PyHIST

PyHIST preprocesses histopathological whole slide images (WSIs) by segmenting tissue regions, generating masks, and extracting tiles at selectable resolutions for machine learning applications.

Key Features:

• Tissue Segmentation: Generates masks that differentiate tissue from background within WSIs.
• Tile Extraction: Overlays a grid on the masked image and evaluates each tile against a content threshold to select foreground tiles for extraction.
• Resolution Flexibility: Extracts selected tiles at user-specified resolutions to support different machine learning model requirements.
• Format Support: Primary support for SVS histopathology slides with experimental support for other image formats.
• Semi-automatic Pipeline: Implements a semi-automatic preprocessing pipeline for segmentation and tile generation.

Scientific Applications:

• Digital pathology: Preprocesses large-scale WSIs for digital pathology analyses.
• Machine learning preprocessing: Produces segmented, content-filtered tiles suitable for training and inference of machine learning models on histological images.
• Cancer diagnostics: Facilitates preparation of histological tiles used in cancer diagnostics and predictive-model development.
• Tissue-level biological analysis: Enables extraction of tissue regions for studies aiming to derive biological insights from histological samples.

Methodology:

Generates a tissue/background mask from the input WSI; overlays a grid of tiles on the masked image and assesses each tile's content against a minimum threshold to classify foreground; extracts tiles that meet the criteria at the desired resolution.