摘要
通过核磁共振设备获得多个离散间距的磁共振切片图像,采用CARESU_NET卷积神经网络对图像进行分割,获取胎儿大脑区域图像。采用CARESU_NET卷积神经网络对间断切片进行边缘重构,恢复完整的边缘信息。对边缘重构后的图像组提取边缘像素,生成三维点云,运用泊松重建方法重建点云表面,得到胎儿大脑三维表面模型。结果表明:基于核磁共振图像的三维表面模型直观生动,提高诊断效率和准确性。
Multiple discrete space magnetic resonance slice images are obtained using a nuclear magnetic resonance device,a CARESU_NET convolutional neural network is used to segment image to extract the fetal brain region images.A CARESU_NET convolutional neural network is used to reconstruct edge on discontinuous slices,complete edge information is restored.A three-dimensional point clouds are generated by extracting edge pixels from the edge-reconstructed images,and the point cloud surface is reconstructed using the Poisson reconstruction method to obtain a three-dimensional surface model of the fetal brain.The results show that the three-dimensional surface model based on nuclear magnetic resonance images is intuitive and vivid,the diagnostic efficiency and accuracy are improved.
作者
蔡凯雄
王强
陈添峰
郑力新
CAI Kaixiong;WANG Qiang;CHEN Tianfeng;ZHENG Lixin(College of Engineering,University of Huaqiao,Quanzhou 362021,China;Children′s Hospital,Quanzhou Maternal and Child Health Hospital,Quanzhou 362000,China)
出处
《华侨大学学报（自然科学版）》
CAS
2024年第1期78-85,共8页
Journal of Huaqiao University(Natural Science)
基金
福建省科技计划项目(2020Y0039)
福建省华侨大学院校联合创新项目(2022YX008)。
关键词
胎儿大脑
三维重建
边缘重构
点云处理
核磁共振
fetal brain
three-dimensional reconstruction
edge reconstruction
point cloud processing
nuclear magnetic resonance
作者简介
通信作者:郑力新(1967-),男,教授,博士,主要从事图像分析、机器视觉、深度学习方法、大数据分析、机器人与视觉一体化技术、网络控制、机电一体化系统等的研究。E-mail:zlx@hqu.edu.cn。
