Maximum Entropy on Erroneous Predictions (MEEP): Improving model calibration for medical image segmentation
This addresses the issue of unreliable models in medical image segmentation, which is critical for clinical decision-making, but it is an incremental improvement as it builds on existing segmentation methods without a paradigm shift.
The paper tackles the problem of overconfident and poorly calibrated deep neural networks in medical image segmentation by introducing Maximum Entropy on Erroneous Predictions (MEEP), a training strategy that selectively penalizes overconfident predictions on misclassified pixels, leading to improvements in both model calibration and segmentation quality in tasks like white matter hyperintensity and atrial segmentation in MRI.
Modern deep neural networks achieved remarkable progress in medical image segmentation tasks. However, it has recently been observed that they tend to produce overconfident estimates, even in situations of high uncertainty, leading to poorly calibrated and unreliable models. In this work we introduce Maximum Entropy on Erroneous Predictions (MEEP), a training strategy for segmentation networks which selectively penalizes overconfident predictions, focusing only on misclassified pixels. Our method is agnostic to the neural architecture, does not increase model complexity and can be coupled with multiple segmentation loss functions. We benchmark the proposed strategy in two challenging segmentation tasks: white matter hyperintensity lesions in magnetic resonance images (MRI) of the brain, and atrial segmentation in cardiac MRI. The experimental results demonstrate that coupling MEEP with standard segmentation losses leads to improvements not only in terms of model calibration, but also in segmentation quality.