Microsoft CEO Satya Nadella announced GigaTIME on March 15, a multimodal AI model designed to transform standard pathology slides into detailed spatial proteomics maps of cancer cells. The model was developed in collaboration with Providence Health and the University of Washington, and has been made publicly available through Azure AI Foundry Labs and Hugging Face.
GigaTIME was trained on a Providence dataset containing 40 million cells with paired imaging data across 21 protein channels. The model has already been applied to pathology slides from 14,256 cancer patients across 51 hospitals and over a thousand clinics within the Providence system, generating approximately 300,000 virtual multiplex immunofluorescence images spanning 24 cancer types and 306 cancer subtypes.
The practical significance is that spatial proteomics - which maps the protein activity in and around tumor cells - currently requires expensive, specialized equipment and processes that most hospitals do not have. GigaTIME aims to generate equivalent data from the standard hematoxylin and eosin stained slides that every pathology lab already produces, potentially democratizing access to advanced cancer diagnostics.
Initial results are promising. The virtual population generated by GigaTIME uncovered 1,234 statistically significant associations linking protein activations with clinical attributes such as biomarkers, disease staging, and patient survival outcomes. By releasing the model as open-source, Microsoft is betting that the broader research community can validate and extend these findings faster than any single institution could alone.