Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University in Skopje, 1000 Skopje, North Macedonia.
Faculty of Medicine, Ss. Cyril and Methodius University in Skopje, 1000 Skopje, North Macedonia.
Sensors (Basel). 2023 Oct 25;23(21):8697. doi: 10.3390/s23218697.
Heart rate variability (HRV) parameters can reveal the performance of the autonomic nervous system and possibly estimate the type of its malfunction, such as that of detecting the blood glucose level. Therefore, we aim to find the impact of other factors on the proper calculation of HRV. In this paper, we research the relation between HRV and the age and gender of the patient to adjust the threshold correspondingly to the noninvasive glucose estimator that we are developing and improve its performance. While most of the literature research so far addresses healthy patients and only short- or long-term HRV, we apply a more holistic approach by including both healthy patients and patients with arrhythmia and different lengths of HRV measurements (short, middle, and long). The methods necessary to determine the correlation are (i) point biserial correlation, (ii) Pearson correlation, and (iii) Spearman rank correlation. We developed a mathematical model of a linear or monotonic dependence function and a machine learning and deep learning model, building a classification detector and level estimator. We used electrocardiogram (ECG) data from 4 different datasets consisting of 284 subjects. Age and gender influence HRV with a moderate correlation value of 0.58. This work elucidates the intricate interplay between individual input and output parameters compared with previous efforts, where correlations were found between HRV and blood glucose levels using deep learning techniques. It can successfully detect the influence of each input.
心率变异性（HRV）参数可以揭示自主神经系统的性能，并可能估计其故障类型，例如检测血糖水平。因此，我们旨在寻找其他因素对 HRV 正确计算的影响。在本文中，我们研究了 HRV 与患者年龄和性别之间的关系，以便相应地调整我们正在开发的无创血糖估计器的阈值，从而提高其性能。虽然迄今为止大多数文献研究都针对健康患者和短期或长期 HRV，但我们通过包括健康患者和心律失常患者以及不同 HRV 测量长度（短、中、长）来采用更全面的方法。确定相关性所需的方法是：（i）点二项相关，（ii）皮尔逊相关，（iii）斯皮尔曼等级相关。我们开发了一个线性或单调依赖函数的数学模型，以及一个机器学习和深度学习模型，构建了分类检测和水平估计器。我们使用了来自 4 个不同数据集的心电图（ECG）数据，这些数据集包含 284 个对象。年龄和性别对 HRV 的影响中等，相关值为 0.58。与以前使用深度学习技术发现 HRV 与血糖水平之间存在相关性的工作相比，这项工作阐明了个体输入和输出参数之间的复杂相互作用。它可以成功地检测到每个输入的影响。
