Related Experiment Video
Updated: Jul 29, 2025
09:34
A Method to Estimate Cadaveric Femur Cortical Strains During Fracture Testing Using Digital Image Correlation
Published on: September 14, 2017
7.5K
Decomposition of musculoskeletal structures from radiographs using an improved CycleGAN framework
Naoki Nakanishi1, Yoshito Otake2, Yuta Hiasa3
1Division of Information Science, Graduate School of Science and Technology, Nara Institute of Science and Technology, Ikoma, Nara, 630-0192, Japan.
Scientific Reports
|May 25, 2023
Summary
This study introduces a novel method to decompose X-ray images into individual bone and muscle structures. This technique enhances diagnostic capabilities for musculoskeletal conditions using advanced AI and standard X-rays.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Biomedical Engineering
Background:
- Current methods for musculoskeletal structure decomposition from radiographs are limited, often requiring dual-energy scans and primarily focusing on high-contrast structures like bones.
- Existing techniques struggle with superimposed muscles that exhibit subtle contrast, hindering comprehensive analysis of musculoskeletal conditions.
Purpose of the Study:
- To develop and validate a method for decomposing musculoskeletal structures (bones and multiple superimposed muscles) from single, standard X-ray images.
- To address the limitations of existing methods by focusing on subtle muscle contrasts and enabling analysis without dual-energy scans.
Main Methods:
- The decomposition problem was framed as an image-to-image translation task using the CycleGAN framework with unpaired training.
- A training dataset was generated via automatic computed tomography (CT) segmentation and virtual projection of musculoskeletal structures.
- The CycleGAN model was enhanced with hierarchical learning and a gradient correlation similarity metric for improved accuracy and resolution.
Main Results:
- Experiments with 475 hip disease patients demonstrated significant enhancement in decomposition accuracy with the added features.
- The method successfully decomposed superimposed muscles and bones from single X-ray images, outperforming existing approaches.
- Validation using a new muscle asymmetry metric showed potential for diagnostic and therapeutic assistance from plain X-ray images.
Conclusions:
- The enhanced CycleGAN framework enables accurate decomposition of musculoskeletal structures from single radiographs, including challenging superimposed muscles.
- The proposed method offers a promising, non-invasive approach for assessing muscle asymmetry and aiding diagnosis and treatment planning in musculoskeletal disorders.
- This technique expands the utility of standard X-ray imaging for detailed musculoskeletal analysis.
