Method to generate a large cohort in-silico for type 1 diabetes

Area of Science:

• Biomedical Engineering
• Computational Biology
• Endocrinology

Background:

• Artificial pancreas systems are crucial for managing type 1 diabetes (T1DM).
• Control algorithms for these systems require extensive validation using simulation tools.
• Current simulators often lack a sufficient number of virtual subjects, limiting testing variability.

Purpose of the Study:

• To develop a novel computational method for expanding virtual subject cohorts in diabetes mathematical models.
• To enhance the physiological representativeness of subjects within simulation environments.
• To improve the robustness of artificial pancreas control algorithm validation.

Main Methods:

• Utilized linear regression on parameter covariances from existing mathematical models (Hovorka's model).
• Generated larger virtual cohorts with distinct physiological characteristics.
• Employed clustering to ensure distinct glucose-insulin dynamics within generated cohorts.

Main Results:

• Successfully generated two larger virtual cohorts of 256 subjects each, significantly increasing population size.
• Demonstrated improved variability in in-silico testing through expanded, physiologically plausible cohorts.
• Validated the method's ability to create distinct, non-overlapping subject dynamics.

Conclusions:

• The proposed methodology effectively generates large, diverse virtual cohorts for T1DM simulations.
• This approach significantly enhances the number of subjects available in existing mathematical models.
• The method is adaptable for expanding cohorts in other diabetes models or scientific domains.