{"architecture":{"backbone":"CTransPath tile encoder + attention aggregation + site text embedding","family":"weakly supervised slide-level model over tile features","input":"WSI tiles + anatomical site","output":"slide-level embedding / task logits","pretraining":"self-supervised tile pretraining + weakly supervised slide-level training on 60,530 WSIs"},"article":null,"cancer_slugs":["pan-cancer"],"category":"pathology","confidence":"medium","datasets":[{"name":"TCGA \u2014 The Cancer Genome Atlas (via NCI Genomic Data Commons)","note":"","role":"training","slug":"tcga"}],"developer":"Yu Lab, Harvard Medical School","evaluation":[{"benchmark":"cancer detection, 11 cancer types, external validation","external":true,"metric":"AUROC","source":"https://www.nature.com/articles/s41586-024-07894-z","value":"~0.94 (paper)"},{"benchmark":"survival prediction across cancer types","external":true,"metric":"c-index / log-rank","source":"https://www.nature.com/articles/s41586-024-07894-z","value":"reported in paper"}],"hf":null,"kind":"foundation","license":"non-commercial (see repository)","limitations":"- Tile encoder (CTransPath) is weaker than newer ViT-L/H foundation encoders; CHIEF's strength is the slide-level training.\n- TCGA is in the training set \u2014 external evaluation must avoid it.","links":{"demo":null,"docs":null,"doi":"10.1038/s41586-024-07894-z","github":"https://github.com/hms-dbmi/CHIEF","huggingface":null,"paper":"https://www.nature.com/articles/s41586-024-07894-z","pmid":null},"modalities":["histopathology"],"name":"CHIEF","notable_uses":"Cancer detection across 11 cancer types with an area under the curve near 0.94 on external validation in the paper; survival prediction across cancer types.","openness":"open-weights","regulatory":{"intended_use_en":"Research.","intended_use_pl":"Badania.","source_url":"https://github.com/hms-dbmi/CHIEF","status":"research-only"},"regulatory_status":"research-only","release_date":"2024-09-04","run_snippet":"","settings":["diagnosis","prognosis","basic-research"],"slug":"chief","sources":[{"label":"Wang X et al. A pathology foundation model for cancer diagnosis and prognosis prediction. Nature 2024","url":"https://www.nature.com/articles/s41586-024-07894-z"},{"label":"GitHub \u2014 hms-dbmi/CHIEF","url":"https://github.com/hms-dbmi/CHIEF"}],"summary":"Clinical Histopathology Imaging Evaluation Foundation model: trained on 60,530 whole-slide images across 19 anatomical sites and validated on 19,491 slides from 24 hospitals for cancer detection, tumour-origin prediction, genomic profiling and survival.","tasks":["feature-extraction","classification","prognosis","survival-analysis"],"training":{"institutions":"Harvard Medical School and partner hospitals; public cohorts","size":"60,530 WSIs (training)","summary_en":"60,530 WSIs from 19 anatomical sites (public cohorts including TCGA, PAIP and others plus institutional data); validation on 19,491 slides from 24 hospitals worldwide.","summary_pl":"60 530 preparat\u00f3w z 19 lokalizacji anatomicznych (kohorty publiczne, w tym TCGA i PAIP, oraz dane instytucjonalne); walidacja na 19 491 preparatach z 24 szpitali na \u015bwiecie."},"updated_at":"2026-09-05T22:26:01.490391","url":"/ai-oncology/models/chief","usage":{"library":"PyTorch (repo code)","notes_en":"Weights are distributed from the GitHub repository under non-commercial terms; the pipeline expects CTransPath tile features."},"verified_at":"2026-09-05T22:26:01.490013","verified_by":"editorial","version":null,"what_it_does":"Slide-level representations built with weak supervision on top of tile features, plus a text embedding of the anatomical site; one model then serves detection, origin, molecular-profile and prognosis heads."}