Related Experiment Video
Updated: May 29, 2025
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
22.4K
Needle tracking and segmentation in breast ultrasound imaging based on spatio-temporal memory network
Qiyun Zhang1, Jiawei Chen1, Jinhong Wang2
1College of Engineering, Shantou University, Shantou, Guangdong, China.
Frontiers in Oncology
|February 3, 2025
Summary
This study introduces a novel Spatio-Temporal Memory Network for improved ultrasound-guided breast tumor biopsy. The AI model enhances needle positioning accuracy and tracking stability, leading to safer clinical outcomes.
Area of Science:
- Medical Imaging
- Computer Vision
- Biomedical Engineering
Background:
- Ultrasound-guided needle biopsy is crucial for pathological analysis but faces challenges due to low signal-to-noise ratio and complex imaging.
- Accurate needle segmentation and tracking are vital to minimize complications and damage during biopsies.
- Existing computer-aided methods struggle with poor image quality, high computational demands, and variable needle shapes.
Purpose of the Study:
- To develop an advanced computer-aided system for precise needle segmentation and tracking in ultrasound-guided breast tumor biopsy.
- To improve the accuracy and stability of needle localization, aiding physicians in complex biopsy procedures.
- To address limitations of current techniques by enhancing reliability in challenging ultrasound environments.
Main Methods:
- Introduction of a novel Spatio-Temporal Memory Network incorporating a hybrid CNN-Transformer encoder.
- Integration of an optical flow estimation method for enhanced temporal analysis.
- Development of a real-time segmentation dataset using ultrasound biopsy video data from 11 breast tumor patients.
Main Results:
- The proposed network significantly outperforms existing methods in needle positioning accuracy and tracking stability.
- Achieved performance metrics include IoU: 0.731, Dice: 0.817, Precision: 0.863, Recall: 0.803, and F1 score: 0.832.
- Demonstrated robust performance in segmenting and tracking biopsy needles within challenging ultrasound images.
Conclusions:
- The Spatio-Temporal Memory Network provides dependable technical support for computer-aided tracking and segmentation in ultrasound-guided breast tumor biopsy.
- The model enhances the reliability and precision of needle localization, contributing to safer and more effective clinical outcomes.
- This advancement offers significant potential for improving the quality of care in interventional radiology and pathology.
