{"id":"W4387978272","doi":"10.1016/j.media.2023.103015","title":"Do we really need dice? The hidden region-size biases of segmentation losses","year":2023,"lang":"en","type":"article","venue":"Medical Image Analysis","topic":"Advanced Neural Network Applications","field":"Computer Science","cited_by":45,"is_retracted":false,"has_abstract":false,"ca_institutions":"Centre Hospitalier de l’Université de Montréal; École de Technologie Supérieure","funders":"H2020 Marie Skłodowska-Curie Actions; Fonds de recherche du Québec – Nature et technologies; HORIZON EUROPE Marie Sklodowska-Curie Actions; Natural Sciences and Engineering Research Council of Canada","keywords":"Dice; Ground truth; Segmentation; Computer science; Artificial intelligence; Perspective (graphical); Matching (statistics); Upper and lower bounds; Pattern recognition (psychology); Mathematics; Statistics","routes":{"ca_aff":true,"ca_fund":true,"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":[0.0003797788,0.0001112117,0.0002409998,0.0002039112,0.0001547959,0.00007250532,0.001096624,0.00004849229,0.000113647],"category_scores_gemma":[0.0009157283,0.00007466985,0.0001908196,0.006314483,0.0003539267,0.0002835449,0.0003030213,0.0001362416,0.00007581594],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002024289,"about_ca_system_score_gemma":0.00005230381,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005088501,"about_ca_topic_score_gemma":0.00003645064,"domain_scores_codex":[0.9981149,0.0001368736,0.0003705906,0.0003494514,0.0008041617,0.0002239977],"domain_scores_gemma":[0.9968041,0.001925341,0.0001953382,0.000818621,0.0001236282,0.0001329876],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00002490573,0.0003180348,0.007869409,0.00005996187,0.001703002,0.0003743443,0.002842904,0.003599856,0.007917535,0.01400554,0.09711391,0.8641706],"study_design_scores_gemma":[0.001435387,0.0001814655,0.04162998,0.0002506007,0.002355617,0.00005635267,0.002106516,0.8764092,0.01328238,0.05125412,0.01006369,0.0009747354],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02082541,0.0002989491,0.9305919,0.0474908,0.00005865275,0.0001811392,0.000004643392,0.0002613198,0.0002871383],"genre_scores_gemma":[0.9699131,0.003587669,0.02397981,0.001173637,0.000198114,0.0001204805,0.00003816747,0.00001694377,0.0009720399],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9490877,"threshold_uncertainty_score":0.3044948,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03027854292138389,"score_gpt":0.3202295275761869,"score_spread":0.289950984654803,"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."}}