Representation learning of resting state fMRI with variational autoencoder

• Published in: NeuroImage (Orlando, Fla.) Vol. 241; p. 118423
• Main Authors: Kim, Jung-Hoon, Zhang, Yizhen, Han, Kuan, Wen, Zheyu, Choi, Minkyu, Liu, Zhongming
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.11.2021; Elsevier Limited; Elsevier
• Subjects: Brain - diagnostic imaging; Brain - physiology; Connectome - methods; Databases, Factual; Deep generative model; Deep learning; Functional magnetic resonance imaging; Humans; Individuality; Latent gradients; Learning; Magnetic Resonance Imaging - methods; Neural networks; Random variables; Rest - physiology; Unsupervised learning; Unsupervised Machine Learning; Variational autoencoder; Variational autoencoder; Deep generative model; Latent gradients; Unsupervised learning
• DOI: 10.1016/j.neuroimage.2021.118423
• ISSN: 1053-8119; 1095-9572

•Variational auto-encoder disentangles generative factors of resting state fMRI activity.•Variational auto-encoder trained with unsupervised learning extracts useful latent representations discriminative across individuals.•Temporal gradients of the latent representations report various types of transition in cortical network activity. Resting state functional magnetic resonance imaging (rsfMRI) data exhibits complex but structured patterns. However, the underlying origins are unclear and entangled in rsfMRI data. Here we establish a variational auto-encoder, as a generative model trainable with unsupervised learning, to disentangle the unknown sources of rsfMRI activity. After being trained with large data from the Human Connectome Project, the model has learned to represent and generate patterns of cortical activity and connectivity using latent variables. The latent representation and its trajectory represent the spatiotemporal characteristics of rsfMRI activity. The latent variables reflect the principal gradients of the latent trajectory and drive activity changes in cortical networks. Representational geometry captured as covariance or correlation between latent variables, rather than cortical connectivity, can be used as a more reliable feature to accurately identify subjects from a large group, even if only a short period of data is available in each subject. Our results demonstrate that VAE is a valuable addition to existing tools, particularly suited for unsupervised representation learning of resting state fMRI activity.