{"id":"W4410382902","doi":"10.1136/gutjnl-2025-335091","title":"Large language models for detecting colorectal polyps in endoscopic images","year":2025,"lang":"en","type":"article","venue":"Gut","topic":"Colorectal Cancer Screening and Detection","field":"Medicine","cited_by":8,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal","funders":"Norges Forskningsråd; European Commission","keywords":"Colorectal Polyp; Medicine; Computer science; Colonoscopy; Rectal Polyp; Radiology; Artificial intelligence; Natural language processing; Colorectal cancer; Rectum; Internal medicine; Cancer","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":[0.0001665422,0.00008059299,0.0001688233,0.0001707767,0.00006092296,0.00001340956,0.00003935932,0.00006569838,0.00001327878],"category_scores_gemma":[0.0001658479,0.00007788665,0.00006121336,0.0002434921,0.00001235524,0.00005386684,0.00002905026,0.0001373817,0.00000360698],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001023474,"about_ca_system_score_gemma":0.00007013061,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002992762,"about_ca_topic_score_gemma":0.0007206177,"domain_scores_codex":[0.9993672,0.00001721965,0.0001379572,0.0001845287,0.00007260773,0.0002205201],"domain_scores_gemma":[0.9997084,0.00009248858,0.000026753,0.0001078937,0.00003185046,0.00003263228],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.02170805,0.0006915078,0.07126734,0.001680202,0.0003366505,0.0001995805,0.01176816,0.0007878232,0.5412372,0.002937128,0.003848384,0.343538],"study_design_scores_gemma":[0.02037876,0.004482164,0.1246586,0.001216241,0.0002523684,0.00006478293,0.007251241,0.1268699,0.7070468,0.003879553,0.003313768,0.000585856],"study_design_candidate":"bench_or_experimental","study_design_consensus":"bench_or_experimental","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9675541,0.0008682435,0.02181856,0.000227512,0.0002429013,0.0004232212,0.00001280794,0.0001164961,0.008736152],"genre_scores_gemma":[0.9970254,0.000006603047,0.001183258,0.0002005962,0.00009146133,0.0001181523,0.000007499441,0.0000101787,0.00135686],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3429521,"threshold_uncertainty_score":0.3176126,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01333324211269625,"score_gpt":0.3012438695644705,"score_spread":0.2879106274517743,"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."}}