Computer processing of CW Doppler trans-valvular spectrograms

The work presented in this paper is directed toward the automated processing of CW Doppler trams-mitral inflow and trams-aortic outflow spectrograms. The multi-stage system consists of four main procedures: data acquisition, pre-processing, pattern recognition and pattern interpretation. The two most important features of the system are included in the stage of pattern recognition: an adaptive algorithm,For eliminating spectrogram background noise and identifying velocity profiles and a hierarchical fuzzy model for estimating the start and end of MDI and ASO waveforms. Comparison of the adaptive algorithm, Otsu's (1979), and Deravi and Pal's (1983) thresholding methods with the results of visual inspection, on a series of 180 in vivo CW Doppler echocardiographic images, shows that the adaptive schema consistently outperforms both other methods. The results of the fuzzy system correlated well with expert manual analysis (r>0.90 with at least 81% match rate). Overall the system is effective, computationally efficient, and overcomes the problems of initialization and velocity profile modeling that were encountered in the literature.