Multimodal Brain Network Jointly Construction and Fusion for Diagnosis of Epilepsy

Abstract

Brain network analysis has been proved to be one of the most effective methods in brain disease diagnosis. In order to construct discriminative brain networks and improve the performance of disease diagnosis, many machine learning–based methods have been proposed. Recent studies show that combining functional and structural brain networks is more effective than using only single modality data. However, in the most of existing multi-modal brain network analysis methods, it is a common strategy that constructs functional and structural network separately, which is difficult to embed complementary information of different modalities of brain network. To address this issue, we propose a unified brain network construction algorithm, which jointly learns both functional and structural data and effectively face the connectivity and node features for improving classification. First, we conduct space alignment and brain network construction under a unified framework, and then build the correlation model among all brain regions with functional data by low-rank representation so that the global brain region correlation can be captured. Simultaneously, the local manifold with structural data is embedded into this model to preserve the local structural information. Second, the PageRank algorithm is adaptively used to evaluate the significance of different brain regions, in which the interaction of multiple brain regions is considered. Finally, a multi-kernel strategy is utilized to solve the data heterogeneity problem and merge the connectivity as well as node information for classification. We apply the proposed method to the diagnosis of epilepsy, and the experimental results show that our method can achieve a promising performance.

A spatially constrained independent component analysis jointly informed by structural and functional network connectivity

“…Our approach involves constructing a new multi-modal ICA model that jointly embeds structural and functional network connectivity. This model is supported by previous studies that indicate that structural connectivity forms the foundation of functional connectivity (Honey et al, 2009; Litwińczuk et al, 2022; Stam et al, 2016; Zhu et al, 2021). In addition, we incorporated prior spatial information into our multi-modal ICA model to increase the generalizability of the estimated ICNs.…”

“…Thus, we assumed regions with higher structural connectivity (more fiber tracts) would exhibit closer functional signal activities because of their shorter distance. Prior works (He & Niyogi, 2003; Zhu et al, 2021), led us to reconstruct an objective function ) to adjust the functional activity of an ICN by considering other structurally connected ICNs. Here, M lj represents structural connectivity weights and FC denotes the correlation between pairs of ICNs ( l , j ).…”

“…To this end, minimizing the structural weighted distances of the paired TC would guide the model. SC weights captured from dMRI can reflect existing physical connectivity between functionally related gray matter regions, this means, more fiber bundles between two brain regions (here is ICNs) caused in more functional correlations and shorter distance between TC [33].…”

“…Inspired by [33] an objective function (min ∑ ‫ܯ‬ ‫ܥܨ‪ฮ‬‬ െ ‫ܥܨ‬ ฮ ଶ ) was reconstructed, where ‫ܯ‬ is SC weights and ‫ܥܨ‬ is correlation between pair ICNs (݈, ݆). Note, FC between ICNs can be maximized by minimizing the distance of the ICNs TC.…”

“…In addition, jointly analyzing models is a growing methodology used to investigate the human brain and how structural and functional relate to each other (Calhoun & Sui, 2016; Puxeddu et al, 2022; Zhu et al, 2021). One may provide variations in brain functions at different regions by multilayer network models (Puxeddu et al, 2020; Shine et al, 2016).…”