Predicting Fatigue in Long Duration Mountain Events with a Single Sensor and Deep Learning Model.

Area of Science:

• Sports Science
• Biomedical Engineering
• Artificial Intelligence

Background:

• Monitoring physical and cognitive fatigue is crucial in sports and occupational settings.
• Traditional fatigue assessment methods can be subjective or require specialized equipment.
• Wearable sensors offer a promising avenue for objective, continuous fatigue monitoring.

Purpose of the Study:

• To evaluate the efficacy of an AI model utilizing data from a single wearable sensor (acceleration and ECG) to quantify cognitive and physical fatigue.
• To assess the model's performance in a practical, self-paced trail running scenario.

Main Methods:

• A field-based protocol involving a trail run and cognitive tests was used to induce fatigue.
• Acceleration and electrocardiogram (ECG) data were collected using a single sensor.
• A Convolutional Neural Network (CNN) model was developed to predict fatigue levels.

Main Results:

• The Finger Tap Test (FTT) and vertical jump height were most sensitive to the fatigue protocol.
• The AI model achieved a mean absolute error of 12.5% for 'walk up' and a range of absolute error of 16.7% for 'run down' using a 200-prediction rolling average.
• The model demonstrated the ability to measure fatigue within specific activity contexts.

Conclusions:

• A single wearable sensor combined with an AI model can effectively measure physical and cognitive fatigue during field-based activities.
• This approach integrates contextual factors and provides a practical method for fatigue research.
• The findings support the application of wearable technology for real-time fatigue monitoring.