njit-etd2003-081 - New Jersey Institute of Technology

njit-etd2003-081 - New Jersey Institute of Technology njit-etd2003-081 - New Jersey Institute of Technology

archives.njit.edu
from archives.njit.edu More from this publisher
20.01.2015 Views

227 5.6 Principal Component Analysis As mentioned in section 3.14, the principal components were arranged in order of decreasing variance with the most informative principal component listed first, and the least informative listed last. The dimensionality of the problem was reduced, i.e., reduces the number of variables without losing much of the information. Thus instead of analyzing a large number of the original variables with complex interrelationships, the data could be analyzed by a small number of uncorrelated principal components. Figure 5.60 presented a principal component plot of the data set containing 48 COPD and 8 normal subjects at rest. The data set gave the following principal components (PC): 1 st PC: LF_coh_HR_rsp — Coherence of HR and respiration signals in the LF range (0.04 - 0.15 Hz). 2nd PC: HF _ coh _ BP _rsp — Coherence of BP and respiration signals in the HF range (0.15 - 0.4 Hz). 3 rd PC: rsp — The respiration rate. This indicated that the largest variance between COPD and normal subjects was the interrelationships between heart rate and respiration. The second largest variance was the interrelationships between blood pressure and respiration. The third largest variance was the respiration rate itself.

228 Figure 5.60 Normal and COPD classification using principal component analysis (PCA). Note: (o: Normal; +: COPD) Table 5.6 Principal Components from Normal and COPD Data P1 P2 P3 Categories 0.0351 -0.3625 0.3099 rsp 0.2960 -0.1943 0.0928 HR 0.2729 -0.2367 0.1637 BP 0.3800 0.0271 -0.0793 LF_coh HR_rsp 0.2725 -0.0886 -0.3849 LF_coh_HR_BP 0.3349 0.2178 -0.0244 LF coh BP rsp 0.3416 0.0691 -0.0556 HF_coh_HR_rsp 0.1389 0.1252 -0.5128 HF cohHRBP 0.1924 0.4961 0.0194 HF_coh BP rsp 0.3386 -0.2607 0.0482 LF_pcoh_HR_rsp 0.0449 0.0540 -0.0351 LFpcoh_HR_BP 0.3341 0.1433 0.0701 LFpcoh_BP rsp 0.2954 -0.2978 0.2039 HF_pcoh_HR_rsp -0.0170 -0.2281 -0.5558 HF_pcoh_HR_BP 0.1308 0.4691 0.2969 HF_pcoh_BP_rsp

227<br />

5.6 Principal Component Analysis<br />

As mentioned in section 3.14, the principal components were arranged in order <strong>of</strong><br />

decreasing variance with the most informative principal component listed first, and the<br />

least informative listed last. The dimensionality <strong>of</strong> the problem was reduced, i.e.,<br />

reduces the number <strong>of</strong> variables without losing much <strong>of</strong> the information. Thus instead <strong>of</strong><br />

analyzing a large number <strong>of</strong> the original variables with complex interrelationships, the<br />

data could be analyzed by a small number <strong>of</strong> uncorrelated principal components.<br />

Figure 5.60 presented a principal component plot <strong>of</strong> the data set containing 48<br />

COPD and 8 normal subjects at rest. The data set gave the following principal<br />

components (PC):<br />

1 st PC: LF_coh_HR_rsp — Coherence <strong>of</strong> HR and respiration signals in the LF<br />

range (0.04 - 0.15 Hz).<br />

2nd PC: HF _ coh _ BP _rsp — Coherence <strong>of</strong> BP and respiration signals in the HF<br />

range (0.15 - 0.4 Hz).<br />

3 rd PC: rsp — The respiration rate.<br />

This indicated that the largest variance between COPD and normal subjects was<br />

the interrelationships between heart rate and respiration. The second largest variance was<br />

the interrelationships between blood pressure and respiration. The third largest variance<br />

was the respiration rate itself.

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

Saved successfully!

Ooh no, something went wrong!