#18 · ML / AI methods

Open software tools are packaging spatial and whole-slide image analysis steps for reuse

The dominant thread is open, reusable software for processing spatial and imaging data rather than new analytical models. STiLE is a coordinate-based tool that automatically dearrays tissue microarray cores in spatial transcriptomics data using connectivity and HDBSCAN clustering. Binary-SPA is a reference-free two-stage method for cell-type annotation in high-resolution spatial transcriptomics, using marker-based high-confidence cells. LazySlide is an open-source, scverse-based Python package for whole-slide image analysis that bridges histopathology with omics workflows. GOLDMARK provides a benchmarking framework for H&E whole-slide image biomarker prediction, and the macaque atlas offers a harmonized resource built with UCE embeddings and human foundation model annotation. What is new is interoperability with established ecosystems such as scverse and the use of reference-free or foundation-model-based annotation. The summaries do not report quantitative performance for the software tools, so accuracy claims remain unverified.