PDAC Tumor Segmentation
About
- Generic Medical Image
- Fixed Mask (Mask of the locations in the fixed image to consider)
- Gross Tumor Volume Segmentation (Gross Tumor Volume Segmentation with indices 1: Primary Gross Tumor Volume, 2: Lymph Node Gross Tumor Volume.)
Model Facts
Summary
This algorithm applies a three-steps pipeline for segmenting Pancreatic Ductal Atenocarcinoma in pancreatic whole-slide images. Firstm the imput WSI is passed through a tissue segmentation which segments the tissue from the backgroung, then an epithelium segmentation network is applied on the WSI and lasta convex hull is created around the segmented tumor epithelium to create a tumor bulk mass.
Mechanism
The WSI is processed by two different networks, both Unet based. The first processes the slide at a spacing of 4.0 for quick segmentation of the tissue. The second processes the WSI at the spacing of 1.0(using the tissue mask for faster inference) segmenting both healthyand tumor epithelium. This last class (tumor epithelium) is then used by a different algorithm which creates a convex hull of the tumor area.
The output of this pipeline is double: - a mask with the tissue (1) - healthy epithelium (2) and tumor epithelium (3) - a mask with the tumor bulk area (1)
Validation and Performance
Uses and Directions
This algorithm was developed for research purposes only.
Warnings
Common Error Messages
Information on this algorithm has been provided by the Algorithm Editors, following the Model Facts labels guidelines from Sendak, M.P., Gao, M., Brajer, N. et al. Presenting machine learning model information to clinical end users with model facts labels. npj Digit. Med. 3, 41 (2020). 10.1038/s41746-020-0253-3