Research on breast tumor segmentation based on the Mamba architecture

Area of Science:

• Medical imaging
• Artificial intelligence in healthcare
• Oncology

Background:

• Accurate medical image segmentation is vital for breast cancer diagnosis.
• Low resolution in breast tumor ultrasound images hinders precise lesion identification.
• Existing segmentation methods face challenges with image detail preservation.

Purpose of the Study:

• To develop a novel segmentation model for breast tumor ultrasound images using the Mamba architecture.
• To address the limitations of low-resolution ultrasound images in lesion localization.
• To enhance segmentation accuracy and preserve image details in breast tumor detection.

Main Methods:

• Integration of the Mamba architecture with foundational models for long-sequence processing.
• Development of a novel segmentation model incorporating VMamba blocks.
• Experimental validation on BUSI and BUS-BRA ultrasound datasets.

Main Results:

• The Mamba architecture model demonstrated superior performance in segmenting breast ultrasound images.
• Enhanced segmentation accuracy and improved processing of image details were observed.
• The model showed effectiveness even under small-sample training conditions.

Conclusions:

• The Mamba architecture offers a promising approach for advancing medical image segmentation in breast cancer detection.
• The developed model provides a foundation for future research in AI-driven diagnostic tools.
• Improved segmentation accuracy can lead to more reliable breast cancer diagnosis.