Learning from Partially Overlapping Labels: Image Segmentation under Annotation Shift

TL;DR

This paper addresses the challenge of training accurate image segmentation models with limited annotated data by proposing strategies to effectively learn from datasets with partially overlapping labels, improving performance in abdominal organ segmentation.

Contribution

It introduces a semi-supervised approach combined with an adaptive cross entropy loss to leverage heterogeneously annotated datasets for better segmentation accuracy.

Findings

Combining datasets with partially overlapping labels improves segmentation performance.

The proposed method outperforms baseline and alternative approaches.

Adaptive cross entropy effectively handles annotation heterogeneity.

Abstract

Scarcity of high quality annotated images remains a limiting factor for training accurate image segmentation models. While more and more annotated datasets become publicly available, the number of samples in each individual database is often small. Combining different databases to create larger amounts of training data is appealing yet challenging due to the heterogeneity as a result of differences in data acquisition and annotation processes, often yielding incompatible or even conflicting information. In this paper, we investigate and propose several strategies for learning from partially overlapping labels in the context of abdominal organ segmentation. We find that combining a semi-supervised approach with an adaptive cross entropy loss can successfully exploit heterogeneously annotated data and substantially improve segmentation accuracy compared to baseline and alternative approaches.