A High-Speed, High-Accuracy Pathology Image Segmentation and Tumor Microenvironment Feature Extraction Tool
SegPath-YOLO addresses critical challenges in pathology image analysis, particularly in handling overlapping and high-density cellular structures and ensuring rapid processing without sacrificing precision. The novelty of SegPath-YOLO lies in its Segmentation and Overlapping-Aware Loss, which utilizes a binary overlap mask to identify and enhance the loss in overlapping regions. In conjunction with PathNuclei attention mechanisms, SegPath-YOLO not only refines segmentation results but also contributes to a deeper characterization and quantification of the tumor microenvironment, significantly aiding in survival outcome predictions.
- Python 3.8+
- PyTorch 1.8.0+
- Gradio 4.20.+
- Other Python libraries as specified in requirements.txt
- Downlaod SegPath-YOLO from Gihtub
git clone https://github.com/yaober/SegPath-YOLO.git- Pip install the ultralytics package
pip install -r requirements.txt- The source code of SegPath-YOLO will be released when the paper is accepted.
Run Gradio for the interactive demo:
python app_gradio.pyContributors names and contact info
- Mr. Jia Yao
- Dr. Ruichen Rong
- Dr. Tao Wang
- Dr. Guanghua Xiao