Spatiotemporal Nonlocal Means Based Denoising for Ultrasound Microvascular Imaging

Ultrasound microvascular imaging (UMI) has been developed to visualize microvasculature to reveal the conditions and pathologies of diseases. The key of UMI is extraction of blood flow signals by suppressing tissue clutters and background noise. However, the suppression of noise is unsatisfactory when we only use common clutter filtering methods without additional processing of denoising. Spatiotemporal non-local means (stNLM) filtering method has been used to effectively reject noise in ultrasound localization microscopy. In this study, we implemented stNLM in UMI to better suppress the background noise for improved microvascular visualization using the spatiotemporal discontinuity of noise. For comparison, the proposed stNLM-based UMI method was compared with three methods, i.e., method without NLM filtering (conventional UMI), NLM filtering on power Doppler image (PDI), and NLM filtering in the spatial domain (sNLM) of filtered blood flow RF data. The denoising superiority of the proposed method was validated in a human spinal cord tumor in vivo. Quantitative indexes including contrast-to-noise ratio (CNR) and peak-to-side level (PSL) were calculated. Results show that stNLM-based UMI can achieve nearly 15 dB improvement in CNR and nearly 30 dB improvement in PSL compared with conventional UMI. The proposed stNLM-based UMI has the potential of achieving high-quality UMI.