Noise cancellation technique for Doppler ultrasound spectrogram enhancement

Doppler ultrasound (DU) automatic detection of clinically relevant blood flow spectral parameters may be a difficult task when the spectrogram is corrupted with different types of noise. This paper introduces a technique to eliminate the spectrogram background noise without damaging the blood flow signals´ content. Tests were performed on simulated signals with known spectral characteristics. Different signal to noise ratios were tested. Results of 100 signals were averaged. Based on the STFT spectrogram the center frequency and bandwidth of the simulated noisy signals were computed and evaluated against the deterministic curve. Results obtained by defining the background area of DU simulated signals and applying 7% of Cancellation level on signals with 10 dB of signal-to-noise ratio (SNR) produced mean frequency estimations with a maximum bias of 106 Hz. The maximum bias of the bandwidth estimations in this case was 124 Hz. Since the sampling frequency of the DU simulated signals were 12800 Hz, usage of the proposed noise cancelation technique enabled spectral parameters estimations with meaningless errors demonstrating the effectiveness of the proposed technique.