Study On The Detection Of Tissue Lesion During HIFU Therapy Based On Ultrasonic Image Features

High intensity focused ultrasound(HIFU)is a new noninvasive treatment technique with great potential,which is suitable for the treatment of malignant and benign tumors of tissues and organs.High focused ultrasound can make the temperature of the tumor target instantly reach 65?100 ?,at the same time combined ultrasonic cavitation effect,thermal effect and other comprehensive effects,which can lead to irreversible lesion of target tissue,but no obvious lesion to the surrounding tissue,and the lesion tissue can be gradually absorbed or scarring.In order to monitor and evaluate the therapeutic effect of HIFU in time during the HIFU treatment,it is necessary to detect the tissue lesion.This paper was based on the experiment of fresh pork tissue irradiated by fo-cused ultrasonic transducer in vitro.The ultrasonic images before and after HIFU radiation were collected by means of B-mode ultrasonic scanner and the computer configured with image acquisition card.Then the characteristic parameters were extracted from the pretreated ultrasonic images.Finally,the tissue lesion was detected.Main work was:Firstly,the principle of treatment,the mechanism and the research back-ground and situation at home and abroad of HIFU were elaborated.Then the feasibility of using ultrasonic image feature parameters to detect tissue lesion was put forward.A large number of experiments were carried out with the image acquisition experimental system.The preprocessing methods of ultrasonic image were studied,including filtering,color space transformation and image difference processing.Secondly,the pretreated ultrasonic images were analyzed with different meth-ods.4 feature parameters such as wavelet fractal dimension,information entropy,conditional entropy and correlation coefficient were extracted.Detection to the tis-sue lesion condition and the corresponding tissue lesion threshold interval of each parameter were studied with support vector machine(SVM).The results showed that to a certain extent each characteristic parameter could identify tissue lesion or not,and the identification effect of it was different.Thirdly,the combination of linear discriminant analysis(LDA)and ultrasonic image feature parameters was proposed to detect tissue lesion.Parameters from 2 characteristic parameters processed by LDA were input to SVM.Then the test samples were detected after training.Compared to 2 feature parameters directly combined and input SVM,the total identification rate of tissue lesion was signifi-cantly improved.In this paper,the method of detecting tissue lesion during HIFU treatment was studied by using the ultrasonic image feature parameters and SVM,and a new idea was proposed for the monitoring of HIFU treatment process.