Providence, based in Renton, Wash., Microsoft Research and the University of Washington have developed an AI pathology model that researchers say can map how immune cells interact with tumors on a scale not previously possible.
The model, called GigaTIME, was detailed in a study published in Cell. Researchers said the tool can generate virtual spatial proteomics images from standard histopathology slides, offering a more accessible alternative to multiplex immunofluorescence — a costly, low-throughput imaging technique typically used to identify tumor and immune cell types.
Using multiplex immunofluorescence to train the system, the team built GigaTIME to produce accurate virtual mIF images from routine pathology slides, according to a Dec. 9 news release. The model then created about 300,000 virtual images from patients across 24 cancer types and 306 subtypes. The study identified more than 1,200 statistically significant associations between immune-related proteins and clinical factors such as biomarkers, staging and survival.
Researchers said the findings suggest the tool could help assess tumor-immune interactions, predict immunotherapy response, and identify mechanisms of immune evasion that inform strategies to overcome resistance.
GigaTIME builds on GigaPath, a model developed by the same institutions and published in Nature in 2024. GigaPath analyzed whole-slide patterns to predict mutations and cancer subtypes, while GigaTIME extends this work by generating virtual spatial proteomics images that reveal individual immune cell interactions with tumors.
