Automated Segmentation of Hepatic Vessels and Lobules in Whole-Slide Images Using U-Net Models

The repository contains curated datasets and trained models developed for the automated analysis of liver histopathology whole slide images (WSIs), specifically focused on the detection and segmentation of liver lobules and vascular structures. The dataset includes six ZIP files containing patch-level training and testing data for each task. Each dataset is organized into training/ and testing/ folders, with corresponding images/ and masks/ subdirectories. The masks provide pixel-level annotations for supervised segmentation. Additionally, the "Training WSIs" and "Testing WSIs" folders contain original WSIs used for generating training and testing samples. A separate Zip file archive includes the PyTorch .pth model files trained on these datasets, enabling reproducibility and further experimentation. An additional ZIP file titled "completeStitchedOverlays" provides complete segmented whole slide liver images used to qualitatively evaluate model performance on unseen data.