND:Pre-surgical assessment with MRI/CT is required to determine if there is intracranial extensionof nasal dermoid sinus cysts, reported to occur in 4-45% of cases. Poorly powered or qualitativeCT studies have suggested that defects in the cribriform plate may predict intracranial extension.The purpose of this investigation was to determine if increased elliptical area of the foramencecum or if the presence of a bifid crista galli predict intracranial extension of nasal dermoidsinus cysts.METHODS:A retrospective analysis of all pediatric patients with pathologically proven nasal dermoid sinuscysts was performed, age 0-18 years, from Jan.1999-Jan. 2006 at a tertiary care pediatrichospital. Elliptical areas of foramen cecum and the presence of a bifid crista galli were obtainedfrom 2D CT examinations of nasal dermoid cases and age-matched controls.RESULTS:Thirty-three cases, (19 male and 14 female, mean age 29.5 months), met the study criteria.There were insufficient cases to compare intracranial extension versus extracranial nasal dermoidcyst data. Therefore, the two groups were combined and compared with age-matched controls.The mean area of ellipse for the nasal dermoid group was 7.4+/-4.6, (95% CI 5.2 – 9.5), and themean area of ellipse for the controls was 3.0+/-3.9, (95% CI 1.5 – 4.5), p=0.004 (student t-test).Regarding the association of bifid crista galli with nasal dermoids approached significance atp=0.056 (Fisher’s exact test).CONCLUSION:An enlarged foramen cecum correlates with the presence of a nasal dermoid cyst; however, dueto low sample size, it is not possible to demonstrate an association with intracranial extension.As well, a bifid crista galli shows some association with nasal dermoid cysts, but further datagathering is required, as well.49 of 58

University of Toronto Department of Medical Imaging Research Day 2006Vascular andInterventional Radiology50 of 58

University of Toronto Department of Medical Imaging Research Day 2006Radiofrequency ablation of renal cell carcinoma using a multitined expandable electrode:preliminary experienceKachura JRDivision of Vascular and Interventional Radiology, Department of Medical Imaging, University HealthNetwork and Mount Sinai Hospital, University of TorontoPurpose:To evaluate the short-term efficacy and safety of radiofrequency ablation (RFA) of renal cellcarcinoma (RCC) using a multitined expandable electrode.Materials and Methods:The medical records and imaging of all patients with RCC treated with RFA in 2004 and 2005were reviewed. The patient cohort consisted of 28 patients (23 male, 5 female) with a mean ageof 62 years (range 34-87 years). Each patient had a single RFA procedure under combinedultrasound and CT guidance under conscious sedation using a 3.5 cm or 4 cm array LeVeenelectrode. All patients except one had significant comorbidities (n=20), renal dysfunction (n=14),previous nephrectomy (n=6), and/or von Hippel Lindau syndrome (n=2). In total, 31 tumorswere ablated with a mean size of 2.6 cm (range 1.0-5.0 cm). There were 16 pathologicallyprovenclear cell RCCs, 5 papillary RCCs, 2 cystic RCCs, 2 oncocytic neoplasms, and 6 tumorswithout pathological confirmation. Successful ablation was defined as no tumor enhancementand no growth on follow-up contrast-enhanced CT and/or MRI.Results:All 31 tumors were successfully ablated, with a mean follow-up of 15 months (range 5 months -2 years). Two major complications (one abscess, another patient admitted for pain control) andtwo minor complications (two patients with leg pain from psoas injury despite thermal protectionwith air or dextrose) were encountered. One patient’s renal dysfunction continued to worsen atan unchanged rate, eventually requiring dialysis.Conclusions:RFA of RCC using a multitined expandable electrode is safe and effective. Long-term follow-upis required.51 of 58

University of Toronto Department of Medical Imaging Research Day 2006Use of Prophylactic Antibiotics for Implanted Chest Port Insertions: Is There ReducedRisk of Infection?R N Raha*, University Health Network, Toronto, Ontario, Canada. J R Beecroft, D K Rajan, MSimons, J R Kachura and K W Sniderman. Et alPurpose: To retrospectively review our experience with implanted chest wall venous accessdevices (ports) to determine if the use of preprocedural antibiotics influenced the infectiouscomplication rate.Materials and Methods: A retrospective study was conducted from January 1997 to July 2005for all chest ports placed by Interventional Radiologists. Data collected included patientdemographics, diabetic status, WBC count at insertion, indication for insertion, insertion site,date of first access and infectious complications within 30 days. The patient cohort consisted ofpatients ranging in age from 18 to 83 years (mean: 51.7 yrs). A total of 778 ports were includedwith most placed for chemotherapy for solid tumors (88.4%). Any documented infection within30 days with positive blood, skin/wound or port cultures was considered an infectiouscomplication resulting from the insertion procedure. Determinants for infection were evaluatedusing logistic regression analysis with a p-value of 0.05 considered the threshold of significance.Results: Technical success of insertions was 100%. A total of 11 (1.4%) culture-positiveinfections were documented within 30 days of insertion. The number of infectious complicationswas 9/680 for patients who did not receive antibiotics (1.3%) and 2/98 patients who did receiveantibiotics (2.%). The difference in infection rate for use of antibiotics was not statisticallysignificant (Fisher’s exact test p = 0.64). Among ports which were first accessed at the time ofinsertion, 1/356 (0.2%) developed an infection within 30 days, compared to 5/111 (4.5%) ofthose accessed between 1-5 days and 5/311 (1.6%) of those accessed after greater than 5 days.This difference was statistically significant (Fisher’s exact test p

University of Toronto Department of Medical Imaging Research Day 2006Ultrahigh vs. high-pressure PTA of venous anastomotic stenosis in HD grafts: Is there adifference in patency?DK Rajan, T Platzker, TW Clark, CE Lok, ME Simons, JR Beecroft, KT Tan, KW Sniderman.Department of Medical Imaging, UHN, Mt. Sinai Hospital, University of Toronto and New YorkUniversity Medical Center (TWC)Purpose:Ultrahigh-pressure (UHP) balloons (Conquest; Bard, Covington, Georgia) are becoming apreferred treatment for stenoses in dialysis grafts despite limited data demonstrating theirsuperiority to conventional angioplasty balloons. We retrospectively evaluated our experiencewith UHP compared to high-pressure (HP) balloon catheters to determine if there was adifference in intervention patency when performing percutaneous transluminal angioplasty(PTA) on venous anastomotic stenosis.Materials and Methods:A retrospective study was conduced from January 2001 to September 2005 and included patientswith synthetic hemodialysis grafts who underwent PTA of venous anastomotic stenosis. Patientswith stenosis in other portions of the venous access circuit or the arterial inflow were excluded.The study cohort consisted of 22 patients who underwent an average of 5 PTA s for a total of 110PTA events. Data collected included graft configuration and location, percent stenosis, balloontype used, residual stenosis, and transonic flow before and after intervention. Patency from timeof initial percutaneous angioplasty to next intervention was estimated using the Kaplan-Meier(KM) technique.Results:A total of 55 PTA s were performed in each group. Technical success was 96% (106/110) andclinical success was 100%. All 4 failures were identified as >30% residual stenosis with twofailures in the UHP group and two in the HP group. Using KM analysis, median survival for theUHP cohort was 4.6 months to next intervention and the HP cohort was 5.4 months to nextintervention (p=0.014, log rank test). The hazard ratio was 1.638 (95% CI 1.126-2.835)suggesting that use of the UHP balloon was associated with a 63% higher risk of loss of patency.The mean change in Qa (blood flow measured with Transonics; available for 71 events)following intervention was 264ml/min with UHP and 560ml/min with HP PTA (p

University of Toronto Department of Medical Imaging Research Day 2006IMAGE-GUIDED SURGERY IN THE PRESENCE OF MOTIONMarshall S. Sussman, Sheena Luu, Gal Sela, Bill Lin, Walter KucharczykDepartment of Medical Imaging, University of Toronto, UHNIn conventional surgical techniques, medical images are typically acquired prior to theprocedure. During the procedure, the surgeon uses the information in the images to guide thesurgery. In the surgeon’s mind, an association must be made between the anatomy seen in theimage, and the actual anatomy in the physical space of the body. Unfortunately, this mentalassociation is prone to error, particularly for small structures. Moreover, some structures seen inthe images may not be visible to the unaided eye, or may be obscured by overlying tissue.In image-guided surgery, a computer algorithm automatically makes the association betweenmedical images and physical anatomy. In theory, this reduces many of the sources of errordiscussed above. In practice, one remaining source of error that is not reduced is motionoccurring between the time of image acquisition, and the time of the surgical procedure. Motionmay be due to a variety of sources, including bulk patient movement, physiological motion (e.g.respiration), or surgery-induced tissue deformation. One possible method for addressing thisproblem would be to use intra-operative imaging. In this case, any changes in anatomicalorientation could be visualized directly. The problem with intra-operative imaging is that, ingeneral, images are of lower quality than the diagnostic scans normally used. As a result, it canbe difficult to visualize accurately the surgical target.In this work, a potential solution to the problem of motion will be presented. It combineselements of both diagnostic and intra-operative imaging. In particular, high-quality diagnosticimages are acquired prior to the surgical procedure. These are used to visualize the target.Lower-quality intra-operative images are acquired during the procedure. However, these imagesare used only visualize motion. An algorithm has been developed to quantify this motion, andthen subsequently to correct the high-quality images. Preliminary ex-vivo data will be presentedusing this technique for a biopsy procedure.54 of 58

University of Toronto Department of Medical Imaging Research Day 2006Distribution of Intraplaque Hemorrhage in Carotid Complicated Plaques Defined by High-Resolution MagneticResonance Direct Thrombus Imaging (hiresMRDTI)Richard Bitar, Alan R Moody, Sean Symons, General Leung, Jagdish Butany, Susan Crisp, Alexander Kiss, D James Sahlas,David Gladstone, Robert MaggisanoBackground: Intraplaque hemorrhage (IPH) is increasingly being recognized as one of the markers that definesatherosclerotic plaques as being at increased risk of causing symptomatic disease, as well as being a potential stimulus for theprogression of atherosclerosis (1-2). Classically, IPH has been thought to occur due to plaque ulceration, allowing flowingblood from the lumen to enter the plaque, resulting in IPH/thrombosis (2). However, a large amount of evidence currentlysuggests that IPH might be the result of rupture/leakiness of the fragile vasa vasorum (2), neovessels which originate from theadventitia of the artery and extend in the direction of the vessel lumen. We have successfully developed high-resolution MRItechniques that, by exploiting the T1 shortening effects of methemoglobin, directly visualize hemorrhage/thrombus inatherosclerotic plaque with 500μm isotropic resolution, allowing very good delineation of the exact location of intraplaquehemorrhage (3). Our hypothesis was that in carotid complicated plaques, IPH is due to vasa vasorum rupture/leakage, and themajority of IPH would be present deep to the vessel lumen(vessel wall/adventitial side). The purpose of this study was toevaluate the distribution of IPH within complicated carotidAFigure 1. Example of 16-segment comparative grid. HighresolutionMRDTI (hiresMRDTI). (A) Original image withoutplacement of grid. IPH is seen as regions of high signal intensity.(B) Image with grid placement. White-dashed lines outline thevessel wall and luminal boundaries. The dashed black linedivides the deeper (D) region from the superficial region (S).BSDatherosclerotic plaques.Methods: Fifteen patients (13 male, 2 female, mean age 71 ±6.1 years [59-80 years]) with symptomatic or asymptomaticcarotid artery stenosis were imaged at 1.5T (GE Twin SpeedMR scanner, USA) using high-resolution Magnetic ResonanceDirect Thrombus Imaging (hiresMRDTI) and a dedicatedcarotid surface coil (SCANMED, USA). The scanningparameters were: TR 11.2ms, TE 3.3 ms, flip angle 15 o , FOV80mm 2 , matrix 160 2 , slice thickness 1mm, spatial resolution0.5mm x 0.5mm x 0.5mm (IP). Fat suppression was achievedusing SPECIAL, a GE proprietary technique. The location ofIPH was assessed by tracing the boundaries of the vessel walland lumen. The region within these two boundaries,representing atherosclerotic plaque, was divided in half. Segments closer to the vessel wall represented the deep portions ofthe plaque, and segments closer to the lumen the superficial portions (Fig. 1). A 16-segment template was used for analysis.Some patients underwent carotid endarterectomy, and their excised plaques were analyzed. Endarterectomy specimens werefixed, decalcified, sectioned and stained with Hematoxylin & Eosin and Silver stain (to assess for plaque surface disruptions).Matching of MRI and histology slices employed thedistance from the bifurcation, and vessel/plaque70.0morphology. Statistical comparisons were performed usingchi-square.Results: A total of 158 axial images were reviewed. Nine-hundred and eighty-four segments contained IPH and were60.050.0used in the analysis. IPH was more frequently present in the40.0deep segments of the plaque (603/984 segments, 61.3%)%compared to the superficial segments (381/984 segments,30.0Deep38.7%) (p

University of Toronto Department of Medical Imaging Research Day 2006Disproportionate increase in cardiac output with exercise in patients with high flowperipheral arterio-venous malformationsS M CHENNAPRAGADA*,D PREISS, A MARDIMAE, K T TAN, D K RAJAN and M E SIMONS. Et al UNIVERSITYHEALTH NETWORK, UNIVERSITY OF TORONTOPurpose:Expand patient assessment by adding non-invasive respiratory CO 2 -based Fick cardiac output(CO) and oxygen consumption (VO 2 ) at rest and with exercise.Background:The indications for embolization of large peripheral arterio-venous malformations (AVM) arecontroversial, particularly if patients are only limited by mild to moderate symptoms.Materials and Methods:Seven patients with peripheral high flow arteriovenous malformations previously embolizedelsewhere were referred for repeat treatment and were assessed pre- and post embolization. COand VO 2 were also measured in a cohort of 10 healthy subjects. Patients performed mild exercisein the form of straight leg raising or walking slowly on a treadmill. Four patients were studiedpre- and post embolization. CO and VO 2 was repeated at least 4 times at 2 minute intervals andaveraged for each reading. Average CO was normalized for level of exercise by dividing by VO 2(nCO).Results:Other than one patient who presented pre embolization in high output failure (CO 17.9, and 5.9pre- and post embolization respectively), nCO at rest, pre- and post embolization, was the sameas the healthy cohort (p = 0.07). There was no difference in nCO pre- and post embolization (p =0.53). However, in all patients, nCO increased more with exercise than normal (slope CO/VO 2 =0.026 0.016 vs 0.013 0.003, p< 0.03).Conclusions:We use this observational study, to hypothesize that exercise intolerance in AVM patients is dueto a disproportional increase in CO in response to exercise compared to healthy subjects. COresponse to exercise appears to more sensitive test than CO at rest for AVM patients and mayeven provide a means to quantitate the shunt flow (as the difference in CO from that expected inhealthy subjects) and thus response to embolization.56 of 58

University of Toronto Department of Medical Imaging Research Day 2006Iatrogenic renal trauma: patterns of arterial injury and endovascular managementTaneja M, Kachura JR, Beecroft JR, Tan KT, Hayeems EB, Rajan DK, Simons ME, SnidermanKW, Ho CSDivision of Vascular and Interventional Radiology, Department of Medical Imaging, UniversityHealth Network and Mount Sinai Hospital, University of TorontoPurpose:To study patterns of iatrogenic renal arterial injuries, and to evaluate efficacy of embolotherapyin their management.Materials and Methods:Clinical presentation, angiographic findings, and outcome were retrospectively reviewed for 20patients (mean age 64.6 years, range 29-86, M=10, F=10) presenting with iatrogenic renalarterial injury from 1997 to 2005. The procedures causing injury were renal biopsy (n=9), partialnephrectomy (n=4), nephrostomy (n=3), percutaneous nephrolithotomy (n=2), and renalangioplasty (n=2).Results:Patients presented within one week of a percutaneous procedure, whereas post partialnephrectomy patients presented an average of 15 days after surgery. Solitary arterial injury wasseen in 13/20 patients. Multiple abnormalities were present in 7/20 patients. The commonestangiographic abnormality was localized contrast extravasation (n=11). Other vascular injuriesincluded pseudoaneurysms (n=8) and arteriovenous fistulae (n=7). All patients were treated withcoil embolization. Additional embolization material was required in 4 cases. Embolization wassuccessful in 19/20 (95%) patients. Success rate following coil embolization alone was 80%. Noprocedure related morbidity or mortality was identified on follow up.Conclusions:Embolotherapy is an effective treatment for iatrogenic renal arterial injuries. Coil embolizationalone is sufficient in most patients.57 of 58

University of Toronto Department of Medical Imaging Research Day 2006Department of Medical Imaging Research Day 2006has been made possible bythe generous support of GE Healthcare.58 of 58