摘要
由于设计医用3D打印外固定支具时,通常要求医生具有丰富的临床经验和一定的CAD专业知识。针对该问题,提出一种基于MeshCNN(mesh convolutional neural network)的外固定支具网格模型分割方法,用于外固定支具构型的自动生成,在MeshCNN网络结构中添加1×1网格卷积以提升准确率。实验结果表明,基于实际收集的人体手腕部位三维网格模型数据集,使用MeshCNN框架训练出的模型,能够实现三维网格模型的分割,从而自动生成腕关节外固定支具的构型。
Designing medical 3D-printed external fixators usually requires doctors to manually model the 3D data of a patient’s limb using specialized CAD software,demanding significant clinical and CAD expertise.To address this,a mesh model segmentation method based on MeshCNN is proposed for the automatic generation of external fixation configurations.By incorporating 1x1 mesh convolution into the MeshCNN network structure,accuracy is improved.Experimental results on 3D mesh model data of human wrist parts demonstrate that the MeshCNN framework can effectively segment 3D mesh models,enabling automatic configuration generation for wrist external fixators.
作者
马赞飞
石志良
肖朝洋
MA Zanfei;SHI Zhiliang;XIAO Chaoyang(Mechanical and Electronic Engineering,Wuhan University of Technology,Wuhan Hubei 430070,China)
