MedElevator
Войти в исследования Neurobiology of disease · 2026
Multiscale multimodal graph convolutional networks for identifying essential tremor and dystonic tremor.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Wang Hongyu, Tao Li, Zhao Xiaole, Xiao Pan, Gui Honge, Xu Bintao, Chen Huiyue, Wang Hansheng, Lv Fajin, Luo Tianyou, Cheng Oumei, Luo Jing, Man Yun, Xiao Zheng, Fang Weidong
Аннотация
BACKGROUND: Essential tremor (ET) and Dystonic tremor (DT) share overlapping motor symptoms, leading to frequent misdiagnoses. Multimodal Graph Convolutional Network (GCNs) with transformer architecture hold promise for early diagnosis and pathogenesis exploration.
OBJECTIVES: To develop a multiscale multimodal GCN (MM-GCN) using structural and functional connectivity features, and identify imaging biomarkers of ET and DT.
METHODS: We collected rs-fMRI, DTI, and sMRI data and constructed eight inter-regional similarity matrices (GM, FA, MD, RD, AD, ReHo, DC, and fALFF) for each subject using Jensen-Shannon divergence. These matrices were input into the GCN architecture with multimodal attention fusion and applied to binary classification tasks (ET vs. HC, DT vs. HC, ET vs. DT) at three spatial scales. Finally, we used Grad-CAM to estimate node and edge importance for interpretability, while graph theory and correlation analyses were conducted for post-hoc validation of discriminative brain regions identified by MM-GCN.
RESULTS: All MM-GCN models demonstrated strong classification performance, with the highest mean accuracies of 95.24% for DT vs. HC, 85.45% for ET vs. HC, and 97.27% for ET vs. DT. The most discriminative brain regions were predominantly localized in the thalamus and basal ganglia, as well as in cerebellar motor and non-motor cortical networks. Correlation analysis revealed that the nodal efficiency of the two salient brain regions was significantly correlated with clinical characteristics.
CONCLUSION: The MM-GCN model demonstrates strong diagnostic capability for differentiating ET and DT. Our findings reveal cerebello-thalamo-cortical circuit involvement in both disorders, advancing multimodal imaging insights into their pathophysiology.
OBJECTIVES: To develop a multiscale multimodal GCN (MM-GCN) using structural and functional connectivity features, and identify imaging biomarkers of ET and DT.
METHODS: We collected rs-fMRI, DTI, and sMRI data and constructed eight inter-regional similarity matrices (GM, FA, MD, RD, AD, ReHo, DC, and fALFF) for each subject using Jensen-Shannon divergence. These matrices were input into the GCN architecture with multimodal attention fusion and applied to binary classification tasks (ET vs. HC, DT vs. HC, ET vs. DT) at three spatial scales. Finally, we used Grad-CAM to estimate node and edge importance for interpretability, while graph theory and correlation analyses were conducted for post-hoc validation of discriminative brain regions identified by MM-GCN.
RESULTS: All MM-GCN models demonstrated strong classification performance, with the highest mean accuracies of 95.24% for DT vs. HC, 85.45% for ET vs. HC, and 97.27% for ET vs. DT. The most discriminative brain regions were predominantly localized in the thalamus and basal ganglia, as well as in cerebellar motor and non-motor cortical networks. Correlation analysis revealed that the nodal efficiency of the two salient brain regions was significantly correlated with clinical characteristics.
CONCLUSION: The MM-GCN model demonstrates strong diagnostic capability for differentiating ET and DT. Our findings reveal cerebello-thalamo-cortical circuit involvement in both disorders, advancing multimodal imaging insights into their pathophysiology.