BehaviorDEPOT

BehaviorDEPOT performs automated behavioral classification and quantitative analysis of rodent behaviors from video to support studies linking behavior with neurophysiological manipulations such as optogenetics and calcium imaging.

Key Features:

• MATLAB implementation: The software is implemented in MATLAB.
• Machine learning–based pose/position estimation: Leverages machine learning algorithms to estimate animal positions over time and space from behavioral video recordings.
• Hard-coded classifiers: Provides built-in classifiers for specific behaviors including freezing and rearing.
• Spatiotemporal filtering: Applies spatiotemporal filters to behavioral signals for analysis across experimental designs including cued tasks and optogenetic manipulations.
• Custom classifier creation: Modules facilitate creation of custom behavioral classifiers and definition of ground-truth behavior labels.
• Data organization and alignment: Produces organized behavioral data arrays and aligns them with neurophysiological recordings for joint analysis.
• Modular architecture and experiment support: Comprises six independent modules, including an Analysis Module and an Experiment Module that supports Arduino-based fear conditioning interfacing with commercial hardware.

Scientific Applications:

• Fear conditioning: Quantification of freezing and other behaviors during fear conditioning experiments.
• Avoidance tasks: Classification and quantification of behaviors in avoidance assays.
• Decision-making tasks: Measurement of task-relevant behaviors in decision-making assays to relate behavior to neural data.
• Compatibility with head-mounted devices: Effective behavior classification across camera types and in mice and rats equipped with optogenetic patch cables or head-mounted Miniscopes.

Methodology:

Implemented in MATLAB, BehaviorDEPOT uses machine learning algorithms for pose/position estimation from video, applies spatiotemporal filters, employs hard-coded classifiers (e.g., freezing, rearing), allows creation of custom classifiers with user-defined ground-truth labels, and aligns organized behavioral data arrays with neurophysiological recordings.