Aleatoric uncertainty estimation with test-time augmentation for medical image segmentation with convolutional neural networks

TL;DR

This paper introduces a novel test-time augmentation method to estimate aleatoric uncertainty in CNN-based medical image segmentation, improving uncertainty quantification and reducing overconfidence in predictions.

Contribution

It provides a theoretical framework for test-time augmentation and demonstrates its effectiveness in estimating aleatoric uncertainty in medical image segmentation tasks.

Findings

Test-time augmentation-based uncertainty outperforms dropout-based methods.

Proposed method reduces overconfident incorrect predictions.

Improves segmentation reliability in MRI brain imaging.

Abstract

Despite the state-of-the-art performance for medical image segmentation, deep convolutional neural networks (CNNs) have rarely provided uncertainty estimations regarding their segmentation outputs, e.g., model (epistemic) and image-based (aleatoric) uncertainties. In this work, we analyze these different types of uncertainties for CNN-based 2D and 3D medical image segmentation tasks. We additionally propose a test-time augmentation-based aleatoric uncertainty to analyze the effect of different transformations of the input image on the segmentation output. Test-time augmentation has been previously used to improve segmentation accuracy, yet not been formulated in a consistent mathematical framework. Hence, we also propose a theoretical formulation of test-time augmentation, where a distribution of the prediction is estimated by Monte Carlo simulation with prior distributions of parameters in an image acquisition model that involves image transformations and noise. We compare and combine our proposed aleatoric uncertainty with model uncertainty. Experiments with segmentation of fetal brains and brain tumors from 2D and 3D Magnetic Resonance Images (MRI) showed that 1) the test-time augmentation-based aleatoric uncertainty provides a better uncertainty estimation than calculating the test-time dropout-based model uncertainty alone and helps to reduce overconfident incorrect predictions, and 2) our test-time augmentation outperforms a single-prediction baseline and dropout-based multiple predictions.