Skip navigation
Please use this identifier to cite or link to this item: http://arks.princeton.edu/ark:/88435/dsp01cn69m431r
Title: Connecting the Connectome Automating segmentation of neurons from tissue imaging
Authors: Donnelly, Neal
Advisors: Funkhouser, Thomas
Contributors: Seung, Sebastian
Department: Computer Science
Class Year: 2014
Abstract: Recent studies in neuroscience have shown the great promise of an approach known as connectomics, which seeks to map the brain by the connections between neurons. Attempts to study the brain in this way extract brain tissue and capture sequential cross-sectional images with an electron microscope. These projects are currently limited by the image processing stage of the pipeline, which turns these images into digitally reconstructed neurons, as state-of-the-art computer vision systems are unable to segment neurons with sufficient reliability. The SNEMI3D challenge is an open con- test that offers a benchmark on the effectiveness of automated neuron segmentation systems and invites researchers to compete in their ability to segment the challenge dataset. GALA is an open source software project to perform automated segmentation created by researchers at the Janelia Farm research campus. In this paper, I detail my efforts to improve the performance of GALA, which was narrowly in second place in the SNEMI3D contest when I began the project. I succeeded in improving GALA's performance and have now decisively taken the top spot in the contest.
Extent: 62 pages
URI: http://arks.princeton.edu/ark:/88435/dsp01cn69m431r
Type of Material: Princeton University Senior Theses
Language: en_US
Appears in Collections:Computer Science, 1988-2016

Files in This Item:
File SizeFormat 
Donnelly_Neal_Thesis.pdf3.88 MBAdobe PDF    Request a copy


Items in Dataspace are protected by copyright, with all rights reserved, unless otherwise indicated.