Automatic MRI segmentation of masticatory muscles using deep learning enables large-scale muscle parameter analysis

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Anatomy

Background:

• Patient-specific implants are crucial for mandibular reconstruction, but current designs neglect individual biomechanical needs.
• Manual segmentation of masticatory muscle parameters is labor-intensive, hindering personalized implant development.

Purpose of the Study:

• To develop and validate a deep learning model for automatic segmentation of eight masticatory muscles from MRI data.
• To enable large-scale analysis of muscle parameters for improved biomechanical considerations in implant design.

Main Methods:

• A deep learning model was trained using 40 T1-weighted MRI scans with manual or pseudo-labelled segmentation.
• The model underwent 5-fold cross-validation and was tested on 10 manually segmented scans.
• Performance was evaluated using Dice similarity coefficient (DSC), intersection over union (IoU), precision, and recall.

Main Results:

• The model achieved high segmentation accuracy with a mean DSC of 0.88, IoU of 0.79, precision of 0.87, and recall of 0.89.
• Demonstrated feasibility for reproducible, large-scale analysis of muscle volume, direction, and estimated forces.
• Indicated the potential for integrating these parameters into patient-specific implant design.

Conclusions:

• The developed deep learning model accurately segments masticatory muscles, facilitating biomechanical analysis.
• This approach advances personalized surgical strategies for mandibular reconstruction.
• Offers a pathway toward improved implant success rates and patient care in reconstructive surgery.