Classification Of Common Carotid Arterial Plaques Based On Ultrasonic Characteristic Parameters Of The Gamma Hybrid Model

In the course of evolution in recent years,carotid atherosclerosis has gradually evolved into a significant health hazard.Atherosclerosis has a low cure rate and a high recurrence rate,and serious cardiovascular diseases such as stroke and myocardial infarction are often life-threatening.Therefore,early identification and diagnosis of atherosclerotic diseases is of great significance to prevent the disease from worsening.At present,there are many methods for detecting atherosclerosis,among which Ultrasound(US)technology is the most widely used.Tissue microstructures have features such as structure and texture.These structured textures appear as speckle noises with different distributions in the US image,and have similar distribution characteristics.These speckles can be extracted by clustering.In ultrasound imaging,the presence of differently distributed scattering within the resolution unit results in a mixed echo response.The commonly used probability distribution model is a single parameter distribution,which can only be used to simply describe the original signal envelope.In these cases,the mixed probability model becomes a more suitable method for statistically describing the characteristics of speckle clustering.This study is based on the classification of carotid plaques based on the characteristic parameters of the Gamma hybrid model of ultrasound images.First,the interest frames are selected for the collected ultrasound images,and the selected interest frames are clustered using the Gamma hybrid model.The obtained parameters form the feature vector,and then the ten-fold cross-validation support vector machine is used based on the four support vector machines.The kernel functions(linear kernel,polynomial kernel,Sigmoid kernel and Gauss radial basis kernel)are used to train the feature vectors,and then the average accuracy of the classification is obtained.In order to compare with other processing methods,this study also uses Rayleigh mixing.The model classifies the same image area of interest,and then uses the same method to study the accuracy of the Rayleigh hybrid model clustering,and compares it with the Gamma hybrid model.In order to verify the effectiveness of this research method,40 atherosclerosis samples were collected from the clinic as the research object,and each frame was taken as 40 frames.Experts are required to use many years of clinical experience to select 40 frames from the 40 samples.Select 5 interest areas(including soft spots,hard spots,mixed spots,and other tissues)in the frame,and perform clustering processing according to the above test method to obtain three parameters:expectations,variance,and the weight of the interest area occupied by the classification(Switzerland).The profit mixture model is variance and weight.)Each type of parameter set is made into a feature vector,and four kernel functions are used to cross-verify it to obtain the clustering accuracy of the two mixture models.The results show that the Gamma hybrid model has a good classification effect on plaque classification when clustering arterial plaques in ultrasound images.Except for the polynomial kernel function,the accuracy of the other three kernel functions is between 70% and 90%,of which Gauss Radial Basis Kernel Support Vector Machine training has the highest accuracy rate,reaching more than 80%,while the Rayleigh hybrid model clustering is not as effective as the Gamma hybrid model.It can be concluded from the above experiments that plaque classification using Gamma mixed model parameter estimation can obtain more accurate results and can be used for clinical computer-aided diagnosis.