Anisotropic Diffusion for Ultrasound Images

Anisotropic Diffusion for Ultrasound Images reduces speckle noise in ultrasound scans while preserving tissue-specific diagnostic features by applying anisotropic diffusion augmented with a probabilistic-driven memory mechanism.

Key Features:

• Probabilistic-Driven Memory Mechanism: Incorporates a delay differential equation for the diffusion tensor that adapts diffusion dynamics based on speckle and tissue statistics, accelerating smoothing in noise-dominated regions and preserving structures in tissue-rich regions.
• Tissue Selective Philosophy: Tailors diffusion behavior according to the statistical properties of different tissues to retain diagnostically relevant features while reducing speckle noise.

Scientific Applications:

• Speckle noise reduction in ultrasound imaging: Reduces multiplicative speckle noise to improve image quality for visual inspection and diagnostic interpretation.
• Preservation of diagnostic tissue structure: Maintains clinically relevant boundaries and tissue characteristics that are important for diagnosis.
• Preprocessing for automated analysis: Serves as a preprocessing step that preserves features needed for automatic analysis methods.
• Validation on datasets: Demonstrated effectiveness on synthetic and real ultrasound images, showing improved balance between noise reduction and detail preservation.

Methodology:

Implements anisotropic diffusion with a memory mechanism expressed via a delay differential equation for the diffusion tensor, leveraging speckle-pattern statistics and tissue statistical properties to adapt local diffusion.