Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Uncertainty Quantification for Machine Learning in Healthcare: A Survey

L. Julián Lechuga López, Shaza Elsharief, Dhiyaa Al Jorf, Firas Darwish, Congbo Ma, Farah E. Shamout

TL;DR

The paper surveys uncertainty quantification (UQ) in machine learning for healthcare, emphasizing safety, reliability, and clinician trust. It develops a pipeline-centered framework summarizing UQ methods across data preprocessing, model training, and evaluation, and provides a taxonomy of approaches from probabilistic, non-probabilistic, and hybrid families. It aggregates open healthcare datasets and studies, analyzes comparative performance, and discusses deployment, fairness, and regulatory considerations. The paper concludes with a roadmap toward standardized evaluation, integrated UQ across the ML lifecycle, and collaboration with clinicians to enable trustworthy clinical AI.

Abstract

Uncertainty Quantification (UQ) is pivotal in enhancing the robustness, reliability, and interpretability of Machine Learning (ML) systems for healthcare, optimizing resources and improving patient care. Despite the emergence of ML-based clinical decision support tools, the lack of principled quantification of uncertainty in ML models remains a major challenge. Current reviews have a narrow focus on analyzing the state-of-the-art UQ in specific healthcare domains without systematically evaluating method efficacy across different stages of model development, and despite a growing body of research, its implementation in healthcare applications remains limited. Therefore, in this survey, we provide a comprehensive analysis of current UQ in healthcare, offering an informed framework that highlights how different methods can be integrated into each stage of the ML pipeline including data processing, training and evaluation. We also highlight the most popular methods used in healthcare and novel approaches from other domains that hold potential for future adoption in the medical context. We expect this study will provide a clear overview of the challenges and opportunities of implementing UQ in the ML pipeline for healthcare, guiding researchers and practitioners in selecting suitable techniques to enhance the reliability, safety and trust from patients and clinicians on ML-driven healthcare solutions.

Uncertainty Quantification for Machine Learning in Healthcare: A Survey

TL;DR

The paper surveys uncertainty quantification (UQ) in machine learning for healthcare, emphasizing safety, reliability, and clinician trust. It develops a pipeline-centered framework summarizing UQ methods across data preprocessing, model training, and evaluation, and provides a taxonomy of approaches from probabilistic, non-probabilistic, and hybrid families. It aggregates open healthcare datasets and studies, analyzes comparative performance, and discusses deployment, fairness, and regulatory considerations. The paper concludes with a roadmap toward standardized evaluation, integrated UQ across the ML lifecycle, and collaboration with clinicians to enable trustworthy clinical AI.

Abstract

Uncertainty Quantification (UQ) is pivotal in enhancing the robustness, reliability, and interpretability of Machine Learning (ML) systems for healthcare, optimizing resources and improving patient care. Despite the emergence of ML-based clinical decision support tools, the lack of principled quantification of uncertainty in ML models remains a major challenge. Current reviews have a narrow focus on analyzing the state-of-the-art UQ in specific healthcare domains without systematically evaluating method efficacy across different stages of model development, and despite a growing body of research, its implementation in healthcare applications remains limited. Therefore, in this survey, we provide a comprehensive analysis of current UQ in healthcare, offering an informed framework that highlights how different methods can be integrated into each stage of the ML pipeline including data processing, training and evaluation. We also highlight the most popular methods used in healthcare and novel approaches from other domains that hold potential for future adoption in the medical context. We expect this study will provide a clear overview of the challenges and opportunities of implementing UQ in the ML pipeline for healthcare, guiding researchers and practitioners in selecting suitable techniques to enhance the reliability, safety and trust from patients and clinicians on ML-driven healthcare solutions.
Paper Structure (45 sections, 3 figures)

This paper contains 45 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Motivation
  3. Uncertainty Quantification in Healthcare: Applications Across the Machine Learning Pipeline
  4. Data Preprocessing
  5. Applications
  6. Opportunities
  7. Label Handling.
  8. Noise Reduction and Augmentation.
  9. Input Transformations.
  10. Model Training
  11. Applications
  12. Opportunities
  13. Architecture-Level Approaches.
  14. Loss Modification.
  15. Model Evaluation
  16. ...and 30 more sections

Figures (3)

  • Figure 1: Overview of distribution and characteristics of reviewed papers.(a) Prevalence of different uncertainty quantification methods across the surveyed papers. (b) Distribution of studies according to the machine learning pipeline stages: data processing, model training, and evaluation. (c) Code availability rates across papers published in various conferences and journals. (d) Medical domains represented in the reviewed studies, alongside their corresponding code availability.
  • Figure 2: UQ in the Clinical Machine Learning Pipeline.(a) Key sources of uncertainty identified at each stage of the pipeline. (b) Expected outcomes of implementing UQ methods for clinical tasks. (c) Relevant UQ techniques applied during data processing, model training, and evaluation.
  • Figure A1: Uncertainty quantification across the ML pipeline. Key UQ methods from different domains applied at each stage: data processing, model training, and evaluation.