A-Eval

A-Eval数据集是由上海交通大学和上海人工智能实验室联合创建，旨在评估腹部多器官分割模型的跨数据集泛化能力。该数据集整合了四个大型公共数据集（FLARE22, AMOS, WORD, TotalSegmentator）的训练集，以及BTCV数据集的训练集，总计包含1432个标记的CT图像和2000个未标记的CT图像，涵盖多种器官和疾病类型。数据集创建过程中，采用了多种数据处理和模型训练策略，以确保模型的泛化能力。A-Eval数据集主要用于解决模型在不同数据集上的泛化问题，特别是在医疗影像分析领域，提高模型的可靠性和准确性。

The A-Eval dataset was jointly developed by Shanghai Jiao Tong University and Shanghai AI Laboratory, with the primary goal of evaluating the cross-dataset generalization performance of abdominal multi-organ segmentation models. It integrates the training splits of four large-scale public datasets (FLARE22, AMOS, WORD, and TotalSegmentator) as well as the training set of the BTCV dataset. In total, the dataset contains 1432 labeled CT images and 2000 unlabeled CT images, covering a wide range of organs and disease categories. During the construction of A-Eval, multiple data processing and model training strategies were employed to ensure the generalization capability of the models. The A-Eval dataset is mainly designed to tackle the cross-dataset generalization problem of models, particularly in the field of medical image analysis, to enhance the reliability and accuracy of such models.