Chest X-ray-based opportunistic screening of sarcopenia using deep learning

Area of Science:

• Radiology
• Gerontology
• Artificial Intelligence

Background:

• Sarcopenia detection and management are crucial for clinical practice.
• Current methods for sarcopenia assessment can be complex and time-consuming.

Purpose of the Study:

• To develop a deep learning model utilizing chest X-ray images for sarcopenia prediction.
• To create a SARC-CXR score integrating predicted muscle parameters and estimation uncertainty.

Main Methods:

• A deep learning model was trained on chest X-ray data to predict appendicular lean mass (ALM), handgrip strength (HGS), and chair rise test performance.
• A machine learning model (SARC-CXR score) was constructed using age, sex, BMI, predicted muscle parameters, and estimation uncertainty.
• Internal and external validation cohorts were used to assess model performance.

Main Results:

• The SARC-CXR score demonstrated good discriminatory performance for sarcopenia in both internal (AUROC 0.813) and external (AUROC 0.780) test sets.
• Predicted low ALM from chest X-ray was the most significant predictor of sarcopenia.
• The SARC-CXR score outperformed the SARC-F score in the internal test set (AUROC 0.813 vs. 0.691).

Conclusions:

• A chest X-ray-based deep learning model can effectively aid in sarcopenia detection.
• The SARC-CXR score shows potential as a non-invasive tool for sarcopenia risk assessment.
• Further research is warranted to explore the clinical utility of this approach.