Research On Segmentation Algorithm In Brain Structral MRI And Its Application In Alzheimer's Disease

Alzheimer's disease(AD)is a chronic neurodegenerative disease associated with various cognitive dysfunction and is the main cause of age-related dementia.So far,the etiology of AD is still not clear.Once the symptoms appear,the condition is irreversible and there is no effective treatment to stop the progression of AD.Therefore,the key to defeat AD is its early diagnosis and early intervention.The 7 pairs of subcortical nuclei in brain,including nucleus accumbens,amygdala,caudate nucleus,hippocampus,globus pallidus,putamen and thalamus,are associated with memory,language expression,spatial working memory and learning,and so play an important role in the progression of AD.Exploring the atrophy patterns of subcortical nuclei in AD and Mild Cognitive Impairment(MCI)can provide additional auxiliary information for the early diagnosis of AD to some extent.Based on this purpose,a segmentation algorithm based on multi-altas registration was proposed,which combined ITK registration framework with image fusion algorithm to realize the segmentation of 7 pairs of subcortical nuclei and finally applied it to the research on AD.The detail works are as follows:(1)Brain tissue was extracted combining histogram analysis with three-dimensional morphological approach,which was named as SSA(Skull Stripping Algorithm).SSA,BET(Brain Extraction Tool)and ITK-Skull Stripping algorithm were respectively applied to extract the brain tissue in 40 brain MR images.The segmentations were scored using five-scores system by a doctor.The segmentations of 4 score or above are:40 cases for SSA,21 cases for BET and 40 cases for ITK-SkullStripping.The results show that the segmentation of SSA is consistent with the ITK-SkullStripping algorithm and is superior to the BET algorithm.(2)ITK image registration framework was used to achieve affine registration and B-spline deformation registration,registrating the atlas to the target image.Then,the image fusion algorithms such as MV(Majority voting),STAPLE(Simultaneousn Truth and Performance Level Estimation)and SBA(Shape-Based Averaging)were applied to achieve the label fusion,obtaining the final segmentation result.Ten subjects of 3DT1 W image were segmented by using the above three fusion algorithms,and the volumes of all subcortical nuclei were calculated.The volumes were compared with those using FMRIB's Integrated Registration and Segmentation Tool(FIRST)which was regarded as the gold standard of segmentation.The result shows that comparing with the FIRST algorithm,there are significant differences between the bilateral nucleus accumbens and thalamus using MV algorithm(p<0.05);the bilateral thalamus segmented by STAPLE algorithm were significantly different;volumes of 7 pairs of subcortical nuclei were not significantly different using SBA algorithm(p>0.05).(3)Segmentation of subcortical nuclei and their volume measurement were performed on 90 subjects combining ITK registration with SBA algorithm.90 subjects were grouped into AD,MCI and normal control(NC)with 30 subjects for each group.SPSS software was then applied to analyze the volume differences between the three groups and the correlation between volumes and the clinical test scores(MMSE and MOCA).The result shows that the volume of the right nucleus accumbens and the left hippocampus in MCI were significantly atrophy compared to NC,and this atrophy deteriorated in AD group,namely,volumes of bilateral hippocampus and nuclei accumbens were significantly decreased in AD compared to NC.This atrophy was strongly correlated with impaired cognitive function,reflected by the result that volume of bilateral nucleus accumbens and hippocampus was significantly correlated with MMSE and MOCA scores.In addition,applying vertex-based analysis of FIRST,it shows that the dorsal-medial left hippocampus,ventral right hippocampus and left nucleus accumbens,posterior right nucleus accumbens and the medial-ventral part of right thalamus shows atrophy most prominently in AD,and the shape distortions in these areas are also correlated with cognitive status.The experimental results show that bilateral nucleus accumbens and hippocampus can play an important role in the cognitive decline of AD and MCI.