Minimizing and quantifying uncertainty in AI-informed decisions: Applications in medicine

• 成果类型: Article
• 署名作者: Curtis, Samuel D.; Panda, Sambit; Li, Adam; Xu, Haoyin; Bai, Yuxin; Ogihara, Itsuki; O'Reilly, Eliza; Wang, Yuxuan; Dobbyn, Lisa; Popoli, Maria; Ptak, Janine; Nehme, Nadine; Silliman, Natalie; Tie, Jeanne; Gibbs, Peter; -Pham, Lan T. Ho; Tran, Bich N. H.; Tran, Thach S.; Nguyen, Tuan V.; Irajizad, Ehsan; Goggins, Michael; Wolfgang, Christopher L.; Wang, Tian-Li; Shih, Ie-Ming; Fader, Amanda; Lennon, Anne Marie; Hruban, Ralph H.; Bettegowda, Chetan; Gilbert, Lucy; Kinzler, Kenneth W.; Papadopoulos, Nickolas; Vogelstein, Bert; Vogelstein, Joshua T.; Douville, Christopher
• 署名单位: Johns Hopkins University; Johns Hopkins University; Johns Hopkins University; Johns Hopkins University; Johns Hopkins University; Johns Hopkins Medicine; Johns Hopkins University; Johns Hopkins University; Columbia University; Johns Hopkins University; Johns Hopkins University; Howard Hughes Medical Institute; Walter & Eliza Hall Institute; University of Melbourne; Peter Maccallum Cancer Center; University of Melbourne; Western Health; University of Technology Sydney; University of Technology Sydney; University of New South Wales Sydney; Johns Hopkins University; Johns Hopkins University; Johns Hopkins Medicine; Johns Hopkins University; Johns Hopkins Medicine; Pennsylvania Commonwealth System of Higher Education (PCSHE); University of Pittsburgh; Pennsylvania Commonwealth System of Higher Education (PCSHE); University of Pittsburgh; Johns Hopkins University; McGill University; Johns Hopkins University
• 刊物名称: PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
• ISSN/ISSBN: 0027-13539
• DOI: 10.1073/pnas.242420312
• 发表日期: 2025-08-26
• 关键词: strong universal consistency; circulating dna; random forests; plasma dna; cancer; aneuploidy

摘要

AI is now a cornerstone of modern dataset analysis. In many real world applications, practitioners are concerned with controlling specific kinds of errors, rather than minimizing the overall number of errors. For example, biomedical screening assays may primarily be concerned with mitigating the number of false positives rather than false negatives. Quantifying uncertainty in AI-based predictions, and in particular those controlling specific kinds of errors, remains theoretically and practically challenging. We develop a strategy called multidimensional informed generalized hypothesis testing (MIGHT) which we prove accurately quantifies uncertainty and confidence given sufficient data, and concomitantly controls for particular error types. Our key insight was that it is possible to integrate canonical cross-validation and parametric calibration procedures within a nonparametric ensemble method. Simulations demonstrate that while typical AI based-approaches cannot be trusted to obtain the truth, MIGHT can be. We apply MIGHT to answer an open question in liquid biopsies using circulating cell-free DNA (ccfDNA) in individuals with or without cancer: Which biomarkers, or combinations thereof, can we trust? Performance estimates produced by MIGHT on ccfDNA data have coefficients of variation that are often orders of magnitude lower than other state of the art algorithms such as support vector machines, random forests, and Transformers, while often also achieving higher sensitivity. We find that combinations of variable sets often decrease rather than increase sensitivity over the optimal single variable set because some variable sets add more noise than signal. This work demonstrates the importance of quantifying uncertainty and confidence-with theoretical guarantees-for the interpretation of real-world data.