Large scale neuronal type classification using graph neural networks Public

Zhang, Allen (Spring 2023)

Permanent URL: https://etd.library.emory.edu/concern/etds/9w0324583?locale=fr
Published

Abstract

Neuron classification is an important task in contemporary neuroscience, providing a deeper understanding of the intricate structure-function relationships of the brain. As more and more data are collected with advancements across domains, machine learning has emerged as an essential tool for automated data handling and analysis. While recent machine learning and traditional approaches to neuron classification have often relied on morphological or electrophysiological features, graph neural networks (GNNs) have not been extensively explored despite their effectiveness in analyzing complex and irregularly structured data. In this study, we show that supervised classification with GNNs on primary brain regions and cell types performs remarkably well across four large datasets. Our findings indicate that GNNs offer a distinctive and promising approach to neuron classification, with numerous potential avenues for future research.

Table of Contents

1 Introduction

2 Methodology

2.1 Dataset and Preprocessing

2.2 GNN Models

2.3 Experiments and Procedure

3 Results

3.1 Accuracy and ROC AUC

3.2 Confusion Matrices

3.3 Data Analysis

4 Discussion

Bibliography

About this Honors Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School
Department
Degree
Submission
Language
  • English
Research Field
Mot-clé
Committee Chair / Thesis Advisor
Committee Members
Dernière modification

Primary PDF

Thumbnail Title Date Uploaded Actions
Large scale neuronal type classification using graph neural networks () 2023-04-10 11:47:42 -0400

Supplemental Files

Thumbnail Title Date Uploaded Actions