Model based analytical approach for physical activity quantification in people with type 1 diabetes

Area of Science:

• Endocrinology and Exercise Physiology
• Biomedical Signal Processing

Background:

• Physical activity (PA) significantly impacts glucose levels in type 1 diabetes (T1D).
• Effective PA monitoring is crucial for optimizing glycemic control, especially with automated insulin delivery systems.
• Distinguishing between aerobic and anaerobic PA is key due to their contrasting effects on blood glucose.

Purpose of the Study:

• To present a novel state-space model utilizing the heart rate (HR) signal.
• To quantify and differentiate between aerobic and anaerobic physical activity.
• To enhance glucose management strategies for individuals with T1D during exercise.

Main Methods:

• Developed a state-space model analyzing HR signal features (mean HR, max HR, fluctuations).
• The model requires no prior training, offering interpretability and intuitive tuning.
• Validated the model using two clinical trials: T1DEXI and a pilot study in Argentina.

Main Results:

• The model successfully quantified and differentiated between aerobic and resistance PA.
• Demonstrated the model's capacity to distinguish exercise types with contrasting influences on glucose levels.
• Confirmed the model's applicability in real-world clinical settings.

Conclusions:

• The developed HR-based state-space model is effective for quantifying and distinguishing aerobic and resistance PA in T1D.
• This approach offers a promising, explainable, and tunable tool for improving exercise management and glycemic control in T1D.
• The findings support the integration of this model into future diabetes management technologies.