Related Experiment Video
Updated: Mar 29, 2026
06:45
Automated Joint Space Detection Improves Bone Segmentation Accuracy
Published on: November 28, 2025
270
External Validation of an Open-Source Model for Automated Muscle Segmentation in CT Imaging of Cancer Patients
Hendrik Erenstein1,2,3, Jona Van den Broeck4, Annemieke van der Heij-Meijer1
1Department of Medical Imaging and Radiation Therapy, Hanze University of Applied Sciences, 9714 CA Groningen, The Netherlands.
Journal of Imaging
|March 27, 2026
Summary
This study validates an AI model for automated L3 muscle segmentation using CT scans. The artificial intelligence model shows high accuracy but requires broader validation for diverse populations and low BMI individuals.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Computed tomography (CT) at the third lumbar vertebra (L3) is crucial for muscle quantification.
- Manual segmentation of L3 muscles is time-consuming and labor-intensive.
Purpose of the Study:
- To externally validate an artificial intelligence (AI) model for automated L3 muscle segmentation.
- To assess the AI model's performance on an independent cohort and analyze subgroup characteristics.
Main Methods:
- AI model trained on 900 public CT scans with expert annotations.
- Validation performed on 232 PET CT scans with manual expert segmentation.
- Post-processing included density-based clustering and Hounsfield unit (HU) thresholding.
- Performance evaluated using Dice Similarity Coefficient (DSC) and Segmentation Surface Error (SSE).
Main Results:
- The AI model achieved a median DSC of 0.978 and a median SSE of 3.863 cm².
- Higher accuracy was observed compared to manual segmentation.
- Model showed slight overestimation at lower BMI values and errors in abdominal wall muscles.
- No significant difference in accuracy based on arm positioning.
Conclusions:
- The AI model provides accurate automated L3 muscle segmentation, supporting large-scale body composition studies.
- Broader validation is needed for low BMI individuals and diverse demographics.
- AI-driven segmentation offers efficiency gains in clinical research.
