Which deep learning models are integrated into PathML?

PathML includes built-in support for HoVer-Net for simultaneous nucleus segmentation/classification and HACTNet for hierarchical cell-type classification.

Can I use this skill for CODEX or multiplexed imaging?

Yes, PathML features specialized modules for multiparametric data analysis, supporting platforms like CODEX, Vectra, and MERFISH for spatial proteomics.

What image formats does the PathML skill support?

PathML supports over 160 proprietary and open formats, including Aperio SVS, Hamamatsu NDPI, Leica SCN, Zeiss ZVI, DICOM, and OME-TIFF.

How does PathML handle large pathology datasets?

It utilizes HDF5 storage structures to efficiently manage tiles, masks, and metadata, ensuring optimized performance for machine learning training and batch processing.

PathML Computational Pathology

Name: PathML Computational Pathology
Author: henriquescastilho

byhenriquescastilho

•

Ciencia de Datos y ML

Powers advanced whole-slide image analysis and machine learning workflows for computational pathology and spatial proteomics.

PathML is a comprehensive Python toolkit designed for high-throughput computational pathology, enabling researchers and developers to process whole-slide images (WSI) across 160+ proprietary formats. It provides modular pipelines for stain normalization, nucleus segmentation using pre-trained models like HoVer-Net, spatial tissue graph construction, and specialized analysis for multiparametric data such as CODEX or multiplex immunofluorescence. By integrating with PyTorch and HDF5, PathML streamlines the transition from raw histology slides to deep learning-driven spatial analysis and cellular quantification.

Características Principales

01Spatial graph construction for cellular and tissue-level relationship analysis

02Supports 160+ whole-slide image formats including SVS, NDPI, and OME-TIFF

03Comprehensive multiparametric imaging support for CODEX and Vectra workflows

04Advanced nucleus segmentation and classification using HoVer-Net and HACTNet

051 GitHub stars

06Automated preprocessing pipelines with stain normalization and artifact detection

Casos de Uso

01Training and deploying deep learning models for automated cancer cell detection and grading

02Managing large-scale pathology datasets with efficient HDF5-based storage and tile management

03Analyzing spatial proteomics and marker expression in multiplexed immunofluorescence data

Características Principales

01Spatial graph construction for cellular and tissue-level relationship analysis

02Supports 160+ whole-slide image formats including SVS, NDPI, and OME-TIFF

03Comprehensive multiparametric imaging support for CODEX and Vectra workflows

04Advanced nucleus segmentation and classification using HoVer-Net and HACTNet

051 GitHub stars

06Automated preprocessing pipelines with stain normalization and artifact detection

Casos de Uso

01Training and deploying deep learning models for automated cancer cell detection and grading

02Managing large-scale pathology datasets with efficient HDF5-based storage and tile management

03Analyzing spatial proteomics and marker expression in multiplexed immunofluorescence data