Characterization of local white matter microstructural alterations in Alzheimer’s disease: A reproducible study

• Published in: Computers in biology and medicine 2024-09, Vol.179, p.108750, Article 108750
• Main Authors: Wen, Caiyun, Zeng, Qingrun, Zhou, Ronghui, Xie, Lei, Yu, Jiangli, Zhang, Chengzhe, Wang, Jingqiang, Yu, Yan, Gu, Yixin, Cao, Guoquan, Feng, Yuanjing, Wang, Meihao
• Format: Article
• Language: English
• Subjects: Abnormalities; Alzheimer's disease; Anisotropy; Automated fiber clustering; Automatic control; Automation; Biology; Cingulum; Cluster analysis; Clustering; Corpus callosum; Correlation analysis; Datasets; Dementia; Diffusion magnetic resonance imaging; Disease; Education; Gender; Magnetic resonance imaging; Neurodegenerative diseases; Neuroimaging; Quantitative analysis; Segments; Substantia alba; Tensors; White matter
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2024.108750

Alzheimer’s disease (AD) is a neurodegenerative disease with a close association with microstructural alterations in white matter (WM). Current studies lack the characterization and further validation of specific regions in WM fiber tracts in AD. This study subdivided fiber tracts into multiple fiber clusters on the basis of automated fiber clustering and performed quantitative analysis along the fiber clusters to identify local WM microstructural alterations in AD. Diffusion tensor imaging data from a public dataset (53 patients with AD and 70 healthy controls [HCs]) and a clinical dataset (27 patients with AD and 19 HCs) were included for mutual validation. Whole-brain tractograms were automatically subdivided into 800 clusters through the automatic fiber clustering approach. Then, 100 segments were divided along the clusters, and the diffusion properties of each segment were calculated. Results showed that patients with AD had significantly lower fraction anisotropy (FA) and significantly higher mean diffusivity (MD) in some regions of the fiber clusters in the cingulum bundle, uncinate fasciculus, external capsule, and corpus callosum than HCs. Importantly, these changes were reproducible across the two datasets. Correlation analysis revealed a positive correlation between FA and Mini-Mental State Examination (MMSE) scores and a negative correlation between MD and MMSE in these clusters. The accuracy of the constructed classifier reached 89.76% with an area under the curve of 0.93. This finding indicates that this study can effectively identify local WM microstructural changes in AD and provides new insight into the analysis and diagnosis of WM abnormalities in patients with AD. •The quantitative analysis was performed on the substructure of fiber tracts.•The proposed method has the potential to study the subfunction of the fiber tracts.•Our findings in AD patients are consistent across clinical and public datasets.