基于超声图像处理的HIFU所致组织损伤自动检测方法:实验研究

目的 探讨基于超声图像处理的HIFU所致组织损伤的自动检测方法。方法 针对HIFU辐照后新鲜离体猪肉声像图中的ROI，通过搜索灰度极大区域自动定位图像中的所有亮斑，结合数学形态学、连通域标记和Canny边缘检测算法提取测试对象的边缘轮廓；根据亮斑中心至边缘轮廓的欧式距离去除边缘附近的亮斑噪声，获取HIFU损伤候选区；而后提取候选区特征参数，并结合支持向量机（SVM）识别HIFU损伤。结果 最大灰度值和矩形度两个特征参数的识别率分别为86.25%和93.33%。选用识别率更高的矩形度，可正确识别单处、多处HIFU损伤或无HIFU损伤的图像。结论 采用此法可直接分析HIFU辐照后超声声像图而自动检测HIFU损伤，无需依靠图像配准技术，可减少手动定位带来的误差。

Automatic detection method of tissue injuries caused by HIFU based on ultrasonic image processing: An experimental study

Objective To explore an automatic method to detect tissue injuries during HIFU treatment for ultrasonic monitor. Methods Based on ROIs on the ultrasonic image of fresh pork obtained after HIFU radiation, all bright spots were located through the scale of maximum gray value search, and the edges of tested objects were extracted by combining the mathematical morphology and the connected component labeling algorithm,as well as the Canny edge detection algorithm. The bright noises near the edges were eliminated according to the Euclidean distance between bright spot center and edges, and HIFU injured candidate areas were acquired sequentially. Then, the characteristic parameters of the candidate areas were extracted, and HIFU injuries were recognized with support vector machine (SVM). Results The recognition rates of the maximum gray value and the rectangular shape was 86.25% and 93.33%, respectively. The ultrasonic images with single and multiple HIFU injuries or not were identified accurately using the rectangular shape. Conclusion The ultrasonic image obtained after HIFU radiation was analyzed directly using this method, made it unnecessary to rely on image registration technology and might reduce error caused by manual localization.