Reduction of the noise and speckle in Doppler blood flow spectrograms by using matching pursuit with pulse coupled neural network

To reduce noise and speckles in the spectrograms of Doppler blood flow signals, a novel method, called matching pursuit with pulse coupled neural network (MPPCNN), has been proposed. The method considered is an iterative decomposition algorithm, which decomposes the Doppler ultrasound signals into linear expansion of atoms in a time-frequency dictionary using the matching pursuit (MP) for de-noising the Doppler ultrasound signal. Then, a simplified unidirectional pulse coupled neural network is used to calculate the firing matrix of the de-noised spectrogram. The Doppler speckles of the de-noised spectrogram are located and removed through analyzing and processing the PCNN firing matrix. Experiments were conducted on simulation signals whose SNRs are 0dB, 5dB and 10dB. The result shows that the MPPCNN performs effectively in reducing noise, eliminating Doppler speckles, and enhancing the Doppler spectrograms.