Related Experiment Video
Updated: Oct 4, 2025
Non-invasive Skeletal Muscle Quantification in Small Animals Using Micro-computed Tomography
Published on: November 8, 2024
A Fully Automated Deep Learning Pipeline for Multi-Vertebral Level Quantification and Characterization of Muscle and
Christopher P Bridge1, Till D Best1, Maria M Wrobel1
1Massachusetts General Hospital and Brigham and Women's Hospital Center for Clinical Data Science (C.P.B., J.K.C., K.P.A.); Martinos Center for Biomedical Imaging, Department of Radiology (C.P.B, K.P.A.); Division of Thoracic Imaging and Intervention (T.D.B., M.M.W., J.P.M., F.J.F.), Department of Radiology, Massachusetts General Hospital; and Department of Radiology, Brigham and Women's Hospital, (K.P.A.), 55 Fruit St, Boston, MA 02114; Charité-Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Radiology, Berlin, Germany (T.D.B.); Department of Radiology, Berlin Institute of Health, Berlin, Germany (T.D.B.); Department of Radiology, Ludwig Maximilian University, Munich, Germany (M.M.W.); Department of Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, Calif (K.M.); Mallinckrodt Institute of Radiology, School of Medicine, Washington University, St Louis, Mo (C.J.); and Departments of Medicine and Radiology, University of Chicago, Chicago, Ill (J.H.C.).
This study presents an automated method using convolutional neural networks (CNNs) to analyze body composition, specifically muscle and adipose tissue, from chest CT scans, offering reliable quantification.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Medicine
- Biomedical Engineering
Background:
- Body composition analysis using imaging biomarkers is crucial for patient assessment.
- Routine chest CT scans offer a valuable, yet often underutilized, source for body composition data.
- Accurate quantification of skeletal muscle and adipose tissue is essential for understanding patient health and prognosis.
Purpose of the Study:
- To develop and validate a fully automated pipeline for multi-vertebral level body composition analysis on chest CT scans.
- To assess the performance of convolutional neural networks (CNNs) in segmenting and quantifying muscle and adipose tissue.
- To establish a reliable method for body composition analysis in patients undergoing chest CT.
Main Methods:
- Retrospective training of two CNNs on 629 chest CT scans from patients prior to lung cancer lobectomy.
- Development of a slice-selection network to identify key vertebral levels (T5, T8, T10).
- Implementation of a U-Net segmentation network for muscle and adipose tissue, with radiologist-guided ground truth for validation.
Main Results:
- The automated pipeline demonstrated high accuracy in assessing cross-sectional area (CSA) with a median absolute error of 3.6%.
- Intraclass correlation coefficients for CSA ranged from 0.959 to 0.998, indicating excellent reliability.
- Median attenuation measurements showed a low error of 1.0 HU, with intraclass correlation coefficients between 0.95 and 0.99.
Conclusions:
- The developed automated pipeline provides accurate and reliable quantification of muscle and adipose tissue at multiple vertebral levels on routine chest CT scans.
- This AI-driven approach enhances the utility of chest CT for comprehensive body composition analysis.
- The findings support the integration of automated body composition assessment into clinical workflows for improved patient evaluation.
