Optimization in the analysis of social interaction test in rodents - development of a new complex vision-based tool for the detection of complex behaviors
(2020) BINP50 20192Degree Projects in Bioinformatics
- Abstract
- Behavior and social interactions are key features to determine the phenotype of an organism. Observing behavioral manifestations give insides on a fundamental level of how inner mechanisms work. But how is it possible to predict behavior in a way, such that it can be quantify and analyzed? One traditional strategy is to annotate the behavior by hand. Observing the animals and making notes about their behavior and the time its displayed, is a tedious and time consuming task. Furthermore doing it by hand leads to human error and inconsistency. To circumvent these inconsistencies a computer based approach is employed. The development of a computer tool to evaluate consistent and accurate video recordings is fundamental in the neuroscience... (More)
- Behavior and social interactions are key features to determine the phenotype of an organism. Observing behavioral manifestations give insides on a fundamental level of how inner mechanisms work. But how is it possible to predict behavior in a way, such that it can be quantify and analyzed? One traditional strategy is to annotate the behavior by hand. Observing the animals and making notes about their behavior and the time its displayed, is a tedious and time consuming task. Furthermore doing it by hand leads to human error and inconsistency. To circumvent these inconsistencies a computer based approach is employed. The development of a computer tool to evaluate consistent and accurate video recordings is fundamental in the neuroscience field. For this reason, the aim of this thesis was to describe and implement the methods and procedures to develop an analysis tool to investigate the behavioral outcomes in the social interaction test in rodents. During this thesis fundamental concepts of Image Analysis and Machine Learning will be explained. A tool based on these concepts will be presented which automatically evaluates video footage of mice, creating predictions of their behavior frame by frame. We believe that this approach, may help the researcher in the analysis task and allow them to interpret data instead of mindlessly hand annotating recordings. (Less)
- Popular Abstract
- How Computers Predict Behavior
Behavior is something inherent to every living animal. The world is perceived by the sensory system and this perception is transmitted to the brain. Depending on this perception and the internal state of the animal, an action is performed and a behavior is manifested. A behavioral manifestations can have many forms. From a frog catching a fly after seeing it, to a deer fleeing from a predator after hearing a tree branch crack, behavior defines all interactions with the surrounding. But how can one investigate an abstract concept such as behavior? One solution is to observe and analyze the behavioral manifestations of an animal.
Mice are mammals of the order Rodents. They show a complex behavior by... (More) - How Computers Predict Behavior
Behavior is something inherent to every living animal. The world is perceived by the sensory system and this perception is transmitted to the brain. Depending on this perception and the internal state of the animal, an action is performed and a behavior is manifested. A behavioral manifestations can have many forms. From a frog catching a fly after seeing it, to a deer fleeing from a predator after hearing a tree branch crack, behavior defines all interactions with the surrounding. But how can one investigate an abstract concept such as behavior? One solution is to observe and analyze the behavioral manifestations of an animal.
Mice are mammals of the order Rodents. They show a complex behavior by interacting with other mice and forming strict hierarchies and social rankings. Their behavior can be broken down into social interactions, such as fighting, mounting or sniffing and non-social actions, such as grooming, rearing or exploring. All these behavioral manifestations can be identified, counted and measured to investigate their behavior.
But how is it possible to detect and identified behavioral manifestations? An obvious approach is to record mice and annotate their behavior by hand. Recording the mice has the advantage of being a passive method to observe mice. It has little influence on their behavior. However annotating the behavior by hand leads to some problems. Human error and different preferences during the annotation process lead to inconsistencies. For example how tired or rested a person is, influences his/her performance. In science all data should be reproducible, therefore humans are not well suited to annotate data by hand. A computer based approach circumvents inconsistencies. It always performs the same!
Detecting a behavioral manifestation is not a trivial task. One approach to create a tool which is able to predict behavior, is using machine Learning to train an Artificial Intelligence (AI). An AI can be trained by showing it a high number of images depicting a behavioral manifestation. After it saw enough images it starts to understand the patterns defining a behavioral manifestation. For example that a mouse is stretching its body when it is rearing. This way the AI is able to predict a behavioral manifestation when shown an image of a mouse. The tool uses the AI to predict behavioral manifestations for each frame (image) of the video.
The data produced by such a tool can be used to identify behavioral dysfunctions. Putting a mice under stress or changing its genetic material can lead to behavioral dysfunction. Observing this gives us a better understanding of behavior and reveals internal processes in mice. We believe that this kind of tool, may help the researcher in the analysis task and allow them to interpret data instead of mindlessly hand annotating recordings.
Master’s Degree Project in Bioinformatics 30 credits 2020
Department of Biology, Lund University
Advisor: Sara Bachiller
Lund University, Faculty of Medicine, Neuroinflammation (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9011365
- author
- Köhler, Robin
- supervisor
- organization
- course
- BINP50 20192
- year
- 2020
- type
- H2 - Master's Degree (Two Years)
- subject
- language
- English
- id
- 9011365
- date added to LUP
- 2020-05-29 11:07:31
- date last changed
- 2020-06-02 08:08:55
@misc{9011365,
abstract = {{Behavior and social interactions are key features to determine the phenotype of an organism. Observing behavioral manifestations give insides on a fundamental level of how inner mechanisms work. But how is it possible to predict behavior in a way, such that it can be quantify and analyzed? One traditional strategy is to annotate the behavior by hand. Observing the animals and making notes about their behavior and the time its displayed, is a tedious and time consuming task. Furthermore doing it by hand leads to human error and inconsistency. To circumvent these inconsistencies a computer based approach is employed. The development of a computer tool to evaluate consistent and accurate video recordings is fundamental in the neuroscience field. For this reason, the aim of this thesis was to describe and implement the methods and procedures to develop an analysis tool to investigate the behavioral outcomes in the social interaction test in rodents. During this thesis fundamental concepts of Image Analysis and Machine Learning will be explained. A tool based on these concepts will be presented which automatically evaluates video footage of mice, creating predictions of their behavior frame by frame. We believe that this approach, may help the researcher in the analysis task and allow them to interpret data instead of mindlessly hand annotating recordings.}},
author = {{Köhler, Robin}},
language = {{eng}},
note = {{Student Paper}},
title = {{Optimization in the analysis of social interaction test in rodents - development of a new complex vision-based tool for the detection of complex behaviors}},
year = {{2020}},
}