Parameter Estimation in a Model of the Human Circadian Pacemaker Using a Particle Filter.

Area of Science:

• Chronobiology
• Biomedical Signal Processing
• Personalized Medicine

Background:

• Accurate real-time estimation of individual circadian clock states is crucial for optimizing medical treatments and human performance.
• Human-centric lighting and biodynamic solutions aim to support circadian rhythms by synchronizing light with the time of day.
• Mathematical models of human circadian physiology exist but require individual parameterization for improved accuracy.

Purpose of the Study:

• To investigate a novel method for enhancing the tracking of individual circadian processes.
• To improve the accuracy of circadian state estimation by individualizing model parameters.
• To demonstrate the benefits of personalized modeling over generalized approaches.

Main Methods:

• Utilized ordinary differential equations and Particle Filter signal processing to model human circadian physiology.
• Developed an estimation method that tracks state variables (phase, amplitude) and optimizes individual model parameters, specifically the intrinsic period (τx).
• Employed minimally-invasive light exposure and sleep-wake observations for parameter estimation, quantifying sensing inaccuracies.

Main Results:

• Demonstrated improved prediction accuracy by estimating individual τx values using both simulated and human subject data.
• Showed that prediction accuracy consistently improves with new observational data.
• Found that estimated τx values correlate well with individual chronotypes, similar to the correlation of τ.

Conclusions:

• Individualizing the estimation of circadian model parameters significantly enhances the accuracy of circadian state estimation.
• Personalized circadian models offer substantial improvements over traditional one-size-fits-all approaches.
• This approach holds promise for advancing personalized medicine and human performance optimization.