Introduction: In recent times, an upsurge in the investigation related to the effects of meditation in reconditioning various cardiovascular and psychological disorders is seen. In majority of these studies, heart rate variability (HRV) signal is used, probably for its ease of acquisition and low cost. Although understanding the dynamical complexity of HRV is not an easy task, the advances in nonlinear analysis has significantly helped in analyzing the impact of meditation of heart regulations. In this review, we intend to present the various nonlinear approaches, scientific findings and their limitations to develop deeper insights to carry out further research on this topic. Results: Literature have shown that research focus on nonlinear domain is mainly concentrated on assessing predictability, fractality, and entropy-based dynamical complexity of HRV signal. Although there were some conflicting results, most of the studies observed a reduced dynamical complexity, reduced fractal dimension, and decimated long-range correlation behavior during meditation. However, techniques, such as multiscale entropy (MSE) and multifractal analysis (MFA) of HRV can be more effective in analyzing non-stationary HRV signal, which were hardly used in the existing research works on meditation. Conclusions: After going through the literature, it is realized that there is a requirement of a more rigorous research to get consistent and new findings about the changes in HRV dynamics due to the practice of meditation. The lack of adequate standard open access database is a concern in drawing statistically reliable results. Albeit, data augmentation technique is an alternative option to deal with this problem, data from adequate number of subjects can be more effective. Multiscale entropy analysis is scantily employed in studying the effect of meditation, which probably need more attention along with multifractal analysis. Methods: Scientific databases, namely PubMed, Google Scholar, Web of Science, Scopus were searched to obtain the literature on “HRV analysis during meditation by nonlinear methods”. Following an exclusion criteria, 26 articles were selected to carry out this scientific analysis.

Hypertension is a global challenge to the public health which can easily lead to life-threatening vascular diseases unless control measures are adopted. Considering the prevalence of vascular diseases and their fatality, early detection of high-risk patients is an important problem in the present world. Heart rate variability (HRV) analysis can be an effective prognostic tool to identify the characteristics of vulnerable patients, considering its reliability in predicting sudden cardiac deaths. However, challenge lies in identifying tenuous differences in HRV between the low-risk and high-risk patients at the early stage. With this motivation, we propose a hybrid approach based on dual-tree complex wavelet packet transform (DTCWPT) and linear time domain as well as nonlinear analysis of HRV signal to extract multitudinous features. A key issue before the HRV analysis of such patients is the presence of marked amount of ectopic beats, which is addressed by using time-varying auto regressive (TVAR) technique. The features extracted from TVAR edited HRV signals are shortlisted by minimum redundancy maximum relevance algorithm for an efficient classifier modeling. Furthermore in this study, we propose to use cost-sensitive RUSBoost (CS-RUSBoost) algorithm for handling the class imbalance problem of the data. A comparative performance evaluation of CS-RUSBoost with RUSBoost, SMOTEBoost, asymmetric AdaBoost algorithm shows a superior result by CS-RUSBoost with G-mean of 0.9352 and F1 score of 0.9347.

In this paper, we propose a textural information-based analysis of scalogram image obtained from continuous wavelet transform of heart rate variability (HRV) signal to study its dynamics during meditation. In addition to features from scalogram image, visibility graph-based complexity measures and multiscale permutation entropies (MPEs) from HRV signal are used to elucidate the modulation in autonomic activity of heart during meditative and non-meditative state. Significant changes in the probability distribution of pixel intensities of scalogram image and undiminished permutation entropy at higher scales are observed during the meditative state. From the extracted features, we have selected the top-ranked features based on ReliefF algorithm and minimum threshold weight of importance set at 0.04. Considering the small sample size (12 subjects) of meditation dataset, we have employed a novel data augmentation technique based on averaging of feature sets to overcome the issue of overfitting. To examine the efficacy of the proposed technique, both the non-augmented and augmented data are applied to four different types of classifiers, namely k-nearest neighbor, support vector machine (SVM), logistic regression and random forest classifiers. Experimental results demonstrate that performance of classifiers in distinguishing meditative and pre-meditative state are much superior with the augmented data as compared to that with regular non-augmented data. Out of these classifiers, radial basis function (RBF)-based SVM results in the best performance with an average accuracy of 96.67%, sensitivity of 95.83% and specificity of 97.5%.

Paroxysmal atrial fibrillation (PAF) is a temporary arrhythmic condition which is often a precursor of permanent/chronic atrial fibrillation. As frequent PAF events may easily lead to serious heart conditions, such as stroke, arterial embolism, it is propitious to have an early warning system. To this end, we propose an automated system for early prediction of PAF events based on statistical and nonlinear features extracted from heart rate variability (HRV) signal. We compute multiscale symbolic entropy, visibility graph-based complexity measures and three time-domain measures from the HRV signal. Out of them, the independent discriminative features are selected by Wilcoxon signed-rank test and correlation assessment. Finally, the selected features are applied to support vector machine (SVM), naive Bayes, and logistic regression classifiers to obtain the best prediction model. We achieve the best prediction results using radial basis function based SVM classifier with sensitivity, specificity and accuracy of 96.15%, 97.06%, 96.80% respectively, from the segments 5-10 mins before the onset of PAF events.