Digital and Remote Interventions for Musculoskeletal Aging: Real-Time Muscle Strain Severity Detection Using...

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Digital Health

Background:

• Prolonged digital device use leads to poor posture and musculoskeletal discomfort.
• Posture-related strain is often overlooked, causing pain and potentially affecting sleep and stress.
• Existing solutions lack integrated posture and muscle strain monitoring.

Purpose of the Study:

• To develop a novel, low-cost machine learning framework for advanced muscle strain severity classification.
• To integrate posture monitoring and muscle strain detection into a unified system using Internet of Things (IoT) devices.
• To provide real-time ergonomic feedback through advanced muscle strain classification.

Main Methods:

• Created a novel dataset using real-time electromyography (EMG) and posture signals from hospital and industrial settings.
• Designed a two-part hardware architecture (posture detection and strain detection) with a NodeMCU ESP8266, ultrasonic sensor, and EMG sensor.
• Developed a hybrid machine learning model combining Vision Transformer (ViT) and XGBoost for classifying strain severity into baseline, compensatory strain, and overload categories.

Main Results:

• Achieved a high classification accuracy of 99.0% (95% CI: 98.5-99.5%) for muscle strain severity.
• Demonstrated a low inference latency of 15.2 ms, enabling real-time feedback.
• Validated diverse muscle strain patterns against clinical assessment procedures.

Conclusions:

• The proposed low-cost IoT framework offers an effective solution for advanced muscle strain severity classification.
• This technology has the potential to mitigate musculoskeletal discomfort associated with prolonged digital device use.
• The study provides a foundation for developing more sophisticated ergonomic feedback systems.