Brain Age Prediction based on fMRI images using Graph Neural Networks and Genetic Algorithm

Brain age prediction using fMRI images and graph neural networks is an advanced approach in the field of neuroimaging data analysis. In this method, fMRI images are initially used to extract the functional connectivity matrix and construct the brain graph. Then, using the Graph Convolutional Network with Self-Embedding (GCN-SE), features related to different brain networks and their interconnections are extracted and used for age prediction. After the initial prediction, the brain age bias correction technique is applied to improve prediction accuracy. This step helps to correct any biases in the brain age predictions and increases the model's accuracy. Finally, genetic algorithms are employed to identify the most important brain networks. This algorithm uses intelligent searching in the possible space to find the optimal subset of networks that have the greatest impact on age prediction. The results of this study, with an average absolute error of 2.003 and 1.261 years for brain images of men and women, respectively, demonstrate that this approach can provide more accurate predictions and identify key brain networks involved in the aging process.