EPFL's CellBridge teaches H&E image models cell identity from spatial transcriptomics
mo_lotfollahi · x · 2026-10-26
Researchers at EPFL released CellBridge, a new preprint led by @sevdoche in collaboration with P. Frossard and Dorina Thanou.
The problem: H&E pathology slides reveal tissue architecture and cell appearance, but identifying exact cell types (e.g. telling a T cell from a B cell) usually requires extra molecular markers. Spatial transcriptomics provides that molecular detail per cell but is far harder to obtain routinely than an H&E slide.
CellBridge trains an image model on millions of cells with paired H&E images and Xenium spatial transcriptomics, so molecular measurements teach the model cell identity — helping it recognize cells when appearance alone isn't enough.
More from Research
- Turing-complete malware can run in PyTorch's backward pass via a zero-valued loss — retr0jirachi · 2026-10-26
- Nathan Lambert's first open models lecture: distillation, risks and the ecosystem explained — natolambert · 2026-10-26
- Search researcher on MRL debate: losses are tolerable but benchmarks are too small with stale QRELs — srchvrs · 2026-10-26
- Paper: agent failures stem from context bloat, not reasoning — quadratic token cost, accuracy cliff — import_jmr · 2026-10-26
- EpochAI reviews 8 agent benchmarks: only 1 verified, 3 flawed — and coding agents may break the reward trilemma — May_F1_ · 2026-10-26
- Francesco Locatello shows AI plus causal inference replacing manual experiment analysis — FrancescoLocat8 · 2026-10-26