Functional Space Variational Inference for Uncertainty Estimation in Computer Aided Diagnosis
This work addresses uncertainty estimation for improved patient safety in computer-aided diagnosis, presenting an incremental improvement over existing Bayesian methods.
The paper tackles the problem of generating well-calibrated probabilistic outputs in deep neural networks for medical image analysis, specifically in skin lesion classification, by shifting Bayesian inference to functional space to achieve better uncertainty estimates with lower computational cost.
Deep neural networks have revolutionized medical image analysis and disease diagnosis. Despite their impressive performance, it is difficult to generate well-calibrated probabilistic outputs for such networks, which makes them uninterpretable black boxes. Bayesian neural networks provide a principled approach for modelling uncertainty and increasing patient safety, but they have a large computational overhead and provide limited improvement in calibration. In this work, by taking skin lesion classification as an example task, we show that by shifting Bayesian inference to the functional space we can craft meaningful priors that give better calibrated uncertainty estimates at a much lower computational cost.