Direct Uncertainty Prediction for Medical Second Opinions
This work addresses the challenge of reducing diagnostic errors and improving patient care by targeting cases with high expert disagreement, though it is incremental as it builds on existing uncertainty prediction methods.
The paper tackles the problem of expert disagreements in medical diagnoses by training machine learning models to predict uncertainty scores for patient cases, identifying those that would benefit most from a second opinion. It demonstrates that Direct Uncertainty Prediction (DUP) outperforms Uncertainty Via Classification, with results validated through theoretical analysis and large-scale medical imaging evaluations.
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.