Using Deep-Learning Based Approaches to Quantify Drosophila Behaviors Open Access

Abstract

Deep learning is an emergent theme in the field of computational neuroscience, and the increasing amount of behavioral data in the video form calls for better tools to quantify behaviors. Building on previous research by Cande et al. (2018) that used an unsupervised method to quantify Drosophila behaviors, we investigate the incorporation of a new animal tracking tool and an autoencoder, both deep-learning-based methods, to define animal behaviors with greater precision and accuracy. Comparing to published results, the behavioral representation from the new analysis pipeline is able to reproduce many aspects of the previous work but has limitations that require further investigation. These results show promising future directions towards linking behavior and the neural circuitry underlying it. 

Table of Contents

Chapter 1: Introduction....................................................................................1

1.1 Significance of Descending Neurons in Animal Motor Control

1.2 Drosophila Descending Neurons as the Model System

1.3 Using Optogenetics to Study Drosophila Descending Motor Control

1.4 Recent Advances in Experimental Analysis Tool and Application 1.5 Hypotheses

Chapter 2: Method.........................................................................................8

2.1 Dataset

2.2 SLEAP Model for Animal Tracking

2.3 Joint Angle Calculation

2.4 Autoencoder and Median Filter for Data Denoising

2.4.1 Autoencoder Training

2.4.2 Joint Angle Data Processing Pipeline 2.5 Motion Mapper for Building the Behavior Map

Chapter 3: Results.........................................................................................19

3.1 Choosing the Behavior Map

3.2 Labeling the Behavior Map

3.3 Optogenetic Analysis

Chapter 4: Discussion.....................................................................................25

4.1 Limitations

4.2 Future Directions

References.................................................................................................30 

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	Emory College
Department	Neuroscience and Behavioral Biology
Degree	B.S.
Submission	Honors Thesis
Language	English
Research Field	Biology, Neuroscience Computer Science
Keyword	behavior descending neurons computational neuroscience Drosophila
Committee Chair / Thesis Advisor	Gordon Berman, Emory University
Committee Members	Davide Fossati, Emory University Astrid Prinz, Emory University Anita Devineni, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Using Deep-Learning Based Approaches to Quantify Drosophila Behaviors ()	2022-04-07 14:10:35 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions