Department of Computer Science, University of Bremen, Bremen, Germany.
Fraunhofer Institute for Digital Medicine MEVIS, Max-von-Laue-Str. 2, 28359, Bremen, Germany.
Int J Comput Assist Radiol Surg. 2024 Feb;19(2):253-260. doi: 10.1007/s11548-023-03001-1. Epub 2023 Aug 16.
Deep neural networks need to be able to indicate error likelihood via reliable estimates of their predictive uncertainty when used in high-risk scenarios, such as medical decision support. This work contributes a systematic overview of state-of-the-art approaches for decomposing predictive uncertainty into aleatoric and epistemic components, and a comprehensive comparison for Bayesian neural networks (BNNs) between mutual information decomposition and the explicit modelling of both uncertainty types via an additional loss-attenuating neuron.
Experiments are performed in the context of liver segmentation in CT scans. The quality of the uncertainty decomposition in the resulting uncertainty maps is qualitatively evaluated, and quantitative behaviour of decomposed uncertainties is systematically compared for different experiment settings with varying training set sizes, label noise, and distribution shifts.
Our results show the mutual information decomposition to robustly yield meaningful aleatoric and epistemic uncertainty estimates, while the activation of the loss-attenuating neuron appears noisier with non-trivial convergence properties. We found that the addition of a heteroscedastic neuron does not significantly improve segmentation performance or calibration, while slightly improving the quality of uncertainty estimates.
Mutual information decomposition is simple to implement, has mathematically pleasing properties, and yields meaningful uncertainty estimates that behave as expected under controlled changes to our data set. The additional extension of BNNs with loss-attenuating neurons provides no improvement in terms of segmentation performance or calibration in our setting, but marginal benefits regarding the quality of decomposed uncertainties.
当在高风险场景（例如医疗决策支持）中使用时，深度神经网络需要能够通过可靠地估计其预测不确定性来指示错误可能性。这项工作系统地概述了将预测不确定性分解为随机和认知分量的最新方法，并针对贝叶斯神经网络（BNN）在互信息分解和通过附加的损失衰减神经元对两种不确定性类型进行显式建模之间进行了全面比较。
在 CT 扫描中的肝脏分割上下文中进行实验。定性评估所得不确定性图中不确定性分解的质量，并针对不同的实验设置（具有不同的训练集大小、标签噪声和分布转移）系统地比较分解不确定性的定量行为。
我们的结果表明，互信息分解能够稳健地产生有意义的随机和认知不确定性估计，而损失衰减神经元的激活似乎具有复杂的收敛特性，噪声较大。我们发现，添加异方差神经元不会显著提高分割性能或校准，而略微提高不确定性估计的质量。
互信息分解易于实现，具有数学上令人愉悦的性质，并产生有意义的不确定性估计，这些估计在我们的数据集中受控变化下表现如预期。在 BNN 中添加损失衰减神经元在我们的设置中不会在分割性能或校准方面提供任何改进，但在分解不确定性的质量方面有一些边际收益。
