Adaptive Doppler blood flow estimation in ultrasound with enhanced spectral resolution and contrast using limited observation windows

Abstract

Measuring blood velocities during the acceleration and deceleration phases of the systolic period can be challenging due to the trade-off between spectral and temporal resolution. This can significantly affect the accuracy of spectrogram reproduction. When temporal samples are reduced, the spectral width may broaden over time, especially during systole. Additionally, shorter observation windows negatively impact factors such as frequency resolution and contrast. This study hypothesizes that a more accurate ultrasound spectrogram can be generated using a new blood velocity estimator with a minimal observation window length of $N=2$. The spectrogram’s accuracy is assessed using various criteria, including spectral resolution, contrast, and spectral broadening over time. The proposed adaptive method integrates a new coherence-based post-filter with the Eigenspace-based Forward-Backward Amplitude Spectrum Capon (ESB-FBASC) technique. The method’s performance was evaluated in different conditions, including simulations of the femoral artery, stationary and complex flow, and in vivo data. Under rapid flow conditions simulated over three heartbeats in 0.2 s, the proposed method demonstrated better temporal resolution compared to the Welch-Ref estimator, effectively capturing rapid velocity changes and reducing spectral broadening, despite using only $N=2$ slow-time samples. For clinical data on the hepatic vein, the proposed estimator improved spectral resolution by 24 %, 44 %, and 67 %, and increased contrast by 79.8 dB, 120.8 dB, and 155.5 dB compared to MASC, Pr.Capon, and Capon, respectively, for $N=2$. Furthermore, the narrowest power spectrum width at 40 dB was achieved with the proposed method, showing an improvement of 38 % and 75 % compared to MASC and Pr.Capon, respectively. As a result, the proposed method effectively reduces power spectrum width and enhances spectrogram accuracy by improving spectral resolution and contrast, all while using the limited observation window length of $N=2$.