Microsoft and Providence Health have released GigaTIME — an open-source multimodal AI model that converts standard, widely available pathology slides into high-resolution tumor immune microenvironment data. Analysis that previously required specialized laboratory assays costing more than $500 per slide and days of processing can now be performed in approximately 20 minutes on a standard GPU, using a tissue slide that costs $10 to $15 to produce. The model is freely available on Microsoft Foundry Labs and Hugging Face.
What GigaTIME Does
To understand why GigaTIME matters, it helps to understand the problem it solves. Cancer research depends on understanding the tumor immune microenvironment (TIME) — how immune cells interact with tumor cells, which proteins are active, and how these patterns predict patient outcomes and treatment responses. The gold standard for this analysis is multiplex immunofluorescence (mIF) imaging — a technique that reveals which immune cells are present and active at the single-cell level across tumor tissue.
The problem: mIF assays are expensive, slow, and require specialized laboratory infrastructure. Standard assays cost over $500 per sample and can only be run on a tiny fraction of available tissue. Most cancer research institutions don’t have access to the technology at scale. The result is that the vast majority of cancer patients’ tissue samples — available as cheap standard pathology slides in every hospital — contain biological information that no one has been able to access.
GigaTIME is a cross-modal translator: it takes a standard hematoxylin and eosin (H&E) slide — the $10 tissue slide available in virtually every pathology lab worldwide — and generates a virtual mIF image that approximates what an expensive lab assay would show. For each of 21 targeted protein channels, the AI evaluates individual pixels in the H&E image and assigns protein activation states across the entire tissue.
The Scale of the Training and Validation
GigaTIME was trained on a dataset of 40 million cells with paired H&E and mIF images across 21 protein channels, collected from Providence Health’s patient population. It was then applied to 14,256 cancer patients across 51 hospitals and over 1,000 clinics, generating a virtual population of approximately 300,000 mIF images spanning 24 cancer types and 306 cancer subtypes.
The analysis uncovered 1,234 statistically significant associations linking protein activation patterns to clinical attributes including biomarkers, cancer staging, and patient survival — associations that were previously impossible to study at this scale because the mIF data didn’t exist. The findings were independently validated against 10,200 patients from The Cancer Genome Atlas, achieving a correlation of 0.88.
This is described as the first population-scale study of tumor immune microenvironments based on spatial proteomics — a category of research that simply didn’t exist at this scale before AI made it computationally feasible.
What It Means for Cancer Research
The practical implication is a structural change in the cost of precision oncology research. Hospitals and research institutions worldwide have archives of thousands to millions of H&E slides from cancer patients — routine slides taken as part of standard care. With GigaTIME, those slides become a vast untapped dataset for studying tumor biology, identifying treatment predictors, and stratifying patients for clinical trials.
Retrospective analysis that was economically infeasible becomes routine. Drug trials can be analyzed for tumor microenvironment patterns using existing tissue archives. Patient stratification models can be built at scales previously impossible. Microsoft CEO Satya Nadella highlighted the model on X, describing it as translating “routine pathology slides into spatial proteomics.”
Open Source and Freely Available
GigaTIME is available now at no cost on Microsoft Foundry Labs and Hugging Face, along with the underlying notebooks and workflows for customization. Researchers can deploy it as an endpoint for production workflows or use it for exploratory analysis. The model is released for research purposes; clinical diagnostic use would require regulatory clearance that the team has not yet pursued.
Conclusion
GigaTIME is one of the clearest examples to date of AI genuinely democratizing access to capabilities that were previously limited to well-funded research institutions. By making advanced tumor analysis accessible from a $10 slide and a standard GPU, it opens a category of cancer research that didn’t previously exist at population scale. Browse our directory to explore the AI tools transforming healthcare and scientific research alongside GigaTIME.
