The Method And Technology Of Repair Of Magnetic Resonance Neuroimages Based On Knowledge And Reasoning

The contrast of magnetic resonance imaging(MRI)data is determined by the combination of relaxation properties and imaging parameters.However,the same experiential imaging parameters used in the MRI examinations may not be suitable for each participant due to individual difference.Therefore,it would be ideal if we can individually customize the imaging parameters in the imaging sequences for personalized imaging,or optimize the contrast of images through post-processing.In this study,we proposed a knowledge-based method that can repair MRI data of poor contrast as if a new scan were acquired using imaging parameters that were individually optimized.The proposed framework of repairing MRI data involves the following steps knowledgebase establishment,intensity normalization,relaxation time deducing,tissue segmentation,and imaging parameter individualization,which finally realized individualized deducing of MRI data and individualized imaging of MRI data.This dissertation demonstrated the contributions of MRI repair method from three dimensionsFirst,we introduced a knowledge-based method of image deducing that recovers the tissue contrast on the conventional MRI data having low contrast.This study proposed an iterative deducing method,which deduced the relaxation properties of individual tissues from knowledgebase via analogical reasoning,and further deduced the weighted images with individually optimized contrast.MRI data of adults verified that our method could effectively deduce the relaxation time maps of whole brain tissues and significantly improve the tissue contrast of conventional weighted images originally having low contrast.MRI data of ischemic rats verified that our method could effectively improve the contrast between lesion and normal tissues in the initial stage of disease.Second,we introduced a knowledge-based method of tissue segmentation that improves the segmentation accuracy on MRI data having low contrast.The knowledge-based image deducing method was first used to optimize the tissue contrast of MRI data,and then a transformed algorithm based on local binary pattern was performed to calculate the tissue probabilities of target voxels.The segmentation results for MRI data with high quality demonstrated that our method could work correctly as those popular methods.The segmentation results for MRI data with low contrast demonstrated that our method could effectively improve the segmentation accuracy compared with those popular methods.The segmentation results for MRI data from babies demonstrated that our method can actually maintain the stability of segmentation on babies' MRI data.Third,we introduced an individual MRI scanning method.This study applied the knowledge-based image deducing method on the conventional MRI data to deduce the relaxation time maps of individuals.We calculated the individualized imaging parameters based on both relaxation time maps and physics theory of conventional MRI.Then,the imaging parameters were used for rescanning in order to realize individualized imaging.The results of rescanning showed that our method could effectively improve the contrast-to-noise ratio of MRI data without decreasing the signal-to-noise ratioIn conclusion,this dissertation applied the MRI repair method on the conventional MRI data from three cases,including 5 healthy adults,9 cerebral ischemic rats and several premature infants.By running statistical tests,we confirmed the contributions of our method in various applications.Our method can effectively improve the contrast of MRI data without extra scanning,which benefits the post-processing of MRI data with low contrast.The deduced MRI data generated via virtual scanning can accurately repeat the results of existing research,which demonstrates the practicability of our method in the research.Our method can effectively increase the visual recognition of lesions,which is clinically useful.