Feature Recognition Method For MRI Images Of Degenerative Diseases

With the rapid development of MRI technology,neuroimaging has made great contributions to the diagnosis of degenerative neurological diseases.Computer technology used in MRI image analysis and processing has been gradually developed and popularized.The deep learning method can provide an automatic feature recognition technology for the imaging diagnosis of common multiple degenerative diseases.In this paper,two feature recognition networks based on deep learning methods are constructed for degenerative diseases in the brain and lumbar spine.The research contents are as follows:(1)For the brain MRI image,a convolutional neural network is conducted for classifying each pixel.Firstly,three sizes of inputs are set.Layers in the network need to be slightly adjusted for each input size.Finally,the optimal input size is selected according to the experimental result.This method can segment the brain tissue,and depict the edge and shape of each tissue.According to the shape of each tissue,it can also assist the doctor in diagnosing whether the brain is degraded.(2)For the lumbar MRI image,A convolutional neural network is constructed for learning lumbar intervertebral similarity.The network needs to be provided with a lumbar vertebra as a labeled image.The rest lumbar images acting as search images,are compared with the labeled image to find L1-L5.Since that bounding boxes of lumbar vertebras are horizontal,but some vertebras have a rotation angle,in the improved experiment,five rotation angles are added to the lumbar search image.As a result,the bounding box can better fit the lumbar shape.Based on the final in rotational angle between adjacent lumbar vertebrae,it can be determined whether the sample has lumbar degenerative disease and an abnormal lumbar position.The above two networks can detect the edge and shape of the brain in the MRI image,as well as the position and type of the lumbar spine.The experimental results show that the recognition accuracy is satisfying.Based on the detected characteristics,the doctor can be assisted to judge more quickly whether the tissue is degraded.