Using deep learning to study emotional behavior in rodent models

Abstract

Quantifying emotional aspects of animal behavior (e.g., anxiety, social interactions, reward, and stress responses) is a major focus of neuroscience research. Because manual scoring of emotion-related behaviors is time-consuming and subjective, classical methods rely on easily quantified measures such as lever pressing or time spent in different zones of an apparatus (e.g., open vs. closed arms of an elevated plus maze). Recent advancements have made it easier to extract pose information from videos, and multiple approaches for extracting nuanced information about behavioral states from pose estimation data have been proposed. These include supervised, unsupervised, and self-supervised approaches, employing a variety of different model types. Representations of behavioral states derived from these methods can be correlated with recordings of neural activity to increase the scope of connections that can be drawn between the brain and behavior. In this mini review, we will discuss how deep learning techniques can be used in behavioral experiments and how different model architectures and training paradigms influence the type of representation that can be obtained.

A modular, cost-effective, versatile, open-source operant box solution for long-term miniscope imaging, 3D tracking, and deep learning behavioral analysis

“…Quantifying behavior with deep learning approaches is a key component in modeling longitudinal animal behavior [46] . This can be done with models such as DeepLabCut, which tracks location and posture across video frames [47] .…”

Exploring the Dynamics of Social Interactions During the Juvenile Stage in a Mouse Model of Fragile X Syndrome

“…38 Deep learning approaches, as displayed by the software used in the present study (REF), enabled the detection and tracking of changes in the positions of defined body parts, that allowed for the investigation of social interactions in a more natural setting. 39 In this study utilizing the FXS mouse model, the length of social interaction following initiation of an approach depends on the willingness of both mice, as discrepancies in the cooperation or interest of a mouse will result in a shorter interaction. Furthermore, the number of social interaction events a mouse engages in correlates to their interest levels: more counts of events may correlate to increased social interest.…”

Data management strategy for a collaborative research center

“…These datasets are saved in a variety of file formats, including video files (.avi), the original raw ASCII log files, text-based file formats (.csv), and detailed stimulus material (e.g., .wav and .png files), among others. We are utilizing deep learning–based approaches for behavior data acquisition, analysis, pose estimation, and so on for both human and rodent experimental models [ 183 , 184 ]. This includes software packages tools such as Noldus EthoVision XT [ 185 ] (behavior data acquisition and analysis from rodents), Bonsai [ 186 ] (behavioral tracking and closed-loop experiments), ANY-maze ( RRID:SCR_014289 [ 187 ] (automated video-tracking software), PsychoPy [ 188 ] (data acquisition and analysis from humans), SPSS (Statistical Package for the Social Sciences), MATLAB and GraphPad Prism [ 189 ] (data analysis and visualization), and DeepLabCut (markerless pose estimation) [ 190 , 191 ] for measuring rodent behavior [ 192 ].…”