•  
  •  
 

Rehabilitation Practice and Science

Abstract

Introduction

Sarcopenia is a systemic condition characterized by the loss of muscle mass, decreased strength and physical performance, leading to increased morbidity and mortality. To address this issue, we developed a rapid, cost-effective, and accurate screening model that integrated basic patient characteristics with a one-dimensional Convolutional Neural Network (1DCNN) analyzing Timed Up and Go (TUG) test data, captured using a simple light detection and ranging (LiDAR) device.

Methods

This study included 145 participants (38 diagnosed with sarcopenia and 107 without). During the TUG test, a LiDAR sensor semi-automatically recorded distance-time data, which was processed using a 1DCNN and combined with patient characteristics in an ensemble model. To address a class imbalance, we employed the synthetic minority oversampling technique (SMOTE). We incorporated a max-pooling layer to optimize our model's performance and implemented a 5-fold cross-validation technique, ensuring robustness and reliability in our results.

Results

The ensemble model achieved an accuracy of 0.79, a sensitivity of 0.63, a specificity of 0.86, an F1-score of 0.63, and an AUROC (Area under Receiver operating characteristic) of 0.76, respectively.

Conclusions

This study proposed an innovative screening tool for sarcopenia that aligned with smart tele-healthcare trends. By utilizing deep learning techniques, it allowed for early detection and better management of sarcopenia patients.

Download

Included in

Physiotherapy Commons

Share

COinS