Boundary loss for highly unbalanced segmentation

TL;DR

This paper introduces a boundary loss function for CNN segmentation that improves training stability and performance in highly unbalanced segmentation tasks by focusing on contour interfaces rather than regions.

Contribution

The paper proposes a novel boundary loss based on contour distance metrics, effectively handling class imbalance and complementing existing regional loss functions.

Findings

Significant performance improvements on unbalanced segmentation tasks.

Enhanced training stability with the boundary loss.

Easy integration with existing deep learning architectures.

Abstract

Widely used loss functions for CNN segmentation, e.g., Dice or cross-entropy, are based on integrals over the segmentation regions. Unfortunately, for highly unbalanced segmentations, such regional summations have values that differ by several orders of magnitude across classes, which affects training performance and stability. We propose a boundary loss, which takes the form of a distance metric on the space of contours, not regions. This can mitigate the difficulties of highly unbalanced problems because it uses integrals over the interface between regions instead of unbalanced integrals over the regions. Furthermore, a boundary loss complements regional information. Inspired by graph-based optimization techniques for computing active-contour flows, we express a non-symmetric $L_2$ distance on the space of contours as a regional integral, which avoids completely local differential computations involving contour points. This yields a boundary loss expressed with the regional softmax probability outputs of the network, which can be easily combined with standard regional losses and implemented with any existing deep network architecture for N-D segmentation. We report comprehensive evaluations and comparisons on different unbalanced problems, showing that our boundary loss can yield significant increases in performances while improving training stability. Our code is publicly available: https://github.com/LIVIAETS/surface-loss .