期刊
MEDICAL IMAGE ANALYSIS
卷 73, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.media.2021.102163
关键词
Dynamic functional connectivity; Single-sided deafness; Different tem poral term; Nodal efficiency-based correlation matrix; rs-fMRI
类别
资金
- National Natural Science Foundation of China [62071049, 61801026]
- National Key Research and Develop-ment Program of China [2020YFC2005200]
This study proposed a dynamic window approach with random length and position to extract participants' dynamic characteristics under different temporal terms, and utilized a linear support vector machine model to identify SSD patients and healthy controls. Findings indicated that FCs related to the frontoparietal, somatomotor, dorsal attention, limbic and default mode networks played significant roles in SSD patients.
As an extreme type of partial auditory deprivation, single-sided deafness (SSD) has been demonstrated to lead to extensive neural plasticity according to multimodal neuroimaging studies. Among them, resting-state functional magnetic resonance imaging (rs-fMRI) offers valuable information on functional connec-tivities (FCs). However, most previous SSD rs-fMRI studies assumed that the extracted FC remains sta-tionary during the entire fMRI scan and neglected dynamic functional activities. Existing fixed window-based dynamic FC analysis also ignores dynamic functional activities under different tem poral terms. Ad-ditionally, due to the cost constraints of using MRI machines, using data-driven methods for unbiased hypothesis investigations may require more effective sample data augmentation techniques. To tackle these challenges and problems together, in this study, we proposed a dynamic window with a random length and position to extract participants' dynamic characteristics under different temporal terms and to extract more information from the dataset. Then, we proposed a nodal efficiency-based correlation matrix to describe the relationships of synergism between regions as features and applied a linear sup-port vector machine (SVM) model to learn the importance of the features, which helped to identify SSD patients and healthy controls. A total of 68 participants (including 23 with left SSD, 20 with right SSD and 25 healthy controls) were enrolled. Our proposed approach with a random window showed clear improvement compared with traditional static and fixed window-based dynamic FC by using the linear SVM model. FCs related to the frontoparietal, somatomotor, dorsal attention, limbic and default mode networks played significant roles in differentiating SSD patients from healthy controls. Additionally, FCs between the somatomotor and frontoparietal networks made the greatest contribution to the classifica-tion model. Regarding brain regions, FCs related to the superior frontal gyrus, superior parietal lobule, superior temporal gyrus, amygdala, and orbital gyrus played significant roles. These findings suggest that networks and regions related to higher-order cognitive functions showed the most significant FC alter-ations in SSD, which may represent a compensatory collaboration of cognitive resources in SSD. (c) 2021 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据