{"id":"W4416887695","doi":"10.1016/j.esmorw.2025.100653","title":"Proof-of-concept: AI-assisted, natural-language-guided survival analysis achieves concordance with human-conducted results in glioblastoma","year":2025,"lang":"en","type":"article","venue":"ESMO Real World Data and Digital Oncology","topic":"Glioma Diagnosis and Treatment","field":"Medicine","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Women and Children’s Health Research Institute; University of Alberta","funders":"","keywords":"Censoring (clinical trials); Survival analysis; Concordance; Statistical analysis; Hazard ratio; Proportional hazards model; Cohort; Preprocessor; Statistical model","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":[0.0001920539,0.0002600887,0.001026802,0.0007093261,0.00006617416,0.00005943963,0.0002516925,0.00008119267,0.00001470313],"category_scores_gemma":[0.0001893535,0.0001871063,0.00008177123,0.002078957,0.0004518609,0.0003281103,0.0002803165,0.0002485963,0.000001791953],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001332516,"about_ca_system_score_gemma":0.0002453622,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001301552,"about_ca_topic_score_gemma":0.04186767,"domain_scores_codex":[0.9980333,0.00008541551,0.0006524802,0.0007060536,0.0002094159,0.0003133718],"domain_scores_gemma":[0.9982153,0.0004409071,0.0002631908,0.0008389035,0.0001210137,0.0001207338],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.006409136,0.003695125,0.7690966,0.0001495276,0.00396008,0.002716538,0.0005906655,0.00003940235,0.0007915669,0.002434708,0.007589392,0.2025273],"study_design_scores_gemma":[0.01566106,0.002333552,0.9624548,0.0003134115,0.00171046,0.0000617863,0.0008213457,0.0004988189,0.006126576,0.00007876614,0.009575139,0.0003642997],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9809914,0.0006113336,0.00001459892,0.002126793,0.000104152,0.0004758165,0.0007749648,0.00004184368,0.0148591],"genre_scores_gemma":[0.9956949,0.00001659447,0.0001578368,0.0002115436,0.00003920975,0.00003047597,0.002379025,0.0000129466,0.001457485],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.202163,"threshold_uncertainty_score":0.9756157,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0335235731364263,"score_gpt":0.3617598371786505,"score_spread":0.3282362640422242,"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."}}