{"architecture":{"backbone":"3D ResNet-style encoder with attention pooling; ensemble of 5 models","family":"3D convolutional network","input":"full LDCT series (DICOM), resampled","output":"cumulative risk at years 1\u20136","pretraining":"supervised on NLST LDCT with future-cancer labels"},"article":null,"cancer_slugs":["lung-bronchus"],"category":"radiology","confidence":"high","datasets":[{"name":"NLST \u2014 National Lung Screening Trial","note":"training and held-out test","role":"training","slug":"nlst"}],"developer":"MIT Jameel Clinic / Massachusetts General Hospital","evaluation":[{"benchmark":"NLST held-out test","dataset_slug":"nlst","external":false,"metric":"AUC, 1-year risk","source":"https://ascopubs.org/doi/10.1200/JCO.22.01345","value":"0.92"},{"benchmark":"Massachusetts General Hospital","external":true,"metric":"AUC, 1-year risk","source":"https://ascopubs.org/doi/10.1200/JCO.22.01345","value":"0.86"},{"benchmark":"Chang Gung Memorial Hospital (Taiwan)","external":true,"metric":"AUC, 1-year risk","source":"https://ascopubs.org/doi/10.1200/JCO.22.01345","value":"0.94"}],"hf":null,"kind":"task-model","license":"MIT","limitations":"- Trained on a screening population of heavy smokers; performance in never-smokers rests on one Taiwanese cohort.\n- Sensitive to acquisition protocol; validate locally before any programme use.","links":{"demo":null,"docs":null,"doi":"10.1200/JCO.22.01345","github":"https://github.com/reginabarzilaygroup/Sybil","huggingface":null,"paper":"https://ascopubs.org/doi/10.1200/JCO.22.01345","pmid":"36634294"},"modalities":["radiology-ct"],"name":"Sybil","notable_uses":"Validated on three independent cohorts (NLST test set, Massachusetts General Hospital, Chang Gung Memorial Hospital in Taiwan); one-year AUCs of 0.92, 0.86 and 0.94 respectively in the JCO paper.","openness":"open-weights","regulatory":{"intended_use_en":"Research; not cleared for clinical use.","intended_use_pl":"Badania; brak dopuszczenia do u\u017cytku klinicznego.","source_url":"https://github.com/reginabarzilaygroup/Sybil","status":"research-only"},"regulatory_status":"research-only","release_date":"2023-01-12","run_snippet":"","settings":["screening"],"slug":"sybil","sources":[{"label":"Mikhael PG et al. Sybil: A Validated Deep Learning Model to Predict Future Lung Cancer Risk From a Single Low-Dose Chest CT. J Clin Oncol 2023","url":"https://ascopubs.org/doi/10.1200/JCO.22.01345"},{"label":"GitHub \u2014 reginabarzilaygroup/Sybil","url":"https://github.com/reginabarzilaygroup/Sybil"}],"summary":"Deep-learning model that predicts an individual's risk of lung cancer over the next one to six years from a single low-dose chest CT, without radiologist annotations or clinical variables.","tasks":["risk-prediction","screening"],"training":{"institutions":"NLST (USA)","populations":"NLST: US current/former heavy smokers aged 55\u201374; external: MGH (USA), CGMH (Taiwan, including never-smokers)","size":"NLST training split (thousands of participants; see paper)","summary_en":"Trained on low-dose CT scans from the National Lung Screening Trial (NLST) with follow-up cancer outcomes; no radiologist labels beyond the trial's outcome data.","summary_pl":"Trenowany na niskodawkowych TK z badania National Lung Screening Trial (NLST) z wynikami obserwacji; bez etykiet radiologicznych poza danymi wynikowymi badania."},"updated_at":"2026-09-05T22:26:01.509480","url":"/ai-oncology/models/sybil","usage":{"hardware":"GPU recommended; CPU inference works per scan","library":"PyTorch (pip package sybil)","notes_en":"pip install sybil; the package downloads the ensemble weights. Input must be a full LDCT series \u2014 thick-slice or contrast CTs are outside the validated domain."},"verified_at":"2026-09-05T22:26:01.509204","verified_by":"editorial","version":null,"what_it_does":"Takes the full LDCT volume and outputs calibrated risk for years 1\u20136. It is a risk stratifier for screening programmes (who needs a shorter interval), not a nodule detector."}