The ULS23 Challenge: a Baseline Model and Benchmark Dataset for 3D Universal Lesion Segmentation in Computed Tomography

在癌症患者的随访CT检查中，肿瘤表现的大小测量对于评估治疗结果至关重要。高效的病灶分割可以加快这些放射学工作流程。虽然许多基准和挑战赛专门针对肝脏、肾脏和肺部等特定器官的病灶分割，但临床实践中遇到的更多种类的病灶需要更普遍的方法。为了填补这一空白，我们引入了ULS23挑战赛，用于在胸腹盆CT检查中进行3D通用病灶分割。ULS23训练数据集包含该区域的38,693个病灶，包括具有挑战性的胰腺、结肠和骨骼病变。为了评估目的，我们精选了一个数据集，包括284名患者的775个病灶。这些病灶在临床背景下被确定为目标病灶，确保了数据集的多样性和临床相关性。ULS23挑战赛可通过uls23.grand-challenge.org公开访问，使全球研究人员能够评估其分割方法的性能。此外，我们已经开发并公开发布了我们的基线半监督3D病灶分割模型。该模型在挑战测试集上的平均Dice系数为0.703±0.240。我们邀请持续提交，以推进未来ULS模型的发展。

ULS23 benchmark for 3D universal lesion segmentation in chest-abdomen-pelvis CT examinations aims to address the gap in lesion segmentation for a larger variety of lesion types encountered in clinical practice.

The paper introduces ULS23 benchmark dataset containing 38,693 lesions across the chest-abdomen-pelvis region, including challenging pancreatic, colon, and bone lesions. The paper also presents a baseline semi-supervised 3D lesion segmentation model that achieved an average Dice coefficient of 0.703 ± 0.240 on the challenge test set. The ULS23 benchmark is publicly accessible via uls23.grand-challenge.org.

The ULS23 benchmark dataset is diverse and clinically relevant, with each lesion identified as a target lesion in a clinical context. The paper provides a baseline model and invites ongoing submissions to advance the development of future ULS models. The dataset and model are publicly accessible.

Other benchmarks and challenges address lesion segmentation in specific organs like the liver, kidneys, and lungs, but the ULS23 benchmark aims to provide a more universal approach to lesion segmentation. No specific related works were mentioned in the abstract.