IntrA

IntrA 是一个开放获取的 3D 颅内动脉瘤数据集，可应用基于点和基于网格的分类和分割模型。该数据集可用于诊断颅内动脉瘤并提取颈部以进行医学和其他深度学习领域的剪裁操作，例如正常估计和表面重建。通过重建扫描的患者 2D MRA 图像收集了 103 个整个脑血管的 3D 模型（由于医学伦理，未发布原始 2D MRA 图像）。从完整的模型中自动生成1909个血管片段，其中1694个健康血管片段和215个动脉瘤片段用于诊断。 116个动脉瘤段由医学专家手动分割和注释；每个动脉瘤节段的尺度是基于术前检查的需要。为每个带注释的 3D 片段计算并包含测地距离矩阵，因为根据血管的形状，测地距离的表达比欧几里得距离更准确。IntrA is an open-access 3D intracranial aneurysm dataset designed for point-based and mesh-based classification and segmentation models. This dataset can be used for intracranial aneurysm diagnosis and aneurysm neck extraction for cropping operations in medical and other deep learning fields, such as normal estimation and surface reconstruction. A total of 103 3D models of the entire cerebral vasculature were collected by reconstructing scanned patient 2D MRA images (the original 2D MRA images were not released due to medical ethical constraints). 1909 vascular segments were automatically generated from the complete models, including 1694 healthy vascular segments and 215 aneurysm segments for diagnostic purposes. 116 of the aneurysm segments were manually segmented and annotated by medical experts; the sizing of each annotated aneurysm segment was based on the requirements of preoperative examinations. Geodesic distance matrices were calculated and included for each annotated 3D segment, as geodesic distance is a more accurate representation than Euclidean distance given the shape of blood vessels.

搜集汇总

背景与挑战

背景概述

IntrA是一个开放获取的3D颅内动脉瘤数据集，专为医学诊断和深度学习任务设计，包含103个脑血管3D模型和1909个血管片段，其中215个为动脉瘤片段，116个由专家手动注释，并附带测地距离矩阵以提高形状分析的准确性。该数据集支持基于点和网格的分类与分割模型，适用于颅内动脉瘤诊断、颈部提取以及正常估计和表面重建等应用。