{"id":"W4399799381","doi":"10.1101/2024.06.19.24309061","title":"An independent, multi-country head-to-head accuracy comparison of automated chest x-ray algorithms for the triage of pulmonary tuberculosis","year":2024,"lang":"en","type":"preprint","venue":"medRxiv","topic":"Tuberculosis Research and Epidemiology","field":"Medicine","cited_by":12,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal; Centre Hospitalier de l’Université de Montréal","funders":"National Institutes of Health; European and Developing Countries Clinical Trials Partnership; National Institute of Allergy and Infectious Diseases; European Commission","keywords":"Triage; Head (geology); Computer science; Pulmonary tuberculosis; Tuberculosis; Medicine; Algorithm; Radiology; Artificial intelligence; Medical emergency; Pathology; Geology","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":[0.003335385,0.0005989097,0.00223601,0.0005986685,0.0001145561,0.00003912762,0.001026722,0.0008625379,0.0001111334],"category_scores_gemma":[0.003856368,0.0004021061,0.0006513501,0.0005989645,0.0003082359,0.00006531339,0.0008881536,0.00194312,0.00004769542],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001563997,"about_ca_system_score_gemma":0.0005165262,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002482217,"about_ca_topic_score_gemma":0.0002819916,"domain_scores_codex":[0.9947865,0.0005979684,0.001806124,0.001129769,0.0008315441,0.0008480971],"domain_scores_gemma":[0.9939011,0.002683708,0.0005290108,0.001851691,0.0005207246,0.0005137711],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.006222742,0.009062896,0.5614737,0.01993435,0.007423114,0.0002738375,0.005067022,0.01247331,0.2601523,0.000141799,0.02946069,0.0883143],"study_design_scores_gemma":[0.00157516,0.0007280749,0.4295681,0.0009641352,0.0008246435,0.00002992343,0.0005004168,0.5546576,0.007087897,0.0001092613,0.003598269,0.0003564993],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9769275,0.005362205,0.001963599,0.008391939,0.001204194,0.004442521,0.001131666,0.0003933009,0.000183099],"genre_scores_gemma":[0.9929771,0.0005256502,0.003753425,0.0004478857,0.0006110048,0.0007734096,0.0006413584,0.0001251552,0.000145054],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5421842,"threshold_uncertainty_score":0.9998431,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1127179901596555,"score_gpt":0.4493965221128523,"score_spread":0.3366785319531967,"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."}}