Deep Learning-Based Grimace Scoring Is Comparable to Human Scoring in a Mouse Migraine Model.

Area of Science:

• Animal behavior research
• Computational neuroscience
• Biomedical imaging

Background:

• Pain assessment in mice is crucial for research.
• The Mouse Grimace Scale (MGS) is a reliable pain indicator.
• Manual MGS scoring is time-consuming and labor-intensive.

Purpose of the Study:

• To develop and validate a deep learning model for automated MGS scoring.
• To assess the performance of the DeepMGS model compared to human scorers.
• To explore the potential of AI in pain quantification from facial expressions.

Main Methods:

• Development of a deep learning model (DeepMGS) for automatic mouse face image analysis.
• DeepMGS predicts individual MGS action units and total scores.
• Comparison of DeepMGS performance against experienced and apprentice human scorers.

Main Results:

• DeepMGS achieved 70-90% accuracy in identifying MGS action units.
• DeepMGS total scores showed high correlation (0.83) with experienced human scorers.
• DeepMGS demonstrated comparable or superior performance to human scorers in pain classification.

Conclusions:

• DeepMGS effectively quantifies spontaneous pain in mice using facial images.
• The model shows potential for broader applications in pain condition prediction.
• Automated facial expression analysis offers a scalable solution for pain research.