Brain Connectivity Analysis in Alzheimer's disease using Graph Convolutional Network

Alzheimer’s disease (AD) can cause a lapse in cognitive capacity along with forgetfulness, and addressing this issue is vital for forthcoming treatment efforts. This study offers another approach to investigating brain links in Alzheimer’s disease with Graph Convolutional Networks (GCN). The GCN technology enables more appropriate utilization of the connections topology that is constructed by brain networks based on fMRI data. The purpose of this paper is to design brain connection networks utilizing modern computational technology. Furthermore, this research interprets the model's predictions using layer-wise relevance propagation, focusing on key brain regions and connections involved in Alzheimer's disease. This paper uses the ADNI dataset for early predictions of AD. The metrics in which the model suggested outperformed the available approaches are, the accuracy is 88.90%, error rate is 11.10% while its precision reached 91, sensitivity of 88 and specificity 89 under NC vs AD; the accuracy of 87.80%, error rate of 12.20% while its precision reached 89, sensitivity of 87 and specificity 88 under NC vs MCI; accuracy is 86.50%, error rate is 13.50% while its precision reached 87, sensitivity of 86 and specificity 87 under AD vs MCI. Hence this work gives a valuable tool for studying brain connections, assisting in the early detection and understanding of AD.