{"id":"W4414120335","doi":"10.1007/s44352-025-00015-0","title":"Replacing experts with student raters for assessing knee recess distension in ultrasound images of adult patients","year":2025,"lang":"en","type":"article","venue":"Discover Imaging.","topic":"Shoulder Injury and Treatment","field":"Medicine","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University; University of Toronto","funders":"","keywords":"Intraclass correlation; Ultrasound; Segmentation; Consistency (knowledge bases); Expert opinion; Medical imaging; Pearson product-moment correlation coefficient","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":[0.00008221489,0.0001677721,0.0003055396,0.0001407962,0.00005908927,0.00005920595,0.00005572322,0.00002555117,0.000004375721],"category_scores_gemma":[0.0001097939,0.0001184981,0.00006669714,0.0001813551,0.00007391482,0.0003031601,0.00003131619,0.00006706733,4.035659e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000130654,"about_ca_system_score_gemma":0.00007752804,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001221305,"about_ca_topic_score_gemma":0.0000167625,"domain_scores_codex":[0.9989198,0.00002919278,0.0002951614,0.0003333444,0.0002143127,0.0002082007],"domain_scores_gemma":[0.9992709,0.0001467069,0.0001030647,0.000280274,0.0001596418,0.00003948538],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0007323613,0.0007056463,0.9883024,0.000183606,0.00009271672,0.00002413944,0.001088067,0.00001623117,0.00390679,0.00003608429,0.0006284585,0.004283553],"study_design_scores_gemma":[0.008586307,0.0002541603,0.8705273,0.001572087,0.0003479496,0.000009223134,0.006263754,0.00005679185,0.1120016,0.00007549212,0.000114029,0.0001912486],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9951148,0.000175719,0.001413271,0.0004858816,0.0002083949,0.0007305182,0.00002038292,0.00002216602,0.001828884],"genre_scores_gemma":[0.998185,0.0000132821,0.0007649523,0.0004738631,0.00001570664,0.00008708314,0.0001085999,0.00001854528,0.0003329918],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.117775,"threshold_uncertainty_score":0.4832214,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.008814316204384057,"score_gpt":0.3291198956246932,"score_spread":0.3203055794203092,"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."}}