从非增强型CT确定脑内急性脑出血

目的快速准确地从非增强型CT图像中检测急性脑出血。方法基于灰度阈值与数学形态学获取脑组织图像；根据图像切片中有无足够的非奇异点而自适应地选择原始图像、脑组织或头颅组织图像来计算中矢状面；基于急性脑出血的高灰度、封中矢状面的较高的灰度不对称性以及比周围脑组织辉度高的特性确定急性可能的急性脑出血区域，去除可能的钙化和线状伪影。并补偿部分容积效应。结果对来自日本及中国的55套出血和15套非出血的临床图像能自动判断是否有急性脑出血，脑出血的检测在视觉上令人满意，在普通PC机上的处理时间小于20秒。结论本算法能快速准确地检测急性脑出血。检测系统可用于临床急性脑出血病人的诊治。

Fast delineation of intracranial hemorrhage from non-enhanced head CT images

Objective This study aims to fast and accurately delineate acute intracranial hemorrhage （AlH） from non-enhanced head CT images. Methods An algorithm is proposed which consists of the following steps： skull peeling through intensity thresholding and mathematical morphology; adaptive selection of the original image, brain structures or head structures for calculating the midsagittal plane （MSP） according to the number of detected inlier symmetrical points on axial slices; determination of AIH candidate regions based on AlH＇s high intensities, high intensity asymmetry with respect to the MSP, and being brighter than their surrounding brain tissues; removal of calcification and linear streaks, and compensation of partial volume effects. Results The algorithm has been validated against 70 clinical non-enhanced CT data sets from multi-centers in China and Japan （55 AIH and 15 non-AIH）. It can differentiate all the AIH from non-AIH, and the delineation of AIH is visually checked with success. Conclusion The proposed algorithm is fast （less than 20 seconds on a PC with 2.21 GHz CPU）, automatic and accurate, and could be employed for the diagnosis and curing of AIH patients.