Fast and Low GPU Memory Abdominal Organ Segmentation in CT

Abdominal organ segmentation plays an important role in clinical practice, and to some extent, it seems to be a solved problem because the state-of-the-art methods have achieved inter-observer performance in several benchmark datasets. However, most of the existing abdominal datasets only contain single-center, single-phase, single-vendor, or single-disease cases, and it is unclear whether the excellent performance can be generalized on more diverse datasets. Moreover, many SOTA methods use model ensembles to boost performance, but these solutions usually have a large model size and cost extensive computational resources, which are impractical to be deployed in clinical practice.

To address these limitations, we will organize a Fast and Low GPU Memory Abdominal Organ Segmentation challenge. We present a large and diverse abdominal CT organ dataset with 500 CT scans from 11 countries, including multi-center, multi-phase, multi-vendor, and multi-disease cases. Participants are required to develop segmentation methods that can segment the liver, kidney, spleen, and pancreas simultaneously, where both accuracy and efficiency will be evaluated.

The challenge has two main features: (1) the dataset is large and diverse, and the submitted methods will be evaluated on the cases from unseen centers; (2) we not only focus on segmentation accuracy but also segmentation efficiency, which are in concordance with real clinical practice and requirements.