Segmentation Loss Odyssey

TL;DR

This paper provides a systematic taxonomy of loss functions for medical image segmentation, revealing their relationships and categorizing them into four meaningful groups, supported by publicly available PyTorch implementations.

Contribution

It introduces a comprehensive taxonomy of segmentation loss functions and explores their fundamental links, aiding better understanding and selection in medical image segmentation.

Findings

Loss functions categorized into four groups

Revealed links between region-based and boundary-based losses

Provided publicly available PyTorch implementations

Abstract

Loss functions are one of the crucial ingredients in deep learning-based medical image segmentation methods. Many loss functions have been proposed in existing literature, but are studied separately or only investigated with few other losses. In this paper, we present a systematic taxonomy to sort existing loss functions into four meaningful categories. This helps to reveal links and fundamental similarities between them. Moreover, we explore the relationship between the traditional region-based and the more recent boundary-based loss functions. The PyTorch implementations of these loss functions are publicly available at \url{https://github.com/JunMa11/SegLoss}.