基于局部区域对应点匹配的MRI图像层间插值方法  

An interlayer interpolation method for MRI images based on matching of corresponding points in local regions

作　　者：李静媛 周孟葳 熊华晖[1,2] 姬长金[1,2] 黄晓清 樊婷婷[1,2] 黄亚奇 LI Jingyuan;ZHOU Mengwei;XIONG Huahui;JI Changjin;HUANG Xiaoqing;FAN Tingting;HUANG Yaqi(School of Biomedical Engineering,Capital Medical University,Beijing 100069;Beijing Key Laboratory of Fundamental Research on Biomechanics in Clinical Application,Capital Medical University,Beijing 100069)

机构地区：[1]首都医科大学生物医学工程学院,北京100069 [2]首都医科大学临床生物力学应用基础研究北京市重点实验室,北京100069

出　　处：《北京生物医学工程》2024年第2期158-164,共7页Beijing Biomedical Engineering

基　　金：国家自然科学基金(34670959、81171422)资助。

摘　　要：目的医学影像层间插值对提高图像分辨率十分重要。基于灰度插值的传统算法在处理结构复杂,层间灰度变化大的区域时常会造成组织细节丢失或边缘模糊。本研究旨在建立一种局部区域对应点匹配插值算法,从而有效改善插值图像质量。方法构建一种局部区域对应点匹配算法。通过将线性插值图像与同位置原始扫描图像之间的对比构建误差图像并确定出高误差区域。对非高误差区域使用线性插值。对于各高误差区分别采用对应点匹配方法处理:首先根据每个区域的局部信息确定其对应于点对匹配度误差函数最小误差的最优参数组,然后为区域内每个插值点找到两侧图像上的最佳匹配点对用于该插值点灰度计算。将此方法应用于轴、矢、冠状位方向系列扫描图像的层间插值,将插值图像与同位置原始扫描图像的均方差作为评价插值质量的标准。结果与其他算法插值结果相比,局部区域对应点匹配算法可显著降低插值误差,矢状位图像插值结果的改善尤为突出。与整体对应点匹配算法插值结果相比,其误差平均值降低达19%,有效改善了原高误差区域的插值质量。结论本研究建立的局部区域对应点匹配插值算法可以有效地解决以往算法在处理复杂组织边界区域插值时的不足,显著提高图像层间插值的精度和质量。Objective Interlayer interpolation in medical images is very important for improving image resolution.Some traditional algorithms based on grayscale interpolation often lead to the loss of tissue details or blurred edges when dealing with regions with complex structures and large grayscale variations between layers.The aim of this study is to establish a local region corresponding point matching interpolation algorithm,so as to effectively improve the quality of the interpolated images.Methods In this study,a local region corresponding point matching algorithm is constructed.The error image is constructed by comparing the image obtained by linear interpolation with the original scanned image at the same location,and high error regions are identified.Linear interpolation is used for the non-high error regions.The corresponding point matching method is used for the high error regions.It first determines the best parameters with the minimum error of the matching error function of the point pairs used for interpolation based on the local information of the region,and then finds the best matching point pair on both sides of the image for each interpolated point in the region for the grayscale calculation of the interpolated points.This method is applied to the interlayer interpolation of axial,sagittal and coronal scanning images,and the mean-square error between the interpolated image and the original scanning image is used as the criterion to evaluate the quality of the interpolation.Results Compared with the interpolation results of other algorithms,the local region corresponding point matching algorithm can significantly reduce the interpolation error,especially in sagittal image interpolation.Compared with the results of the overall correspondence point matching algorithm,the average error can be reduced by 19%,indicating that the quality of the interpolated image in the original high error region is effectively improved.Conclusions The local region corresponding point matching interpolation algorithm established

关 键 词：磁共振影像 医学图像层间插值 对应点匹配插值 局部区域方法 

分 类 号：R318.04[医药卫生—生物医学工程]