Problem. The analysis of massive neuroimaging datasets is made difficult due to 1) the accessibility of adequate storage and computational resources, 2) vast number of branch points in analysis pipelines for preprocessing and model training / validation, 3) maintaining the de-identification of data while building algorithms for predicting individual traits is a catch-22 that potentially challenges tenets set in GPDR and other regulations regarding privacy. Standardization of pipelines and dataset specifications via the open source BIDS specification has addressed issues of dataset reproducibility and sharing. However, issues regarding access to adequate cloud resources and privacy preservation of individuals remain difficult challenges.
Project Lead + Mentor: Shady El Damaty, Ph.D.