Segmentation And Classification Of Hippocampus Based On MRI

The hippocampus,located in the medial temporal lobe of the brain,and with amygdala constitute the limbic system of brain.It is a bridge between the human body and the brain control center,which controls human emotions and is related to a variety of neurological diseases.At present,the aging of the population has been widespread all over the world,the number of elderly population is increasing,and its health problems have also aroused widespread concern in the society.AD is mainly concentrated in the elderly population,and the incidence rate is increasing with age,which seriously threatens the health of the elderly.It is found that the morphological characteristics of the hippocampus are closely related to the incidence of Alzheimer’s disease.If the hippocampus can be accurately segmented from the brain tissue,it will be better to study the disease.Because MRI has the characteristics of high resolution and good imaging effect on the soft tissues including brain,it is usually used as an important data in the study of hippocampus.The hippocampus has the characteristics of small size and irregular shape.The traditional segmentation method is prone to the wrong segmentation,and the segmentation speed is slow,which consumes a lot of time.Therefore,in the process of hippocampal segmentation,we first finished image preprocessing according to the characteristics of the data,then optimized the traditional u-net algorithm,and use the improved algorithm to the preprocessed data to achieve segmentation of hippocampal,and finally use classification method to complete the classification of Alzheimer’s disease according to the characteristics of the segmented image.The specific research work is as follows:1.Data preprocessing.Using the adaptive histogram equalization algorithm of limited contrast(CLAHE)to enhance the contrast of the image data,making the image more natural;using the image denoising method driven by curvature to remove the interference factors in the image data,and clearly display the information of each organization;using the image data enhancement technology,translated,rotated,distorted and other operations to expand the sample data to meet the model training practice requirements.2.Image segmentation.The u-net algorithm model is used to segment the hippocampus of the preprocessed image data.In the algorithm model,the batch normalization(BN)is introduced to improve the problem of slow training speed and gradient disappearance.At the same time,the residual module is introduced after the convolution layer to solve the problem of performance degradation after the network is deepened.The Adam optimization algorithm is used to solve the problems such as the loss of learning rate,slow convergence speed of the network and large change of loss function amplitude.Compared with the original network,the improved u-net algorithm improves the segmentation accuracy and algorithm performance.3.Image classification.SVM-REF feature selection method and SVM classification method were used to classify Alzheimer’s disease.SVM-REF feature selection method is used to select the features of the segmented hippocampus,and then SVM classification method is used to classify the symptoms of Alzheimer’s patients according to the selected features.