TMODFCFNet: Topological Multi-Order Dynamic Functional Connection Fusion Network for the Diagnosis of ADHD in...

Traditional attention-deficit hyperactivity disorder (ADHD) diagnostic methods are subjective and often time-consuming, prompting the need for more objective and efficient diagnostic tools. This paper proposed a novel topological multi-order dynamic functional connectivity fusion network (TMODFCFNet) for the efficient diagnosis of ADHD in children based on fNIRS. The TMODFCFNet constructs a topological three-order functional connectivity (TTOFC) approach to uncover advanced connectivity patterns between brain regions during cognitive tasks. And integrating both second-order and third-order functional connectivity (FC) To further enhance performance. Additionally, the model employs a sliding window (SW) technique to simultaneously compute second-order and third-order dynamic FC, generating a time-varying adjacency matrix that improves diagnostic accuracy. Spatial features of the second-order and third-order dynamic FC are extracted within each SW using dynamic spatial convolution, while temporal dependencies of neural coordination across different SWs are captured through temporal convolution. Experimental results using clinical data from Xi'an People's Hospital (Xi'an Fourth Hospital) demonstrate that TMODFCFNet effectively distinguishes ADHD children from healthy controls. These findings suggest that the model provides a promising tool for ADHD diagnosis and may contribute to further research on ADHD mechanisms and clinical assessment based on brain functional connectivity.