Variability and Complexity of Non-stationary Functions: Methods for Post-exercise HRV.

Area of Science:

• Physiology
• Biomedical Engineering
• Data Science

Background:

• Heart rate variability (HRV) is a key noninvasive marker of cardiac autonomic function.
• HRV analysis typically requires stationary data, limiting its application in dynamic physiological states like post-exercise.
• Non-stationarity in time-series data can significantly bias common HRV indices.

Purpose of the Study:

• To investigate the impact of non-stationarity on popular HRV indices (SDNN, rMSSD, SampEn).
• To evaluate the effectiveness of polynomial detrending methods for analyzing non-stationary HRV data.
• To provide clarity on analyzing HRV in physiological research, particularly in post-exercise scenarios.

Main Methods:

• Simulated non-stationary time-series data with varying physiological relevance, variances, lengths, and trend types (linear, quadratic, cubic).
• Applied polynomial detrending (orders 1-3) and sequential differencing to simulated data.
• Calculated variability (SDNN, rMSSD) and complexity (SampEn) metrics on raw and detrended time-series.

Main Results:

• The impact of trend and data length differed between physiological and non-physiological parameters.
• rMSSD remained robust against trends in physiological data; SDNN and SampEn showed significant bias.
• Polynomial detrending improved the accuracy of HRV indices, bringing them closer to stationary values, especially with optimal fit.

Conclusions:

• Polynomial detrending offers a reliable and accurate approach for assessing HRV variability and complexity in non-stationary time-series.
• This method is particularly valuable for analyzing physiological data, such as HRV immediately following exercise.
• Proper detrending is crucial for accurate interpretation of HRV in dynamic physiological research.