JMIR Publications

ABSTRACT

Background:

Early detection is clinically crucial for the strategic handling of sarcopenia, yet the screening process, which includes assessments of muscle mass, strength, and function, remains complex and difficult to access.

Objective:

This study aims to develop a convolutional neural network (CNN) model based on ultrasound (US) images to simplify the diagnostic process and promote its accessibility.

Methods:

This study prospectively evaluated 357 participants (101 sarcopenia and 256 non-sarcopenia) for training, encompassing three types of data: muscle US images, clinical information, and laboratory information. Three monomodal models based on each data type were developed in the training cohort, respectively. The data type with the best diagnostic performance was selected to develop the bimodal and multimodal model by adding other one or two data types. Subsequently, the diagnostic performance of the above models was compared. The contribution ratios of different data types were further analyzed for the multimodal model. By comprehensive comparison, we finally identified the optimal model (SARCO model) as the convenient solution. Moreover, the SARCO model underwent an external validation by 145 participants (68 sarcopenia and 77 non-sarcopenia) and a proof-of-concept validation by 82 participants (19 sarcopenia and 63 non-sarcopenia) from 2 other hospitals, respectively.

Results:

The monomodal model based on US images achieved the highest area under the receiver operator characteristic curve (AUC) of 0.827 and an F1 score of 0.738 among the three monomodal models. The performance of the multimodal model demonstrated statistical differences compared to the best monomodal model (AUC: 0.845 vs. 0.827, P=0.019), the two bimodal models based on US images + clinical information (AUC: 0.845 vs. 0.826, P=0.031) and US images + laboratory information (AUC: 0.845 vs. 0.832, P=0.035), respectively. On the other hand, US images contributed the most evidence for diagnosing sarcopenia (0.787) and non-sarcopenia (0.823) in the multimodal models. After comprehensive clinical analysis, the monomodal model based on US images was identified as the SARCO model. Subsequently, the SARCO model achieved satisfactory prediction performance in external validation and proof-of-concept, with an AUC of 0.801 and 0.757 and F1-scores of 0.727 and 0.666, respectively.

Conclusions:

All three types of data contributed to sarcopenia diagnosis, while US images played a dominant role in model decision-making. The SARCO model based on US images is potentially the most convenient solution for diagnosing sarcopenia. Clinical Trial: The research protocol was registered at www.chictr.org.cn (ChiCTR2300073651).