Recognition Of Medical Image With A Convolutional Operator Based Hidden Markov Model

In recent years,the development of medical imaging technology has made greater and greater achievements.At the same time,more and more achievements have been made in the field of computer technology.There are more and more researches on medical images using computers,and they have played a role in medical diagnosis.The role of is becoming more and more important.The use of medical image recognition in auxiliary diagnosis and treatment can become one of its key technologies.Recognizing and classifying brain tumors based on the results of the patient’s brain magnetic resonance imaging(MRI)has become an important issue in medical image processing.Computer programs can effectively analyze the patient’s brain MRI images and generate accurate tumor identification and classification results,thereby significantly reducing the time required for diagnosis,which may increase the patient’s chances of survival.This paper mainly uses the method of combining the convolution operator and the hidden Markov model(HMM)to process the medical image of the brain to obtain better recognition results.This thesis mainly includes the following research contents:(1)Develop a new image enhancement algorithm to enhance the image.According to the histogram of the image to be enhanced,the objective function is obtained.The objective function is used as an index to measure the visual contrast of the image,and then the dynamic programming method is used to optimize the objective function,and finally the best enhancement effect is obtained.Since the color image cannot be processed directly in the RGB model,it is necessary to convert the RGB model of the image to the HIS model,process the I component,and finally convert it to the RGB model to obtain the effect of color image enhancement.After experimenting with some gray-scale images and color images,it can be seen that the method in this paper is better than other methods.In the final comprehensive evaluation,after the attenuated image is enhanced,the PSNR score of this method is higher,and in terms of image quality,the number of pictures using this method is better than other methods.The calculation efficiency is also higher.(2)Chapter 4 proposes to combine convolution operator and hidden Markov model for brain medical image segmentation.First,apply the convolution operator to the image to obtain the feature vector of each pixel in the image.Then,based on the feature vectors of all pixels,an HMM model is constructed for the first line of the image,and an adaptive dynamic programming method is used to process the features of all pixels in the image line by line.After the dynamic programming process is over,the labels are assigned to the pixels in the image,and the segmentation result is finally obtained.We process medical images of the brain with multiple sclerosis lesions with different parameters,including white matter,gray matter and CSF in three different regions.The accuracy is evaluated by calculating sensitivity and specificity.Compared with the image segmentation result using HMM,the final segmentation result is significantly improved.(3)Design a set of convolution operators to process each pixel and its neighborhood in the image to obtain the characteristics of the pixels in the tumor area.The feature vector of the pixel is constructed according to the result generated by the convolution operator.The classification of tumors can be regarded as a process of labeling.Pixels from different types of tumor regions receive different labels.The labeling process is modeled by HMM,where the state is the label of the pixel.It is mainly achieved through two processes of training and testing.After the training data is input in the training process,the HMM finally outputs the training parameters of the HMM.After entering the test process,input the MRI image,use the Viterbi algorithm to assign the state to each pixel in the image through the HMM,and then calculate the tumor area HMM to assign each pixel The number of pixels in each state is searched for the state of the maximum number of pixels allocated to the tumor area,and the result of the recognition and classification is finally obtained.After experimenting with images of different tumor types in the benchmark data set,it is concluded that the recognition and classification accuracy of the method used in this article is higher than some of the latest classification methods.And in terms of calculation time,it is shorter and more efficient than other methods.In recent years,the performance of computers has been greatly improved,and artificial intelligence has developed rapidly,pattern recognition technology has gradually matured,and computer vision technology has developed rapidly,prompting analysis and auxiliary detection to be realized through computer technology.The computer-aided diagnosis system provides powerful support for the detection and diagnosis of diseases.The computer can further process the medical image,assist the doctor to analyze the image and judge the result.The use of computers to assist doctors in medically assisted diagnosis can reduce the workload of doctors.When the medical image is a gray-scale image,the human eye has certain restrictions on the gray-scale.The computer can overcome this difficulty,and the doctor’s efficiency has been improved;in addition,the medical image can be judged objectively,making the diagnosis more efficient.Therefore,researching medical images and using computers to assist doctors in auxiliary diagnosis has important research significance and very important research value.