Abstract
Asymptomatic circulating cerebral emboli, which are particles larger than red blood cells, can be detected by Doppler ultrasound. Embolic signals are short duration transient like signals. Embolic signals can be extracted from quadrature Doppler signals, which are obtained at the end of quadrature demodulation. The wavelet transform is an ideal method for analysis and detection of such signals by optimizing time-frequency resolution. In literature systems employing discrete wavelet transform (DWT) for detecting embolic signals exist. Dual-tree complex wavelet transform (DTCWT), which is a shift invariant transform with limited redundancy, is an improved version of DWT. Conventionally, prior to applying the DTCWT to an quadrature signal, first the quadrature signal must be decoded into forward and reverse signals and then two DTCWT should be applied. However, modified dual-tree complex wavelet transform (MDTCWT), which reduces the computational complexity compared to conventional algorithm, can be used. In this study a new emboli detection system based on MDTCWT is proposed. In detection algorithm, embolic signals and artifact signals were decomposed to nine scales and different features were extracted from each scale. Then for each scale, by using extracted features a new feature vector was created and these feature vectors are fed into support vector machines individually and the comparative results of individual feature sets are proposed.