Direct Uncertainty Prediction for Medical Second Opinions

TL;DR

This paper introduces Direct Uncertainty Prediction (DUP), a method for estimating patient case uncertainty directly from raw features, improving identification of cases needing second opinions in medical diagnosis.

Contribution

The paper demonstrates that training models to predict uncertainty directly from raw data outperforms traditional two-step methods, supported by theory and large-scale medical imaging experiments.

Findings

DUP outperforms Uncertainty Via Classification in accuracy.

Direct prediction of uncertainty improves identification of cases needing second opinions.

Theoretical and empirical evidence supports DUP's effectiveness.

Abstract

The issue of disagreements amongst human experts is a ubiquitous one in both machine learning and medicine. In medicine, this often corresponds to doctor disagreements on a patient diagnosis. In this work, we show that machine learning models can be trained to give uncertainty scores to data instances that might result in high expert disagreements. In particular, they can identify patient cases that would benefit most from a medical second opinion. Our central methodological finding is that Direct Uncertainty Prediction (DUP), training a model to predict an uncertainty score directly from the raw patient features, works better than Uncertainty Via Classification, the two-step process of training a classifier and postprocessing the output distribution to give an uncertainty score. We show this both with a theoretical result, and on extensive evaluations on a large scale medical imaging application.