A Data-Adaptive Loss Function for Incomplete Data and Incremental Learning in Semantic Image Segmentation

TL;DR

This paper introduces a novel data-adaptive loss function for semantic image segmentation that effectively handles incomplete data and supports incremental learning, reducing training time while maintaining performance.

Contribution

The paper proposes a new loss function that adapts to available data, enabling learning from incomplete annotations and automatically incorporating new structures in medical image segmentation.

Findings

Performs on par with baseline models in accuracy.

Reduces training time significantly.

Effectively incorporates new structures during training.

Abstract

In the last years, deep learning has dramatically improved the performances in a variety of medical image analysis applications. Among different types of deep learning models, convolutional neural networks have been among the most successful and they have been used in many applications in medical imaging. Training deep convolutional neural networks often requires large amounts of image data to generalize well to new unseen images. It is often time-consuming and expensive to collect large amounts of data in the medical image domain due to expensive imaging systems, and the need for experts to manually make ground truth annotations. A potential problem arises if new structures are added when a decision support system is already deployed and in use. Since the field of radiation therapy is constantly developing, the new structures would also have to be covered by the decision support system. In the present work, we propose a novel loss function, that adapts to the available data in order to utilize all available data, even when some have missing annotations. We demonstrate that the proposed loss function also works well in an incremental learning setting, where it can automatically incorporate new structures as they appear. Experiments on a large in-house data set show that the proposed method performs on par with baseline models, while greatly reducing the training time.