Doppler ultrasound spectral enhancement using the Gabor transform-based spectral subtraction

Most of the important clinical indices of blood flow are estimated from the spectrograms of Doppler ultrasound (US) signals. Any noise may degrade the readability of the spectrogram and the precision of the clinical indices, so the spectral enhancement plays an important role in Doppler US signal processing. A new Doppler US spectral enhancement method is proposed in this paper and implemented in three main steps: the Gabor transform is used to compute the Gabor coefficients of a Doppler US signal, the spectral subtraction is performed on the magnitude of the Gabor coefficients, and the Gabor expansion with the spectral subtracted Gabor coefficients is used to reconstruct the denoised Doppler US signal. The different analysis and synthesis windows are examined in the Gabor transform and expansion. The signal-to-noise ratio (SNR) improvement together with the overall enhancement of spectrograms are examined on the simulated Doppler US signals from a femoral artery. The results show the denoising method based on the orthogonal-like Gabor expansion achieves the best denoising performance. The experiments on some clinical Doppler US signals from umbilical arteries confirm the superior denoising performance of the new method.