Direct Uncertainty Prediction for Medical Second Opinions

By Maithra Raghu, Katy Blumer, Rory Sayres, Z. Obermeyer, Robert D. Kleinberg, Sendhil Mullainathan, Jon M. Kleinberg. 2018

156 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Family Practice, Health Informatics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Family Practice

Percy Liang

In Health Informatics

David A. Sontag

Citations in the corpus

Measuring Calibration in Deep Learning (2019, background): “The reliability of a machine learning model’s confidence in its predictions is critical for high risk applications, such as deciding whether to trust a medical diagnosis prediction (Crowson et al., 2016; Jiang et al., 2011; Raghu et al., 2018).”
Transfusion: Understanding Transfer Learning with Applications to Medical Imaging (2019, unknown): “Because each block’s nal layer has a di erent number of neurons and channels, the CCA similarities for each of the heatmaps is at a slightly di erent scale Raghu et al. (2017). Nevertheless, we can observe some interesting patterns. Firstly, we note that in lower layers (Block 1 nal, Block 2 nal), t”
Selective Classification Can Magnify Disparities Across Groups (2020, methodology): “Others incorporate abstention into model training (Bartlett & Wegkamp, 2008; Geifman & El-Yaniv, 2019; Feng et al., 2019) and learn to abstain on examples human experts are more likely to get correct (Raghu et al., 2019; Mozannar & Sontag, 2020; De et al., 2020).”
A survey of uncertainty in deep neural networks (2021, methodology): “the training process [44] or external components that have to be trained and evaluated additionally [46].”
Consistent Estimators for Learning to Defer to an Expert (2020, background): “…expert agreement with the target, this can be done by learning a model where the target is whether the expert agrees with the task label and extracting conﬁdence scores from this model [RBS + 19], and ﬁnally 3) compare who between the classiﬁer and the expert is more conﬁdent and accordingly defer.”