Date of Award

Summer 2016

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

Committee Director

Shuiwang Ji (Director)

Committee Member

Andrey Chernikov

Committee Member

Jing He

Committee Member

Dean Krusienski

Abstract

Understanding how the brain functions and quantifying compound interactions between complex synaptic networks inside the brain remain some of the most challenging problems in neuroscience. Lack or abundance of data, shortage of manpower along with heterogeneity of data following from various species all served as an added complexity to the already perplexing problem. The ability to process vast amount of brain data need to be performed automatically, yet with an accuracy close to manual human-level performance. These automated methods essentially need to generalize well to be able to accommodate data from different species. Also, novel approaches and techniques are becoming a necessity to reveal the correlations between different data modalities in the brain at the global level.

In this dissertation, I mainly focus on two problems: automatic segmentation of brain electron microscopy (EM) images and stacks, and integrative analysis of the gene expression and synaptic connectivity in the brain. I propose to use deep learning algorithms for the 2D segmentation of EM images. I designed an automated pipeline with novel insights that was able to achieve state-of-the-art performance on the segmentation of the \textit{Drosophila} brain. I also propose a novel technique for 3D segmentation of EM image stacks that can be trained end-to-end with no prior knowledge of the data. This technique was evaluated in an ongoing online challenge for 3D segmentation of neurites where it achieved accuracy close to a second human observer. Later, I employed ensemble learning methods to perform the first systematic integrative analysis of the genome and connectome in the mouse brain at both the regional- and voxel-level. I show that the connectivity signals can be predicted from the gene expression signatures with an extremely high accuracy. Furthermore, I show that only a certain fraction of genes are responsible for this predictive aspect. Rich functional and cellular analysis of these genes are detailed to validate these findings.

ISBN

9781369170580

Available for download on Saturday, October 14, 2017

Share

COinS