torchmil/RSNA_ICH_MIL

RSNA-ICH多实例学习数据集是一个改编自RSNA Intra-Cranial Hemorrhage Detection数据集的版本，专为多实例学习(MIL)设计。数据集包含经过预处理的CT扫描特征，每个CT扫描的切片都经过特征提取。如果一个切片包含出血的证据，则该切片被标记为阳性。如果一个CT扫描中至少有一个阳性切片，那么该CT扫描也被标记为阳性。数据集提供了训练和测试数据的划分，方便进行标准化实验。

The RSNA-ICH Multiple Instance Learning (MIL) dataset is an adapted version of the RSNA Intra-Cranial Hemorrhage Detection dataset, tailored for Multiple Instance Learning (MIL). It contains preprocessed CT scan features, with each slice of a CT scan undergoing feature extraction. A slice is labeled as positive if it contains evidence of hemorrhage, and a CT scan is labeled as positive if it has at least one positive slice. The dataset provides train/test splits for standardized experimentation.