{"id":"W3190176307","doi":"10.1136/bjophthalmol-2021-319030","title":"Accuracy of automated machine learning in classifying retinal pathologies from ultra-widefield pseudocolour fundus images","year":2021,"lang":"en","type":"article","venue":"British Journal of Ophthalmology","topic":"Retinal Imaging and Analysis","field":"Medicine","cited_by":56,"is_retracted":false,"has_abstract":true,"ca_institutions":"Centre Hospitalier de l’Université de Montréal; École de Technologie Supérieure; Hôpital Maisonneuve-Rosemont","funders":"","keywords":"Artificial intelligence; Interpretability; Bespoke; Computer science; Machine learning; Medicine; Receiver operating characteristic; Pattern recognition (psychology)","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":[0.0006634967,0.0001515746,0.0008885289,0.0002274255,0.00006337008,0.00005156889,0.0001623364,0.0001691454,0.0006901072],"category_scores_gemma":[0.005567438,0.0001592537,0.0002844651,0.0003858261,0.0001502373,0.000158662,0.00005639784,0.00112232,0.000004248613],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004735916,"about_ca_system_score_gemma":0.0002313796,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007910678,"about_ca_topic_score_gemma":0.000007411618,"domain_scores_codex":[0.9977454,0.0004972936,0.0009404399,0.0002446054,0.0002747174,0.0002975734],"domain_scores_gemma":[0.9977059,0.0007988011,0.0007302654,0.0001509519,0.0005094603,0.0001046577],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0002497917,0.0002990572,0.6622292,0.00009293346,0.000185629,0.1977222,0.0001319048,0.0001863899,0.1273974,0.000002659222,0.0005855547,0.0109172],"study_design_scores_gemma":[0.001948018,0.0005144657,0.548551,0.001436461,0.0002332421,0.4251731,0.001049464,0.0006149939,0.01997245,0.000228413,0.0001143612,0.0001640739],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9849667,0.01110667,0.00007828663,0.00143631,0.0001491989,0.00005483216,0.0000184259,0.00003514247,0.002154446],"genre_scores_gemma":[0.9928698,0.00140209,0.0050655,0.00007848441,0.0001041552,0.000002017127,0.0000328181,0.00002000675,0.0004251059],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2274509,"threshold_uncertainty_score":0.7556189,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03033893365081698,"score_gpt":0.3248056799866705,"score_spread":0.2944667463358536,"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."}}