Microsurgical Aneurysm Clipping Surgery (MACS) Dataset with image-level aneurysm presence/absence annotations

We collected and present a microsurgical aneurysm clipping surgery (MACS) dataset comprising of 16 videos of MACS procedures with aneurysm presence/ absence annotations at frame-level (~350k frames) conducted by expert neurosurgeons. We extract frames at 5 fps and establish a dataset of 356,165 images. The MACS dataset is composed of FHD (1920x1080) surgical videos from the operative microscope of 16 patients during surgical repair of intracerebral aneurysms. The study was registered with the hospital’s (National Hospital for Neurology and Neurosurgery, UCLH NHS) local audit committee and data sharing was approved by the information governance lead. All patients provided written informed consent for their video to be collected for research. The MACS dataset was blindly reviewed by two senior vascular neurosurgeons in duplicate. Frames were classified as follows: Type-X : No aneurysm in microscope’s view, Type-Y : Aneurysm in microscope’s view (including both visible and clipped aneurysms), Type-Z: Frame excluded from analysis.

本数据集汇集并呈现了一组由资深神经外科专家执行的显微神经外科动脉瘤夹闭手术（MACS）视频，共包含16个手术过程视频，其中对帧级别的动脉瘤是否存在进行了标注（约350,000帧）。我们以每秒5帧的频率提取帧，构建了一个包含356,165幅图像的数据集。MACS数据集由16位患者在颅内动脉瘤手术修复过程中，手术显微镜所拍摄的FHD（1920x1080）高清手术视频组成。该研究已由医院（国家神经病学和神经外科医院，UCLH NHS）的本地审计委员会注册，数据共享已获得信息治理负责人的批准。所有患者均提供了书面知情同意，允许收集其视频用于研究。MACS数据集经两位资深血管神经外科专家进行了盲法双重审阅。帧的分类如下：类型-X：显微镜视野中无动脉瘤，类型-Y：显微镜视野中存在动脉瘤（包括可见和已夹闭的动脉瘤），类型-Z：帧被排除在分析之外。