Microsoft Unveils Efficient Pathology Foundation Models for Large-Scale Research
Microsoft has developed two new foundation models that make large-scale pathology research more accessible and practical. The Flash family extends GigaPath and GigaTIME, both of which were previously demonstrated to support whole-slide analysis and tumor microenvironment modeling from routinely collected pathology data.
The new models, GigaPath-Flash and GigaTIME-Flash, are built on a common efficient backbone that preserves useful pathology representations while substantially reducing computational resources. They achieve competitive performance at substantially lower computational cost than other whole-slide pretrained models.
GigaPath-Flash provides efficient tile and slide encoders, while GigaTIME-Flash replaces the CNN backbone of the original GigaTIME with the GigaPath-Flash ViT-S encoder. The efficiency gains of the Flash models are substantial, with GigaPath-Flash reducing compute requirements by 50 times and GigaTIME-Flash being 6 times faster.
The open models support population-scale discovery, helping researchers investigate disease biology, biomarkers, and clinical outcomes across diverse cancer datasets. They can be applied repeatedly and affordably across large patient populations, making them a valuable tool for researchers in the field of pathology.