地图集(解剖学)
计算机科学
模式
失智症
神经影像学
模态(人机交互)
人工智能
脑图谱
机器学习
模式识别(心理学)
神经科学
疾病
痴呆
心理学
医学
社会学
病理
解剖
社会科学
作者
Vincent Le Du,Charley Presigny,Arabella Bouzigues,Valérie Godefroy,Bénédicte Batrancourt,Richard Lévy,Fabrizio De Vico Fallani,Raffaella Migliaccio
标识
DOI:10.1109/biosmart54244.2021.9677866
摘要
Multilayer networks (MNs) constitute an elegant and insightful multidimensional or multimodal framework. Bimodal MNs made from brain functional and structural networks extracted from neuroimaging modalities commonly lay the ground for truly emergent multimodal analysis. Thus far, they are computed using the same atlas for both layers. However, different atlases are required for specific imaging modalities. Depending on which atlas is chosen for a specific modality, this can lead to information from the other modalities being compromised. In this paper, we propose a new way to build such networks using specific atlases suited to each modality. The new technique is based on the computation of spatial overlaps between regions from different parcellations used for each available modality. We generalized the multiplex core-periphery method used to distinguish core and peripheral brain regions to apply it to such MNs, and to evaluate the approach and compare it to previous versions. We applied this new method in behavioral variant frontotemporal dementia (bvFTD) patients and healthy controls. First, we chose two specific atlases, the AAL2 and Schaefer100-Yeo17, for our DWI and fMRI data respectively. Subsequently, we computed richness and coreness for each subject. Finally, we benchmarked our results to evaluate the technique. We obtained higher peaks of significance and Fishers Criterion than with the previous method in the conditions that replicates previous findings. This highlights the potential of our multi-atlas MNs as well as their usefulness in MN analysis.
科研通智能强力驱动
Strongly Powered by AbleSci AI