The discrimination of stenosed carotid bifurcation models with smooth and irregular plaque surface. Part II. The multivariate statistical analysis of ultrasonic Doppler velocity data

According to the conclusion of the first part of the paper the Doppler spectral indices were analyzed in a detailed manner to provide a means to differentiate bifurcation models with different irregularities. Autoregressive (AR) models using the Akaike FPE criterion were fitted to the curves of the spectral broadening index, turbulence intensity index and coefficient of symmetry. These curves were also submitted to the data reduction significant point extraction algorithm. The multivariate discriminant analysis was performed on parameter vectors containing coefficients of AR models, data reduction rate and data reduction error. The discriminant analysis allowed distinguishing models with the same degree of stenosis and different plaque surface characteristics on the basis of these vectors. The highest discrimination efficiency was observed for parameter vectors obtained from TBI and SKEW curves. The efficiency of discrimination was slightly higher for more severely stenosed models.