ThighMRIseg

WARNING

This project V1.0 has been archived and is no longer updated.

ThighMRIseg is a deep learning-based thigh muscle MRI segmentation tool trained on a multimodal, multiethnic dataset.

Overview

ThighMRIseg automatically segments 11 individual thigh muscles from MRI scans using a SegResNet architecture deployed via MONAI Label. It supports multiple MRI modalities and has been trained on data from three ethnic populations.

Segmentation Demo

Target Muscles (11 muscles)

Muscle	Compartment
Biceps Femoris (long head)	Posterior
Biceps Femoris (short head)	Posterior
Semitendinosus	Posterior
Semimembranosus	Posterior
Adductor Magnus	Medial
Vastus Intermedius	Anterior
Vastus Lateralis	Anterior
Vastus Medialis	Anterior
Rectus Femoris	Anterior
Gracilis	Medial
Sartorius	Anterior

Supported MRI Modalities

• IDEAL fat/water separation
• T1-weighted
• T2-weighted
• STIR (Short Tau Inversion Recovery)

Integration

• MONAI Label server deployment
• Compatible with OHIF Viewer and 3D Slicer
• One-click "Auto Segmentation" in clinical workflows

OHIF Viewer

3D Slicer

Citation

Multimodal Multiethnic Thigh Muscle MRI Analysis. GitHub: Hirriririir/Multimodal-Multiethnic-Thigh-Muscle-MRI-analysis