Breakthrough in Breast Ultrasound Image Segmentation with BS-Mamba Architecture

Strategic Insights -

Researchers from Shanghai Maritime University have made significant strides in breast cancer diagnosis by developing a novel architecture called BS-Mamba, which utilizes a unique combination of modules to enhance segmentation performance and speed. According to a new report published in the International Journal of Pattern Recognition and Artificial Intelligence, the BS-Mamba architecture has demonstrated exceptional segmentation accuracy, outperforming current state-of-the-art methods. The research proposes a BreastSegMamba block, a hybrid attention block, and a data-sensitive fuzzy logic function to capture global and local high-resolution information, strengthen feature extraction, and enhance feature edges.

Key Takeaways:

  • The BS-Mamba architecture is a novel framework for breast ultrasound image segmentation that incorporates three new modules to improve performance and speed.
  • The BreastSegMamba block captures both global and local high-resolution information and processes complex features quickly.
  • The hybrid attention block is designed to strengthen the extraction of positional and channel features for accurate segmentation.
  • The data-sensitive fuzzy logic function focuses on the edges of ultrasound images to enhance features in critical areas for higher segmentation performance.
  • The BS-Mamba architecture has been extensively validated through comprehensive experiments and ablation studies on the BUSI and Dataset B breast ultrasound datasets.
  • It significantly outperforms current state-of-the-art methods in terms of segmentation accuracy, achieving a Dice coefficient of 92.13% and an IoU of 86.42% on the BUSI dataset, and a Dice coefficient of 92.62% and an IoU of 86.67% on Dataset B.
  • BS-Mamba achieves a training time acceleration of up to 2.88x in comparison to other established models, further substantiating its computational efficiency.

Statistics:

  • Dice coefficient: 92.13% on the BUSI dataset, 92.62% on Dataset B
  • IoU: 86.42% on the BUSI dataset, 86.67% on Dataset B
  • Training time acceleration: up to 2.88x
  • Number of modules: 3 (BreastSegMamba block, hybrid attention block, data-sensitive fuzzy logic function)
  • Research funding: Shanghai Pujiang Program

Sources:

  • Bs-mamba: a Robust Network for Breast Ultrasound Image Segmentation Using Mamba Architecture. International Journal of Pattern Recognition and Artificial Intelligence, 2025.
  • VerticalNews. New Pattern Recognition and Artificial Intelligence Findings from Shanghai Maritime University Discussed (Bs-mamba: a Robust Network for Breast Ultrasound Image Segmentation Using Mamba Architecture). Robotics & Machine Learning, June 30, 2025; p 277.