STFNet: A spatial-temporal feature aggregation network for breast lesion segmentation in ultrasound videos
Chuansong Fan1, Lei Wu2, Yu Liu3
1School of Computer Science and Information Security, Guilin University of Electronic Technology, Guilin, 541004, China; Department of Radiology, Guangdong Provincial People's Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangzhou, 510080, China; Guangdong Provincial Key Laboratory of Artificial Intelligence in Medical Image Analysis and Application, Guangzhou, 510080, China.
Accurate breast cancer diagnosis is improved using STFNet, a novel network for segmenting lesions in ultrasound videos. This method enhances early detection by analyzing dynamic imaging data, outperforming current techniques.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Accurate segmentation of breast lesions in ultrasound is crucial for early breast cancer detection and diagnosis.
- Existing methods primarily focus on static 2D ultrasound images, neglecting the dynamic nature of clinical ultrasound examinations.
- Ultrasound videos offer richer information but pose segmentation challenges due to complex spatial-temporal variations.
Purpose of the Study:
- To introduce STFNet (Spatial-Temporal Feature Aggregation Network) for robust automatic segmentation of breast lesions in ultrasound videos.
- To address the limitations of static image analysis by incorporating temporal dynamics into lesion segmentation.
- To improve the accuracy and reliability of automated breast cancer diagnosis using ultrasound video data.
Main Methods:
- STFNet integrates a Spatial Feature Extraction Module (SFEM) and a Temporal Feature Extraction Module (TFEM) to capture both local details and temporal dependencies.
- A Multi-Scale Feature Fusion (MSFF) module hierarchically combines multi-resolution spatial and temporal features for enhanced boundary delineation.
- A hybrid loss function with boundary-aware optimization is employed to mitigate speckle noise and low contrast issues.
Main Results:
- STFNet achieved high performance in segmenting breast lesions in ultrasound videos, with Dice scores of 80.27% on Dataset A and 78.68% on Dataset B.
- The proposed method demonstrated superior performance compared to existing state-of-the-art segmentation techniques.
- Validation was conducted across two academic hospitals involving 490 patients and 550 videos.
Conclusions:
- STFNet provides accurate and automated segmentation of breast lesions in ultrasound videos, aligning with clinical practice.
- The integration of spatial and temporal feature aggregation enhances the robustness of the segmentation model.
- This approach holds significant potential for improving early detection and diagnosis of breast cancer through advanced video analysis.
