{"id":"W4412171536","doi":"10.1109/memea65319.2025.11068011","title":"Comparative Analysis of Automated Rules-Based Scoring Versus Human Scoring of Sleep Apnea","year":2025,"lang":"en","type":"article","venue":"","topic":"Context-Aware Activity Recognition Systems","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa; Bruyère; Royal Ottawa Mental Health Centre; Carleton University","funders":"","keywords":"Sleep apnea; Computer science; Apnea; Artificial intelligence; Data mining; Medicine; Internal medicine","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01441798,0.0009746504,0.001202257,0.002320483,0.0006089903,0.002173305,0.0009989942,0.0008554046,0.001143774],"category_scores_gemma":[0.0413531,0.000223798,0.0006683975,0.001113424,0.0006150166,0.0009194568,0.0009999083,0.0004011979,0.0006954733],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000634868,"about_ca_system_score_gemma":0.0006970646,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003193814,"about_ca_topic_score_gemma":0.00327203,"domain_scores_codex":[0.9798496,0.01182005,0.001315611,0.002379113,0.004171395,0.0004641753],"domain_scores_gemma":[0.9408106,0.03933952,0.002327997,0.004299647,0.01249668,0.0007255896],"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":[0.009619545,0.001079813,0.1994633,0.0009271601,0.00198367,0.0006126618,0.00197892,0.05561594,0.0259364,0.001143408,0.005097147,0.696542],"study_design_scores_gemma":[0.0005018101,0.004018493,0.2520496,0.0001854958,0.0008549716,0.001351552,0.001623981,0.6924918,0.03780732,0.002261129,0.006548868,0.0003050644],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9404681,0.0008282644,0.05145174,0.0001380418,0.0002027186,0.0002988094,0.0006754162,0.001446381,0.004490517],"genre_scores_gemma":[0.9605905,0.0001492043,0.03696898,0.00008297454,0.00005796391,0.0001230936,0.00108659,0.0001260079,0.000814799],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01441798,"threshold_uncertainty_score":0.07625043,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06373990278501908,"score_gpt":0.3505173365575344,"score_spread":0.2867774337725154,"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."}}