{"id":"W4409285424","doi":"10.2196/67706","title":"Extracting Pulmonary Embolism Diagnoses From Radiology Impressions Using GPT-4o: Large Language Model Evaluation Study","year":2025,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Venous Thromboembolism Diagnosis and Management","field":"Medicine","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Medical diagnosis; Workflow; Wilcoxon signed-rank test; Medicine; Computer science; Artificial intelligence; Machine learning; Medical physics; Radiology; Internal medicine; Database","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003853846,0.002856108,0.001053245,0.001750725,0.0004854579,0.001394331,0.002088401,0.001399455,0.003239829],"category_scores_gemma":[0.01055969,0.0005354099,0.001863685,0.00110776,0.0005779179,0.001762345,0.001734316,0.002557565,0.001981196],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001728383,"about_ca_system_score_gemma":0.001897238,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02104991,"about_ca_topic_score_gemma":0.01993285,"domain_scores_codex":[0.9980787,0.0006603036,0.0002287563,0.0005489683,0.0003242628,0.000159002],"domain_scores_gemma":[0.9932175,0.004861619,0.0002482538,0.0005526977,0.000899086,0.0002207701],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00300272,0.002453342,0.02893952,0.001521646,0.001204048,0.00160426,0.0004768988,0.2710646,0.01308888,0.001075427,0.03813092,0.6374377],"study_design_scores_gemma":[0.0001570075,0.0006514926,0.005491823,0.00005329085,0.0001918629,0.0003611729,0.0001726309,0.9837264,0.006061151,0.0006366985,0.002447153,0.00004929712],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7779707,0.006705358,0.1667455,0.002320274,0.0009226218,0.001267364,0.01490659,0.023213,0.005948683],"genre_scores_gemma":[0.8222576,0.001696061,0.1157939,0.0008521759,0.0001919303,0.0007542829,0.0526622,0.0005663726,0.005225451],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.02104991,"threshold_uncertainty_score":0.0418548,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03213257730883608,"score_gpt":0.3940973742237627,"score_spread":0.3619647969149266,"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."}}