{"id":"W4413422599","doi":"10.1016/j.acra.2025.08.014","title":"Evaluating Large Language Models for Radiology Systematic Review Title and Abstract Screening","year":2025,"lang":"en","type":"article","venue":"Academic Radiology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":2,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of British Columbia","funders":"National Institute of Biomedical Imaging and Bioengineering; National Institutes of Health; American Roentgen Ray Society; U.S. Department of Defense","keywords":"Computer science; Medical physics; Medicine; Radiology","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":[],"consensus_categories":[],"category_scores_codex":[0.000949911,0.00005915055,0.000314269,0.0001095338,0.00004777406,0.000001482306,0.00005001821,0.0001659505,0.00005592276],"category_scores_gemma":[0.001113228,0.00004792651,0.00003365427,0.0001083873,0.00002880372,0.00002494478,0.00001099914,0.0002144259,0.00002531875],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002363361,"about_ca_system_score_gemma":0.00006637401,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002298746,"about_ca_topic_score_gemma":0.000001415618,"domain_scores_codex":[0.9992706,0.00006070645,0.0003315523,0.0001413448,0.00003572558,0.0001600737],"domain_scores_gemma":[0.9992683,0.0004563057,0.00007707809,0.0001189526,0.00004467282,0.00003471698],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"systematic_review","study_design_gemma":"systematic_review","study_design_scores_codex":[0.0001998615,0.00007012053,0.002943899,0.5560547,0.000417925,0.00001594761,0.004478683,0.00009393758,0.006574337,0.07715242,0.2321107,0.1198875],"study_design_scores_gemma":[0.00268266,0.002768531,0.004041005,0.4663485,0.008967126,0.004606865,0.01636116,0.2926349,0.007557879,0.1448052,0.04686297,0.002363121],"study_design_candidate":"systematic_review","study_design_consensus":"systematic_review","genre_codex":"review","genre_gemma":"empirical","genre_scores_codex":[0.147923,0.772986,0.03513871,0.02072982,0.001443391,0.005648675,0.00002273081,0.0001736735,0.01593401],"genre_scores_gemma":[0.9346324,0.03818291,0.004367909,0.01641326,0.0005907801,0.0004377911,0.0001000639,0.00002498497,0.005249952],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7867094,"threshold_uncertainty_score":0.1954386,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.243666170429819,"score_gpt":0.5292139635232682,"score_spread":0.2855477930934492,"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."}}