27.02.2013 Views

Lower kV Imaging and Noise Reduction Techniques - Research ...

Lower kV Imaging and Noise Reduction Techniques - Research ...

Lower kV Imaging and Noise Reduction Techniques - Research ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

120 <strong>kV</strong><br />

CTDIvol = 24.5<br />

mGy<br />

<strong>Lower</strong> <strong>kV</strong> <strong>Imaging</strong> <strong>and</strong> <strong>Noise</strong><br />

<strong>Reduction</strong> <strong>Techniques</strong><br />

J. G. Fletcher, MD<br />

Mayo Clinic Rochester<br />

80 <strong>kV</strong><br />

CTDIvol = 5.2<br />

mGy<br />

With Imagebased<br />

Denoising


Disclosures<br />

• Grant support - Siemens, Genentech,<br />

Abbott, Centocor<br />

• Lifeng Yu<br />

• Shuai Leng<br />

Acknowledgments<br />

• Cynthia McCollough<br />

• Garrett Spencer<br />

• Amy Hara<br />

• Jeff Fidler<br />

• Jim Huprich<br />

• David Hough


http:// mayoresearch.mayo.edu/ctcic/educational-<br />

resources.csm


Overview<br />

•Dose <strong>Reduction</strong><br />

• mAs<br />

• Manual adaptation<br />

• AEC<br />

• <strong>kV</strong> selection<br />

•<strong>Noise</strong> reduction methods<br />

• Different approaches<br />

• Observer performance


The Gr<strong>and</strong> Scheme<br />

<strong>kV</strong> Selection & <strong>Noise</strong> <strong>Reduction</strong><br />

• Will not reduce dose as much as<br />

eliminating…<br />

� unjustified exams<br />

� superfluous acquisitions (e.g.,<br />

unenhanced, delayed)<br />

• Should facilitate (not hinder)<br />

accomplishment of diagnostic task<br />

• Performed with mAs reduction


Differences in Commercial<br />

Implementation of AEC<br />

From Yu et al. <strong>Imaging</strong> in Medicine 2009


Tube Current <strong>Reduction</strong><br />

Original Full Dose<br />

50% Dose Simulation<br />

75% Dose Simulation<br />

25% Dose Simulation


mAs <strong>Reduction</strong>:<br />

CT Enterography<br />

Routine Image Reconstruction<br />

∆ Qual Ref mAs 240 => 160 mAs<br />

Thicken Slices 2 => 3 mm


mAs <strong>Reduction</strong>:<br />

CT Enterography<br />

Routine Image Reconstruction<br />

Prior protocols uses 400 & 440 mA => NI 31<br />

Thicken slice 2.5 => 3.7 mm


Low <strong>kV</strong><br />

80 <strong>kV</strong> 120 <strong>kV</strong><br />

• Majority of abdominal CT scans: 120 <strong>kV</strong><br />

• It is possible to reduce to 80-90 <strong>kV</strong><br />

• Why should we use it?<br />

• When should we use it?<br />

• How can I use lower <strong>kV</strong> to reduce<br />

radiation dose?


(CT Number)<br />

Contrast<br />

300<br />

250<br />

200<br />

150<br />

100<br />

50<br />

0<br />

How does <strong>kV</strong> affect iodine<br />

enhancement?<br />

Iodine Contrast vs. <strong>kV</strong>p<br />

60 80 100 120 140 160<br />

<strong>kV</strong>p<br />

Patient<br />

size<br />

Small<br />

Medium<br />

Large<br />

From Bodily et al. RSNA 2008<br />

• Iodine att’n at 80<br />

<strong>kV</strong> twice that of<br />

140 <strong>kV</strong><br />

• Relative to iodine<br />

att’n at 120 <strong>kV</strong><br />

• 70% higher at<br />

80 <strong>kV</strong><br />

• 25% higher at<br />

100 <strong>kV</strong>


linear attenuation coefficient<br />

(1/cm)<br />

100000.0<br />

10000.0<br />

1000.0<br />

100.0<br />

10.0<br />

1.0<br />

0.1<br />

How does <strong>kV</strong> affect water<br />

enhancement?<br />

80<br />

<strong>kV</strong><br />

140<br />

<strong>kV</strong><br />

0 50 100 150<br />

x-ray energy (keV)<br />

Water<br />

Cortical Bone<br />

Pure Iodine<br />

Courtesy Dr. Lifeng Yu<br />

Bruesewitz et al. RSNA 2009<br />

• Relative contrast<br />

changes only hold<br />

for high atomic<br />

number<br />

substances<br />

• Iodine, barium<br />

• NOT water,<br />

soft tissue,<br />

calcium


Increased Contrast<br />

2 cc/s, 120 <strong>kV</strong> 2 cc/s, 80 <strong>kV</strong> (DS)<br />

2 weeks apart


Relative Contrast Differences due to<br />

Iodine Also Increase at Low <strong>kV</strong><br />

120 <strong>kV</strong> 100 <strong>kV</strong>


Improved Disease Conspicuity<br />

140 <strong>kV</strong> 80 <strong>kV</strong><br />

Macari M et al. AJR 2010


Relative Contrast Differences due to<br />

Iodine Also Increase at Low <strong>kV</strong><br />

80 <strong>kV</strong><br />

45 HU diff lesion<br />

lesion-liver liver<br />

120 <strong>kV</strong><br />

21 HU diff lesion<br />

lesion-liver liver


Reduced Radiation Dose<br />

120 <strong>kV</strong>, CTDI vol=5.18 mGy 100 <strong>kV</strong>, CTDI vol=3.98 mGy<br />

<strong>Lower</strong>-<strong>kV</strong> Benefits –<br />

Reduced Radiation Dose


Increased <strong>Noise</strong> <strong>and</strong>/or Artifacts<br />

140<strong>kV</strong><br />

80<strong>kV</strong>


Increased <strong>Noise</strong> & Artifacts<br />

40 cm lateral width 40 cm lateral width<br />

80 <strong>kV</strong> CTDI vol = 21.4 mGy 120 <strong>kV</strong> CTDI vol = 21.4 mGy


Benefits<br />

• Increased iodine<br />

signal<br />

• Increased relative<br />

contrast differences<br />

(conspicuity)<br />

• Ability to compensate<br />

for suboptimal IV<br />

contrast<br />

• Radiation dose<br />

reduction<br />

Low <strong>kV</strong><br />

Risks<br />

• Increased noise<br />

• Increased potential<br />

for artifacts<br />

• At 80 <strong>kV</strong>, tube<br />

current limits or pitch<br />

can be a problem<br />

• Minor effect – focal<br />

spot size is slightly<br />

larger<br />

Individual assessment of each patient <strong>and</strong> diagnostic task


<strong>kV</strong> <strong>Reduction</strong>: Clinical Use<br />

• Limited IV access<br />

• Limited contrast dose<br />

• Subtle attenuation differences<br />

===================================<br />

• Young patients<br />

• Small <strong>and</strong> medium-sized adult patients<br />

(lateral width from CT scout ≤ 41 cm)<br />

Can I do both?<br />

Dosematched<br />

Dosereduced


Low <strong>kV</strong><br />

Patient Selection<br />

• Issue is noise (patient size)<br />

• Organ of interest<br />

• Measurements of size<br />

• 116 pts undergoing 80 <strong>kV</strong> CT<br />

• 2 – 3 mm slices<br />

• IQ, artifact, confidence<br />

• Multiple pt size measures


Low <strong>kV</strong><br />

Patient Selection<br />

Patient Size <strong>and</strong> Acceptability<br />

• Lateral width the best predictor of acceptable<br />

image quality<br />

< 36 cm => 80 <strong>kV</strong> imaging acceptable<br />

< 41 cm => 100 <strong>kV</strong> imaging acceptable<br />

• Larger patients may not be able to undergo low<br />

<strong>kV</strong> imaging<br />

• Patient size selection only insures good quality


Dose-matched Low <strong>kV</strong><br />

To Preserve Image Quality<br />

Restaging unresectable Islet Cell tumor<br />

Chest CT at 80 seconds (to avoid compromise of abdominal timing)<br />

100 <strong>kV</strong> Chest


Dose-matched Low <strong>kV</strong><br />

To Preserve Image Quality<br />

Unresectable pancreatic cancer with foot swelling<br />

3 minute delayed 80 <strong>kV</strong> images show CFV clot<br />

Portal phase image


Dose-matched Low <strong>kV</strong><br />

To Preserve Image Quality<br />

Triple-phase 100 <strong>kV</strong><br />

enterography with 80 cc<br />

Omnipaque 300 due to solitary<br />

kidney<br />

Reduced IV Contrast<br />

80 <strong>kV</strong> CT enterography<br />

CR = 1.4, 80 cc Omnipaque 300 w/ NS flush


Dose-matched Low <strong>kV</strong><br />

To Preserve Image Quality<br />

Reduced Contrast Injection Rate<br />

80 <strong>kV</strong> enterography 2° to compromised venous pedal access (2 cc / s)<br />

SCAN AT 85 seconds


Dose-matched Low <strong>kV</strong><br />

To Preserve Image Quality<br />

• Locate a dose-matched <strong>kV</strong> table for your<br />

scanner (two are located at the end of this presentation)<br />

• Perform 120 <strong>kV</strong> topo & insure width at<br />

liver < 41 cm<br />

• Prescribe scan volume & record CTDI vol<br />

• Change <strong>kV</strong> to 100 <strong>kV</strong><br />

• Plug in dose-matched mA from table<br />

• If 2 cc/s, delay scanning to 85 sec for<br />

portal phase


Low <strong>kV</strong> to <strong>Lower</strong> Radiation Dose<br />

120 <strong>kV</strong><br />

17.3 mGy<br />

100 <strong>kV</strong><br />

7.71 mGy<br />

• <strong>kV</strong> <strong>and</strong> dose-matched mAs to get the lower dose<br />

• Iodine CNR allows you to give up dose & tolerate noise


Low <strong>kV</strong> <strong>Imaging</strong> While Reducing Dose<br />

• Need to consider both patient size <strong>and</strong><br />

diagnostic task into <strong>kV</strong> selection process<br />

• Greater the iodine contrast differences,<br />

the greater ability to reduce dose for<br />

smaller pts<br />

• <strong>kV</strong> selection combined with lowering mAs<br />

(from dose-matched mAs)<br />

• Creates a new technique chart for each<br />

diagnostic task<br />

L. Yu, H. Li, J. Fletcher, C. McCollough, Medical Physics, 37(1), 2010.


General Strategy for <strong>kV</strong> selection<br />

• Iodine CNR is a not a sufficient image quality<br />

index<br />

• Must also constrain image noise<br />

• We already know what noise levels we can<br />

tolerate<br />

CNR ≥<br />

CNRref<br />

, <strong>and</strong> σ ≤ α •<br />

ref<br />

L. Yu, H. Li, J. Fletcher, C. McCollough, Medical Physics, 37(1), 2010.<br />

σ


<strong>Lower</strong> <strong>kV</strong> for Dose <strong>Reduction</strong><br />

• 80 <strong>kV</strong> prescribed using vendor <strong>kV</strong> selection tool<br />

• mA chosen to stay within noise limits<br />

• 25% dose reduction


General Strategy for <strong>kV</strong> selection<br />

Consider 80 <strong>kV</strong> imaging<br />

Contrast by 70%<br />

Here’s the idea<br />

Improved contrast permits the noise level to increase<br />

L. Yu, H. Li, J. Fletcher, C. McCollough, Medical Physics, 37(1), 2010.


General Strategy for <strong>kV</strong> selection<br />

Consider 80 <strong>kV</strong> imaging<br />

Contrast by 70%<br />

<strong>Noise</strong> by 70%<br />

Here’s the idea<br />

L. Yu, H. Li, J. Fletcher, C. McCollough, Medical Physics, 37(1), 2010.


General Strategy for <strong>kV</strong> selection<br />

Consider 80 <strong>kV</strong> imaging<br />

Contrast by 70%<br />

<strong>Noise</strong> by 70%<br />

Here’s the idea<br />

≥ Contrast 120<br />

<strong>Noise</strong> 120<br />

L. Yu, H. Li, J. Fletcher, C. McCollough, Medical Physics, 37(1), 2010.


General Strategy for <strong>kV</strong> selection<br />

Consider 80 <strong>kV</strong> imaging<br />

Contrast by 70%<br />

<strong>Noise</strong> by 70%<br />

Here’s the idea<br />

≥ Contrast 120<br />

<strong>Noise</strong> 120<br />

L. Yu, H. Li, J. Fletcher, C. McCollough, Medical Physics, 37(1), 2010.


General Strategy for <strong>kV</strong> selection<br />

Consider 80 <strong>kV</strong> imaging<br />

Contrast by 70%<br />

<strong>Noise</strong> by 70%<br />

Here’s the idea<br />

≥ Contrast 120<br />

<strong>Noise</strong> 120<br />

Increased noise permits the dose reduction<br />

L. Yu, H. Li, J. Fletcher, C. McCollough, Medical Physics, 37(1), 2010.


General Strategy for <strong>kV</strong> selection<br />

Consider 80 <strong>kV</strong> imaging<br />

Contrast by 70%<br />

<strong>Noise</strong> by 60%<br />

Here’s the idea<br />

≥ Contrast 120<br />

<strong>Noise</strong> 120<br />

As patients get larger (or task requires less noise), the<br />

acceptable increase noise (σ) becomes smaller<br />

L. Yu, H. Li, J. Fletcher, C. McCollough, Medical Physics, 37(1), 2010.


General Strategy for <strong>kV</strong> selection<br />

Consider 80 <strong>kV</strong> imaging<br />

Contrast by 70%<br />

<strong>Noise</strong> by 50%<br />

Here’s the idea<br />

≥ Contrast 120<br />

<strong>Noise</strong> 120<br />

As patients get larger (or task requires less noise), the<br />

acceptable increase noise (σ) becomes smaller<br />

L. Yu, H. Li, J. Fletcher, C. McCollough, Medical Physics, 37(1), 2010.


General Strategy for <strong>kV</strong> selection<br />

Consider 80 <strong>kV</strong> imaging<br />

Contrast by 70%<br />

<strong>Noise</strong> by 40%<br />

Here’s the idea<br />

The dose reduction will also decline<br />

≥ Contrast 120<br />

<strong>Noise</strong> 120<br />

As patients get larger (or task requires less noise), the<br />

acceptable increase noise (σ) becomes smaller


• Math is involved<br />

Low <strong>kV</strong><br />

Reduce Radiation Dose<br />

• Two pathways<br />

• Technique charts that prescribe both<br />

<strong>kV</strong> <strong>and</strong> mA for a given task<br />

• Commercial <strong>kV</strong> selection tools<br />

• Over a large number of patients, dose is<br />

reduced by about 20 - 30%


CTDIvol (mGy) per 100 effective mAs<br />

14<br />

12<br />

10<br />

8<br />

6<br />

4<br />

2<br />

0<br />

GE VCT CTDIvol/100mAs on scanner<br />

GE VCT CTDIvol/100 mAs from square conversion (use 120<strong>kV</strong><br />

as reference)<br />

Siemens Flash CTDIvol/100mAs on scanner<br />

Siemens Flash CTDIvol/100mAs from square conversion (use<br />

120<strong>kV</strong> as reference)<br />

<strong>kV</strong>2 <strong>kV</strong>2 <strong>kV</strong>2 <strong>kV</strong>2 <strong>kV</strong>2 <strong>kV</strong>2 <strong>kV</strong>2 <strong>kV</strong>2 80 100 120 140<br />

Tube Potential (<strong>kV</strong>)<br />

The widely used relation “Radiation output CTDIvol is proportional to <strong>kV</strong>p 2 for the same<br />

mAs” is not accurate.<br />

As shown above, the actual CTDIvol at 80 <strong>kV</strong>p is about ~50% lower on both GE <strong>and</strong><br />

Siemens scanners for the lower <strong>kV</strong>’s


Low <strong>kV</strong><br />

Reduce Radiation<br />

Dose<br />

Dose reduction percentage<br />

Practice<br />

Implementation<br />

40.0<br />

35.0<br />

30.0<br />

25.0<br />

20.0<br />

15.0<br />

10.0<br />

5.0<br />

0.0<br />

28 33 38 43 48<br />

Later width (cm)<br />

16 – 30% dose savings<br />

Depending on Pt size<br />

<strong>kV</strong> 80<br />

<strong>kV</strong> 100<br />

Right to Left<br />

Patient Width <strong>kV</strong> Relative Dose Factor<br />

25 80 0.65<br />

26 80 0.66<br />

27 80 0.68<br />

28 80 0.69<br />

29 80 0.71<br />

30 80 0.73<br />

31 80 0.73<br />

32 100 0.74<br />

33 100 0.74<br />

34 100 0.75<br />

35 100 0.75<br />

36 100 0.78<br />

37 100 0.81<br />

38 100 0.83<br />

39 100 0.86<br />

40 100 0.89<br />

41 100 0.92<br />

42 100 0.94<br />

43 100 0.97<br />

44 120 1<br />

45 120 1<br />

46 120 1<br />

47 120 1<br />

48 120 1<br />

49 120 1<br />

50 120 1<br />

51 120 1<br />

52 120 1<br />

53 120 1<br />

54 140 0.96<br />

55 140 0.87


Auto <strong>kV</strong> Look-up Table<br />

CT Enterography, 80 <strong>kV</strong><br />

CTDIvol 5.5 mGy<br />

Patient lateral width: 25 cm (liver)<br />

30 cm (pelvis)<br />

Dose reduction 31%


Commercial <strong>kV</strong> Selection<br />

• Based on iCNR <strong>and</strong> task-specific noise<br />

levels<br />

• Takes complex relationship of pitch, <strong>kV</strong>,<br />

generator size (kW), <strong>and</strong> tube current<br />

(mA) limits into account


Radiation dose reduction<br />

Commercial <strong>kV</strong> Selection<br />

50%<br />

40%<br />

30%<br />

20%<br />

10%<br />

0%<br />

-10%<br />

-20%<br />

80 <strong>kV</strong> 100 <strong>kV</strong> 120 <strong>kV</strong> 140 <strong>kV</strong><br />

Routine Abdominal CTA<br />

• <strong>kV</strong> changed in 73% cases<br />

•Overall 30% dose reduction, but depends on<br />

patient size<br />

• iCNR <strong>and</strong> EQC identical in subset with<br />

comparisons @ 120 <strong>kV</strong> despite dose savings<br />

Courtesy Dr. Lifeng Yu<br />

50<br />

45<br />

40<br />

35<br />

30<br />

25<br />

Patient lateral width (cm)<br />

Radiation dose reduction<br />

Routine Abdominal CT<br />

•Overall 20% dose reduction, but depends on<br />

patient size<br />

•iCNR <strong>and</strong> EQC identical in subset with<br />

comparisons @ 120 <strong>kV</strong> despite dose savings<br />

60%<br />

50%<br />

40%<br />

30%<br />

20%<br />

10%<br />

0%<br />

43<br />

scans<br />

75<br />

scans<br />

44<br />

scans<br />

162<br />

scans<br />

80 <strong>kV</strong> 100 <strong>kV</strong> 120 <strong>kV</strong> Overall


100 <strong>kV</strong>, 23% dose reduction c/w 120 <strong>kV</strong><br />

100 <strong>kV</strong>, 26% dose reduction c/w 120 <strong>kV</strong>


120 <strong>kV</strong><br />

(CTDIvol 18.89 mGy)<br />

100 <strong>kV</strong><br />

(CTDIvol 7.13 mGy)<br />

Routine Reconstruction<br />

More dramatic dose reductions can be achieved if<br />

we permit noise levels to increase further


mAs <strong>Reduction</strong>:<br />

CT Enterography<br />

Sensation 64<br />

Siemens Scanners<br />

∆ Qual Ref mAs 240 => 160 mAs<br />

Thicken Slices 2 => 3 mm<br />

Table replaced with Care<strong>kV</strong><br />

(vendor <strong>kV</strong> selection tool) where<br />

available


mAs <strong>Reduction</strong>:<br />

CT Enterography<br />

GE Scanners<br />

Prior protocols uses 400 & 440 mA<br />

Thicken slice 2.5 => 3.7 mm


<strong>Noise</strong> <strong>Reduction</strong><br />

to Reduce Image <strong>Noise</strong><br />

(<strong>and</strong> make lower dose more pleasant)<br />

120 <strong>kV</strong><br />

18.89 mGy<br />

100 <strong>kV</strong><br />

7.13 mGy<br />

100 <strong>kV</strong> + IRIS<br />

7.13 mGy<br />

Routine dose <strong>and</strong> noise<br />

Excessive noise<br />

<strong>Lower</strong> dose <strong>and</strong> similar<br />

noise<br />

Image-based noise reduction (IRIS, Siemens Medical Solutions)


<strong>Noise</strong> <strong>Reduction</strong><br />

to Reduce Image <strong>Noise</strong><br />

(<strong>and</strong> make lower dose more pleasant)<br />

Half-Dose Denoised B43 Full Dose


Reduce Dose <strong>and</strong> Make Diagnosis<br />

Adapt for Patient <strong>and</strong> Diagnostic Task<br />

mA, <strong>kV</strong>, <strong>Noise</strong> <strong>Reduction</strong><br />

Full dose – 120 <strong>kV</strong> & 240 Qual Ref mAs Half dose – 120 <strong>kV</strong> & 120 Qual Ref mAs<br />

+ <strong>Noise</strong> <strong>Reduction</strong>


<strong>Noise</strong> <strong>Reduction</strong><br />

Increases Fidelity to a Higher Dose Image<br />

Routine Dose + FBP<br />

Daniel 69305159<br />

Low Dose + ASIR<br />

15 mSv 6 mSV<br />

Iterative <strong>Noise</strong> <strong>Reduction</strong> Method (ASIR, GE Healthcare)<br />

Courtesy Dr. Amy Hara


<strong>Noise</strong> <strong>Reduction</strong> Methods<br />

Iterative <strong>Noise</strong> <strong>Reduction</strong> Methods (ASIR)<br />

• Singh, Kalra, Hsieh et al. Radiology 2010; 257:<br />

373 - 383<br />

• 22 pts<br />

• 4 additional scans @ 50 – 200 mAs,<br />

reconstructed with FBP & 30 – 70 % ASIR<br />

• No lesions missed on FBP or ASIR images<br />

• Significant improvement in noise, IQ,<br />

conspicuity at lowest dose level<br />

• No loss of contrast or sharpness


Image <strong>Noise</strong> <strong>Reduction</strong><br />

Common Themes<br />

• <strong>Noise</strong> reduction can improve image quality &<br />

confidence*<br />

• There are always trade-offs<br />

• <strong>Noise</strong> <strong>and</strong> dose reduction<br />

• Performance can be preserved for a variety<br />

of tasks at very low doses**<br />

• Unclear if CT noise reduction improves<br />

observer performance<br />

* Singh Radiology 2010, Sagara AJR 2010, Prakash Inves Radiol 2010<br />

** Allen AJR 2009, Kambadakone AJR 2009, Siddiki Inflamm Bowel Dz 2009


Finding the Low-dose Sweet Spot<br />

• Finding the low dose “sweet spot” the radiologist’s responsibility<br />

• Reducing the noise without artifacts or minimizing conspicuity of<br />

subtle structures <strong>and</strong> lesions is the manufacturer responsibility<br />

Courtesy Dr. Cynthia McCollough


Finding the Low-dose Sweet Spot<br />

?<br />

Courtesy Dr. Cynthia McCollough


The Challenge<br />

Original Dose ½ Dose + Half <strong>Noise</strong> Dose <strong>Reduction</strong><br />

Are we missing anything?<br />

Is <strong>Noise</strong> <strong>Reduction</strong> Improving Diagnosis?


Denoising <strong>and</strong> Dose <strong>Reduction</strong><br />

Observer Performance<br />

Lee, Park, et al. AJR 2011 (in press)<br />

• 92 pts<br />

• FD, ½ dose, ½ dose with<br />

noise reduction<br />

• ½ dose = 3.5 mGy CTDIvol<br />

• Evaluated imaging findings<br />

of inflammation at TI<br />

• 1/2 dose with <strong>and</strong> without<br />

noise reduction found<br />

agreement with full dose in ><br />

85% of cases


There are other reasons to<br />

consider or adopt noise<br />

reduction<br />

Image quality,<br />

Confidence & Fatigue


<strong>Noise</strong> <strong>Reduction</strong><br />

Trade-offs<br />

<strong>Noise</strong><br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

Full Dose CT + FBP<br />

0<br />

0<br />

-0.1<br />

1 2 3 4 5 6 7 8 9 10<br />

Spatial Frequency (lp / cm)<br />

Courtesy Dr. Amy Hara


<strong>Noise</strong><br />

<strong>Noise</strong> <strong>Reduction</strong><br />

Trade-offs<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

Half Dose CT + FBP<br />

Full Dose CT + FBP<br />

0 1 2 3 4 5 6 7 8 9 10<br />

-0.1<br />

Spatial Frequency (lp / cm)<br />

Courtesy Dr. Amy Hara


<strong>Noise</strong><br />

<strong>Noise</strong> <strong>Reduction</strong><br />

Trade-offs<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

-0.1<br />

Half Dose CT + FBP<br />

Half Dose CT + 40% ASIR<br />

Full Dose CT + FBP<br />

0 1 2 3 4 5 6 7 8 9 10<br />

Spatial Frequency (lp / cm)<br />

Courtesy Dr. Amy Hara


Effect of % ASIR<br />

No ASIR = FBP 40-50% ASIR 100% ASIR<br />

NOISE<br />

Same dataset, CTDI=9<br />

ASIR<br />

Courtesy Dr. Amy Hara


<strong>Noise</strong> <strong>Reduction</strong><br />

Trade-offs<br />

• <strong>Noise</strong> texture – blotchiness or pixelated<br />

• Can be minimized by<br />

• Strength or blending settings<br />

• Selection of dose<br />

• Depends on method, but often minor <strong>and</strong><br />

reader effect idiosyncratic<br />

Singh et al. Radiology 2010<br />

Prakash Invest Radiol 2010<br />

Lee et al. AJR 2011 (in press)


<strong>Noise</strong> <strong>Reduction</strong> Strategies<br />

• Reconstruction kernel<br />

• Image <strong>and</strong> projection-space<br />

denoising<br />

• Iterative reconstruction<br />

• Iterative noise reduction<br />

methods


Image Reconstruction<br />

Acquisition<br />

Projection Data<br />

PS Filter<br />

Image<br />

Reconstruction<br />

• Filtered Back<br />

Projection<br />

(reconstruction<br />

kernel)<br />

• Iterative<br />

Reconstruction<br />

Iterative <strong>Noise</strong> <strong>Reduction</strong> Methods (ASIR,<br />

SAFIRE)<br />

CT image<br />

Imagespace<br />

<strong>Noise</strong><br />

<strong>Reduction</strong><br />

Filtered CT<br />

image


<strong>Noise</strong> <strong>Reduction</strong><br />

Reconstruction Kernel in Analytical Reconstruction<br />

Original 3D Edge Preserving<br />

<strong>Noise</strong> <strong>Reduction</strong><br />

61.5 HU<br />

- 41%<br />

36.3 HU<br />

Courtesy of R. Raupach


<strong>Noise</strong> <strong>Reduction</strong><br />

Reconstruction Kernel in Analytical Reconstruction<br />

B43 B40


<strong>Noise</strong> <strong>Reduction</strong><br />

Image-space Denoising<br />

• Linear or non-linear filters directly applied<br />

on the reconstructed images<br />

• Tend to change the appearance of CT<br />

images (noise texture)<br />

• Potentially to sacrifice low-contrast<br />

detectability, if the filter strength is not<br />

well controlled<br />

• Performance requires careful evaluation for<br />

each diagnostic task


Integration of <strong>Noise</strong> <strong>Reduction</strong> on a<br />

Departmental Basis<br />

PACS<br />

<strong>Noise</strong> <strong>Reduction</strong><br />

Within Image<br />

Reconstruction System


Integration of <strong>Noise</strong> <strong>Reduction</strong> on a<br />

Departmental Basis<br />

Image-based <strong>Noise</strong> <strong>Reduction</strong><br />

PACS


Comparison of 4 image-based noise reduction methods<br />

(1/2 dose CTE)<br />

Original B40<br />

IRIS I40<br />

SharpView A81<br />

IRIS2 B30


Differences in Image-based <strong>Noise</strong> Filters<br />

SharpView Careful <strong>Noise</strong> <strong>Reduction</strong> SharpView Subtraction<br />

Mayo Image-based <strong>Noise</strong> Filter* Mayo Image-based <strong>Noise</strong> Filter*


Image-based Denoising<br />

Image Quality Improvement


Projection Space Denoising<br />

Bilateral filtering - smoothes the image,<br />

with weights determined according to<br />

• Proximity & intensity similarity<br />

• Models <strong>and</strong> predicts noise<br />

• Useful to consider benefit of noise<br />

reduction with lower <strong>kV</strong><br />

M<strong>and</strong>uca et al. Med Phys 2009


Projection Space Denoising<br />

80 <strong>kV</strong> Non-Denoised CTE<br />

80 <strong>kV</strong> Projection-space Denoised<br />

CTE<br />

Diagnostic quality achieved by reducing dose by half with <strong>kV</strong>/mAs reduction<br />

followed by noise reduction<br />

Guimaraes et al. Acad Radiol 2010


Projection Space Denoising<br />

B43 Denoised 80<strong>kV</strong><br />

Low <strong>kV</strong> Low dose + PS DN<br />

B40 Mixed <strong>kV</strong><br />

3/3 readers rated conspicuity same/greater for ½ dose low <strong>kV</strong> with PS DN


Full dose blended Half dose 80 <strong>kV</strong><br />

80 <strong>kV</strong> + PS 80 <strong>kV</strong> + PS/IS<br />

80 <strong>kV</strong> + PS/MBF<br />

Paulsen et al ARC 2010


Iterative Reconstruction<br />

• Models CT system geometry (e.g., beam spectrum,<br />

noise, beam hardening effect, scatter <strong>and</strong> incomplete<br />

data sampling).<br />

• Different degrees of credibility among projection data<br />

are allowed<br />

• <strong>Noise</strong> models based on photon statistics <strong>and</strong> other<br />

physical properties of the data acquisition<br />

• May improve spatial resolution <strong>and</strong> reduce image<br />

artifacts such as beam hardening, windmill, <strong>and</strong> metal<br />

artifacts<br />

• Can be independent of analytical reconstruction<br />

• High computation load<br />

• Usually used in a misleading fashion for marketing


Iterative reconstruction<br />

Projection<br />

Data<br />

Compare<br />

& update<br />

Simulated<br />

Projection data<br />

Iteration loop:<br />

Optimize an<br />

objective function<br />

Projection<br />

(System model)<br />

Modified<br />

Data<br />

Image<br />

Backprojection<br />

Backprojection<br />

Final<br />

Image


Iterative Reconstruction<br />

Advantages:<br />

Physical effects <strong>Noise</strong> Models<br />

Beam spectrum<br />

<strong>Noise</strong><br />

Beam hardening<br />

Scatter<br />

Incomplete sampling<br />

More accurate<br />

<strong>Lower</strong> image noise<br />

Improve spatial<br />

resolution<br />

Reduce artifacts


69895464<br />

Ultra low dose with MBIR<br />

MBIR = Model based iterative reconstruction<br />

Example: CT at 10 mAs (routine = 200 mAs)<br />

FBP 50% ASIR<br />

MBIR<br />

64 x 0.625, helical pitch 1, 120<strong>kV</strong>p<br />

Same pt, same scan<br />

(MBIR, GE Healthcare) Courtesy Dr. Amy Hara


Iterative reconstruction<br />

Potential for Practice<br />

Full Dose 50% Dose 60% Dose 80% Dose<br />

Pilot study<br />

21 pts, proven liver <strong>and</strong> pancreatic lesions, 3 readers<br />

IQ <strong>and</strong> lesion conspicuity scores overall better using Prototype IR reconstruction<br />

at 50% dose<br />

Ehman et al. SGR 2010


Iterative <strong>Noise</strong> <strong>Reduction</strong> Methods<br />

that Utilize Projection Data<br />

• ASIR (Adaptive Statistical Iterative<br />

Reconstruction, GE Healthcare)<br />

• Takes into account photon statistics<br />

<strong>and</strong> image noise<br />

• Blends ASIR & FBP image to userprescribed<br />

degree<br />

• FDA approved <strong>and</strong> greatest clinical<br />

experience


Iterative <strong>Noise</strong> <strong>Reduction</strong><br />

Impact on Implementation & Image Quality<br />

• Sagara, Hara, Pavlicek, et al. AJR 2010; 195:<br />

713 – 719<br />

• 53 pts with prior CT exams<br />

• Compared dose & image quality utilizing a<br />

lower-dose acquisition (NI: 22 → 31)<br />

followed by recon using 40% ASIR<br />

• Overall 33% reduction in dose (25 → 17 mGy<br />

CTDIvol)<br />

• <strong>Lower</strong>-dose ASIR: ↓noise, ↓sharpness, =<br />

diagnostic acceptability


Iterative <strong>Noise</strong> <strong>Reduction</strong><br />

Impact on Implementation & Image Quality<br />

17 mGy CTDIvol<br />

40% ASIR<br />

3.75 mm slices<br />

27 mGy CTDIvol<br />

FBP<br />

From Sagara, Hara, Pavlicek, et al. AJR 2010; 195: 713 – 719


Low Dose CTE with ASIR: 4 mSv<br />

43 yo F BMI = 19<br />

Courtesy Dr. Amy Hara


Iterative <strong>Noise</strong> <strong>Reduction</strong><br />

Impact on Implementation & Image Quality<br />

• Prakash, Kalra, Kambadakone et al. Invest Radiol 2010;<br />

45: 202 – 210<br />

• Retrospective, case-control design<br />

• Used weight-based AEC settings, with NI increased<br />

for the patients on whom 40% ASIR was performed<br />

• 156 ASIR patients, 66 FBP patients<br />

• Overall 26% dose reduction for lower AEC settings<br />

• ↓noise with ASIR<br />

• At this level of dose reduction, ASIR images had ↑<br />

IQ in large pts, = IQ in smaller pts


Iterative <strong>Noise</strong> <strong>Reduction</strong><br />

Impact on Implementation & Image Quality<br />

Flicek et al. AJR 2010<br />

• Phantom & human study<br />

(18 pts)<br />

• 50 mAs supine vs. 25 mAs<br />

prone + ASIR<br />

• <strong>Lower</strong> dose ASIR<br />

acquisition – no difference in<br />

2D or 3D IQ


Iterative <strong>Noise</strong> <strong>Reduction</strong> Methods<br />

that Utilize Projection Data<br />

• SAFIRE (Sinogram-affirmed IR, Siemens<br />

Medical Solutions)<br />

• Iteratively compares an image in<br />

which noise has been detected &<br />

subtracted with model-based forward<br />

projection<br />

• Under FDA review, not approved<br />

currently


Iterative <strong>Noise</strong> <strong>Reduction</strong><br />

Impact on Implementation & Image Quality<br />

• Krueger et al. SGR 2011<br />

• 40 pts, 3 readers<br />

• Comparison of full-dose, ½ dose, ½ dose with<br />

SAFIRE (CTDIvol = 13.8 <strong>and</strong> 6.8 mGy)<br />

• Assessment: IQ, Organ detection & confidence,<br />

significant findings<br />

• All diagnostically acceptable<br />

• IQ: Full > ½ dose (in pelvis: ↑SAFIRE)<br />

• Reader confidence: idiosyncratic<br />

• Significant Findings: varied by reader, not dose or<br />

reconstruction


Improved Image Quality Does ≠ Improved Observer Performance<br />

Subtle Dz Neo-terminal ileum <strong>and</strong> Perianal Fistula Can be seen on half –dose ± SAFIRE (2 mm slice<br />

thickness, corresponding to 3.3 mGy), even though IQ markedly improved


CT Dose <strong>Reduction</strong>:<br />

Implementation into Practice<br />

2 mm slice<br />

Routine wFBP<br />

2 mm slice<br />

SAFIRE<br />

19 yo female<br />

Crohn’s<br />

CTDI vol 3.49 mGy


Comparison of <strong>Noise</strong> <strong>Reduction</strong> Methods<br />

IRIS I40 SharpView A81<br />

Original B40<br />

Projection Space 1 SAFIRE<br />

Projection Space 2


<strong>Noise</strong> <strong>Reduction</strong><br />

Future<br />

• Observer performance studies at lower<br />

doses<br />

• Improved methods available<br />

38 yo female<br />

Crohn’s<br />

CTDIvol 6.5 mGy<br />

St<strong>and</strong>ard 13 mGy<br />

PS DN


Conclusions<br />

• <strong>Lower</strong> dose abdominal CT can be<br />

achieved through a variety of<br />

techniques<br />

• Technique charts, reduced mAs,<br />

AEC, <strong>kV</strong> selection<br />

• Noisier images can be interpreted<br />

without loss of diagnostic accuracy<br />

• <strong>Noise</strong> reduction methods improve<br />

image quality of low dose images


Conclusions<br />

• Observer performance studies<br />

needed to define<br />

• benefit of noise reduction at low<br />

doses<br />

• extent of dose reduction possible<br />

Routine B40 kernel<br />

CTDIvol = 3.5 mGy<br />

Iterative <strong>Noise</strong> <strong>Reduction</strong> Method


http:// mayoresearch.mayo.edu/ctcic/educational-<br />

resources.csm


Tube Current <strong>Reduction</strong><br />

Original Full Dose<br />

50% Dose Simulation<br />

75% Dose Simulation<br />

25% Dose Simulation

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!