Deep learning in functional brain mapping and associated applications

To understand brain functions and disorders, mapping functional brain networks from functional magnetic resonance images (fMRI) have been under extensive study for decades. In recent years, it has been shown that deep learning models can be applied to fMRI data with superb representation ability over traditional methods. However, due to the high dimensionality of fMRI volumes and the lack of data, deep learning models of fMRI tend to suffer from overfitting in the training process. Besides, it is still challenging for deep learning to model complex spatio-temporal dependencies in fMRI time series. This chapter provides a review of the current literatures on deep learning models of fMRI including deep learning for mapping functional brain networks from fMRI, spatio-temporal models of fMRI, neural architecture search of deep learning models on fMRI, representing fMRI as embeddings, and deep fusion of brain structure-function in brain disorders.