Related Experiment Video
Updated: Nov 21, 2025
06:19
Integration of Animal Behavioral Assessment and Convolutional Neural Network to Study Wasabi-Alcohol Taste-Smell Interaction
Published on: August 16, 2024
675
Automated detection of mouse scratching behaviour using convolutional recurrent neural network.
Koji Kobayashi1, Seiji Matsushita1, Naoyuki Shimizu1
1Department of Animal Radiology, Graduate School of Agricultural and Life Sciences, The University of Tokyo, 1-1-1, Yayoi, Bunkyo-ku, Tokyo, 113-8657, Japan.
Scientific Reports
|January 13, 2021
Summary
Researchers developed a novel deep neural network method to automatically detect scratching behavior in mice, a key indicator of itching and stress. This new technique accurately quantifies scratching, aiding in the study of various disease models.
Area of Science:
- Neuroscience
- Animal Behavior
- Computational Biology
Background:
- Scratching is a crucial behavior in experimental animals, reflecting pruritus and psychological stress.
- Accurate quantification of scratching is essential for understanding these conditions.
Purpose of the Study:
- To develop and validate a novel deep neural network (DNN) method for automated scratching detection in mice.
- To assess the accuracy and applicability of the DNN in both chemically-induced and disease models of scratching.
Main Methods:
- A convolutional recurrent neural network (CRNN) was designed and trained using manually labeled video data of mouse scratching.
- Scratching was induced using lysophosphatidic acid and observed via standard camera recordings.
- The CRNN model was evaluated on its sensitivity, positive predictive rates, and correlation with human observations.
Main Results:
- The CRNN achieved high accuracy in detecting scratching, with sensitivity of 81.6% and a positive predictive rate of 87.9% in initial tests.
- The DNN model's predictions for the number and duration of scratching events closely correlated with human observations.
- The method demonstrated strong performance in a hapten-induced atopic dermatitis mouse model, showing 94.8% sensitivity and 82.1% positive predictive rate.
Conclusions:
- A novel and accurate method for automated scratching detection in mice using CRNN has been established.
- This DNN-based approach provides a reliable tool for studying scratching behavior in various experimental settings and disease models.
- The developed method can significantly aid research into conditions involving itching and stress in animal models.
