Automated analysis of a novel object recognition test in mice using image processing and machine learning.

Area of Science:

• Neuroscience
• Behavioral Science
• Computer Science

Background:

• The novel object recognition test (NORT) is crucial for assessing novelty recognition in animal models.
• Manual NORT analysis is labor-intensive and limits high-throughput studies.
• Automated behavioral analysis is needed to enhance efficiency and objectivity.

Purpose of the Study:

• To develop and validate an automated machine learning-based method for analyzing exploratory behavior in the NORT.
• To compare the performance of the automated method against human observer assessments.
• To enable high-throughput analysis of NORT in experimental animals.

Main Methods:

• Utilized DeepLabCut, a machine learning model, to detect mouse nose and tail base coordinates from video recordings.
• Developed a Support Vector Machine (SVM) classifier trained on mouse feature vectors to distinguish exploratory from non-exploratory behaviors.
• Validated the SVM's performance by comparing its predictions of exploratory duration and novelty discrimination index with human observer data.

Main Results:

• The trained SVM demonstrated high accuracy in detecting exploratory behaviors, correlating strongly with human assessments.
• Automated analysis of NORT using the SVM showed significant correlation with human observer evaluations of exploratory behavior duration.
• The novelty discrimination index derived from SVM predictions closely matched human observations.

Conclusions:

• An automated machine learning approach effectively analyzes exploratory behavior in the NORT.
• This method provides a reliable and objective alternative to manual observation, facilitating high-throughput behavioral phenotyping.
• The developed system enhances the efficiency and scalability of cognitive behavior research in animal models.