Travels through the time-frequency zone: advanced Doppler ultrasound processing techniques

Doppler ultrasound echoes from cardiac structures is rich in detail and highly nonstationary. The goal of time frequency analysis is to extract features from these echo signals for high confidence visual and machine classification of cardiovascular condition. Presented are side-by-side comparisons of numerous time-frequency representations (TFRs) of Doppler ultrasound data. The techniques illustrated include the short-time Fourier transform, the wavelet transform, a fixed cone kernel distribution (CKD), the adaptive cone kernel distribution, the adaptive optimal kernel (AOK) TFR, MUSIC, and model based TFRs. The trade-offs among the various analysis representations are discussed.< >