Related Experiment Video
Updated: Nov 28, 2025
10:37
Author Spotlight: Decellularization-Based Quantification of Skeletal Muscle Fatty Infiltration
Published on: June 9, 2023
3.7K
Deep learning-based thigh muscle segmentation for reproducible fat fraction quantification using fat-water
Jie Ding1,2, Peng Cao1, Hing-Chiu Chang1
1Department of Diagnostic Radiology, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Pok Fu Lam, Hong Kong SAR, China.
Insights Into Imaging
|November 30, 2020
Summary
This study introduces an automated deep learning method for thigh muscle segmentation in MRI, improving fat fraction quantification accuracy and reproducibility compared to manual methods. This technique aids in assessing fat infiltration in thigh muscles.
Area of Science:
- Medical Imaging
- Deep Learning Applications
- Musculoskeletal MRI
Background:
- Accurate whole thigh muscle segmentation is crucial for quantitative MRI analysis.
- Current methods face challenges in time-efficiency and reproducibility.
- Deep learning offers a potential solution for automated segmentation.
Purpose of the Study:
- To develop an automated whole thigh muscle segmentation method using deep learning.
- To enable reproducible fat fraction quantification on fat-water decomposition MRI.
- To compare the accuracy and reproducibility of automated vs. manual segmentation.
Main Methods:
- A U-net deep learning model was trained on thigh MRI datasets.
- Automated segmentation was applied to quadriceps femoris, sartorius, gracilis, and hamstring muscles.
- Segmentation accuracy was assessed using Dice coefficients and volume differences.
- Reproducibility of fat fraction quantification was evaluated using intraclass correlation coefficients (ICCs).
Main Results:
- Automated segmentation achieved average Dice coefficients > 0.85.
- Average volume difference was 7.57% and mean fat fraction difference was 0.17%.
- Automated segmentation demonstrated higher ICCs (0.921) for fat fraction reproducibility than manual segmentation (0.902).
- Significantly higher mean fat fraction was detected in abnormal thighs with fat infiltration.
Conclusions:
- The automated thigh muscle segmentation method is accurate and reproducible.
- It offers superior reproducibility in fat fraction estimation compared to manual segmentation.
- This automated approach facilitates quantitative analysis of fat infiltration in thigh muscles.
