Related Experiment Video
Updated: Aug 26, 2026
Utility of Dissociated Intrinsic Hand Muscle Atrophy in the Diagnosis of Amyotrophic Lateral Sclerosis
Published on: March 4, 2014
Machine learning-based diagnostic models for prevalent metabolic syndrome among patients with disuse muscle atrophy:
Jiaojiao Shu1, Halimire Abudujilili2, Wenhui Xiong1
1School of Nursing, Xinjiang Medical University, Urumqi, China.
Background:
Disuse muscle atrophy (DMA) leads to muscle loss and impaired motor function, while metabolic syndrome (MetS) increases the risk of cardiovascular disease and diabetes. Muscle loss exacerbates insulin resistance, accelerates MetS progression, and creates a vicious cycle. This study aimed to develop and validate an interpretable machine learning (ML) model for identifying prevalent MetS in patients with DMA.
Methods:
This retrospective study was conducted using data from 1,004 patients who visited the Department of Rehabilitation Medicine at four medical institutions in Urumqi, Xinjiang, between January 2018 and June 2024. The data were randomly divided into a training set (n = 704) and a validation set (n = 300). Feature selection was performed using least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation. Seven ML algorithms-Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), XGBoost, LightGBM, Support Vector Machine (SVM), and Artificial Neural Network (ANN)-were used to construct diagnostic models. Model performance was evaluated using receiver operating characteristic curves, calibration plots, and decision curve analysis. The SHapley Additive exPlanations (SHAP) framework was applied to interpret the contributions of feature variables.
Results:
The prevalence of Mets among patients with DMA was 45.72%, eight key predictors were identified. The RF model demonstrated the best discriminatory performance, with an area under the curve (AUC) of 0.961 (95% CI: 0.938-0.979), accuracy of 90.0%, precision of 92.8%, sensitivity of 84.7%, specificity of 94.5%, and F1 score of 88.5%. The top five predictors identified by SHAP analysis were waist circumference, blood glucose, Cardiometabolic Index (CMI), High-density lipoprotein (HDL), and Triglyceride-Glucose index (TyG).
Conclusion:
We developed an interpretable ML model to aid in the early diagnosis of MetS. Among the models, particularly RF, showed outstanding predictive performance and promising application potential. It provides a practical tool for the early identification and targeted intervention, supports clinical decision-making, and promotes more rational allocation of healthcare resources.
