An anatomical knowledge-based MRI deep learning pipeline for white matter hyperintensity quantification associated with cognitive impairment

• Published in: Computerized medical imaging and graphics 2021-04, Vol.89, p.101873-101873, Article 101873
• Main Authors: Liang, Li, Zhou, Pengzheng, Lu, Wanxin, Guo, Xutao, Ye, Chenfei, Lv, Haiyan, Wang, Tong, Ma, Ting
• Format: Article
• Language: English
• Subjects: Aging; Algorithms; Alzheimer's disease; Anatomical knowledge; Brain; Cognition; Cognition & reasoning; Cognitive ability; Cognitive impairment; Configurations; Corpus callosum; Decision making; Deep learning; Geriatrics; Knowledge; Lesions; Localization; Machine learning; Magnetic resonance imaging; Metric space; Neurodegenerative diseases; Older people; Pipelines; Segmentation; Segments; Spatial data; Substantia alba; White matter hyperintensities
• ISSN: 0895-6111; 1879-0771
• DOI: 10.1016/j.compmedimag.2021.101873

•An anatomical knowledge-based MRI deep learning pipeline was developed to simultaneously segment and locate white matter hyperintensities.•This pipeline significantly improved segmentation performance, especially in cohorts with mild lesion burden.•Different white matter hyperintensities patterns were associated with normal aging and cognitive impairments. Recent studies have confirmed that white matter hyperintensities (WMHs) accumulated in strategic brain regions can predict cognitive impairments associated with Alzheimer’s disease (AD). The knowledge of white matter anatomy facilitates lesion-symptom mapping associated with cognition, and provides important spatial information for lesion segmentation algorithms. However, deep learning-based methods in the white matter hyperintensity (WMH) segmentation realm do not take full advantage of anatomical knowledge in decision-making and lesion localization processes. In this paper, we proposed an anatomical knowledge-based MRI deep learning pipeline (AU-Net), handcrafted anatomical-based spatial features developed from brain atlas were integrated with a well-designed U-Net configuration to simultaneously segment and locate WMHs. Manually annotated data from WMH segmentation challenge were used for the evaluation. We then applied this pipeline to investigate the association between WMH burden and cognition within another publicly available database. The results showed that AU-Net significantly improved segmentation performance compared with methods that did not incorporate anatomical knowledge (p