The RSNA Intracranial Aneurysm (RSNA-ICA) Dataset
This study addresses the high miss rate of intracranial aneurysms, particularly minute lesions, in non-angiographic imaging and the scarcity of high-quality, multimodal public datasets. To overcome these limitations, a consortium of 21 institutions across five continents constructed a large-scale multimodal dataset comprising 7,202 CTA, MRA, and MRI scans. Data quality was rigorously ensured through dual expert annotation and three-dimensional voxel-level segmentation. This pioneering dataset fills a critical gap in the field by serving as the benchmark resource for the 2025 challenge, thereby substantially enhancing the cross-device generalization performance of artificial intelligence models for intracranial aneurysm detection.
0 citationsRead paper