Research on modeling and exploring of the normal brain maturity, such as in vivo study of the anatomy of the developing brain, can provide references for developmental pathologies. In this paper, we model and explore brain development by learning a discriminative representation of the cortical brain data (T1 MRI) with a class-wise non-negative dictionary learning (NDDL) approach. For each class, the proposed approach performs data modeling by first projecting the data into non-negative low-rank encoding coefficients with an analysis dictionary and then applying the coefficients onto an orthogonal synthesis dictionary to reconstruct the data. It also uses additional regularizers to enforce distal classes to fit into different analysis dictionaries. The learning problem is formulated as a sparse and low rank optimization problem, and solved with an alternating direction method of multipliers(ADMM). The effectiveness of the proposed approach is tested on brain age prediction problems by exploring the cortical status, and the experiments are conducted on the PING dataset. The proposed approach produces competitive results. Further, we were able for the first time to capture the status of brain thickness of specific cortical surface area with aging.
Lecture Notes in Computer Science
School of Computer Science

Zhang, M. (Mingli), Guo, Y, Zhang, C. (Caiming), Poline, J.-B. (Jean-Baptiste), & Evans, A. (Alan). (2019). Modeling and Analysis Brain Development via Discriminative Dictionary Learning. In Lecture Notes in Computer Science. doi:10.1007/978-3-030-33843-5_8