{"id":"W4410180988","doi":"10.1038/s41746-025-01594-2","title":"High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis","year":2025,"lang":"en","type":"article","venue":"npj Digital Medicine","topic":"Venous Thromboembolism Diagnosis and Management","field":"Medicine","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"North York General Hospital; St. Michael's Hospital; University of Toronto","funders":"","keywords":"Pulmonary embolism; Medicine; Artificial intelligence; Radiology; Pattern recognition (psychology); Machine learning; Computer science; Surgery","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.000237435,0.0003772156,0.0009947869,0.0003398623,0.0002560377,0.000064416,0.0001777201,0.0001236956,0.0001556877],"category_scores_gemma":[0.0002896672,0.0002741925,0.00009592265,0.0005755887,0.0002128223,0.0003776258,0.0001121623,0.0002772073,0.00001649414],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001522048,"about_ca_system_score_gemma":0.0001240565,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000111963,"about_ca_topic_score_gemma":0.000004080046,"domain_scores_codex":[0.9978956,0.00001986334,0.0004802895,0.0005476336,0.0004728221,0.0005837944],"domain_scores_gemma":[0.9987357,0.000255136,0.0001258899,0.0004027351,0.0002738294,0.000206654],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.002085778,0.004067264,0.3684724,0.006390128,0.002499226,0.0004505332,0.0019311,0.003366097,0.006849899,0.008641479,0.03151859,0.5637275],"study_design_scores_gemma":[0.01973619,0.009373812,0.5964221,0.01817109,0.004555581,0.0002017858,0.003903244,0.02231869,0.00870918,0.000972307,0.3137906,0.001845349],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9727151,0.0008595297,0.001187369,0.006322273,0.0004968004,0.001441624,0.00001344786,0.0001982712,0.01676563],"genre_scores_gemma":[0.9913426,0.0006099163,0.0005120393,0.002296801,0.0005264059,0.0003328666,0.0001697451,0.00006792497,0.004141717],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5618821,"threshold_uncertainty_score":0.999971,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01753628578119739,"score_gpt":0.2670570294513299,"score_spread":0.2495207436701325,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}