Research On The Identification Of Biological Tissue Degeneration Based On Multi-iterative VMD And Ultrasound Signal Features

High-intensity focused ultrasound(HIFU)is used to treat cancer tumors,and the treatment area is located under the skin and cannot be observed directly with the naked eye.Research on the monitoring of HIFU treatment process is great significance to ensure the safety and efficiency of treatment.This article sets up a HIFU experimental system to explore the detection of the target area of surgery by analyzing the ultrasonic echo signals scattered by biological tissues.The main work of this paper is:1.The ultrasonic signals collected by fiber optic hydrophones contain noise.This paper proposes to use multi-iteration variational modal decomposition(MIVMD)to denoise the ultrasonic backscatter signals generated by biological tissues.However,the ultrasonic backscattered signal of biological tissue in the collected ultrasonic echo signal only occupies a small segment.Denoising directly will get undesirable results,so it is necessary to intercept the ultrasonic echo signal of biological tissue from the collected signal according to the biological tissue slice,the sampling frequency of the signal and the signal frequency.In this paper,the differences between empirical modal decomposition(EMD),variational modal decomposition(VMD),and MIVMD are analyzed by using simulation signals.In this paper,the differences between MIVMDand EMD and VMD are compared using simulated signals.The results show that the MIVMD proposed in this paper can better remove the noise in the signal.2.The compounding and multi-scale improvement of dispersion entropy(DE)was improved,and compound multiscale dispersion entropy(CMDE)was proposed as a feature of ultrasonic backscatter signals.In this paper,simulation signals are used to analyze the stability difference between the two improved methods of multi-scale and composite multi-scale,and the influence of signal length on the stability of the two improved methods is also analyzed.After calculating the entropy of the actual ultrasonic backscattered signal after MIVMD processing,GK fuzzy clustering and K-means clustering is used to cluster the extracted ultrasonic backscattered signal features.The approximate entropy,sample entropy and fuzzy entropy are compared with the CMDE,and the evaluation index of clustering with different entropy parameters and the recognition rate of degeneration of biological tissues are calculated.The results show that the CMDE of ultrasonic backscattered signals in degeneration and non-degeneration tissues is significantly different,and the degree of feature aggregation is better.Using CMDE and GK fuzzy clustering can get better recognition results.3.This paper proposes the use of autoregressive models and support vector machines(AR-SVM)to identify degeneration of biological tissues.The AR model was constructed for the ultrasonic backscattered signal after MIVMD processing.The optimal AR model order of ultrasonic backscattered signal and the selection of coefficients as recognition features were studied,and the characteristic parameters obtained were identified by support vector machine.The experimental results show that when the 11 th coefficient of the 20 th order AR model is used as the characteristic parameter,the AR-SVM method can effectively recognize the degeneration of biological tissues,and the recognition rate reaches 93%.