Principal component analysis of multiple noninvasive blood flow derived signals

Ultrasonic measurements of blood velocity waveforms from the femoral and tibial posterior arteries were analyzed by principal component analysis (PCA) simultaneously with the first derivatives of electrical impedance signals (dZ/dt) from the calf of normal subjects and patients with peripheral arterial disease. The three noninvasive measurements were performed separately and recorded on magnetic tape together with the electrocardiogram (ECG). After low-pass filtering at 30 Hz, the signals were digitized at 200 samples/s, synchronized by the R wave of the ECG, and coherently averaged using at least 50 cardiac cycles with uniform heart rates (+or-10%) for each type of signal. For an ensemble of 99 waveforms corresponding to 33 lower limbs of an initial population of 23 patients, the first six principal components account for 96.5% of the total signal variance. Automatic identification of patients with significant arterial occlusions (<50%) was accurate with the coefficients of the first principal component of the blood-velocity signals of the tibial posterior artery and the third principal component of dZ/dt, resulting in a diagnostic precision, sensitivity, and specificity of 100%.