医疗红外线热影像应用于辅助临床冠状动脉血管阻塞率分析

红外线数位热影像非常适合显示受测物体曲面所发射红外线辐射之二维温度分佈，因此红外线心脏热影像又被称为冠状动脉热像血管造影。在本研究中，提出了一种数位影像处理之演算法，以定位血栓阻塞或颈缩狭窄之位置。本演算法包含四个连续之处理过程；首先，针对红外线数位热影像之二维温度灰阶像素进行二值化处理，利用多阈值适性分割来区分心脏(背景)或冠状动脉(前景)像素。接著第二处理过程将分割所得原始动脉像素，经过一连串之胡椒状像素去除(pepper-like pixels removing)、外围像素侵蚀(erosion)、与原区域相减(subtraction)、递迴式邻域探访(recursive neighborhood visiting)、动脉轮廓点串链(point linked list)建立以及短边删除等处理之后，提取出冠状动脉之候选轮廓。在第三个处理过程中，由医生自主点选一冠状动脉分支，利用形态学细化(morphological thinning)以及临域分析(neighborhood analysis)，以便计算最长之中心轴曲线。在最后一个处理过程中，在沿著有向中心轴曲线上移动，将每一个中心轴像素至最近之左右边动脉血管轮廓相加起来，成为该中心轴像素之可变血管管径，并沿此弯曲之中心轴绘製出矫直哩程管径图，以提供医生有用之狭窄处生理资讯。接著本研究定义一血管阻塞率公式，以区域阻塞率最小值来预估未来动脉阻塞最可能发生之位置。

The study of using infrared thermal image to assist surgery of coronary artery bypass

Digital infrared thermography is suitable for monitoring the planar two-dimensional temperature distribution of curved surfaces of objects by sensing their infrared radiation. Cardiac infrared thermography also has a thermal coronary angiography alias. This study proposes a digital image processing methodology for locating blood clot blockage. This methodology contains four consequent processes. The two-dimensional gray scale infrared thermograph pixels are first binarized and classified as background or coronary arteries using multi-thresh adaptive segmentation. The coronary artery contours are extracted from segmented raw pixels using continuous pepper-like pixel removal, erosion, subtraction, recursive neighborhood visiting, contour point-list construction and short edge deletion. In the third process one coronary artery branch is selected by physicians for calculating the longest curved central axis using morphological thinning and neighborhood analysis. In the last process the nearest left and right distances from each pixel along the directional central axis to its corresponding boundary contour are added as the coronary artery variable diameter at the current pixel''s position. A variable diameter versus straighten length diagram along this axial curved path is plotted to provide useful physiological information to the physician. An obstruction rate equation is then defined to calculate the possible vascular blockage positions with the local minimal rates. Finally, preoperative cases are tested to prove the predictive positions are correct in comparison to individual patient myocardial perfusion imaging.