Incremental Learning for Multi-organ Segmentation with Partially Labeled Datasets

TL;DR

This paper introduces an incremental learning approach for multi-organ segmentation that effectively utilizes sequentially-constructed, partially labeled datasets, addressing catastrophic forgetting and improving model adaptability in medical imaging.

Contribution

It is the first to propose an incremental learning framework for multi-organ segmentation with partially labeled datasets, emphasizing the role of distribution differences in catastrophic forgetting.

Findings

The method effectively mitigates catastrophic forgetting in multi-organ segmentation.

Distribution differences are identified as a key factor in catastrophic forgetting.

Experiments on five datasets demonstrate significant performance improvements.

Abstract

There exists a large number of datasets for organ segmentation, which are partially annotated, and sequentially constructed. A typical dataset is constructed at a certain time by curating medical images and annotating the organs of interest. In other words, new datasets with annotations of new organ categories are built over time. To unleash the potential behind these partially labeled, sequentially-constructed datasets, we propose to learn a multi-organ segmentation model through incremental learning (IL). In each IL stage, we lose access to the previous annotations, whose knowledge is assumingly captured by the current model, and gain the access to a new dataset with annotations of new organ categories, from which we learn to update the organ segmentation model to include the new organs. We give the first attempt to conjecture that the different distribution is the key reason for 'catastrophic forgetting' that commonly exists in IL methods, and verify that IL has the natural adaptability to medical image scenarios. Extensive experiments on five open-sourced datasets are conducted to prove the effectiveness of our method and the conjecture mentioned above.