Design And Implementation Of Spinal Cord Recognition And Segmentation System Based On MRI Image

Multiple Sclerosis(MS)is a common auto-immune demyelinating neurodegenerative disease of the central nervous system.MS patients often have tissue damage in their spinal cords.Many clinical studies have shown that spinal cord tissue damage correlates significantly with MS patients’clinical symptoms.Therefore,there is a practical need in quantifying spinal cord tissue damage.As a noninvasive technique,magnetic resonance imaging(MRI)plays a pivotal role in MS diagnosis.Due to the unique structure and complicated surrounding environment of spinal cord,conventional MRI cannot reveal fine details of spinal cord structure.High resolution MRI generates significantly more amount of data.Physicians need to manually evaluate a large amount of MRI images to find tissue damage.This process is labor intensive and carries relatively high error rate.To solve this problem,several software(FreeSurfer,SPM(Statistical parametric mapping),FSL(FMRIB Software Library))have been developed.The available tools generally present the following problems:unable to analyze spinal cord data in a fully automated way,unable to work on MRI images in different contrast,unable to work on images with low signal-to-noise ratio and unable to handle abnormal spinal cords.In order to improve the efficiency and accuracy of MS diagnosis,we aim to build a fully automated software with Python programming language and various image process techniques.The software has the following functions:(1)Detect the location information of spinal cord(including coordinate,orientation and volume),segment the whole spinal cord and validate segmentation results with experiments.(2)Label different levels of the spinal cord in MRI images and output the coordinates and cross-sectional areas of each level.(3)Segment grey matter and white matter in the spinal cord.Through analyzing the spinal cord structure of healthy subjects,we construct an atlas of grey and white matter.The atlas is then registered to MRI images and output coordinates of grey and white matter.The software shows high accuracy,reliability and reproducibility in clinical MRI images.Therefore,it will improve the efficiency and accuracy of MS diagnosis and other related auto-immune disease.