[Classification of systemic lupus erythematosus resting-state functional magnetic resonance imaging data based on the end-to-end model].
Authors
Affiliations (8)
Affiliations (8)
- Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, P. R. China.
- School of Big Data and Information Engineering, Xinjiang University of Technology, Hotan, Xinjiang 848000, P. R. China.
- Brain Cognition and Brain-computer Intelligence Integration Group, Kunming University of Science and Technology, Kunming 650500, P. R. China.
- Hangzhou Seventh People's Hospital, Affiliated Mental Health Center, Zhejiang University School of Medicine, Hangzhou 310063, P. R. China.
- Faculty of Science, Kunming University of Science and Technology, Kunming 650500, P. R. China.
- School of Information Engineering, Engineering University of the Chinese People's Armed Police Force, Xi'an 710000, P. R. China.
- School of Psychology, Shanghai Normal University, Shanghai 200234, P. R. China.
- Department of Rheumatology and Immunology, First Affiliated Hospital of Kunming Medical University, Kunming 650032, P. R. China.
Abstract
Systemic lupus erythematosus (SLE) frequently involves the central nervous system, inducing abnormal alterations in brain functional and structural networks and resulting in cognitive and psychological dysfunction in patients. To identify abnormal brain functional network patterns associated with SLE, this study adopted a dynamic threshold strategy to detect key functional connections and construct sparse brain functional networks. A graph transformation network (GTNet) was utilized to model the optimized networks, capturing local topological features and global dependencies to effectively identify abnormal patterns of SLE-related brain functional networks. Experimental results showed that the proposed model achieved an average classification accuracy of (87.48 ± 6.77)% on the resting-state functional magnetic resonance imaging dataset consisting of 107 SLE patients and 107 healthy controls. Further analysis showed that the difference in small-world properties between the two groups was statistically significant ( <i>t</i> = -2.96, <i>P</i> < 0.01). In conclusion, the model constructed in this study provides a new scheme for the auxiliary diagnosis of SLE, and its automated classification architecture has potential application value.