A White Matter Fiber Tracking Algorithm Based Method For Marking Disease Characteristics And Its Application

The white matter fiber tracking technology based on the Diffusion Weighted Magnetic Resonance Imaging?DW-MRI?is the only way to show the direction of nerve fibers.With the development of imaging technology,the use of white matter fiber tracking technology to explore the relationship between brain regions and various neurological diseases has become a hot spot in clinical applications.Parkinson's disease?PD?,as the second most common long-term neurodegenerative disease,is the main research object of pathological research of neurological diseases.Therefore,this paper focuses on the above two issues,based on machine learning and pathological research indicators for in-depth research and proposes corresponding solutions.The main work and results of this paper are as follows:?1?Aiming at the single problem of disease differentiation characteristics,an anatomical feature extraction method based on white matter fiber tracking algorithm was proposed.The method flow includes four steps:data preprocessing,fiber reconstruction,fiber tracking,and brain anatomical feature extraction.At the same time,the voxel-based whole brain fiber characteristics index is introduced,including the sum of all fiber lengths of each voxel?_vand its length inverse in terms of fiber voxels?_inv,and the fiber characteristics based on brain connectivity-inter-regional connectivity Fiber bias index LI.?2?Aiming at the lack of reliable scientific neurological disease signature method,based on the white matter fiber tracking algorithm,a machine learning method was introduced and a ReliefF-SVM feature classification algorithm was proposed.The method uses the ReliefF algorithm to select the characteristics of the fiber quantification index,and then uses the Support Vector Machine?SVM?to classify the disease and combine the connectivity to study the potential differences between diseases.The method is a fully automated process that eliminates the need for additional manual intervention by the doctor,and the entire process is efficient and concise.?3?In view of the indistinguishable problem between Parkinson's disease and SWEDD,this paper treats 22 patients with SWEDD based on voxel-based fiber index and fiber connectivity information,and distinguishes the two types of diseases by the proposed disease signature method.The experimental results show that the SD_STREAM-based disease signature method based on white matter fiber tracking algorithm can successfully distinguish PD and SWEDD,and the classification accuracy is as high as 81.25%.According to the classification results,the globular pallidus of the left brain,the nucleus accumbens,the ventral caudate body of the right brain,the nucleus accumbens and the posterior parietal thalamus can mark PD and SWEDD.In the experimental results of connectivity,it can be seen that PD and SWEDD differ in the connection between the cingulate gyrus and the putamen and thalamus.The successful differentiation of Parkinson's disease and SWEDD syndrome verifies the feature extraction method based on white matter fiber tracking algorithm and the effectiveness of ReliefF-SVM feature classification algorithm,which can provide a new strategy for disease analysis.