Background: The current significant progress in the use of heart rate variability in the solution of many diagnostic and therapeutic problems is determined by the availability of standardized methods of measurement and physiological interpretation of heart rate variability indices on the one hand and the high technological level of state-of-the-art electronic measuring equipment that is used for automatic registration and computer processing of cardio-signals.
Methods: A retrospective analysis of anonymized cardio screening results of 22,433 adult residents of 565 settlements (cities and villages) across all 20 administrative districts of the Khmelnytskyi Region (Ukraine) was conducted to find a statistically significant connection between individual heart rate variability parameters and the age of people.
Results: Primary statistical analysis and visualization showed a correlation between the selected heart rate variability parameters and the age and sex of the examined persons. The study found values of the predicted age slightly over estimation versus the actual age for very young test subjects and below estimation for elderly subjects.
Conclusion: The use of neural network computations and the modification of the algorithm through the construction of individual training samples for different age intervals, and the creation of individual ensembles of classification neural networks, therefore achieved a prediction of the age of examined persons based on the values of their time and frequency domain heart rate variability indices, with 87% accuracy for women and 85% accuracy for men in the 66-85 years age interval and at least 85% for age groups across the entire sample.
Keywords: ECG studies; cardio screening; neural networks; noninvasive diagnostic methods.
© 2017 Wiley Periodicals, Inc.