Related Experiment Video
Updated: Jul 24, 2025
Segmentation and Linear Measurement for Body Composition Analysis using Slice-O-Matic and Horos
Published on: March 21, 2021
Deep Learning-Based Fully Automated Segmentation of Regional Muscle Volume and Spatial Intermuscular Fat Using CT
Rui Zhang1, Aiting He2, Wei Xia3
1School of Biomedical Engineering (Suzhou), Division of Life Sciences and Medicine, University of Science and Technology of China, Hefei, Anhui, China (R.Z.); Medical Imaging Department, Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, Suzhou, China (R.Z., W.X., J.J., W.S., X.G.).
A new deep learning (DL) system accurately segments gluteus maximus muscle and fat distribution from CT scans. This automated tool shows good agreement with expert radiologists for muscle evaluation.
Area of Science:
- Medical imaging analysis
- Artificial intelligence in radiology
- Musculoskeletal imaging
Background:
- Accurate assessment of gluteus maximus muscle volume and intermuscular fat distribution is crucial for clinical evaluation.
- Manual segmentation of muscle and fat from CT images is time-consuming and subject to inter-observer variability.
Purpose of the Study:
- To develop and validate a deep learning (DL) system for automated segmentation of gluteus maximus muscle and measurement of its fat distribution using CT images.
- To evaluate the performance of the DL system against manual segmentation by a radiologist.
Main Methods:
- A deep learning system utilizing Attention U-Net and Otsu binary thresholding was developed.
- The system was trained and tested on CT images from 472 subjects.
- Segmentation accuracy was assessed using Dice similarity coefficient (DSC), Hausdorff distance (HD), and average surface distance (ASD). Fat fraction agreement was evaluated using intraclass correlation coefficients (ICCs) and Bland-Altman plots.
Main Results:
- The DL system demonstrated strong segmentation performance with DSC values of 0.930 and 0.873 on the test sets.
- The fat fraction measurements from the DL system showed good agreement with the radiologist's assessments (ICC=0.748).
Conclusions:
- The developed DL system provides accurate and fully automated segmentation of the gluteus maximus muscle.
- The system's reliable fat fraction evaluation supports its potential for clinical muscle assessment.
