{"id":"W4405796181","doi":"10.1136/rmdopen-2024-004628","title":"Anatomy-centred deep learning improves generalisability and progression prediction in radiographic sacroiliitis detection","year":2024,"lang":"en","type":"article","venue":"RMD Open","topic":"Spondyloarthritis Studies and Treatments","field":"Medicine","cited_by":9,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"Janssen Biotech; Berlin Institute of Health; Swedish Orphan Biovitrum; Technische Universität München; European Commission; Bundesministerium für Bildung und Forschung; Canon Medical Systems Corporation; Max Kade Foundation; Schering-Plough; AbbVie; Bayern Innovativ; Celgene; Pfizer; Australian Government; Wyeth; Wilhelm Sander-Stiftung; Amgen; Eli Lilly and Company; Bristol-Myers Squibb","keywords":"Medicine; Sacroiliitis; Deep learning; Radiography; Radiographic anatomy; Artificial intelligence; Anatomy; Medical physics; Radiology; Disease; Pathology","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":[0.0001406003,0.00011684,0.0002030116,0.0001234899,0.0001342476,0.0001105927,0.00003997626,0.0000707974,0.00004730757],"category_scores_gemma":[0.00004201323,0.00009602986,0.00004679761,0.0003457457,0.00005962355,0.0002116414,0.00009838805,0.0001709034,0.000007116825],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009946827,"about_ca_system_score_gemma":0.00001823653,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005290793,"about_ca_topic_score_gemma":0.0006404729,"domain_scores_codex":[0.999092,0.00005705783,0.0001795525,0.0003621874,0.0001221186,0.0001870424],"domain_scores_gemma":[0.9997348,0.00002641984,0.00002506631,0.00011555,0.00002467032,0.00007348733],"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":[0.0002447383,0.0001262519,0.4341492,0.0001580867,0.00009934328,0.0001371853,0.0005441138,0.000002951452,0.004054921,0.00002733962,0.00003820539,0.5604177],"study_design_scores_gemma":[0.005544395,0.00116749,0.9630144,0.000731138,0.0001465071,0.0002103672,0.000588891,0.01031922,0.002762851,0.0002347206,0.01509595,0.0001840961],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9796863,0.01837095,0.00001542244,0.0002516789,0.0002215498,0.000880743,0.000006505093,0.00009151561,0.000475289],"genre_scores_gemma":[0.9974166,0.001905582,0.0002132515,0.00001994808,0.00007470346,0.0001403328,0.00003720419,0.00001659805,0.0001757441],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5602335,"threshold_uncertainty_score":0.3915984,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01419501059822151,"score_gpt":0.2989372230836289,"score_spread":0.2847422124854074,"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."}}