Wavelet feature extraction of Doppler blood flow waveforms

The maximum frequency waveform of the Doppler ultrasound signal was analyzed using a multi-scale wavelet transform to extract its maximas variation of wavelet transform modulus under various scales. This maximas variation was then applied to the feature extraction of Doppler signals from common carotid arteries. It was found from clinical experiments that the shape of this variation from cases with normal cerebral vessels differed from those associated with abnormal cases. To diagnose cerebral vessel diseases, the variation was fitted by a polynomial whose coefficients were put into a back-propagation (BP) neural network for the classification. It was shown that this approach had a satisfied performance, and could be a novel means in the cerebral vascular disease diagnosis.