Study On The Detection And Recognition Of HIFU Lesions Based On Ultrasonic Image

High Intensity Focused Ultrasound(HIFU)is a non-invasive local therapy technology which has been developed in recent 20 years,which is mainly used to treat benign and malignant tumors.The principle is to focus ultrasound from outside to internal body,and form local high energy in the focal area,so that the tissue instantaneously reaches the high temperature above 65°C,leading to instantaneously coagulative necrosis of the tissue,so as to achieve the goal of non-invasive treatment of tumors.Because of its thermal mechanism,so it is very important to achieve real-time monitoring of tissue lesions in the course of treatment.In this paper,we adopt the fresh pork tested in vitro as the experimental object.By B-mode ultrasound,we obtain the Ultrasound images after HIFU treatment.The ultrasound images were then de-noised,gray-scale processed and the region of interest was intercepted.Based on the characteristics of the region of interest,tissue lesions was identified and detected.The main tasks are listed as follows:(1)A HIFU lesions detection and recognition method based on pixel search algorithm is proposed which combines rough localization and HIFU lesions identification.Firstly,in rough localization,all the bright spots in the image are located by using the search algorithm of the scale of maximum gray value.Secondly,the edge contour of the test object is obtained by using mathematical morphology and Canny edge detection algorithm,and the candidate areas of HIFU lesions are obtained by European criterion.Lastly,rectangular shape of HIFU lesions candidate regions are extracted and identified by SVM.Experimental results show,The method realizes automatic detection of HIFU lesion regions,and therectangula shape is selected as the characteristic parameter with higher recognition rate than the maximum gray value.(2)A HIFU lesions detection and recognition method of HIFU lesions based on differential block search algorithm is proposed,divided into three parts: rough localization,HIFU lesions identification and lesions segmentation.Firstly,the image is partitioned into blocks by rough localization.Then,the sliding window is used to detect the differential blocks,and the connected differential blocks are classified into connected regions.the edge noise is screened out by the gray information around the connected regions,and the candidate regions for HIFU lesions are obtained.Secondly,the rectangular shape training SVM is used to identify the candidate areas of HIFU lesions.Lastly,the determined HIFU lesion regions is segmented and edge detected to obtain the actual area.The study found that the differential block algorithm of this method has a better ability to describe the characteristics of HIFU lesion regions,and eliminates many intermediate steps and reduces the amount of computation compared to pixel point search algorithms.In this paper,the ultrasound image after HIFU treatment is studied,which reduces the steps of image registration,and realizes the automatic detection and recognition of HIFU lesion region,which provides a new direction for real-time monitoring during HIFU treatment.