Establishment and preliminary application of object recognition system based on DeepLabCut.

Area of Science:

• Neuroscience
• Behavioral Science
• Artificial Intelligence

Background:

• Assessing rodent cognitive function is crucial for understanding aging and disease.
• Traditional methods for analyzing exploratory behavior have limitations in sensitivity and precision.
• Deep learning offers potential for more objective and detailed behavioral analysis.

Purpose of the Study:

• To develop and validate a DeepLabCut (DLC)-based system for precise quantification of mouse exploratory behavior.
• To assess the system's efficacy in detecting cognitive changes in natural aging and periodontitis mouse models.
• To overcome limitations of traditional visual inspection in cognitive assessments.

Main Methods:

• Constructed a custom arena with a high-definition industrial camera.
• Trained the DLC deep learning algorithm to track five mouse body landmarks.
• Quantified 36 behavioral indicators across sniffing frequency, exploration duration, and novelty preference, using dynamic distance thresholds for fine-grained analysis.

Main Results:

• The DLC system detected significant cognitive reductions in aged mice, including decreased exploration frequency and duration of novel objects.
• In periodontitis models, the DLC system identified altered exploration patterns toward old and new objects, and reduced novelty preference.
• DLC analysis revealed subtle cognitive changes missed by traditional visual inspection in both aging and disease models.

Conclusions:

• The developed DLC-based system provides sensitive, precise, and multidimensional quantification of mouse exploratory behavior.
• This system effectively distinguishes cognitive characteristics of aged and disease model mice.
• It offers comprehensive behavioral evidence for elucidating neural mechanisms of cognitive impairment in aging and inflammation.