Speckle Noise Reduction in Ultrasound Images

Speckle Noise Reduction in Ultrasound Images evaluates and quantifies filtering and denoising algorithms to suppress speckle noise in ultrasound imagery and to assess trade-offs among smoothing, edge preservation, and artifact suppression.

Key Features:

• Comparative framework: Systematic comparison of filtering and denoising algorithms using controlled simulations and objective quality metrics.
• Noise-free reference synthesis: Synthesizes noise-free reference ultrasound images (e.g., kidney phantom) for ground-truth evaluation.
• Field II acoustic simulation: Corrupts reference images with realistic speckle patterns using Field II acoustic simulation.
• Algorithm classes compared: Compares compounding-based approaches and post-processing algorithms including spatial-domain adaptive filters and frequency- or wavelet-domain thresholding methods.
• Spatial-domain methods: Includes spatial-domain adaptive filters such as AWMF, ASR, nonlinear diffusion, and MAP estimation.
• Frequency- and wavelet-domain methods: Includes frequency- or wavelet-domain thresholding approaches for denoising.
• Objective metrics: Uses MSE, PSNR, normalized MSE, CNR, LSNR, and SNR measured against the known ground truth.
• Trade-off quantification: Quantifies trade-offs among smoothing, edge preservation, and artifact suppression across filtering strategies.
• Generalizability emphasis: Emphasizes generalizable tendencies observed in simulated experiments to guide interpretation on real ultrasound data.
• Parameter guidance: Provides quantitative insight to inform parameter selection and algorithm choice for clinical and research ultrasound processing workflows.

Scientific Applications:

• Algorithm benchmarking: Provides quantitative benchmarks of speckle reduction methods using simulated ground truth.
• Filter selection for clinical data: Guides selection and interpretation of speckle reduction filters when true noise-free clinical images are unavailable.
• Research workflow optimization: Informs parameter tuning and algorithm choice in clinical and research ultrasound processing workflows.

Methodology:

Synthesizes a noise-free reference ultrasound image (e.g., kidney phantom), corrupts it using Field II acoustic simulation to generate realistic speckle, compares compounding-based approaches and post-processing algorithms (spatial-domain adaptive filters including AWMF, ASR, nonlinear diffusion, MAP estimation, and frequency-/wavelet-domain thresholding), and evaluates performance using MSE, PSNR, normalized MSE, CNR, LSNR, and SNR against the known ground truth.