基于超限学习机2.5D-3D配准的脊椎手术导航

发布时间：2018-06-08 11:23

本文选题：脊椎手术 + 手术导航　；参考：《计算机仿真》2017年11期

【摘要】：基于视觉的自动导航是自动手术系统和手术机器人的关键组成部分。手术前基于CT扫描重建病人组织器官的三维模型,术中采集对应器官的二维影像,使其与三维重建模型进行实时匹配和对准,即实现虚-实配准,即可实现计算机视觉辅助的手术导航。面向脊椎这一刚性组织,研究观察影像与三维重建模型的自动刚性配准。为实现快速、鲁棒的特征提取和匹配,提出使用超限学习机构建的深度学习模型,基于大量训练数据学习2.5D深度图像和3D模型关键点的特征提取,并进行三维配准。人体脊椎数据库的配准实验和猪脊椎模拟手术导航的结果表明,上述方法具有训练时间短、匹配精度高、运行速度快等特点,非常适合于脊椎手术导航。
[Abstract]:Vision-based automatic navigation is a key component of automatic surgery system and surgical robot. Before operation, 3D models of tissue and organs of patients were reconstructed based on CT scanning, and 2D images of corresponding organs were collected during operation, which were matched and aligned with 3D reconstruction models in real time, that is, virtual and real registration was realized. It can realize the operation navigation aided by computer vision. For the rigid tissue of spine, the automatic rigid registration of image and three-dimensional reconstruction model was studied. In order to achieve fast and robust feature extraction and matching, a depth learning model based on over-limit learning machine is proposed. Based on a large amount of training data, the feature extraction of 2.5D depth image and the key points of 3D model is learned, and 3D registration is carried out. The results of the registration experiment of human spine database and the simulated operation navigation of pig spine show that the above methods have the advantages of short training time, high matching accuracy and fast running speed, so they are very suitable for the navigation of spinal surgery.
【作者单位】：中南大学湘雅二医院;国防科技大学计算机学院;
【基金】：国家自然科学基金项目(61540065,61572507)
【分类号】：R687.3;TP391.41

【相似文献】