A Proposal for a Data-Driven Approach to the Influence of Music on Heart Dynamics

Electrocardiographic signals (ECG) and heart rate viability measurements (HRV) provide information in a range of specialist fields, extending to musical perception. The ECG signal records heart electrical activity, while HRV reflects the state or condition of the autonomic nervous system. HRV has been studied as a marker of diverse psychological and physical diseases including coronary heart disease, myocardial infarction, and stroke. HRV has also been used to observe the effects of medicines, the impact of exercise and the analysis of emotional responses and evaluation of effects of various quantifiable elements of sound and music on the human body. Variations in blood pressure, levels of stress or anxiety, subjective sensations and even changes in emotions constitute multiple aspects that may well-react or respond to musical stimuli. Although both ECG and HRV continue to feature extensively in research in health and perception, methodologies vary substantially. This makes it difficult to compare studies, with researchers making recommendations to improve experiment planning and the analysis and reporting of data. The present work provides a methodological framework to examine the effect of sound on ECG and HRV with the aim of associating musical structures and noise to the signals by means of artificial intelligence (AI); it first presents a way to select experimental study subjects in light of the research aims and then offers possibilities for selecting and producing suitable sound stimuli; once sounds have been selected, a guide is proposed for optimal experimental design. Finally, a framework is introduced for analysis of data and signals, based on both conventional as well as data-driven AI tools. AI is able to study big data at a single stroke, can be applied to different types of data, and is capable of generalisation and so is considered the main tool in the analysis.

Introduction

While the electrocardiogram (ECG) is an electrical recording of heart activity, heart rate variability (HRV) is a measurement derived from the ECG signal that provides information about the state or condition of the autonomic nervous system (ANS) (1). Given that HRV is linked to autonomic cardiac control, it has been studied as a marker for diverse diseases (2), both physical (3) and psychological (4)—e.g., for coronary heart disease, myocardial infarction, and stroke (5). HRV has been used to observe, among other indicators, the effects of medicines (6, 7), emotional responses (8), and the impact of exercise (9). Research has shown that HRV is affected by stress, as a result of low parasympathetic activity; stress reflects a sympathetic dominance (10), while relaxation states are associated with parasympathetic dominance (11, 12). Just as it has been measured to observe the effects of various elements on the human body, HRV has also been considered in assessing the effects of sound and music (13). Studies have been undertaken to look at the effect of music therapy (14) and diverse soundscape conditions (15). Given that it is a measurement related to both ANS and the state of the heart, HRV is a crucial record able to provide information about multiple aspects that might react to music. These include changes in stress levels (16), emotions (17), anxiety (18), subjective sensations (19), and blood pressure (13).

Although ECG and HRV are widely used in research related to health and perception, the studies vary greatly in their methodology. Koelsch and Jäncke therefore affirm the need to develop high-quality systematic research to study the effects of music on the heart—implying the implementation of standardised methodologies in this type of research (13, 20, 21). From experimental design to methods of data analysis and reports, huge variation is found. Researchers have therefore proposed several recommendations about measuring HRV for experiment planning, data analysis, and data reporting (22). Nevertheless, as the recommendations generally address an global focus, it becomes necessary to develop methodologies focused on specific stimuli such as sound and music and their characteristics (21). These methodologies should be guided towards how stimuli are able to affect physiological signals such as HRV (23) and autonomic control of the heart (24). Bearing in mind these conditions, this document proposes a methodological framework to design new experiments to study the effects of such sounds as musical structures and noise on ECG and HRV signals; the need for this framework arises as a result of the difficulty of making comparisons between many of the studies carried out on this topic.

The framework reveals just how complex any research examining the effects of sound on the heart might be. Shown within the framework are as many elements to be considered as possible, however clearly not all of these elements are presented in their entirety. Instead, providing general guidelines in this type of research and understanding that the topic is very complex, the framework makes no claim to be a set of unbreakable rules. Rather, the guidelines should be adapted to each study and the possibilities featured therein, the ultimate goal being that research on this topic might begin to be standardised. Future research could thus be more productive and even more conclusive as different studies will have a better chance of being able to be compared. Similar findings, as well as points of disagreement might then be illustrated, based on firmer foundations.

In the first instance, this document puts forward some elements of experimental design, such as definition of the experimental procedure, selection of subjects and sounds according to the research aims, and elements required for data collection. It then presents a framework for data analysis, in which it introduces the selection of data analysis techniques, methods of analysis of demographic, perception, ECG and HRV data, and elements to associate stimuli with ECG and HRV signals. Finally, some recommendations for the report of outcomes are presented.

Framework Description/Details: Materials and Equipment

The framework comprises three main components (Figure 1): experimental design and procedure; data analysis; and report of outcomes. In the following sections, each is described along with the elements it contains.

FIGURE 1

Figure 1. Block diagram of framework.

Experimental Design

The first phase involves experimental design. It is necessary here to clearly establish the aim of the study, according to which the experimental procedure will be defined, and subjects and sound stimuli selected. Devices for signal/data acquisition should be selected and measurement conditions optimised. Ideally, an interdisciplinary team should be formed to design the experiments, including health professionals (physicians, psychologists), experts in data analysis, computing science and statistics, and professionals in music or music therapy, with the resulting design benefiting from research from these disciplines. It is important to consider that, depending on the study aim, the experimental design ought to allow researchers to perform causality analysis, i.e., to establish cause-effect relationships.

Defining Experimental Procedure

Definition of the experimental procedure is critical and involves carrying out as many pilots as necessary to fully adjust all procedures. Note that at this point, given the multiple design options, only general indications are provided.

Experimental Design and Procedure

In the design, it is necessary to select between existing experimental designs (25) or a mixture of several, taking account of advantages and disadvantages and the aim of the study (Table 1).

TABLE 1

Table 1. Experimental designs and their description (25, 26).

Other relevant elements to consider in experimental design are physiological and psychological measurements, baseline measurement, stimuli presentation, possible carryover effects, and consideration of a control group. All physiological and psychological variables should be defined to achieve the research aim. Baseline measurement is recommended in all observable variables and may become a useful information source in the data analysis phase. The method of presentation of stimuli should be selected in such a way that study subjects will be as affected by them as much as possible. Experimental design should consider any source of carryover effects such as learning, fatigue, habituation, sensitisation, contrast, and adaptation (25). Design should reduce these unwanted effects. Finally, the design should establish consideration of inclusion or exclusion of a control group, based on the aim and hypothesis of the study, and conditions or limitations in resources and in the sample available.

It is good practise to consider a control group. The control group should not be in a state of silence and subjects in this group should receive a type of stimuli (13) different to that of the study group(s). Where a control is not possible, effects could be established in respect of subject baseline measurements.

System Selection

Subject Selection

Given the study goal, the research population is selected (children, young people, men, women, with a particular disease or in good health) and inclusion and exclusion criteria (39) defined. A general flow process is proposed below (Figure 2):

FIGURE 2

Figure 2. A general flow diagram to select and apply inclusion and exclusion criteria.

Regardless of application of pre-determined inclusion and exclusion criteria, the following steps are proposed:

1. Selection of subjects according to age range and gender.

2. Definition of whether the research concerns musicians/music students or non-musicians, or a mix. Note that perception of music and noise in musicians could produce an outcome bias, as it may differ from that of non-musicians (40– 42).

3. Assessment of physical and psychological condition through medical evaluation or tests such as General Health Questionnaire (GHQ) (43), Medical Outcomes Study 36-item Short-Form Health Survey (SF-36) (44), Instrumental Activity of Daily Living (IADL) (45), Patient Health Questionnaire (PHQ) (46), and State-Trait Anxiety Inventory (STAI) (47). These evaluations can be applied in the inclusion and exclusion of subject; moreover, if performed after experimental procedures or as part of them, they could be considered as a source of information that might be susceptible to analysis as a response to stimuli or experimental procedures.

4. Selection of subjects based on their health conditions. According to the previous evaluation, healthy subjects or subjects with a specific disease should be chosen; in the latter case, it is essential to control the type of medicine taken, as some medicines could affect outcomes unexpectedly.

5. Providing subjects with all necessary indications before the experiments. Indications might include avoiding consumption of heavy meals, psychoactive substances, alcoholic beverages, stimulants, caffeine, and tobacco during at least 24 h before experiments or data acquisition. These substances could affect experimental measurements since they could produce unwanted effects both in physiological (heart rate, blood pressure) and psychological variables (anxiety) (48). Moreover, refraining from the practise of sport and exhaustive exercise is suggested between 24 and 48 h before data acquisition (49–51). Finally, on compliance with given instructions, subjects may be included for participation in the experimental phase.

Selection and Conditioning of Stimuli

It should be decided if both music and noise are to be used. Nevertheless, where it is important to consider more than one type of stimuli, the use of an acoustical control stimulus is recommended (13). If the study is linked to music, music with lyrics and stimuli with verbal content should be avoided unless the experimental design requires it. Verbal content could produce additional effects (52), deviate outcomes or perhaps overcome effects produced by stimuli with no verbal content (52, 53). A general flow process is proposed to select sound stimuli (Figure 3).

FIGURE 3

Figure 3. A general flow diagram to select the sound stimuli.

Regarding the selection of music, it is first necessary to define if the research is to involve simple structures—scales, simple melodic lines, harmonic intervals, chords and simple progressions—for an in-depth analysis of the bare essence of music, or more complex structures—complex melodic lines, chord progressions, sections or complete songs or musical pieces—for a more general perspective. Having thus defined musical structures, it should be established whether these will be synthetic or natural. Where design permits, the use of synthetic sounds is recommended, since is possible to exert more control over several stimuli characteristics—frequency content (timbre), homogeneity in the range of sounds, and precision in playback. Synthetic sounds are also suggested if the study focuses on more profound or specific elements of music. In contrast, natural sounds might be preferred if the research has a more general approach.

Apropos of sound synthesis, it is possible to select between several methods according to sound type (54). For instance, to synthesise pitched sounds, it is recommended to use additive synthesis, concatenative/granular synthesis, frequency modulation synthesis, or oscillators (55). Alternatively, for acoustic instrument sounds, wavetables, waveguides/physical models, concatenative/granular synthesis, or additive synthesis (55) are encouraged. For textures and soundscapes, concatenative/granular synthesis, linear predictive coding, stochastic and wavelet-based methods (55) are favoured.

Concerning noise selection, noise sounds can be chosen principally by considering their frequency content, selected from six main types (56): brown, pink, white, grey, blue, and purple. Selection is made using frequency bands. For instance, to study low frequencies, pink or brown noise should be selected; for higher frequencies, blue or purple noise, while white or grey noise is suitable when considering all of the frequencies. Once sounds are selected, stimuli are presented in at least standard CD quality, i.e., sample rate of 44.1 or 48 KHz and 16 bits per sample (57). Similarly, synthetic sounds should meet sound/acoustic research standards, such as ISO 226: Acoustics—Normal equal-loudness-level contours (58); ISO 16: Acoustics—Standard tuning frequency (Standard musical pitch) (59); and ISO 266: Acoustics—Preferred frequencies (60). If there is no focus on effects of sound level, volume, intensity, or loudness, the loudness of all stimuli should be normalised, applying methods such as ReplayGain (61); the sound equivalent level in dB should be the same for all the considered stimuli. Furthermore, it is vital to consider sound levels according to the dose of sound energy. Following the WHO-ITU standard: Safe listening devices and systems (62) is advised.

Data Collection

Consent Form

All procedures ought to be conducted under the Helsinki Declaration, where it should be ensured that subjects will not be at risk. A consent form with a description of the research and experimental procedures must be approved by a competent ethical committee and the form presented to be voluntarily signed by each subject before beginning the experimental process.

Measurement Conditions

For data acquisition, it is important to ensure study subjects are as comfortable as possible. In this way, outcomes are much less likely to be influenced by tiredness or discomfort. Maintain light exposure and influence of external sounds as low as possible. Subjects should be in a comfortable posture, e.g., in a supine position or sitting in a comfortable chair, to be more focused on the proposed task(s). They might also reduce body movement, improving the process of physiological signal acquisition. It is important to emphasise that the posture of the subjects needs to be standardised (supine or sitting) since gravity influences the baroceptors, affecting the HRV and the analysis that could be carried out.

Elsewhere, subjects should be alone during experiments. Where procedures allow, one subject per room is preferable, to avoid dealing with conditions involving more subjects. Researchers themselves should stay out of the experiment room. Where monitoring the subjects is necessary, remote use of a video camera is suggested.

Physiological and Psychological Measurements

For physiological measurements, in addition to the variables ECG and HRV, it is recommended to maintain a record of other variables likely to change. The acquisition of several electroencephalographic (EEG) channels is invaluable, as well as electrodermal activity (EDA) or galvanic skin response (GSR) measurement. Measurements of blood pressure (BP), respiration, and photoplethysmography (PPG), likewise. Measurement of psychological and perception variables (13) is also recommended; the STAI test, and recording perception using scales of valence, arousal, and dominance; to measure emotion, non-verbal assessment is recommended—Geneva emotion wheel (63) or Self-assessment manikin (SAM) where subjects can evaluate valence, arousal, and dominance on a pictorial scale (64); while a questionnaire on musical preferences of subjects and their listening habits (13) is recommended.

Devices for Signal/Data Acquisition

Hardware implementation enables measurement of the variables of interest. Having defined the physiological variables to be acquired, the devices that obtain the best quality measurements should be selected. Devices should also be selected to ensure the safety of study subjects and all personnel involved. All hardware devices should therefore comply with or be based on elements that comply with standards such as International Standardisation Organisation (ISO) or International Electrotechnical Commission (IEC). Devices or components should thus be made based on IEC 60601 for Medical electrical equipment: General requirements for basic safety and essential performance (65, 66); and ISO 13485: Quality management for medical devices (67).

Having ensured the safety of the subjects, the minimum sample acquisition rate (SAR) is determined, taking into account the frequency content of the physiological variables to be measured. According to the Nyquist principle, a sampling rate is required of at least twice the highest frequency of the signals to be recorded (68). An acquisition system with a minimum sample rate is therefore recommended based on the bandwidth of the signals (69): 250 Hz for ECG; 200 Hz for EEG; 10–100 Hz for EDA (70); 100 Hz for BP; 25 Hz for respiratory movements; and 1 Hz for PPG.

A final consideration regarding hardware implementation relates to standardisation of the sample rate. If the study requires a multisignal acquisition, all signals should be acquired with the same sample rate. In the event that signals are acquired with different sample rates, then in order to perform comparative analysis all of them must be resampled at the same sample rate. Finally, the acquisition system needs to be synchronised with the presentation of stimuli in such a way that it is possible to know the concordance between the presented stimuli and the acquired signals.

Presentation of Stimuli

A double-blind study design is recommended (13), in which sounds should be presented using headphones, preferably with noise-cancelling in order to reduce possible effects of external sounds. Another possibility is the consideration of a triple-blind design, in which a blind statistical analysis might be applied to complement an AI analysis. It is suggested to consider stimulus duration of at least 10 s; this interval of time is recommended since it allows the establishment of a pattern behaviour in some aspects of the HRV analysis (22, 71). The inclusion of a period of silence between successive stimuli is suggested, to reduce a carry-over effect from the previous stimulus. The duration of this period of silence should be set, considering the general experimental design: the total duration of the experiment should be as short as possible while enabling the stated objective to be achieved. Shorter experiments are preferred to lengthier ones to reduce fatigue in the study subjects, minimising the impact of tiredness in the latter part of the experimental phases. Where possible, it is advised that study subjects keep their eyes closed as they listen to stimuli during the experiment; the use of a mask to cover the eyes would assist in avoiding the influence of visual stimuli on the measurements. This would also favour the capture of electroencephalographic signals as eyelid movements are reduced, minimising a source of noise.

Baseline Measurement

Once informed consent has been signed, and before beginning the experimental procedures, a period of at least 15 min is suggested to allow all physiological variables to stabilise. After this period, a record of physiological variables should be performed to establish the baseline. Baseline measurements must be carried out under the same conditions as the rest of the experiment in order to compare outcomes. The baseline is used to determine if there was an influence of stimuli, in relation to it.

Data Analysis

This section is focused on the more recent techniques of data analysis, specifically techniques of artificial intelligence: machine and deep learning. These types of technique have shown significant potential in data analysis and modelling. They are particularly useful and present advantages in comparison with traditional techniques when large time series with many variables are considered. Their use is encouraged since they can deal with a significant number of features. As shown in this paper, research on this topic has several sources of features or data; features can be extracted from signals such as ECG, HRV, and other physiological data. Moreover, data from perception, psychological variables and demographic data might be available. Analysis of sound stimuli would provide more features; with music stimuli, among other features it is possible to take account of pitch, tempo, loudness, and melodic and harmonic relations. Finally, the analysis of different subjects according to age or gender groups should be undertaken carefully during all stages of data analysis since it is important to bear in mind the physiological differences between them.

Selection of Analysis Techniques

It is mostly necessary to use some kind of software to implement the experimental procedure (Table 2), to implement psychological tests or questionnaires with demographic information as well as to get information about listening habits or musical preferences. Additionally, software could be used to control the flow of actions in the experimental process—to implement an audio player with selected stimuli while at the same time presenting a graphical interface with which to obtain the perception scores of subjects.

TABLE 2

Table 2. Software to analyse data and implement experiments in the behavioural sciences.

Demographic and Perception Data Analysis

Demographic data are related to measurements of the size and composition of a particular population. This data is usually made up of descriptive information, including gender, age, marital status, household composition, ethnic background, state of health, education and training, employment status, income, and urban and rural residence (85). It is advisable to complete this information with preferences related to music knowledge and the listening habits of study subjects (86). Equally, it could be complemented with information about the subjects' perception or reported emotion.

Commonly, demographic data could be classified as nominal, ordinal, interval, or ratio. Its analysis might be performed with a descriptive or inferential character. Estimation of percentages, measures of central tendency and dispersion could be computed to enable a descriptive study (85). Measures of central tendency could be determined through arithmetic, geometric and harmonic means, median, and mode; measures of dispersion could be performed by means of variance, standard deviation, and quantiles. The relationships between groups or population characteristics could be studied if an inferential approach is sought (85). These relationships could be studied with tools such as correlation and regression. A review of the book “Methods of Demographic Analysis” is recommended for a complete overview of demographic data analysis tools (85).

ECG and HRV Analysis

This section of the text presents the general methods used to process ECG and HRV data. However, it does not include the analysis of other physiological signals since it is not within the scope of this document. If the study to be undertaken includes other biosignals, revision of some documents related to EEG (87), EDA (64, 88), PPG (89), and BP (90) is suggested. This section therefore presents especially information related to digital signal processing, since ECG and HRV are generally recorded as digital signals. Additionally, AI techniques such as machine and deep learning are described, since they are considered as the main analysis tools in this proposal.

Introduction to Methodology of Analysis Using AI

- Digital signal processing

A signal is a variable phenomenon that changes with time (though it may vary with another parameter, such as space) and can be measured (91). Signals are processed through different tools or methodologies, which include statistics and digital signal processing (DSP). DSP is an instrument composed of a set of numerous mathematical tools designed for extracting, enhancing, storing, and transmitting useful information from a signal (92).

In many biomedical and bioengineering applications, data are collected using polls, standardised tests, medical or laboratory exams, and sensors. Once data is collected, it is analysed through different stages such as data pre-processing, feature extraction and selection, and data modelling (Figure 4) (93). In biomedical signal processing, pre-processing is commonly regarded as signal filtering; feature selection is also related to dimension reduction; and data modelling is associated with classification and, to a lesser extent, prediction. The modelling stage is also complemented by expert knowledge and metadata. These stages aim to perform the detection, prediction, and decision-making tasks, in which AI techniques such as machine and deep learning are used as part of digital signal analysis. In this regard, within the deep learning gamut, transfer learning is considered a very useful and powerful technique.

FIGURE 4

Figure 4. Stages of data and signal processing (93).

- Artificial intelligence

Artificial intelligence, or AI, is focused largely on tasks such as problem-solving and learning. For instance, a common application is the development of computational algorithms to distinguish between healthy and unhealthy subjects. AI seeks to develop algorithms to sort out real problems more efficiently than can humans (94) and has applications in several areas; however, in the context of this document the main implementations are related to expert systems, data mining or knowledge extraction, and knowledge representation (95).

Approaches based on AI present advantages compared to traditional analysis. AI might be used to study big data as a whole (96), it has the ability to “learn” features from a large volume of data (97), and it could also be utilised to study the influences of many features, initially, without having an idea which feature might be the most important one. AI also has the potential to study different types of data—features, time series, signals, and images. In general AI systems work without the need for human experts (expert systems: after training or development) (98). Other benefits of AI are related to efficiency, accuracy, and precision in analysis (99); competence to identify, classify and extract features from complex, high-dimensional and noisy data (100); the capability of generalisation; robustness; and the possibility to integrate expert knowledge (101).

Methodology of Analysis

Preprocessing: Signal Conditioning

- Signal detrending

Some non-stationary signals such as EEG, EMG, and HRV signals (102) are usually pre-processed in such a way that they could be considered as stationary signals. Detrending techniques are implemented so that the mean and standard deviation of the signals are almost constant (103). This procedure is widely used in treatment of signals such as EEG and ECG, and sometimes in HRV. Another way to deal with non-stationary signals is to analyse them by taking small parts in which they could be considered as stationary (103). Baseline wander is produced by different sources of low-frequency noise (104), among them a deficient contact of the electrodes to the skin and some effects of respiration as changes in impedance (105). Other techniques for detrending biosignals include cascaded moving average filter (105), quadratic variation reduction (104), wavelet transform (106), and adaptive filtering (107).

- Signal filtering

After removing baseline wander, a filtering stage is commonly performed to remove the noise of medium or high frequencies from several sources such as physiological variability, environmental noise or interference, transducer artifact, and electronic noise (103). A conventional filter stage is implemented to cancel interference from the power-line; in this case, it is possible to use anything from a notch filter with a cut-off frequency of 50 or 60 Hz to an adaptive filter (108). Concerning ECG signals, since ECG is affected by diverse noise sources (109, 110), the filtering process is a crucial stage that will influence systems analysis stages such as feature extraction and classification. Some techniques used are FIR and IIR filters (111), least mean square filters (112), wavelet transform (113), and Kalman filtering (114). Finally, after signal filtering, the signal to noise ratio (SNR) (115) is often used to assess the performance of the filtering stage.

- R-peak/QRS-complex detection: HRV computation

An important step following signal pre-processing is segmentation of R-peaks or QRS complexes. This segmentation is necessary in HRV computation since the analysis requires to know with precision the moment (occurrence) of each R-peak or QRS-complex (116). The element most used for detecting R-peaks is perhaps the Pan-Tompkins algorithm (117), other methods being linear regression algorithm (118), adaptive Hermite functions (119), adaptive bandpass filters and wavelet analysis (120), and CNN (121, 122). Once the R-peaks or QRS complexes are segmented, it is possible to compute the HRV signal. Thus, HRV is related to the variation between consecutive R-peaks, namely RR or NN intervals (116). HRV is normally presented as a time series expressed as the difference in milliseconds between successive R-peaks or QRS complexes.

ECG Feature Extraction

Some important information in ECG is represented by characteristics which are named features, and they are used for several purposes such as ECG filtering (123), ECG quality assessment (124), and disease classification (110). Many tools have been used to extract features from ECG signals, such as wavelet transform (113), PCA (125), statistics (126), analysis-based autocorrelation (127), Fourier transform (128), singular value decomposition SVD, variational mode decomposition VMD (129), Hilbert transform (130), and morphological methods (131). Several features have been extracted (132), for example, morphological features (commonly P, Q, R, S, T, and U waves) (121, 126, 133), statistical features (energy, mean, standard deviation, maximum, minimum, kurtosis and skewness) (126), wavelet features (coefficients and metrics extracted from continuous WT, Dual-Tree complex WT, tunable Q factor WT, flexible analytic WT and dyadic DWT) (134–136) and others, such as Lyapunov Exponents (137), the ratio of power spectrum (138), power spectral density (138), Kolmogorov Sinai entropy (137), and Kolmogorov complexity (137).

In addition to these elements, other tools could contribute to new approaches in ECG signal processing and analysis; e.g., fractal analysis. Thus, fractal geometry and multifractals allow analysing and processing complex shapes and signals (139). It is useful to look at the fractal or chaotic nature of ECG signals to inspect how the cardiac mechanism works, and so design modern approaches for analysis (140). More recent publications have shown that fractal analysis has contributed to the study of heart activity (141, 142). In addition to analysis with fractals, it is pertinent to bear in mind that this analysis can be used for the feature extraction process (143).

HRV Feature Extraction

HRV parameters are normally extracted from four different approaches—geometrical/non-linear analysis, fractal analysis, time domain, and frequency domain (Table 3) (150).

TABLE 3

Table 3. Analysis of HRV signals.

- HRV and time-frequency domain

Frequently, continuous wavelet transform (CWT) is used as a time-frequency representation of ECG signals (151). However, this tool is not often used in analysing HRV (152); instead, Lomb-Scargle periodograms have occasionally been used (153). As such, this technique is proposed as an element to link HRV analysis to new techniques such as those related to deep learning. Several tasks could be performed through CWT by using algorithms such as convolutional neural networks.

Feature Selection—Dimension Reduction

After feature extraction, feature selection is required; here, several useful techniques have been used: wrappers (154) as a recursive feature elimination method (155), filters (156) as information gain (157, 158), and embedded, such as least absolute shrinkage and selection operator LASSO (159). It is sometimes necessary to reduce the number of selected features from the last stage by selecting the principal variables that best represent the signals and suppressing the ones with redundant information (dimensionality reduction). Moreover, PCA, LDA, Fuzzy C-mean, divergence analysis, ICA, and Fisher score are the most widely used dimension reduction tools in the applications with the best performance (160). In case of ECG analysis, among the most common tools are clustering methods such as k-means and hierarchical clustering (161), matrix factorisation methods such as singular value decomposition (SVD) (131), PCA (162), LDA (163), independent component analysis (ICA) (164), and other methods such as genetic algorithms (GAs) (164) and canonical correlation analysis (165).

ECG/HRV Modellin—Classification

- Machine learning

Classification is one of the final stages in analysing ECG signals. Most research develops systems for several tasks, such as disease classification (166), patient classification (167), ECG simulation (168), and emotion recognition (169). With this aim, supervised methods such as naive Bayes (170), random forest (171), genetic algorithms (128), linear and quadratic discriminants (172), SVM (173, 174), decision trees (175), discriminant analysis (138), and ANN (173, 174) have been used. In the same way, unsupervised methods such as hierarchical clustering (161), Gaussian mixture models (176), self-organising maps (177), and kNN (178) have been used. Modern methods of deep learning (179) such as CNN (180), long short-term memory (LSTM) (181), deep neural network (DNN) (182), robust deep dictionary learning (RDDL) (183) and restricted Boltzmann machine (RBM) (184) have been implemented.

Machine learning has contributed to various elements such as detection or classification of heartbeats (185), arrhythmias (129, 186), and unexpected changes in heart morphology (187, 188). Aspects related to the diagnosis of cardiac diseases and the analysis and classification of a considerable volume of ECG recordings have been improved. Real-time analysis and ECG simulations are growing topics (189). Regarding ECG classification techniques such as support vector machine, ANN, hidden Markov model, linear discriminant analysis, naive Bayes, and hybrid methods have been implemented (190). From the literature review, probabilistic neural network and support vector machine are observed as the classification algorithms with the most accuracy, higher than 98%, for cardiac arrhythmia detection (160).

Machine learning has been used to analyse ECG data (191), while these techniques have been used to a lesser degree to analyse HRV data. The extreme gradient boost (XGBoost) algorithm was thus implemented to find a connexion between HRV and long-term cardiovascular (192). Some algorithms such as logistic regression, support vector machine, random forest and AdaBoost were trained to classify between healthy and pulmonary patients from their HRV data (193). In addition to classification, XGBoost has been implemented to predict cardiovascular events from HRV parameters (194). Moreover, a Q-learning algorithm was implemented to associate HRV with the avoidance of negative emotional events (195).

- Deep learning

Among new machine learning methodologies, deep learning is being used in several areas related to biomedical engineering (196), perhaps due to the fact that deep learning performs better for large, diverse datasets than standard classification and analysis tools (179). Regarding ECG analysis, deep learning has been implemented in several applications (197), among them heartbeat classification, detection of coronary artery disease, myocardial infarction and congestive heart failure, detection and classification of arrhythmia, and detection and monitoring of atrial fibrillation (179). Analysis of ECG signals using deep learning has been implemented in applications related to sleep, to classify sleep stages and to detect obstructive sleep apnea (179, 196).

Even though deep learning is a new tool, published results have shown that it is possible to model several attributes of ECG signals and extract their particular features through deep learning (179). Convolutional neural networks (CNN) are the most used deep learning algorithms in analysis of several biological signals such as EEG, EMG, and ECG (179). One example is observed in Al Rahhal et al. (180), where the authors used CNN with continuous wavelet transform for detection and classification of arrhythmia. Another example is observed in (198), where deep CNN were used to detect atrial fibrillation.

Deep learning techniques have also been used to analyse HRV signals, though less often than for ECG signals. Deep neural networks were used to measure stress levels from HRV records (199). CNNs were used in emotion recognition tasks, where HRV was integrated with multiple physiological signals (200). An LSTM approach was implemented to identify sleep-wakes from acceleration data and HRV (201) and to detect congestive heart failure (202).

Association Between Sound Stimuli and ECG/HRV Signals

Having observed the advantages of using AI, and in particular the benefits that machine learning and deep learning bring to cardiovascular medicine (203) on including ECG and HRV analysis, it is clear these tools represent a valid method to associate these signals to sound stimuli. As such, it is proposed to consider classification tasks as an element of correlation between signal features, in a machine learning approach, or time-frequency representation of the signals, in a deep learning approach, and stimuli characteristics as target classes (Figure 5). In this approach, high performance in the classification process represents a good correlation between the signals measured and the stimuli presented. The association between stimuli and their effects on the heart should be carried out by means of HRV/ECG comparison between participants exposed to the same auditory stimulation.

FIGURE 5

Figure 5. Methodology for the association between sound stimuli and ECG/HRV signals.

Correspondingly, as described above, following signal acquisition, conditioning of the signal is necessary. Depending on the power source, applying a notch filter with a cut-off frequency at 50 or 60 Hz may well be required. In any event, some pre-processing techniques will also require to be applied (filtering techniques might be applied, for example, to remove the noise from different sources). The techniques should be chosen depending on the sources of noise present during the experiment as well as the acquisition system employed. Once signals have been suitably conditioned, if few data are found, a data augmentation process may be necessary. Augmentation may be required on applying machine learning techniques, whereas with deep learning techniques, it is almost always necessary. In contrast, the possibility exists of applying transfer learning techniques (204), where it is possible to deal with restricted data. Augmentation techniques suited to ECG signals might include noise addition (205), wavelet-based shrinkage filtering (206), and signal windowing or segmentation (206) with or without overlap. After signal pre-processing, if the research is focused on HRV, this should be extracted from the captured ECG signal. R-peaks must be segmented and subsequently, computing the time difference between successive peaks is required. With both ECG and HRV signals conditioned, the next phases of the study can begin.

At this point, the method for analysing the data is selected determined by the main aim of research and the amount of data available (machine or transfer learning if data are scarce; deep learning techniques where data is more abundant). If determining which data features or elements most affect outcomes is relevant (i.e., an exploratory, descriptive, or explanatory approach), machine learning techniques should be applied. Where the focus is on the results (i.e., without considering their provenance, in a relational, or applicative approach), transfer or deep learning tools apply. It is next necessary to prepare the data before applying it; if a machine learning technique was selected, some techniques of feature extraction from the data are likely required. After this procedure, where there is a large set of features, applying a feature selection method is recommended. If a transfer or deep learning technique was selected, applying a transformation of data into a two-dimensional representation—a time-frequency or autoencoder algorithm transformation—might be required (207).

The dataset should next be split into sets to train and test the selected algorithm; a validation set may be included. The training set should always be the larger set, usually 60–80% of total data. The remainder is assigned to the test set, or distributed between validation and test sets. The k-fold cross-validation method can reduce overfitting (208). As data is split into training sets, the algorithm to analyse data is chosen. If this algorithm is supervised, the target class could be established from different characteristics of the sound stimuli: an association between the training data and the selected characteristics could thus be performed. Following algorithm selection, the algorithm structure must be selected taking training data and research aims into account.

Performance Evaluation

It is then time to train the algorithm using the training set or k-fold cross-validation method. Performance evaluation with the test set follows, measured using such elements as the confusion matrix (209), and derived metrics such as F1 score, accuracy (208), and receiver operation characteristic (ROC) curve (210). Although these metrics are used widely, most have drawbacks (13). F1, for instance only takes into account positive classes, carrying bias by disregarding negative classes (13). Similarly, AUC does not take a classification threshold, which classifiers generally require (13).

Among other possible metrics, the Kappa coefficient (211) and Matthews correlation coefficient (MCC) have emerged strongly (212). By considering all elements of the confusion matrix, these provide a better idea of the general performance of machine learning algorithms (13, 213). The Kappa coefficient can take values between 0 and 1, for which Landis and Koch have proposed a scale for interpreting the coefficient (211), divided into intervals 0 to 0.20, 0.21 to 0.40, 0.41 to 0.60, 0.61 to 0.80, and 0.81 to 1.00 that indicate slight, fair, moderate, substantial, and almost perfect agreement, respectively. MCC is usually interpreted as a correlation coefficient because it can produce values between minus one and one (214). A value of one relates to a perfect classification; minus one indicates discrepancy between observations and prediction; and zero indicates an uncorrelated or random prediction (13). It must be noted that MCC can be used even with imbalanced datasets (214).

Given the framework aims, MCC is highly recommended for validating the performance of the selected machine learning tools. It can equally provide a quantitative measurement of the correlation between sound stimuli and the responses observed in the variables, since it can be interpreted as a correlation coefficient.

Metric for Applications Related to Harmonic Musical Intervals and Noise

In addition to using MCC, developing new metrics for each specific application is recommended (215, 216). One possibility involves implementing a cost matrix. In classification applications, the cost matrix provides information about the cost of wrong classifications. Cost is incremented with each instance incorrectly classified. This metric can consequently describe system performance and further be adapted to each specific application. Unlike most standard metrics, a cost matrix provides the fullest idea of right and wrong classifications.

If performance is evaluated as not sufficiently high, improvements are made to the system. Adjustments to the algorithm structure could be performed, or a different algorithm employed. Finally, when performance is high enough, data analysis can proceed. An example of application of this framework is again found in “Assessment of heart rate variability and heart response to harmonic music interval stimuli using a transfer learning approach.”

Building a cost matrix as evaluation metric is proposed, based on the following rules:

1. A hierarchical order must be established between classes to be classified, dependent on each specific case: e.g., in classifying harmonic musical intervals, a hierarchical order might be determined by degree of consonance or dissonance. The established order will allow assigning classification costs. Sorting classes into adjacent orders will cost less than sorting between more hierarchically distant classes.

2. When seeking to classify more than one general class, a hierarchy must be established for each general class: e.g., for two general classes A and B, with subclasses A1, A2, B1, and B2, two hierarchies are established, one for general class A and one for B. Maximum cost to pay is set equal to 1 (C = 1). This cost will be distributed among all possible classification options for each class. For more than one general class, a sub cost (SC) is defined for each class. This sub cost will be established dividing the maximum total cost by the total number of classes.

$SC=\frac{C}{\#GeneralClasses}$

3. SC is then distributed among all subclass elements. If an element belonging to a subclass is classified within another subclass, a higher cost will be paid than if classified as an element of its same subclass. Taking the previous example, if an element in class A1 is classified within class A2, the cost to be paid will be less than if classified within any class of general class B, B1 or B2. In the case of the SC of different classes, this ought to be distributed equally among classes. Thus, in the case of classification of elements of general class A, classification of these subclasses as elements of subclass B will cost SC/2 where 2 corresponds to the number of classes, B1 and B2, of subclass B in this example.

4. For the SC of classes corresponding to the same subclass, the cost of correctly classified classes will have the minimum value, corresponding to zero. To distribute SC between erroneous classifications within the same subclass, three methods are proposed:

$\begin{array}{l}\text{ n}=\text{number of subclasses}\\ \text{SC}=\text{sub cost}\end{array}$

a. Linear distribution: in this case a minimum cost is assigned, given by Equation 1, and the costs corresponding to the other classes according to Equation 2:

$\begin{array}{ll}Cos{t}_{Minimum}=\frac{SC}{{\displaystyle {\sum }_{i=1}^{n-1}i}}& (1)\end{array}$

$\begin{array}{ll}S{C}_i=\{\begin{array}{c}0,i=1\\ \left(i-1\right)\ast Cos{t}_{Minimum},i=\left[2,n\right]\end{array}& (2)\end{array}$

b. Distribution based on the inverse-square law. Since various physical phenomena—gravity, electrostatics, and the radiation of light and sound—vary inversely to the square of the distance, allocation of costs is proposed taking this approach into account. Equation 3 is proposed to assign SC values:

$\begin{array}{ll}S{C}_i=\frac{1-\frac{1}{i^2}}{\#GeneralClasses\ast {\displaystyle {\sum }_{j=1}^{n}1-}\frac{1}{j^2}},i=\left[1,n\right]& (3)\end{array}$

c. Distribution based on Gompertz function (217). As this function is widely used in the field of biology (description of growth of plants, animals, cells, bacteria, etc.), it is proposed as a tool for cost allocation. To assign SC values, Equation 4 is proposed:

$\begin{array}{l}S{C}_i=\frac{e^{-\beta e^{-\gamma i}}-e^{-\beta e^{-\gamma }}}{\#GeneralClasses\ast {\displaystyle {\sum }_{j=1}^n{e}^{-\beta e^{-\gamma i}}}-e^{-\beta e^{-\gamma }}},\\ i=\left[1,n\right]& (4)\end{array}$

In this case β and γ represent constants that modify horizontal displacement of the function and growth rate, respectively.

Linear distribution is recommended where the cost of wrong classifications does not have great consequences, i.e. when the cost to pay for wrong classifications is not very high. This method could be used in classifying the same type of stimulus, but would be recommended in classifying heart disease and normal signals. Distribution based on the inverse square law is recommended for stricter applications, e.g. medical applications where a high cost is paid for wrong classifications. Finally, use of the Gompertz-based distribution is generally recommended. This function can be adapted according to each specific application, through constants β and γ and can be used for both flexible and strict cost applications.

Once established, the cost matrix is multiplied by the confusion matrix to determine the classification cost matrix. All values in the classification cost matrix are then added together to obtain the total cost. Finally, to standardise the value between 0 and 1, total cost is divided by maximum possible cost. Maximum possible cost is determined considering the case where each instance of the classes is classified as the class with the highest cost in the cost matrix.

Framework Validation

Following implementation, framework validation is a vital step (218), to determine whether it was well-adapted to the study in question and used to the full, or might be improved. Just as it was suggested above that an interdisciplinary team carry out the experiment design, so a multidisciplinary team is recommended for framework validation, which should be conducted from the viewpoint of each discipline involved in the research. The framework is evaluated according to application requirements and its implementation is then optimised under research conditions (219). Validation should confirm that the framework is within ethical considerations, all its procedures ensuring high quality scientific practise and should further assure that results can be compared with outcomes of other research performed under similar conditions. Such validation increases possibilities to obtain meaningful data. In this regard, some critical characteristics of testing and clinical validation presented in Beniczky and Ryvlin (220) were adapted to this framework as follows (Table 4):

TABLE 4

Table 4. Stage of framework validation.

Validation should assess the quality, reliability, robustness, and consistency or reproducibility of the outcomes (221). As a validation protocol of this framework, carrying out the following steps is suggested (Figure 6):

1. Present the adapted framework to an expert committee.

2. Determine if the framework and the experimental design are correct. Otherwise, it is necessary to perform adjustments and go back to step 1.

3. Cheque if all experimental materials are working well: devices, stimuli, laboratory, or experimental place in optimal conditions.

4. Develop a set of pilots or validation experiments.

FIGURE 6

Figure 6. Framework validation protocol.

With the outcomes of the pilots, confirm if the hypothesis is verifiable and assess the quality, reliability, robustness, and consistency or reproducibility of outcomes. If this evaluation does not produce good results, it is necessary to perform adjustments and go back to step 1.

One method of verifying the validation protocol is by means of a checklist. Constructing a checklist of items included in the research is recommended (Table 5).

TABLE 5

Table 5. Checklist of research elements.

This framework was inspired in the studies reported in the literature and also in our own research and experience in the fields of artificial intelligence, perception, music, and health. One example of the application of this framework can be observed in our most recent publication “Recognition of valence judgments in music perception using electrocardiographic signals and machine learning” (222)¹.

Some Suggestions to Present a Report of Outcomes

On completion of result analysis, it is essential to provide a report as a scientific paper regarding the findings of the research. Initially, it is necessary to make clear the objective of the study and the methodology with which to achieve it. Provide as much information as possible about the methodological procedure, describing the most critical aspects of the experimental procedure. Stimuli, method of presentation of stimuli, and physical conditions of the place of experimentation should also be described, together with the considered population and the methods employed to acquire data.

It is furthermore relevant to provide sufficient information about the methods used to analyse the data. Outcomes of the analysis should be presented clearly and precisely. Finally, it is essential to demonstrate the significance of the results and their relevance. A possible additional element is the presentation of limitations in the research and possible ways of improving similar research in the future. Wherever possible, it is a good idea to present new conceivable ideas to deal with new studies related to the research topic or the study aim. If it is desired to obtain a complete reference about research in music and its written reports, review of the book “Music in words: a guide to researching and writing about music” is recommended (223).

Other useful tools to bear in mind for reporting outcomes are: STROBE statement (STrengthening the Reporting of OBservational studies in Epidemiology) (224, 225); CONSORT statement (Consolidated Standards of Reporting Trials) (226, 227); SQUIRE stands (Standards for QUality Improvement Reporting Excellence) (228, 229); STARD initiative (Standards for Reporting of Diagnostic Accuracy Studies) (230, 231); and STREGA initiative (STrengthening the REporting of Genetic Association Studies) (232, 233).

Discussion

A methodological framework to design new experiments to study the effects of musical structures and noise on ECG and HRV signals was presented. This framework has three main components—experimental design and procedure, data analysis, and report of outcomes—and is able to be generalised to research related to other types of sound. Its objective is to provide guidelines to standardise new studies and thus facilitate comparison between study outcomes. AI techniques were considered as the main analysis tool as AI has recently revealed its advantages in studying different types of data, showing great capacity to deal with big data at a single stroke with efficiency, accuracy, and precision in analysis. Moreover, AI allows extracting features from complex, high-dimensional and noisy data, providing a high capability of generalisation and robustness.

Considering that new trends in AI suggest developing new metrics for each specific application, a cost matrix was introduced as an evaluation metric. Additionally, validation of the framework is presented, for which a clinical validation was adapted. As future work, this is a new proposal that needs to be validated in future experiments; moreover, it requires a series of studies with different types of patients and different sound techniques in order to be improved. This framework was developed to contribute to quality improvement of research associated with sound and music. This version constitutes an initial perusal of the perception field; it is expected that researchers can contribute with their own vision and experience to develop new, enhanced versions of the present framework.

Data Availability Statement

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

Author Contributions

EI-Á: conceptualisation, methodology, literature review, writing—original draft, and writing—review and editing. HL-C and FM-B: supervision and reviewing. LN: reviewing and editing. RV-C: conceptualisation, methodology, supervision, and reviewing and editing. All authors contributed to the article and approved the submitted version.

Funding

This work was supported by Universidad del Cauca, Universidad del Valle, and Colciencias, Colombia (Funding call No. 727 of 2015). The sponsors (did not have direct participation) had no role in the study design, data collection and analysis, neither decision to publish, nor preparation of the manuscript.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher's Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Acknowledgments

We are especially grateful to Colin McLachlan for suggestions relating to the English text.

Footnotes

1. ^Available in http://ceur-ws.org/Vol-2747/paper6.pdf

References

1. Christodoulou G, Salami N, Black DS. The utility of heart rate variability in mindfulness research. Mindfulness. (2020) 11:554–70. doi: 10.1007/s12671-019-01296-3

CrossRef Full Text | Google Scholar

2. Schwerdtfeger AR, Schwarz G, Pfurtscheller K, Thayer JF, Jarczok MN, Pfurtscheller G. Heart rate variability (HRV): from brain death to resonance breathing at 6 breaths per minute. Clin Neurophysiol. (2020) 131:676–93. doi: 10.1016/j.clinph.2019.11.013

PubMed Abstract | CrossRef Full Text | Google Scholar

3. Kirizawa JM, Garner DM, Arab C, Valenti VE. Is heart rate variability a valuable method to investigate cardiac autonomic dysfunction in subjects with leukemia? A systematic review to evaluate its importance in clinical practice. Supp Care Cancer. (2020) 28:35–42. doi: 10.1007/s00520-019-05047-x

PubMed Abstract | CrossRef Full Text | Google Scholar

4. An H, Han JW, Jeong H-G, Kim TH, Lee JJ, Lee SB, et al. Parasympathetic predominance is a risk factor for future depression: a prospective cohort study. J Affect Disord. (2020) 260:232–7. doi: 10.1016/j.jad.2019.09.015

PubMed Abstract | CrossRef Full Text | Google Scholar

5. Ernst G. Heart rate variability. Circulation. (2014) 93:5. doi: 10.1161/01.CIR.93.5.1043

CrossRef Full Text | Google Scholar

6. Niehoff J, Matzkies M, Nguemo F, Hescheler J, Reppel M. The effect of antiarrhythmic drugs on the beat rate variability of human embryonic and human induced pluripotent stem cell derived cardiomyocytes. Sci Rep. (2019) 9:14106. doi: 10.1038/s41598-019-50557-7

PubMed Abstract | CrossRef Full Text | Google Scholar

7. Lilleberg HS, Cichosz SL, Svensson M, Christensen JH, Fleischer J, Eide I, et al. The effect of marine n-3 polyunsaturated fatty acids on heart rate variability in renal transplant recipients: A randomized controlled trial. Nutrients. (2019) 11:22847. doi: 10.3390/nu11122847

PubMed Abstract | CrossRef Full Text | Google Scholar

8. Bornas X, Fiol-Veny A, Balle M. Differences in autonomic flexibility in adolescents with distinct emotion regulation styles during acute stress. Nonlinear Dyn Psychol Life Sci. (2020) 24:1–21.

PubMed Abstract | Google Scholar

9. Gronwald T, Hoos O. Correlation properties of heart rate variability during endurance exercise: A systematic review. Ann Noninvasive Electrocardiol. (2020) 25:12697. doi: 10.1111/anec.12697

PubMed Abstract | CrossRef Full Text | Google Scholar

10. Kim H-G, Cheon E-J, Bai D-S, Lee YH, Koo B-H. Stress and heart rate variability: a meta-analysis and review of the literature. Psychiatry Investig. (2018) 15:235–45. doi: 10.30773/pi.2017.08.17

PubMed Abstract | CrossRef Full Text | Google Scholar

11. Tyagi A, Cohen M. Yoga and heart rate variability: A comprehensive review of the literature. Int J Yoga. (2016) 9:97–113. doi: 10.4103/0973-6131.183712

PubMed Abstract | CrossRef Full Text | Google Scholar

12. Komori T. The relaxation effect of prolonged expiratory breathing. Ment Illn. (2018) 10:7669. doi: 10.4081/mi.2018.7669

PubMed Abstract | CrossRef Full Text | Google Scholar

13. Koelsch S, Jäncke L. Music and the heart. Eur Heart J. (2015) 36:3043–9. doi: 10.1093/eurheartj/ehv430

PubMed Abstract | CrossRef Full Text | Google Scholar

14. Palma S, Keilani M, Hasenoehrl T, Crevenna R. Impact of supportive therapy modalities on heart rate variability in cancer patients - a systematic review. Disabil Rehabil. (2020) 42:36–43. doi: 10.1080/09638288.2018.1514664

PubMed Abstract | CrossRef Full Text | Google Scholar

15. Erfanian M, Mitchell AJ, Kang J, Aletta F. The psychophysiological implications of soundscape: a systematic review of empirical literature and a research Agenda. Int J Environ Res Public Health. (2019) 16:93533. doi: 10.3390/ijerph16193533

PubMed Abstract | CrossRef Full Text | Google Scholar

16. Kunikullaya KU, Goturu J, Muradi V, Hukkeri PA, Kunnavil R, Doreswamy V, et al. Music versus lifestyle on the autonomic nervous system of prehypertensives and hypertensives–a randomized control trial. Complement Ther Med. (2015) 23:733–40. doi: 10.1016/j.ctim.2015.08.003

PubMed Abstract | CrossRef Full Text | Google Scholar

17. Nardelli M, Valenza G, Greco A, Lanata A, Scilingo EP. Recognizing emotions induced by affective sounds through heart rate variability. IEEE Trans Affect Comput. (2015) 6:385–94. doi: 10.1109/TAFFC.2015.2432810

CrossRef Full Text | Google Scholar

18. Chiu HW, Lin LS, Kuo MC, Chiang HS, Hsu CY. Using heart rate variability analysis to assess the effect of music therapy on anxiety reduction of patients. Comput Cardiol. (2003) 2003:469–72. doi: 10.1109/CIC.2003.1291194

CrossRef Full Text | Google Scholar

19. Chuang C-Y, Han W-R, Li P-C, Young S-T. Effects of music therapy on subjective sensations and heart rate variability in treated cancer survivors: a pilot study. Complement Ther Med. (2010) 18:224–6. doi: 10.1016/j.ctim.2010.08.003

PubMed Abstract | CrossRef Full Text | Google Scholar

20. Dumont E, Syurina E V, Feron FJM, van Hooren S. Music interventions and child development: a critical review and further directions. Front Psychol. (2017) 8:1694. doi: 10.3389/fpsyg.2017.01694

PubMed Abstract | CrossRef Full Text | Google Scholar

21. de Witte M, Spruit A, van Hooren S, Moonen X, Stams G-J. Effects of music interventions on stress-related outcomes: a systematic review and two meta-analyses. Health Psychol Rev. (2019) 36:1–31. doi: 10.1080/17437199.2019.1627897

PubMed Abstract | CrossRef Full Text | Google Scholar

22. Laborde S, Mosley E, Thayer JF. Heart rate variability and cardiac vagal tone in psychophysiological research – recommendations for experiment planning, data analysis, and data reporting. Front Psychol. (2017) 8:213. doi: 10.3389/fpsyg.2017.00213

PubMed Abstract | CrossRef Full Text | Google Scholar

23. Bretherton B, Deuchars J, Windsor WL. The effects of controlled tempo manipulations on cardiovascular autonomic function. Music Sci. (2019) 2:2059204319858281. doi: 10.1177/2059204319858281

CrossRef Full Text | Google Scholar

24. Ribeiro MKA, Alcântara-Silva TRM, Oliveira JCM, Paula TC, Dutra JBR, Pedrino GR, et al. Music therapy intervention in cardiac autonomic modulation, anxiety, and depression in mothers of preterms: randomized controlled trial. BMC Psychol. (2018) 6:57. doi: 10.1186/s40359-018-0271-y

PubMed Abstract | CrossRef Full Text | Google Scholar

25. Bordens KS, Abbott BB. Research Design and Methods: A Process Approach. 10th ed. New York, NY: McGraw-Hill Education (2011).

Google Scholar

26. Cooper HM, Hedges LV, Valentine JC. Handbook of Research Synthesis and Meta-Analysis, 3rd edition (2019), 129–468.

Google Scholar

27. Herget A-K. Well-known and unknown music as an emotionalizing carrier of meaning in film. Media Psychol. (2020) 24:385–415. doi: 10.1080/15213269.2020.1713164

CrossRef Full Text | Google Scholar

28. Cordi MJ, Ackermann S, Rasch B. Effects of relaxing music on healthy sleep. Sci Rep. (2019) 9:9079. doi: 10.1038/s41598-019-45608-y

PubMed Abstract | CrossRef Full Text | Google Scholar

29. Rogers K, Jörg M, Weber M. Effects of background music on risk-taking and general player experience. In: Institute of Media Informatics, Ulm University. Ulm: Association for Computing Machinery, Inc. (2019). doi: 10.1145/3311350.3347158

CrossRef Full Text | Google Scholar

30. Baylan S, McGinlay M, Macdonald M, Easto J, Cullen B, Haig C, et al. Participants' experiences of music, mindful music, and audiobook listening interventions for people recovering from stroke. Ann N Y Acad Sci. (2018) 1423:349–59. doi: 10.1111/nyas.13618

PubMed Abstract | CrossRef Full Text | Google Scholar

31. Heene E, De Raedt R, Buysse A, Van Oost P. Does negative mood influence self-report assessment of individual and relational measures?: An experimental analysis. Assessment. (2007) 14:86–93. doi: 10.1177/1073191106293980

PubMed Abstract | CrossRef Full Text | Google Scholar

32. Zelenak MS. Predicting music achievement from the sources of self-efficacy: An exploratory study. Bull Counc Res Music Educ. (2019) 22:63–77. doi: 10.5406/bulcouresmusedu.222.0063

CrossRef Full Text | Google Scholar

33. Faus S, Matas A, Elósegui E. Music and regaining calm when faced with academic stress. Cogent Arts Humanit. (2019) 6:1634334. doi: 10.1080/23311983.2019.1634334

CrossRef Full Text | Google Scholar

34. Liu M-H, Zhu L-H, Peng J-X, Zhang X-P, Xiao Z-H, Liu Q-J, et al. Effect of personalized music intervention in mechanically ventilated children in the PICU: a pilot study. Pediatr Crit Care Med. (2020) 21:e8–14. doi: 10.1097/PCC.0000000000002159

PubMed Abstract | CrossRef Full Text | Google Scholar

35. Sepúlveda-Vildósola AC, Herrera-Zaragoza OR, Jaramillo-Villanueva L, Anaya-Segura A. Music as an adjuvant treatment for anxiety in pediatric oncologic patients. Rev Med Inst Mex Seguro Soc. (2014) 52:S50–4.

PubMed Abstract | Google Scholar

36. Gebhardt S, Dammann I, Loescher K, Wehmeier PM, Vedder H, von Georgi R. The effects of music therapy on the interaction of the self and emotions—An interim analysis. Complement Ther Med. (2018) 41:61–6. doi: 10.1016/j.ctim.2018.08.014

PubMed Abstract | CrossRef Full Text | Google Scholar

37. Kingsford-Smith A, Evans P. A longitudinal study of psychological needs satisfaction, value, achievement, and elective music intentions. Psychol Music. (2019) 25:143–60. doi: 10.1177/0305735619868285

CrossRef Full Text | Google Scholar

38. Taxis JC, Rew L, Jackson K, Kouzekanani K. Protective resources and perceptions of stress in a multi-ethnic sample of school-age children. Pediatr Nurs. (2004) 30:477–82.

Google Scholar

39. Garg R. Methodology for research I. Indian J Anaesth. (2016) 60:640–5. doi: 10.4103/0019-5049.190619

PubMed Abstract | CrossRef Full Text | Google Scholar

40. Arndt C, Schlemmer K, van der Meer E. Same or different pitch? Effects of musical expertise, pitch difference, and auditory task on the pitch discrimination ability of musicians and non-musicians. Exp Brain Res. (2020) 238:247–58. doi: 10.1007/s00221-019-05707-8

PubMed Abstract | CrossRef Full Text | Google Scholar

41. Sorati M, Behne DM. Musical expertise affects audiovisual speech perception: findings from event-related potentials and inter-trial phase coherence. Front Psychol. (2019) 10:2562. doi: 10.3389/fpsyg.2019.02562

PubMed Abstract | CrossRef Full Text | Google Scholar

42. Zhang F, Roland C, Rasul D, Cahn S, Liang C, Valencia G. Comparing musicians and non-musicians in signal-in-noise perception. Int J Audiol. (2019) 58:717–23. doi: 10.1080/14992027.2019.1623424

PubMed Abstract | CrossRef Full Text | Google Scholar

43. Murabayashi N, Akahoshi T, Ishimine R, Saji N, Takeda C, Nakayama H, et al. Effects of music therapy in frail elderlies: controlled crossover study. Dement Geriatr Cogn Dis Extra. (2019) 9:87–99. doi: 10.1159/000496456

CrossRef Full Text | Google Scholar

44. Innes KE, Selfe TK, Kandati S, Wen S, Huysmans Z. Effects of mantra meditation versus music listening on knee pain, function, and related outcomes in older adults with knee osteoarthritis: an exploratory randomized clinical trial (RCT). Evidence-based Complement Altern Med. (2018) 2018:7683897. doi: 10.1155/2018/7683897

PubMed Abstract | CrossRef Full Text | Google Scholar

45. Simmons-Stern NR, Deason RG, Brandler BJ, Frustace BS, O'Connor MK, Ally BA, et al. Music-based memory enhancement in Alzheimer's disease: promise and limitations. Neuropsychologia. (2012) 50:3295–303. doi: 10.1016/j.neuropsychologia.2012.09.019

PubMed Abstract | CrossRef Full Text | Google Scholar

46. Baker FA, Stretton-Smith P, Clark IN, Tamplin J, Lee Y-EC. A group therapeutic songwriting intervention for family caregivers of people living with dementia: A feasibility study with thematic analysis. Front Med. (2018) 5:151. doi: 10.3389/fmed.2018.00151

PubMed Abstract | CrossRef Full Text | Google Scholar

47. Lieber AC, Bose J, Zhang X, Seltzberg H, Loewy J, Rossetti A, et al. Effects of music therapy on anxiety and physiologic parameters in angiography: A systematic review and meta-analysis. J Neurointerv Surg. (2019) 11:416–31. doi: 10.1136/neurintsurg-2018-014313

PubMed Abstract | CrossRef Full Text | Google Scholar

48. Vinader-Caerols C, Monleón S, Carrasco C, Parra A. Effects of alcohol, coffee, and tobacco, alone or in combination, on physiological parameters and anxiety in a young population. J Caffeine Res. (2012) 2:70–6. doi: 10.1089/jcr.2012.0018

PubMed Abstract | CrossRef Full Text | Google Scholar

49. Huang C-J, Webb HE, Zourdos MC, Acevedo EO. Cardiovascular reactivity, stress, and physical activity. Front Physiol. (2013) 4:314. doi: 10.3389/fphys.2013.00314

PubMed Abstract | CrossRef Full Text | Google Scholar

50. Gifford RM, Boos CJ, Reynolds RM, Woods DR. Recovery time and heart rate variability following extreme endurance exercise in healthy women. Physiol Rep. (2018) 6:e13905. doi: 10.14814/phy2.13905

PubMed Abstract | CrossRef Full Text | Google Scholar

51. Vitale JA, Bonato M, La Torre A, Banfi G. Heart rate variability in sport performance: do time of day and chronotype play a role? J Clin Med. (2019) 8:723. doi: 10.3390/jcm8050723

PubMed Abstract | CrossRef Full Text | Google Scholar

52. Brattico E, Alluri V, Bogert B, Jacobsen T, Vartiainen N, Nieminen S, et al. A functional MRI study of happy and sad emotions in music with and without lyrics. Front Psychol. (2011) 2:308. doi: 10.3389/fpsyg.2011.00308

PubMed Abstract | CrossRef Full Text | Google Scholar

53. Condit-Schultz N, Huron D. Catching the lyrics. Music Percept An Interdiscip J. (2015) 32:470–83. doi: 10.1525/mp.2015.32.5.470

CrossRef Full Text | Google Scholar

54. Moffat D, Selfridge R, Reiss JD. Sound effect synthesis. In: Foundations in Sound Design for Interactive Media (New York, NY: Routledge) (2019). p. 274–99. doi: 10.4324/9781315106342-13

CrossRef Full Text | Google Scholar

55. Misra A, Cook P. Toward synthesized environments: a survey of analysis and synthesis methods for sound designers and composers. In: Proc 2009 Int Comput Music Conf ICMC 2009. Montreal, CA (2009).

Google Scholar

56. Ferri F, Tajadura-Jiménez A, Väljamäe A, Vastano R, Costantini M. Emotion-inducing approaching sounds shape the boundaries of multisensory peripersonal space. Neuropsychologia. (2015) 70:468–75. doi: 10.1016/j.neuropsychologia.2015.03.001

PubMed Abstract | CrossRef Full Text | Google Scholar

57. Reiss JD. A meta-analysis of high resolution audio perceptual evaluation. AES J Audio Eng Soc. (2016) 64:364–79. doi: 10.17743/jaes.2016.0015

CrossRef Full Text | Google Scholar

58. International Organization for Standardization I. ISO 226 Acoustics - Normal Equal-Loudness-Level Contours. (2003). Available online at: https://www.iso.org/obp/ui/#iso:std:iso:226:ed-2:v1:en

59. Degani A, Dalai M, Leonardi R, Migliorati P. Comparison of tuning frequency estimation methods. Multimed Tools Appl. (2015) 74:5917–34. doi: 10.1007/s11042-014-1897-2

PubMed Abstract | CrossRef Full Text | Google Scholar

60. ISO TC. 43, ISO 266: 1997, Acoustics: Preferred Frequencies for Measurements. Geneva: Int Organ Stand (ISO) (1997).

61. Wolters M, Mundt H, Riedmiller J. Loudness normalization in the age of portable media players. in Dolby Germany GmbH. Nuremberg: Germany (2010).

62. Organization WH. Safe listening. Bull World Health Organ. (2015) 93:213. doi: 10.2471/BLT.15.010415

CrossRef Full Text

63. Scherer KR. The geneva emotion wheel. Soc Sci Inf . (2005) 4:625–729. doi: 10.1177/0539018405058216

CrossRef Full Text | Google Scholar

64. Bradley MM, Lang PJ. Measuring emotion: The self-assessment manikin and the semantic differential. J Behav Ther Exp Psychiatry. (1994) 25:49–59. doi: 10.1016/0005-7916(94)90063-9

PubMed Abstract | CrossRef Full Text | Google Scholar

65. International Electrotechnical Commission I. IEC 60601-1:2005 Corrigendum 1 - Amendment 1 - Medical Electrical Equipment - Part 1: General Requirements for Basic Safety and Essential Performance. (2005). Available online at: https://webstore.iec.ch/publication/2604#additionalinfo

66. International Electrotechnical Commission I. IEC 60601-2-37: Medical Electrical Equipment - Part 2-37: Particular Requirements for the Basic Safety and Essential Performance of Ultrasonic Medical Diagnostic and Monitoring Equipment. (2015). Available online at: https://webstore.iec.ch/publication/22634

67. International Organization for Standardization I. ISO 13485:2016 Medical Devices - Quality Management Systems - Requirements for Regulatory Purposes. (2016). Available online at: https://www.iso.org/standard/59752.html

68. Yao Y-C. Nyquist frequency. In Balakrishnan N, Colton T, Everitt B, Piegorsch W, Ruggeri F, Teugels J editors. Wiley StatsRef: Statistics Reference Online. (2014). doi: 10.1002/9781118445112.stat03517

CrossRef Full Text | Google Scholar

69. Penzel T, Kesper K, Becker HF. Biosignal monitoring and recording. in Zieliński K, Duplaga M, Ingram D, editors. Information Technology Solutions for Healthcare. London (2006). doi: 10.1007/1-84628-141-5_13

CrossRef Full Text | Google Scholar

70. Martinsen ØG, Pabst O, Tronstad C, Grimnes S. Sources of error in ac measurement of skin conductance. J Electr Bioimpedance. (2015) 2015:49–53. doi: 10.5617/jeb.2640

CrossRef Full Text | Google Scholar

71. Munoz L, Roon A, Riese H, Thio C, Oostenbroek E, Westrik I, et al. Validity of (Ultra-)Short recordings for heart rate variability measurements. PLoS ONE. (2015) 10:e0138921. doi: 10.1371/journal.pone.0138921

PubMed Abstract | CrossRef Full Text | Google Scholar

72. Wallisch P, Lusignan ME, Benayoun MD, Baker TI, Dickey AS, Hatsopoulos NG. MATLAB for Neuroscientists?: An Introduction to Scientific Computing in MATLAB. Amsterdam: Academic Press (2014). doi: 10.1016/C2009-0-64117-9

CrossRef Full Text | Google Scholar

73. Nagar S editor. Introduction to Octave BT - Introduction to Octave: for Engineers and Scientists. Berkeley, CA: Apress (2008). p. 1–16. doi: 10.1007/978-1-4842-3201-9_1

CrossRef Full Text | Google Scholar

74. Dubey S, Narang N, Negi PS, Ojha VN, editors. LabVIEW Programming BT - LabVIEW based Automation Guide for Microwave Measurements. Singapore: Springer (2017). p. 3–9. doi: 10.1007/978-981-10-6280-3_2

CrossRef Full Text | Google Scholar

75. Doberkat E-E. Python 3. Berlin: De Gruyter Oldenbourg (2018). doi: 10.1515/9783110544138

CrossRef Full Text | Google Scholar

76. Kleiner M, Brainard DH, Pelli DG, Broussard C, Wolf T, Niehorster D. What's new in Psychtoolbox-3? Perception. (2007) 13:1–16. doi: 10.1068/v070821

CrossRef Full Text | Google Scholar

77. Peirce JW. PsychoPy-psychophysics software in Python. J Neurosci Methods. (2007) 162:8–13. doi: 10.1016/j.jneumeth.2006.11.017

PubMed Abstract | CrossRef Full Text | Google Scholar

78. Mathôt S, Schreij D, Theeuwes J. OpenSesame: An open-source, graphical experiment builder for the social sciences. Behav Res Methods. (2012) 44:314–24. doi: 10.3758/s13428-011-0168-7

PubMed Abstract | CrossRef Full Text | Google Scholar

79. Henninger F, Kieslich PJ, Hilbig BE. Psynteract: A flexible, cross-platform, open framework for interactive experiments. Behav Res Methods. (2017) 49:1605–14. doi: 10.3758/s13428-016-0801-6

PubMed Abstract | CrossRef Full Text | Google Scholar

80. Mueller ST, Piper BJ. The psychology experiment building language (PEBL) and PEBL test battery. J Neurosci Methods. (2014) 222:250–9. doi: 10.1016/j.jneumeth.2013.10.024

PubMed Abstract | CrossRef Full Text | Google Scholar

81. Cohen J, MacWhinney B, Flatt M, Provost J. PsyScope: An interactive graphic system for designing and controlling experiments in the psychology laboratory using Macintosh computers. Behav Res Methods Instruments Comput. (1993) 25:257–71. doi: 10.3758/BF03204507

CrossRef Full Text | Google Scholar

82. Testable SRLT. Available online at: https://www.testable.org/ (accessed March 10, 2020).

83. Anwyl-Irvine AL, Massonnié J, Flitton A, Kirkham N, Evershed JK. Gorilla in our midst: an online behavioral experiment builder. Behav Res Methods. (2020) 52:388–407. doi: 10.3758/s13428-019-01237-x

PubMed Abstract | CrossRef Full Text | Google Scholar

84. Stoet G. PsyToolkit: A software package for programming psychological experiments using Linux. Behav Res Methods. (2010) 42:1096–104. doi: 10.3758/BRM.42.4.1096

PubMed Abstract | CrossRef Full Text | Google Scholar

85. Yusuf F, Swanson DA, Martins JM. Methods of Demographic Analysis. Netherlands: Springer (2014). doi: 10.1007/978-94-007-6784-3

CrossRef Full Text | Google Scholar

86. Trappe H-J, Voit G. The cardiovascular effect of musical genres. Dtsch Arztebl Int. (2016) 113:347–52. doi: 10.3238/arztebl.2016.0347

PubMed Abstract | CrossRef Full Text | Google Scholar

87. Cohen MX. Analyzing Neural Time Series data: Theory and Practice. (2014). Available online at: https://mitpress.mit.edu/books/analyzing-neural-time-series-data (accessed August 4, 2020).

88. Kołodziej M, Tarnowski P, Majkowski A, Rak RJ. Electrodermal activity measurements for detection of emotional arousal. Bull Polish Acad Sci Tech Sci. (2019) 67:813–26. doi: 10.24425/bpasts.2019.130190

CrossRef Full Text | Google Scholar

89. Millán CA, Girón NA, Lopez DM. Analysis of relevant features from photoplethysmographic signals for atrial fibrillation classification. Int J Environ Res Public Health. (2020) 17:498. doi: 10.3390/ijerph17020498

PubMed Abstract | CrossRef Full Text | Google Scholar

90. El-Hajj C, Kyriacou PA. A review of machine learning techniques in photoplethysmography for the non-invasive cuff-less measurement of blood pressure. Biomed Signal Process Control. (2020) 58:101870. doi: 10.1016/j.bspc.2020.101870

CrossRef Full Text | Google Scholar

91. Weeks M. Digital Signal Processing Using Matlab and Wavelets. Massachusetts: Infinity Science Press (2007).

Google Scholar

92. Ingle VK, Proakis JG. Digital Signal Processing Using Matlab. Stamford, CT: Cengage Learning (2012).

Google Scholar

93. Banaee H, Ahmed UM, Loutfi A. Data mining for wearable sensors in health monitoring systems: a review of recent trends and challenges. Sensors. (2013) 13:17472. doi: 10.3390/s131217472

PubMed Abstract | CrossRef Full Text | Google Scholar

94. Schuh G, Reinhart G, Prote JP, Sauermann F, Horsthofer J, Oppolzer F, et al. Data mining definitions and applications for the management of production complexity. in Procedia CIRP. Ljubljana. doi: 10.1016/j.procir.2019.03.217

CrossRef Full Text | Google Scholar

95. Jatin B. Applications of Artificial Intelligence and Associated Technologies. in Emerg Technol Eng Biomed Manag Sci ETEBMS. New York, NY: McGraw-Hill (2016) 5:181–18.30.

Google Scholar

96. Graham S, Depp C, Lee EE, Nebeker C, Tu X, Kim H-C, et al. Artificial intelligence for mental health and mental illnesses: an overview. Curr Psychiatry Rep. (2019) 21:116. doi: 10.1007/s11920-019-1094-0

PubMed Abstract | CrossRef Full Text | Google Scholar

97. Jiang F, Jiang Y, Zhi H, Dong Y, Li H, Ma S, et al. Artificial intelligence in healthcare: past, present and future. Stroke Vasc Neurol. (2017) 2:230–43. doi: 10.1136/svn-2017-000101

PubMed Abstract | CrossRef Full Text | Google Scholar

98. Josefiok M, Krahn T, Sauer J. A survey on expert systems for diagnosis support in the field of neurology BT - intelligent decision technologies. In Neves-Silva R, Jain LC, Howlett RJ, editors. Cham: Springer International Publishing.

Google Scholar

99. Amisha, Malik P, Pathania M, Rathaur VK. Overview of artificial intelligence in medicine. J Fam Med Prim care. (2019) 8:2328–31. doi: 10.4103/jfmpc.jfmpc_440_19

PubMed Abstract | CrossRef Full Text | Google Scholar

100. Wang L, Ding J, Pan L, Cao D, Jiang H, Ding X. Artificial intelligence facilitates drug design in the big data era. Chemom Intell Lab Syst. (2019) 194:103850. doi: 10.1016/j.chemolab.2019.103850

CrossRef Full Text | Google Scholar

101. Kayacan E, Khanesar MA. Type-2 fuzzy neural networks. Butterworth-Heinemann (2012). doi: 10.1016/B978-0-12-802687-8.00004-9

CrossRef Full Text

102. Muthuswamy J. Biomedical signal analysis. Stand Handb Biomed Eng Des. (2004).

Google Scholar

103. Semmlow JL, Griffel B. Biosignal and Medical Image Processing, Third Edition, Boca Raton, FL: CRC Press (2014).

Google Scholar

104. Fasano A, Villani V. Baseline wander removal for bioelectrical signals by quadratic variation reduction. Signal Process. (2014) 99:48–57. doi: 10.1016/j.sigpro.2013.11.033

CrossRef Full Text | Google Scholar

105. Sreekrishna RR, Nalband S, Amalin Prince A. Real time cascaded moving average filter for detrending of electroencephalogram signals. In: Department of Electrical and Electronics Engineering, Birla Institute of Technology and Science Pilani. Goa: Institute of Electrical and Electronics Engineers Inc. (2021). doi: 10.1109/ICCSP.2016.7754244

CrossRef Full Text | Google Scholar

106. Luong DT, Thuan ND, Hoang CD, Trang N V, Due TQ. Study on limitation of removal of baseline noise from electrocardiography signal in measurement using wavelet analysis. In: Dept. of Biomedical Engineering, Hanoi University of Science and Technology. Hanoi: Viet Nam (2013). doi: 10.1109/ICUFN.2013.6614867

CrossRef Full Text | Google Scholar

107. Fedotov AA, Akulova AS. Adaptive filter for eliminating baseline wander of pulse wave signals. In: Samara State Aerospace University, Department of Laser and Bioengineering Systems. Samara: Springer Verlag (2015). doi: 10.1007/978-3-319-19387-8_248

CrossRef Full Text | Google Scholar

108. Liu Y-L, Chang N-C, Hsu S-F, Lin D-L, Lin Y-D. An adaptive algorithm for canceling power-line interference in biopotential measurement. Biomed Eng - Appl Basis Commun. (2004) 16:350–4. doi: 10.4015/S1016237204000487

CrossRef Full Text | Google Scholar

109. Satija U, Ramkumar B, Manikandan MS. A review of signal processing techniques for electrocardiogram signal quality assessment. IEEE Rev Biomed Eng. (2018) 11:36–52. doi: 10.1109/RBME.2018.2810957

PubMed Abstract | CrossRef Full Text | Google Scholar

110. Keshavamurthy TG, Eshwarappa MN. Review paper on denoising of ECG signal. In: 2017 Second International Conference on Electrical, Computer and Communication Technologies (ICECCT). Coimbatore (2017). doi: 10.1109/ICECCT.2017.8117941

CrossRef Full Text | Google Scholar

111. Milchevski A, Gusev M. Performance evaluation of FIR and IIR filtering of ECG signals. Int Conf . (2018) 665:103–12. doi: 10.1007/978-3-319-68855-8_10

CrossRef Full Text | Google Scholar

112. Troudi S, Ktata S, Fadhel Y Ben, Rahmani S, Ghommam J, Al-Haddad K. Analysis and extraction characteristic parameters of ECG signal in real-time for intelligent classification of cardiac arrhythmias. In: 2016 17th International Conference on Sciences and Techniques of Automatic Control and Computer Engineering (STA). Sousse (2017). doi: 10.1109/STA.2016.7952071

CrossRef Full Text | Google Scholar

113. Seena V, Yomas J. A review on feature extraction and denoising of ECG signal using wavelet transform. in 2014 2nd International Conference on Devices, Circuits and Systems (ICDCS). Coimbatore (2014). doi: 10.1109/ICDCSyst.2014.6926190

CrossRef Full Text | Google Scholar

114. Ozkaraca O, Guler I. Denoising and remote monitoring of ECG signal with real-time extended Kalman filter in a wearable system. Biomed Eng - Appl Basis Commun. (2015) 27:1550009. doi: 10.4015/S101623721550009X

CrossRef Full Text | Google Scholar

115. Aiboud Y, Mhamdi JE, Jilbab A, Sbaa H. Review of ECG signal de-noising techniques. In: 2015 Third World Conference on Complex Systems (WCCS). Marrakech (2015). doi: 10.1109/ICoCS.2015.7483313

CrossRef Full Text | Google Scholar

116. Hagmair S, Braunisch MC, Bachler M, Schmaderer C, Hasenau A-L, Bauer A, et al. Implementation and verification of an enhanced algorithm for the automatic computation of RR-interval series derived from 24 h 12-lead ECGs. Physiol Meas. (2017) 38:1–14. doi: 10.1088/1361-6579/38/1/1

PubMed Abstract | CrossRef Full Text | Google Scholar

117. Pan J, Tompkins WJ. A real-time QRS detection algorithm. IEEE Trans Biomed Eng. (1985) 32:230–6. doi: 10.1109/TBME.1985.325532

PubMed Abstract | CrossRef Full Text | Google Scholar

118. Aspuru J, Ochoa-Brust A, Félix RA, Mata-López W, Mena LJ, Ostos R, et al. Segmentation of the ECG signal by means of a linear regression algorithm. Sensors. (2019) 19:40775. doi: 10.3390/s19040775

PubMed Abstract | CrossRef Full Text | Google Scholar

119. Kovács P, Böck C, Meier J, Huemer M. ECG segmentation using adaptive hermite functions. In: Department of Numerical Analysis, Eötvös L. Hungary: Institute of Electrical and Electronics Engineers Inc. (2018). doi: 10.1109/ACSSC.2017.8335601

CrossRef Full Text | Google Scholar

120. Appathurai A, Jerusalin Carol J, Raja C, Kumar SN, Daniel A V, Jasmine Gnana Malar A, et al. A study on ECG signal characterization and practical implementation of some ECG characterization techniques. Meas J Int Meas Confed. (2019) 147:40. doi: 10.1016/j.measurement.2019.02.040

CrossRef Full Text | Google Scholar

121. Moskalenko V, Zolotykh N, Osipov G. Deep learning for ECG segmentation. Adv Neural Comput Mach Learn Cognit Res. (2020) 856:246–54. doi: 10.1007/978-3-030-30425-6_29

CrossRef Full Text | Google Scholar

122. Lee JS, Lee SJ, Choi M, Seo M, Kim SW. QRS detection method based on fully convolutional networks for capacitive electrocardiogram. Expert Syst Appl. (2019) 134:66–78. doi: 10.1016/j.eswa.2019.05.033

CrossRef Full Text | Google Scholar

123. Poungponsri S, Yu X-H. An adaptive filtering approach for electrocardiogram (ECG) signal noise reduction using neural networks. Neurocomputing. (2013) 117:206–13. doi: 10.1016/j.neucom.2013.02.010

CrossRef Full Text | Google Scholar

124. KuŽílek J, Huptych M, Chudáček V, Spilka J, Lhotská L. Data driven approach to ECG signal quality assessment using multistep SVM classification. in 2011 Computing in Cardiology. Hangzhou (2011).

Google Scholar

125. Rodríguez R, Mexicano A, Bila J, Cervantes S, Ponce R. Feature extraction of electrocardiogram signals by applying adaptive threshold and principal component analysis. J Appl Res Technol. (2015) 13:261–9. doi: 10.1016/j.jart.2015.06.008

CrossRef Full Text | Google Scholar

126. Dohare AK, Kumar V, Kumar R. Detection of myocardial infarction in 12 lead ECG using support vector machine. Appl Soft Comput J. (2018) 64:138–47. doi: 10.1016/j.asoc.2017.12.001

PubMed Abstract | CrossRef Full Text | Google Scholar

127. Jung W-H, Lee S-G. ECG identification based on non-fiducial feature extraction using window removal method. Appl Sci. (2017) 7:7111205. doi: 10.3390/app7111205

CrossRef Full Text | Google Scholar

128. Prasad BVP, Parthasarathy V. Detection and classification of cardiovascular abnormalities using FFT based multi-objective genetic algorithm. Biotechnol Biotechnol Equip. (2018) 32:183–93. doi: 10.1080/13102818.2017.1389303

CrossRef Full Text | Google Scholar

129. Chetan A, Tripathy RK, Dandapat S. A diagnostic system for detection of atrial and ventricular arrhythmia episodes from electrocardiogram. J Med Biol Eng. (2018) 38:304–15. doi: 10.1007/s40846-017-0294-5

PubMed Abstract | CrossRef Full Text | Google Scholar

130. Lee S-H, Chung K-Y, Lim JS. Detection of ventricular fibrillation using Hilbert transforms, phase-space reconstruction, and time-domain analysis. Pers Ubiquitous Comput. (2014) 18:1315–24. doi: 10.1007/s00779-013-0735-2

CrossRef Full Text | Google Scholar

131. Yücelbaş S, Yücelbaş C, Tezel G, Özşen S, Yosunkaya S. Automatic sleep staging based on SVD, VMD, HHT and morphological features of single-lead ECG signal. Expert Syst Appl. (2018) 102:193–206. doi: 10.1016/j.eswa.2018.02.034

CrossRef Full Text | Google Scholar

132. Le TH-N, Le TM, Le T, Van Toi V. Feature extraction techniques for automatic detection of some specific cardiovascular diseases using ECG: a review and evaluation study. In: Department of Biomedical Engineering, International University—Vietnam National University. Ho Chi Minh City, Viet Nam: Springer Verlag (2018). doi: 10.1007/978-981-13-5859-3_94

CrossRef Full Text | Google Scholar

133. Roza VCC, De Almeida AM, Postolache OA. Design of an artificial neural network and feature extraction to identify arrhythmias from ECG. in ISCTE Instituto Universitário de Lisboa/IT, Universidade Federal Do Rio Grande Do Norte. Brazil: Institute of Electrical and Electronics Engineers Inc. (2017). doi: 10.1109/MeMeA.2017.7985908

CrossRef Full Text | Google Scholar

134. Sarvan C, Ozkurt N. Multiwavelet feature sets for ECG beat classification. In: Elektrik Ve Elektronik Mühendisligi Bölümü, Yaşar Üniversitesi, Izmir. Turkey: Institute of Electrical and Electronics Engineers Inc. (2017). doi: 10.1109/SIU.2017.7960297

CrossRef Full Text | Google Scholar

135. Qin Q, Li J, Zhang L, Yue Y, Liu C. Combining low-dimensional wavelet features and support vector machine for arrhythmia beat classification. Sci Rep. (2017) 7:6596. doi: 10.1038/s41598-017-06596-z

PubMed Abstract | CrossRef Full Text | Google Scholar

136. Adam M, Oh SL, Sudarshan VK, Koh JE, Hagiwara Y, Tan JH, et al. Automated characterization of cardiovascular diseases using relative wavelet nonlinear features extracted from ECG signals. Comput Methods Programs Biomed. (2018) 161:133–43. doi: 10.1016/j.cmpb.2018.04.018

PubMed Abstract | CrossRef Full Text | Google Scholar

137. Oh SL, Hagiwara Y, Adam M, Sudarshan VK, Koh JE, Tan JH, et al. Shockable versus nonshockable life-threatening ventricular arrhythmias using dwt and nonlinear features of ECG signals. J Mech Med Biol. (2017) 17:48. doi: 10.1142/S0219519417400048

CrossRef Full Text | Google Scholar

138. Bhoi AK, Sherpa KS, Khandelwal B. Multimodal Classification of Arrhythmia and Ischemia Using QRS-ST Analysis. Adv Syst Cont Automat. (2018) 442:679–92. doi: 10.1007/978-981-10-4762-6_65

CrossRef Full Text | Google Scholar

139. Reljin IS, Reljin BD. Fractal geometry and multifractals in analyzing and processing medical data and images. Arch Oncol. (2002) 10:283–93. doi: 10.2298/AOO0204283R

CrossRef Full Text | Google Scholar

140. Arjunan SP, Kumar D, Aliahmad B. Fractals: Applications in Biological Signalling and Image Processing. Boca Raton, FL: CRC Press (2017).

Google Scholar

141. Shekatkar SM, Kotriwar Y, Harikrishnan KP, Ambika G. Detecting abnormality in heart dynamics from multifractal analysis of ECG signals. Sci Rep. (2017) 7:1–11. doi: 10.1038/s41598-017-15498-z

PubMed Abstract | CrossRef Full Text | Google Scholar

142. Namazi H, Kulish V V. Fractal based analysis of the influence of odorants on heart activity. Sci Rep. (2016) 6:1–8. doi: 10.1038/srep38555

PubMed Abstract | CrossRef Full Text | Google Scholar

143. Parastesh Karegar F, Fallah A, Rashidi S. ECG based human authentication with using Generalized Hurst Exponent. In: Biomedical Engineering Faculty, Amirkabir University of Technology. Tehran: Institute of Electrical and Electronics Engineers Inc. (2017). doi: 10.1109/IranianCEE.2017.7985480

CrossRef Full Text | Google Scholar

144. Vanzella LM, Dagostinho DBB, de Figueiredo MPF, Castrillón CIM, Netto Junior J, Vanderlei LCM. Periodized aerobic interval training modifies geometric indices of heart rate variability in metabolic syndrome. Med. (2019) 55:55090532. doi: 10.3390/medicina55090532

PubMed Abstract | CrossRef Full Text | Google Scholar

145. Regis da Costa e Oliveira J, Base LH, Maia LCP, Ferreira de Lima Antão JYF, de Abreu LC, Oliveira FR, et al. Geometric indexes of heart rate variability in healthy individuals exposed to long-term air pollution. Environ Sci Pollut Res. (2020) 27:4170–7. doi: 10.1007/s11356-019-06965-3

PubMed Abstract | CrossRef Full Text | Google Scholar

146. Hsu C-H, Tsai M-Y, Huang G-S, Lin T-C, Chen K-P, Ho S-T, et al. Poincaré plot indexes of heart rate variability detect dynamic autonomic modulation during general anesthesia induction. Acta Anaesthesiol Taiwanica. (2012) 50:12–8. doi: 10.1016/j.aat.2012.03.002

PubMed Abstract | CrossRef Full Text | Google Scholar

147. Faust O, Bairy MG. Nonlinear analysis of physiological signals: a review. J Mech Med Biol. (2012) 12:155. doi: 10.1142/S0219519412400155

CrossRef Full Text | Google Scholar

148. Shaffer F, Ginsberg JP. An overview of heart rate variability metrics and norms. Front public Heal. (2017) 5:258. doi: 10.3389/fpubh.2017.00258

PubMed Abstract | CrossRef Full Text | Google Scholar

149. Draghici AE, Taylor JA. The physiological basis and measurement of heart rate variability in humans. J Physiol Anthropol. (2016) 35:22. doi: 10.1186/s40101-016-0113-7

PubMed Abstract | CrossRef Full Text | Google Scholar

150. Liu J, Xu X, Zhang G, Wu S, Liu M, Zhou Y. Differences analysis of HRV in time and meridians based on multi-lead ECG signals. In: College of Information Science and Engineering, Northeastern University. Shenyang: Association for Computing Machinery (2018). doi: 10.1145/3242840.3242841

CrossRef Full Text | Google Scholar

151. Wachowiak MP, Hay DC, Johnson MJ. Assessing heart rate variability through wavelet-based statistical measures. Comput Biol Med. (2016) 77:222–30. doi: 10.1016/j.compbiomed.2016.07.008

PubMed Abstract | CrossRef Full Text | Google Scholar

152. Pale U, Thurk F, Kaniusas E. Heart rate variability analysis using different wavelet transformations. In: Faculty of Electrical Engineering and Computing, University of Zagreb. Zagreb: Institute of Electrical and Electronics Engineers Inc. (2016). doi: 10.1109/MIPRO.2016.7522403

CrossRef Full Text | Google Scholar

153. Hagad JL, Fukui K-I, Numao M. Modelling naturalistic work stress using spectral HRV representations and deep learning. Proc Annual Confer JSAI. (2020) 1128:267–77. doi: 10.1007/978-3-030-39878-1_24

CrossRef Full Text | Google Scholar

154. Rad AB, Eftestol T, Engan K, Irusta U, Kvaloy JT, Kramer-Johansen J, et al. ECG-Based classification of resuscitation cardiac rhythms for retrospective data analysis. IEEE Trans Biomed Eng. (2017) 64:2411–8. doi: 10.1109/TBME.2017.2688380

PubMed Abstract | CrossRef Full Text | Google Scholar

155. Antink CH, Leonhardt S, Walter M. Fusing QRS detection and robust interval estimation with a random forest to classify atrial fibrillation. in Philips Department for Medical Information Technology, RWTH Aachen University, Helmholtz-Institute, Pauwelsstr. Aachen: IEEE Computer Society (2017). doi: 10.22489/CinC.2017.167-163

CrossRef Full Text | Google Scholar

156. Huda S, Jelinek H, Ray B, Stranieri A, Yearwood J. Exploring novel features and decision rules to identify cardiovascular autonomic neuropathy using a hybrid of wrapper-filter based feature selection. In: CIAO, GSITMS, University of Ballarat, VIC, Australia. (2010). doi: 10.1109/ISSNIP.2010.5706769

CrossRef Full Text | Google Scholar

157. Elsayyad A, Al-Dhaifallah M, Nassef AM. Features selection for arrhythmia diagnosis using Relief-F algorithm and support vector machine. In: College of Engineering at Wadi Addawasir, Sattam Bin Abdulaziz University. Wadi Aldawasir: Institute of Electrical and Electronics Engineers Inc. (2017). doi: 10.1109/SSD.2017.8166920

CrossRef Full Text | Google Scholar

158. Li D, Ma W, Zhao J. A novel J wave detection method based on massive ECG data and mapreduce. BigCom. (2016) 9784:399–408. doi: 10.1007/978-3-319-42553-5_34

CrossRef Full Text | Google Scholar

159. Huang R, Zhou Y. Disease classification and biomarker discovery using ECG data. Biomed Res Int. (2015) 2015:680381. doi: 10.1155/2015/680381

PubMed Abstract | CrossRef Full Text | Google Scholar

160. Kaplan Berkaya S, Uysal AK, Gunal S, Ergin S, Gunal S, Bilginer Gulmezoglu M. A Survey on ECG Analysis. Elsevier (2018). doi: 10.1016/j.bspc.2018.03.003

CrossRef Full Text | Google Scholar

161. Donoso FI, Figueroa RL, Lecannelier EA, Pino EJ, Rojas AJ. Clustering of atrial fibrillation based on surface ECG measurements. Annu Int Conf IEEE Eng Med Biol Soc. (2013) 2013:4203–6. doi: 10.1109/EMBC.2013.6610472

PubMed Abstract | CrossRef Full Text | Google Scholar

162. Annavarapu A, Borra S, Kora P. ECG signal dimensionality reduction-based atrial fibrillation detection. Lect Notes Comput Vis Biomech. (2018) 26:383–406. doi: 10.1007/978-3-319-65981-7_14

CrossRef Full Text | Google Scholar

163. Varatharajan R, Manogaran G, Priyan MK. A big data classification approach using LDA with an enhanced SVM method for ECG signals in cloud computing. Multimed Tools Appl. (2018) 77:10195–215. doi: 10.1007/s11042-017-5318-1

CrossRef Full Text | Google Scholar

164. Kaya Y, Pehlivan H, Tenekeci ME. Effective ECG beat classification using higher order statistic features and genetic feature selection. Biomed Res. (2017) 28:7594–603.

Google Scholar

165. Moeynoi P, Kitjaidure Y. Dimension reduction based on canonical correlation analysis technique to classify sleep stages of sleep apnea disorder using EEG and ECG signals. In: Faculty of Electrical Engineering, King Mangkut's Institute of Technology. Ladkrabang: Institute of Electrical and Electronics Engineers Inc. (2017). doi: 10.1109/ECTICon.2017.8096272

CrossRef Full Text | Google Scholar

166. Halil I, Usta N, Yildiz M. Classification of ECG arrhythmia with machine learning techniques. In: 2017 16th IEEE International Conference on Machine Learning and Applications (ICMLA). (2017). doi: 10.1109/ICMLA.2017.0-104

CrossRef Full Text | Google Scholar

167. Zhang C, Wang G, Zhao J, Gao P, Lin J, Yang H. Patient-specific ECG classification based on recurrent neural networks and clustering technique. In: 2017 13th IASTED International Conference on Biomedical Engineering (BioMed). Innsbruck (2017). doi: 10.2316/P.2017.852-029

CrossRef Full Text | Google Scholar

168. Wang JT, Sehmi AS, Jones NB, Bono DP. A knowledge-based system for qualitative ECG simulation and ECG analysis. in [1991] Proceedings Computers in Cardiology. Venice (1991). doi: 10.1109/CIC.1991.169016

CrossRef Full Text | Google Scholar

169. Cheng Z, Shu L, Xie J, Chen CLP. A novel ECG-based real-time detection method of negative emotions in wearable applications. In: 2017 International Conference on Security, Pattern Analysis, and Cybernetics (SPAC). Shenzhen (2017). doi: 10.1109/SPAC.2017.8304293

CrossRef Full Text | Google Scholar

170. Aljafar L, Alotaiby TN, Al-Yami RR, Alshebeili SA, Zouhair J. Classification of ECG signals of normal and abnormal subjects using common spatial pattern. In: Pince Sultan University, Saudi Arabia: IEEE Computer Society. Ras Al Khaimah (2017). doi: 10.1109/ICEDSA.2016.7818547

CrossRef Full Text | Google Scholar

171. Sridhar GV, Mallikarjuna Rao P. Framework for Classifying Cardiac Arrhythmia. Singapore: Springer (2018). doi: 10.1007/978-981-10-4280-5_16

CrossRef Full Text | Google Scholar

172. Rad AB, Engan K, Katsaggelos AK, Kvaløy JT, Wik L, Kramer-Johansen J, et al. Automatic cardiac rhythm interpretation during resuscitation. Resuscitation. (2016) 102:44–50. doi: 10.1016/j.resuscitation.2016.01.015

PubMed Abstract | CrossRef Full Text | Google Scholar

173. Sansone M, Fusco R, Pepino A, Sansone C. Electrocardiogram pattern recognition and analysis based on artificial neural networks and support vector machines: A review. J Healthc Eng. (2013) 4:465–504. doi: 10.1260/2040-2295.4.4.465

PubMed Abstract | CrossRef Full Text | Google Scholar

174. Ansari S, Farzaneh N, Duda M, Horan K, Andersson HB, Goldberger ZD, et al. A review of automated methods for detection of myocardial ischemia and infarction using electrocardiogram and electronic health records. IEEE Rev Biomed Eng. (2017) 10:264–98. doi: 10.1109/RBME.2017.2757953

PubMed Abstract | CrossRef Full Text | Google Scholar

175. Kumar TS, Kanhangad V. Detection of electrocardiographic changes in partial epileptic patients using local binary pattern based composite feature. Australas Phys Eng Sci Med. (2018) 41:209–16. doi: 10.1007/s13246-017-0605-8

PubMed Abstract | CrossRef Full Text | Google Scholar

176. Pinto JR, Cardoso JS, Lourenço A, Carreiras C. Towards a continuous biometric system based on ECG signals acquired on the steering wheel. Sensors. (2017) 17:17102228. doi: 10.3390/s17102228

PubMed Abstract | CrossRef Full Text | Google Scholar

177. Vanitha L, Suresh GR. Hybrid SVM classification technique to detect mental stress in human beings using ECG signals. In: ECE, Loyola Institute of Technology. Chennai: Institute of Electrical and Electronics Engineers Inc. (2014). doi: 10.1109/ICACCS.2013.6938735

CrossRef Full Text | Google Scholar

178. Saini R, Bindal N, Bansal P. Classification of heart diseases from ECG signals using wavelet transform and kNN classifier. In: International Conference on Computing, Communication & Automation. Noida (2015). doi: 10.1109/CCAA.2015.7148561

CrossRef Full Text | Google Scholar

179. Faust O, Hagiwara Y, Hong TJ, Lih OS, Acharya UR. Deep learning for healthcare applications based on physiological signals: A review. Comput Methods Programs Biomed. (2018) 161:1–13. doi: 10.1016/j.cmpb.2018.04.005

PubMed Abstract | CrossRef Full Text | Google Scholar

180. Al Rahhal MM, Bazi Y, Al Zuair M, Othman E, BenJdira B. Convolutional neural networks for electrocardiogram classification. J Med Biol Eng. (2018) 38:1014–25. doi: 10.1007/s40846-018-0389-7

CrossRef Full Text | Google Scholar

181. Yildirim Ö. A novel wavelet sequences based on deep bidirectional LSTM network model for ECG signal classification. Comput Biol Med. (2018) 96:189–202. doi: 10.1016/j.compbiomed.2018.03.016

PubMed Abstract | CrossRef Full Text | Google Scholar

182. Nanjundegowda R, Meshram VA. Arrhythmia detection based on hybrid features of T-wave in electrocardiogram. Int J Intell Eng Syst. (2018) 11:153–62. doi: 10.22266/ijies2018.0228.16

CrossRef Full Text | Google Scholar

183. Majumdar A, Ward R. Robust greedy deep dictionary learning for ECG arrhythmia classification. In: IIIT Delhi. New Delhi: Institute of Electrical and Electronics Engineers Inc. (2017). doi: 10.1109/IJCNN.2017.7966413

CrossRef Full Text | Google Scholar

184. Wu Z, Ding X, Zhang G. A novel method for classification of ECG arrhythmias using deep belief networks. Int J Comput Intell Appl. (2016) 15:218. doi: 10.1142/S1469026816500218

CrossRef Full Text | Google Scholar

185. Jambukia SH, Dabhi VK, Prajapati HB. ECG beat classification using machine learning techniques. Int J Biomed Eng Technol. (2018) 26:32–53. doi: 10.1504/IJBET.2018.089255

CrossRef Full Text | Google Scholar

186. Raj S, Ray KC. Sparse representation of ECG signals for automated recognition of cardiac arrhythmias. Expert Syst Appl. (2018) 105:49–64. doi: 10.1016/j.eswa.2018.03.038

PubMed Abstract | CrossRef Full Text | Google Scholar

187. Stȩpień K, Grzegorczyk I. Classification of ECG recordings with neural networks based on specific morphological features and regularity of the signal. In: Faculty of Physics, Warsaw University of Technology. Koszykowa: IEEE Computer Society (2017). doi: 10.22489/CinC.2017.356-350

CrossRef Full Text | Google Scholar

188. Soliński M, Perka A, Rosiński J, Łepek M, Rymko J. Classification of atrial fibrillation in short-term ECG recordings using a machine learning approach and hybrid QRS Detection. In: Faculty of Physics, Warsaw University of Technology. Koszykowa: IEEE Computer Society (2017). doi: 10.22489/CinC.2017.337-201

CrossRef Full Text | Google Scholar

189. Lyon A, Mincholé A, Martínez JP, Laguna P, Rodriguez B. Computational techniques for ECG analysis and interpretation in light of their contribution to medical advances. J R Soc Interface. (2018) 15:20170821. doi: 10.1098/rsif.2017.0821

PubMed Abstract | CrossRef Full Text | Google Scholar

190. Saleh H, Bayasi N, Mohammad B, Ismail M. Self-powered SoC Platform for Analysis and Prediction of Cardiac Arrhythmias. Cham: Springer. (2018) doi: 10.1007/978-3-319-63973-4

CrossRef Full Text | Google Scholar

191. Minchole A, Camps J, Lyon A, Rodriguez B. Machine learning in the electrocardiogram. J Electrocardiol. (2019) 57S:S61–4. doi: 10.1016/j.jelectrocard.2019.08.008

PubMed Abstract | CrossRef Full Text | Google Scholar

192. Zhang L, Wu H, Zhang X, Wei X, Hou F, Ma Y. Sleep heart rate variability assists the automatic prediction of long-term cardiovascular outcomes. Sleep Med. (2020) 67:217–24. doi: 10.1016/j.sleep.2019.11.1259

PubMed Abstract | CrossRef Full Text | Google Scholar

193. Rahman MJ, Nemati E, Rahman MM, Nathan V, Vatanparvar K, Kuang J. Automated assessment of pulmonary patients using heart rate variability from everyday wearables. Smart Heal. (2020) 15:100081. doi: 10.1016/j.smhl.2019.100081

CrossRef Full Text | Google Scholar

194. Yan X, Zhang L, Li J, Du D, Hou F. Entropy-based measures of hypnopompic heart rate variability contribute to the automatic prediction of cardiovascular events. Entropy. (2020) 22:20241. doi: 10.3390/e22020241

PubMed Abstract | CrossRef Full Text | Google Scholar

195. Katahira K, Fujimura T, Matsuda Y-T, Okanoya K, Okada M. Individual differences in heart rate variability are associated with the avoidance of negative emotional events. Biol Psychol. (2014) 103:322–31. doi: 10.1016/j.biopsycho.2014.10.007

PubMed Abstract | CrossRef Full Text | Google Scholar

196. Park C, Took CC, Seong J-K. Machine learning in biomedical engineering. Biomed Eng Lett. (2018) 3:58. doi: 10.1007/s13534-018-0058-3

PubMed Abstract | CrossRef Full Text | Google Scholar

197. Rim B, Sung N-J, Min S, Hong M. Deep learning in physiological signal data: A survey. Sensors. (2020) 20:40969. doi: 10.3390/s20040969

PubMed Abstract | CrossRef Full Text | Google Scholar

198. Xia Y, Wulan N, Wang K, Zhang H. Detecting atrial fibrillation by deep convolutional neural networks. Comput Biol Med. (2018) 93:84–92. doi: 10.1016/j.compbiomed.2017.12.007

PubMed Abstract | CrossRef Full Text | Google Scholar

199. Al-Jebrni AH, Chwyl B, Wang XY, Wong A, Saab BJ. AI-enabled remote and objective quantification of stress at scale. Biomed Signal Process Control. (2020) 59:101929. doi: 10.1016/j.bspc.2020.101929

CrossRef Full Text | Google Scholar

200. Oh S, Lee J-Y, Kim DK. The design of CNN architectures for optimal six basic emotion classification using multiple physiological signals. Sensors. (2020) 20:30866. doi: 10.3390/s20030866

PubMed Abstract | CrossRef Full Text | Google Scholar

201. Chen Z, Wu M, Wu J, Ding J, Zeng Z, Surmacz K, et al. A Deep learning approach for sleep-wake detection from hrv and accelerometer data. In: Institute for Infocomm Research, A STAR, 1 Fusionopolis Way #21-01 Connexis, Singapore. Singapore: Institute of Electrical and Electronics Engineers Inc. (2019). doi: 10.1109/BHI.2019.8834502

CrossRef Full Text | Google Scholar

202. Wang L, Zhou X. Detection of congestive heart failure based on LSTM-based deep network via short-term RR intervals. Sensors. (2019) 19:71502. doi: 10.3390/s19071502

PubMed Abstract | CrossRef Full Text | Google Scholar

203. Miller DD. Machine intelligence in cardiovascular medicine. Cardiol Rev. (2020) 28:53–64. doi: 10.1097/CRD.0000000000000294

PubMed Abstract | CrossRef Full Text | Google Scholar

204. Van Steenkiste G, van Loon G, Crevecoeur G. Transfer learning in ECG classification from human to horse using a novel parallel neural network architecture. Sci Rep. (2020) 10:186. doi: 10.1038/s41598-019-57025-2

PubMed Abstract | CrossRef Full Text | Google Scholar

205. Li D, Li X, Zhao J, Bai X. Automatic staging model of heart failure based on deep learning. Biomed Signal Process Control. (2019) 52:77–83. doi: 10.1016/j.bspc.2019.03.009

CrossRef Full Text | Google Scholar

206. Wang C, Yang S, Tang X, Li B. A 12-lead ECG Arrhythmia Classification Method Based on 1D Densely Connected CNN. Cham: Springer (2019).

Google Scholar

207. Nurmaini S, Darmawahyuni A, Mukti ANS, Rachmatullah MN, Firdaus F, Tutuko B. Deep learning-based stacked denoising and autoencoder for ECG heartbeat classification. Electron. (2020) 9:9010135. doi: 10.3390/electronics9010135

CrossRef Full Text | Google Scholar

208. Zheng J, Chu H, Struppa D, Zhang J, Yacoub SM, El-Askary H, et al. Optimal multi-stage arrhythmia classification approach. Sci Rep. (2020) 10:59821. doi: 10.1038/s41598-020-59821-7

PubMed Abstract | CrossRef Full Text | Google Scholar

209. Abdalla FYO, Wu L, Ullah H, Mkindu H, Nie Y, Zhao Y. ECG arrhythmia discrimination using SVM and nonlinear and non-stationary decomposition. In: School of Electronics and Information Engineering, Harbin Institute of Technology. Harbin: Institute of Electrical and Electronics Engineers Inc. (2019). doi: 10.1109/ISSPIT47144.2019.9001889

CrossRef Full Text | Google Scholar

210. El_Rahman SA. Multimodal biometric systems based on different fusion levels of ECG and fingerprint using different classifiers. Soft Comput. (2020) 24:12599–632. doi: 10.1007/s00500-020-04700-6

CrossRef Full Text | Google Scholar

211. Sim J, Wright CC. The kappa statistic in reliability studies: use, interpretation, and sample size requirements. Phys Ther. (2005) 85:257–68. doi: 10.1093/ptj/85.3.257

PubMed Abstract | CrossRef Full Text | Google Scholar

212. Sharma M, Rajendra Acharya U. A new method to identify coronary artery disease with ECG signals and time-Frequency concentrated antisymmetric biorthogonal wavelet filter bank. Pattern Recognit Lett. (2019) 125:235–40. doi: 10.1016/j.patrec.2019.04.014

CrossRef Full Text | Google Scholar

213. Diez P, editor. Chapter 1 - Introduction. In: Smart Wheelchairs and Brain-Computer Interfaces. Academic Press (2018). p. 1–21. doi: 10.1016/B978-0-12-812892-3.00001-7

CrossRef Full Text

214. Song Q, Guo Y, Shepperd M. A comprehensive investigation of the role of imbalanced learning for software defect prediction. IEEE Trans Softw Eng. (2018) PP:1. doi: 10.1109/TSE.2018.2836442

CrossRef Full Text | Google Scholar

215. Ben Ishak M. Toward New Evaluation Metrics for Relational Learning BT - Advances in Integrations of Intelligent Methods: Post-workshop volume of the 8th International Workshop CIMA 2018, Volos, Greece, November 2018. In: conjunction with IEEE ICTAI 2018. Singapore: Springer Singapore (2018). doi: 10.1007/978-981-15-1918-5_4

CrossRef Full Text | Google Scholar

216. Kailkhura B, Gallagher B, Kim S, Hiszpanski A, Han TY-J. Reliable and explainable machine-learning methods for accelerated material discovery. NPJ Comput Mater. (2019) 5:108. doi: 10.1038/s41524-019-0248-2

CrossRef Full Text | Google Scholar

217. Rzadkowski G, Głazewska I, Sawińska K. The gompertz function and its applications in management. Found Manag. (2015) 7:185–90. doi: 10.1515/fman-2015-0035

CrossRef Full Text | Google Scholar

218. United Nations Office on Drugs and Crime. Guidance for the Validation of Analytical Methodology and Calibration of Equipment used for Testing of Illicit Drugs in Seized Materials and Biological Specimens. New York, NY (2013).

219. International Organization for Standardization I, International Electrotechnical Commission I. ISO/IEC 17025 – General requirements for the competence of testing and calibration laboratories. (2018). Available at: https://www.iso.org/publication/PUB100424.html

220. Beniczky S, Ryvlin P. Standards for testing and clinical validation of seizure detection devices. Epilepsia. (2018) 59:9–13. doi: 10.1111/epi.14049

PubMed Abstract | CrossRef Full Text | Google Scholar

221. Huber L. Validation and Qualification in Analytical Laboratories. New York, NY: Informa Healthcare (2007). doi: 10.3109/9780849382680

CrossRef Full Text | Google Scholar

222. Idrobo-Ávila E, Loaiza-Correa H, Muñoz-Bolaños F. Recognition of valence judgments in music perception using electrocardiographic signals and machine learning. In: Rosas-Paredes K, Villalba-Condori KO, editors. VI Iberoamerican Conference of Computer Human Interaction. Arequipa: CEUR Workshop Proceedings (2020).

Google Scholar

223. Herbert T. Music in WORDS: A Guide to Researching and Writing About Music. (2009). Available online at: https://global.oup.com/academic/product/music-in-words-9780195373738?q=music&prevNumResPerPage=20&prevSortField=9&facet_narrowbyproducttype_facet=Print&facet_cet=Paperback&facet_narrowbyprice_facet=25to50&sortField=9&resultsPerPage=60&start=0&lang=en&cc=us (accessed August 4, 2020).

224. University of Bern. STROBE Statement: Available checklists. (2021). Available online at: https://www.strobe-statement.org/index.php?id=available-checklists (accessed April 23, 2021).

225. Cuschieri S. The STROBE guidelines. Saudi J Anaesth. (2019) 13:S31–4. doi: 10.4103/sja.SJA_543_18

PubMed Abstract | CrossRef Full Text | Google Scholar

226. Website C. Consort - Welcome to the CONSORT Website. (2021). Available online at: http://www.consort-statement.org/ (accessed April 23, 2021).

227. Cuschieri S. The CONSORT statement. Saudi J Anaesth. (2019) 13:S27–30. doi: 10.4103/sja.SJA_559_18

PubMed Abstract | CrossRef Full Text | Google Scholar

228. Website S. Squire | Home Page. (2021). Available online at: http://www.squire-statement.org/ (accessed April 23, 2021).

229. Ogrinc G, Davies L, Goodman D, Batalden P, Davidoff F, Stevens D. SQUIRE 2.0 (Standards for QUality Improvement Reporting Excellence): revised publication guidelines from a detailed consensus process. BMJ Qual Saf. (2016) 25:986–92. doi: 10.1136/bmjqs-2015-004411

PubMed Abstract | CrossRef Full Text

230. Website S. STARD initiative. (2021). Available online at: https://ibooked.no/stard-statement.html (accessed April 23, 2021).

231. Cohen JF, Korevaar DA, Altman DG, Bruns DE, Gatsonis CA, Hooft L, et al. STARD 2015 guidelines for reporting diagnostic accuracy studies: explanation and elaboration. BMJ Open. (2016) 6:e012799. doi: 10.1136/bmjopen-2016-012799

PubMed Abstract | CrossRef Full Text | Google Scholar

232. Website S. STrengthening the REporting of Genetic Association Studies (STREGA): An Extension of the STROBE Statement. | The EQUATOR Network. (2021). Available online at: https://www.equator-network.org/reporting-guidelines/strobe-strega/ (accessed April 23, 2021).

233. Little J, Higgins JPT, Ioannidis JPA, Moher D, Gagnon F, von Elm E, et al. STrengthening the REporting of Genetic Association Studies (STREGA)— an extension of the STROBE statement. PLOS Med. (2009) 6:e1000022. doi: 10.1371/journal.pmed.1000022

PubMed Abstract | CrossRef Full Text | Google Scholar