White Matter Hyperintensities Segmentation Based On Convolution Neural Network And Quanlity Analysis

Segmentation and subsequent quantitative assessment of lesions in medical images provides valuable information for neuropathological analysis,and is of great significance for the planning of treatment strategies,the detection of disease progression,and the prognosis of patients.White matter hyperintensities(WMH)are brain regions that exhibit higher than normal tissue levels on T2 weighted magnetic resonance imaging(MRI).The spatial distribution and volume of white matter high signal are closely related to small vessel disease.And there is a significant correlation with the decline in cognitive function caused by neurodegenerative diseases.A number of studies have shown that the appearance of white matter high signal may even be one of the important disease markers of Alzheimer’s disease(AD).The position,size and shape of white matter high signal appearing in the brain region are uncertain.Therefore,it is very difficult to design an effective segmentation model.In clinical practice,the accurate white matter high signal segmentation is mainly manually edited by experts.The method is cumbersome,time consuming,and introduces observer variability.In addition,in order to determine whether a specific area is part of a lesion,multiple brain image modal sequences with different contrasts need to be considered,and the knowledge and level of experts are important factors affecting the accuracy of segmentation,so the development of accurate automatic segmentation algorithm has become a medical An important research endpoint for image calculation.Based on this,this paper proposes to use the multi-modal magnetic resonance image to achieve accurate and reliable white matter high signal segmentation based on the improved neural network U-net network model.In routine clinical practice,high white matter signals are often found in the brain regions of Alzheimer’s disease,neurological regression,and healthy older people.The current visual rating scale for assessing white matter high signals is the Fazekas scale.Fazekas can only approximate the volume and number of white matter high signals in the brain region,and cannot accurately provide accurate information on the spatial location of white matter.Because it can be used to obtain a better correlation between white matter high signal and specific symptoms,or better define the pattern associated with normal pathological aging.Therefore,based on the brain segmentation based on T1-weighted structural magnetic resonance images,the T2 weighted and T1-weighted images are registered to realize the accurate localization of the white space position of white matter high signal,and the longitudinal multi-data analysis is used to quantify the white matter high signal.Pathological analysis.In this paper,the significant segmentation accuracy is improved by comparing with the existing network segmentation.The feasibility of the established network model is verified.The relationship between age and white matter damage,white matter damage and cognitive function are quantitatively analyzed.The relationship between the two sides of the deep white matter,bilateral deep white matter,bilateral subcortical white matter,right suboccipital white matter and other areas have a significant correlation.