Second opinion needed: communicating uncertainty in medical machine learning (2021-01-05T00:00:00.000000Z)

TL;DR

An overview of different approaches to uncertainty quantification and abstention for machine learning is provided and how these techniques could improve the safety and reliability of current ML systems being used in healthcare settings are highlighted.

Abstract

There is great excitement that medical artificial intelligence (AI) based on machine learning (ML) can be used to improve decision making at the patient level in a variety of healthcare settings. However, the quantification and communication of uncertainty for individual predictions is often neglected even though uncertainty estimates could lead to more principled decision-making and enable machine learning models to automatically or semi-automatically abstain on samples for which there is high uncertainty. In this article, we provide an overview of different approaches to uncertainty quantification and abstention for machine learning and highlight how these techniques could improve the safety and reliability of current ML systems being used in healthcare settings. Effective quantification and communication of uncertainty could help to engender trust with healthcare workers, while providing safeguards against known failure modes of current machine learning approaches. As machine learning becomes further integrated into healthcare environments, the ability to say “I’m not sure” or “I don’t know” when uncertain is a necessary capability to enable safe clinical deployment.

Authors

Benjamin Kompa

1 papers

Jasper Snoek

8 papers

A. Beam

1 papers

TL;DR

Abstract

Authors

References75 items

The Convergence of Human and Artificial Intelligence on Clinical Care - Part I

Consistent Estimators for Learning to Defer to an Expert

A Human-Centered Evaluation of a Deep Learning System Deployed in Clinics for the Detection of Diabetic Retinopathy

Clinical prediction models

BatchEnsemble: An Alternative Approach to Efficient Ensemble and Lifelong Learning

Pitfalls of In-Domain Uncertainty Estimation and Ensembling in Deep Learning

The Case for Bayesian Deep Learning

Challenges to the Reproducibility of Machine Learning Models in Health Care.

Deep Gamblers: Learning to Abstain with Portfolio Theory

Quality of Uncertainty Quantification for Bayesian Neural Network Inference

Analyzing the role of model uncertainty for electronic health records

Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift

Reporting of artificial intelligence prediction models

Conformal Prediction Under Covariate Shift

Adversarial attacks on medical machine learning

The limits of distribution-free conditional predictive inference

A systematic review shows no performance benefit of machine learning over logistic regression for clinical prediction models.

SelectiveNet: A Deep Neural Network with an Integrated Reject Option

High-performance medicine: the convergence of human and artificial intelligence

Frequentist uncertainty estimates for deep learning

Bayesian Deep Convolutional Networks with Many Channels are Gaussian Processes

Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: A cross-sectional study

Big Data and Machine Learning in Health Care.

A Flexible and Adaptive Framework for Abstention Under Class Imbalance

Gaussian Process Behaviour in Wide Deep Neural Networks

CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning

Implicit Weight Uncertainty in Neural Networks.

Deep Neural Networks as Gaussian Processes

Selective Classification for Deep Neural Networks

Deep Learning at Chest Radiography: Automated Classification of Pulmonary Tuberculosis by Using Convolutional Neural Networks.

What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?

Multiplicative Normalizing Flows for Variational Bayesian Neural Networks

Dermatologist-level classification of skin cancer with deep neural networks

Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs.

Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles

A calibration hierarchy for risk models was defined: from utopia to empirical data.

Structured and Efficient Variational Deep Learning with Matrix Gaussian Posteriors

Variational Inference: A Review for Statisticians

Report of the First International Workshop on Learning over Multiple Contexts (LMCE 2014)

Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning

Weight Uncertainty in Neural Networks

Probabilistic Backpropagation for Scalable Learning of Bayesian Neural Networks

Calibration of Risk Prediction Models

F1000Prime recommendation of Calibration of risk prediction models: impact on decision-analytic performance.

Risk prediction with machine learning and regression methods

Comparing risk prediction models

Practical Variational Inference for Neural Networks

Dataset Shift in Machine Learning

TWO CULTURES

A tutorial on conformal prediction

Strictly Proper Scoring Rules, Prediction, and Estimation

Mixture Regression for Covariate Shift

Gaussian Processes For Machine Learning

Gaussian Processes in Machine Learning

Statistical Modeling: The Two Cultures (with comments and a rejoinder by the author)

An Optimal Reject Rule for Binary Classifiers

Multiple Classifier Systems

Popular ensemble methods

The Error-Reject Tradeoff

An optimum character recognition system using decision functions

Reprints and permission information

Appendix for: How Good is the Bayes Posterior in Deep Neural Networks Really?

the 2020 CHI Conference on Human Factors in Computing Systems 1 – 12

BatchEnsemble: Efficient Ensemble of Deep Neural Networks via Rank-1 Perturbation

Consistent algorithms for multiclass classification with an abstain option

Arti ﬁ cial intelligence in healthcare

Uncertainty in Deep Learning

A cross-sectional study

Black-box α-divergence Minimization

Patterns of dataset shift

Bayesian Learning for Neural Networks ributed to the conception, writing, and editing

PREDICTIVE INFERENCE

Introduction to Artificial Intelligence: Reasoning with Constraints

UCI machine learning repository-heart disease data set

On optimum recognition error and reject tradeoff

Field of Study