Anisotropic diffusion filter with memory based on speckle statistics for ultrasound images

Area of Science:

• Medical Imaging
• Image Processing
• Biomedical Engineering

Background:

• Speckle noise in ultrasound (US) imaging degrades image quality, hindering visual inspection and automated analysis.
• Existing speckle reduction filters often remove diagnostically relevant tissue information, leading to over-filtering.
• Speckle patterns contain inherent tissue characteristics crucial for clinical diagnosis.

Purpose of the Study:

• To develop an advanced speckle reduction technique for ultrasound images.
• To overcome the over-filtering problem associated with conventional diffusion filters.
• To preserve diagnostically important tissue information during speckle reduction.

Main Methods:

• Proposed an anisotropic diffusion filter incorporating a probabilistic-driven memory mechanism.
• Formulated the memory mechanism using a delay differential equation for the diffusion tensor.
• The filter selectively accelerates diffusion in non-relevant regions while preserving details in diagnostically significant areas based on tissue statistics.

Main Results:

• The proposed filter effectively reduces speckle noise in both synthetic and real ultrasound images.
• Demonstrated superior performance in preserving clinically relevant structures compared to state-of-the-art filters.
• The memory mechanism successfully mitigated the over-filtering issue, retaining essential diagnostic features.

Conclusions:

• The anisotropic diffusion filter with a probabilistic memory mechanism offers an effective solution for speckle reduction in ultrasound imaging.
• This approach enhances image analysis by preserving critical tissue information lost in conventional methods.
• The findings suggest improved potential for both visual inspection and automated analysis of ultrasound data.