Minimizing and quantifying uncertainty in AI-informed decisions: Applications in medicine
Abstract
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.
Article Details
Journal Info
Proceedings of the National Academy of Sciences
National Academy of Sciences
Authors (34)
Samuel D. Curtis
Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine
Sambit Panda
Department of Biomedical Engineering, Johns Hopkins University
Adam Li
Department of Computer Science, Columbia University
Haoyin Xu
Department of Biomedical Engineering, Johns Hopkins University
Yuxin Bai
Department of Biomedical Engineering, Johns Hopkins University
Itsuki Ogihara
Department of Biomedical Engineering, Johns Hopkins University
Eliza O’Reilly
Department of Applied Mathematics and Statistic, Johns Hopkins University
Yuxuan Wang
Lisa Dobbyn
Department of Oncology, the Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine
Maria Popoli
Department of Oncology, the Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine
Janine Ptak
Department of Oncology, the Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine
Nadine Nehme
Department of Oncology, the Sidney Kimmel Cancer Center, Johns Hopkins University School of Medicine
Natalie Silliman
Department of Oncology, the Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine
Jeanne Tie
Division of Personalized Oncology, Walter and Eliza Hall Institute of Medical Research
Peter Gibbs
Division of Personalized Oncology, Walter and Eliza Hall Institute of Medical Research
Lan T. Ho-Pham
BioMedical Research Center, Pham Ngoc Thach University of Medicine
Bich N. H. Tran
Saigon Precision Medicine Research Center
Thach S. Tran
Saigon Precision Medicine Research Center
Tuan V. Nguyen
Saigon Precision Medicine Research Center
Ehsan Irajizad
Department of Biostatistics, The University of Texas MD Anderson Cancer Center
Michael Goggins
Department of Oncology, the Sidney Kimmel Cancer Center, Johns Hopkins University School of Medicine
Christopher L. Wolfgang
Department of Surgery, New York University Langone
Tian-Li Wang
Department of Pathology, Johns Hopkins University School of Medicine
Ie-Ming Shih
Department of Pathology, Johns Hopkins University School of Medicine
Amanda Fader
Department of Gynecology and Obstetrics, Johns Hopkins Medical Institutions
Anne Marie Lennon
Department of Medicine, University of Pittsburgh Medical Center
Ralph H. Hruban
Department of Oncology, the Sidney Kimmel Cancer Center, Johns Hopkins University School of Medicine
Chetan Bettegowda
Lucy Gilbert
Department of Oncology, McGill University Health Centre, Montreal
Kenneth W. Kinzler
Department of Oncology, the Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine
Nickolas Papadopoulos
Department of Oncology, the Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine
Bert Vogelstein
Department of Oncology, the Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine
Joshua T. Vogelstein
Department of Biomedical Engineering, Johns Hopkins University
Christopher Douville
Department of Oncology, the Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine