颅内电极定位中的Zernike矩快速算法研究  

Research on Fast Algorithm for Zernike Moment in Intracranial Electrode Localization

作　　者：刘思伟[1] 董硕[1] 严汉民[1] 

机构地区：[1]首都医科大学宣武医院医学工程处,北京100053

出　　处：《中国医疗设备》2015年第1期23-27,共5页China Medical Devices

基　　金：国家科技支撑计划项目资助(2011BAI02B02);国家自然科学基金项目资助(81372923)

摘　　要：目的减少颅内电极定位中多模图像Zernike矩的配准耗时,提升其配准精度。方法用几何中心法和Hough变换法对人脑CT和MRI图像进行几何参数和配准区域的计算,并根据人脑结构特点建立模型对算法有效性进行测试。本文分别应用骨架算法、颅骨提取法、腐蚀联合大津阈值法、边缘增强法、边缘增强联合腐蚀法、Bottom hat法、Sobel边缘提取法对人脑CT和MRI图像进行预处理。结果对于模型图像,Hough变换法和几何中心法测得的质心与实际质心坐标分别相差±1个像素和±5个像素,主轴误差范围均<3%。对于CT和MRI图像,几何中心法的计算结果更优,其质心点和主轴的计算误差范围分别可控制在周围5邻域内和±6%内。结论几何中心法和Hough变换法均能够很好地计算出模型图像几何参数和配准区域。对于CT和MRI图像,几何中心法的计算效果较好,Hough变换法的计算效果有待提升。Objective To reduce the registration time and improve the registration accuracy of multimodality images with Zernike moment in intracranial electrode localization. Methods The geometric center method and Hough transform method were used in calculating the geometric parameters and registration areas of CT and MRI images of human brain. The validity of the two methods was tested with simulated models which were constructed according to the structural characteristics of human brain. Various methods including skeleton algorithm, skull extraction method, erode combined with OTSU threshold value method, edge enhancement method, edge enhancement combined with erode method, Bottom hat method and Sobel edge extraction method were used in the preprocessing of CT and MRI images. Results The center-of-mass coordinates of stimulated images measured by geometric center method and Hough transform method are different from real center-of-mass coordinates within ±1 pixel and ±5 pixels, respectively. The principle axis errors of the two methods are less than 3%. The results of relevant parameters of CT and MRI images calculated by geometric center method are better that those of CT and MRI images calculated by Hough transform method. The calculation errors of center-of-mass points and principle axis of CT and MRI images calculated by geometric center method can be controlled within 5 neighborhoods and ±6%, respectively. Conclusion Both geometric center method and Hough transform method can calculate the geometric parameters and registration areas of simulated models. However, the calculation effect of geometric center method is better than that of Hough transform method for the geometric parameters and registration areas of CT and MRI images.

关 键 词：几何中心法 Hough变换法 多模图像配准 颅内电极定位 ZERNIKE矩 

分 类 号：R742.1[医药卫生—神经病学与精神病学] TH774[医药卫生—临床医学]