Breast Tumor Ultrasound Image Segmentation Method Based on Improved Residual U-Net Network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Accurate breast tumor recognition is crucial for effective diagnosis and treatment planning.
• Ultrasound imaging is a widely used modality for breast lesion assessment.
• Deep learning models, such as U-Net, show promise for automated image segmentation.

Purpose of the Study:

• To develop an efficient and accurate breast tumor segmentation method for ultrasound images.
• To enhance the U-Net framework by incorporating residual blocks and an attention mechanism.
• To improve the model's ability to extract relevant features for precise tumor recognition.

Main Methods:

• A modified U-Net framework incorporating residual blocks to mitigate gradient disappearance and training difficulties.
• Integration of a fusion attention mechanism, considering both spatial and channel attention, to enhance feature extraction.
• Application of the proposed method to breast tumor ultrasound image segmentation.

Main Results:

• The proposed method achieved a Dice index of 0.921, demonstrating superior image segmentation performance.
• The incorporation of residual blocks improved the stability and training of the deep network.
• The attention mechanism effectively enhanced the model's ability to capture critical image features.

Conclusions:

• The enhanced U-Net framework with residual blocks and attention mechanism provides excellent performance for breast tumor segmentation in ultrasound images.
• This approach offers a promising tool for improving the accuracy and efficiency of breast tumor diagnosis.
• Further research can explore the clinical applicability and validation of this advanced segmentation technique.