Gravelines C. Deep Learning via Stacked Sparse Autoencoders for Automated Voxel

Diploma (Master), University of Western Ontario, 2014, -85 p.Functional brain parcellation – the delineation of brain regions based on functional connectivity – is an active research area lacking an ideal subject-specific solution independent of anatomical composition, manual feature engineering, or heavily labelled examples. Deep learning is a cutting-edge area of machine learning on the forefront of current artificial intelligence developments. Specifically, autoencoders are artificial neural networks which can be stacked to form hierarchical sparse deep models from which highlevel features are compressed, organized, and extracted, without labelled training data, allowing for unsupervised learning. This thesis presents a novel application of stacked sparse autoencoders to the problem of parcellating the brain based on its components’ (voxels’) functional connectivity, focusing on the medial parietal cortex. Various depths of autoencoders are investigated, yielding results of up to (68 ± 3)% accuracy compared with ground truth parcellations using Dice’s coefficient. This data-driven functional parcellation technique offers promising growth to both the neuroimaging and machine learning communities.Introduction and Literature Review
Methodology
Implementation and Results
Discussion and Conclusion