{"id":"W3131158873","doi":"10.21203/rs.3.rs-151985/v1","title":"Medical imaging algorithms exacerbate biases in underdiagnosis","year":2021,"lang":"en","type":"preprint","venue":"Research Square","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":10,"is_retracted":false,"has_abstract":false,"ca_institutions":"Vector Institute; University of Toronto","funders":"","keywords":"Algorithm; Computer science; Medical imaging; Artificial intelligence","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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.0881145,0.001275731,0.001565913,0.003913693,0.001394396,0.006315355,0.002165001,0.003846357,0.0117233],"category_scores_gemma":[0.5062705,0.001206076,0.001825059,0.003483932,0.003300885,0.005119409,0.003248462,0.003485845,0.002830339],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001091489,"about_ca_system_score_gemma":0.001925081,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002849485,"about_ca_topic_score_gemma":0.002440331,"domain_scores_codex":[0.9233096,0.05283532,0.005490089,0.009382547,0.007099968,0.001882502],"domain_scores_gemma":[0.3620707,0.5294366,0.03710771,0.05262988,0.01683388,0.001921255],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.003149381,0.0004136167,0.5750843,0.001729806,0.003202601,0.00181469,0.004313013,0.00606989,0.00353605,0.05542616,0.04398349,0.3012769],"study_design_scores_gemma":[0.0009345441,0.0005747782,0.2688625,0.001723495,0.006446135,0.009671572,0.003039627,0.04431494,0.01817112,0.5687327,0.077306,0.0002226637],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.4684969,0.0245872,0.3094884,0.1082088,0.007293443,0.0008496345,0.006063846,0.002364812,0.07264681],"genre_scores_gemma":[0.9210334,0.002294046,0.04773014,0.01651018,0.004297607,0.0001663208,0.001472479,0.000626768,0.005868983],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9118855,"threshold_uncertainty_score":0.4659996,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4034576381905601,"score_gpt":0.5734717171442163,"score_spread":0.1700140789536562,"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."}}