alterlab-pathml

An end-to-end computational pathology workflow unifies image loading, preprocessing, segmentation, graph construction, and ML model inference into a modular framework. PathML pipelines process large pathology datasets using tile-based processing and export results to HDF5 or AnnData formats.

Multiparametric imaging support for CODEX, Vectra, and MERFISH is integrated into the computational pathology workflow. The framework provides quality control and data management for multiparametric imaging alongside standard whole-slide image processing.

Running PathML graph analysis requires Python and PathML, with optional libraries including PyTorch, OpenSlide, h5py, and AnnData. These dependencies support scalable tile-based processing, graph construction, and HDF5 or AnnData export for large pathology datasets.

The best way to classify cell types is constructing cell graphs from tissue sections and running ML models for inference. This computational pathology approach uses PathML pipelines to segment cells, build graphs, and deploy models for classification tasks.

Scalable analysis for large whole-slide datasets is supported through tile-based processing and distributed computing. The PathML workflow enables efficient handling of large pathology images and exports results to HDF5 or AnnData formats for downstream use.

Loading and preprocessing pathology slides uses predefined PathML pipelines that handle image loading and preprocessing steps. After applying the pipeline, results can be exported to disk for subsequent segmentation, graph construction, and ML model deployment.