中文摘要：

针对脑梗CT图像自动判读问题，文中提出一种基于随机Hough变换和BP网络的脑部异常检测算法。首先通过随机Hough变换，利用不同脑组织CT图像的灰度信息，检测圆形或类圆形区域，初步判断可能异常的区域，利用RHT变换求出该区域中心位置及半径；然后采集可能异常部位的灰度特征信息，利用训练过的BP网络进行比对分析，排除正常区域，确定脑部异常部位。试验证明，该方法能够判读出脑部异常部位，准确率较高，可以为医生判读脑部CT图像提供决策依据。

英文摘要：

Aiming at the problem of infarction CT image automatic interpretation,present a method based on random Hough transform and BP network brain anomaly detection algorithm. First,through the random Hough transform,the use of different brain CT image gray infur- mation,detect a round or oval area,preliminarily judge possibly abnormal regions,using RHT for the region center and radius;Then the acquisition may be the gray feature information of abnormal parts,using the tralned BP network to compare,exclude normal region,define brain abnormalities site. The test shows that this method can read out of the brain abnormalities site,has high accuracy,can provide physician interpretation of brain CT images with basis for decision making.