Multicenter Acute Ischemic Stroke, MRI and Clinical Text Dataset

Study on the topological feature of infarct lesion and corresponding brain function impairment is of clinical and scientific significance for cerebral stroke. High-quality, large-scale imaging and the matching clinical data are essential for the research. We collected a multimodal MRI dataset of 5788 acute ischaemic stroke patients, which, to the best of our knowledge, is the largest stroke dataset that includes detailed and complete clinical textdata. The release of this dataset is instrumental in facilitating the development of artificial intelligence models aimed at diversifying analytical methodologies. These models explore a number of learnable standardised or repetitive tasks that currently rely on humans, including disease-related scores calculation such as the Aspect score, prognostic prediction, selection of treatment regimens, lesion labelling, and lesion segmentation.

脑梗死病灶的拓扑特征及其伴随的脑功能损伤研究，对于脑卒中的临床实践与科学探索均具备重要价值。高质量、大规模的影像学数据与匹配的临床资料，是此类研究开展的必要基础。本研究收集了5788例急性缺血性脑卒中患者的多模态磁共振成像（multimodal MRI）数据集。据我们所知，该数据集是目前收录详细完整临床文本数据的规模最大的脑卒中相关数据集。该数据集的公开发布，将有力推动面向多样化分析范式的人工智能模型研发进程。此类模型可覆盖多项当前依赖人工完成的可学习标准化或重复性任务，包括疾病相关评分计算（如ASPECT评分（Aspect score））、预后预测、治疗方案遴选、病灶标记与病灶分割等。