Music has been shown to influence physiological functions in humans, but its effects on fetal heart rate variability (fHRV) are not well understood. This study aimed to assess the response of classical music exposure on the nonlinear behavior of fetal heart rate fluctuations in fetuses between 32 and 40 weeks of gestation using recurrence quantification analysis (RQA). We collected R–R time series from 37 fetuses in the third trimester following a study into four stages: PRE (baseline), STIM1 (first musical piece), STIM2 (second musical piece), and POST (post-exposure). The fetal R–R time series from each stage were evaluated using RQA indices such as determinism (DET), average diagonal line length (L), maximum line length (LMAX), entropy (ENTR), and trapping time (TT), as well as conventional linear indices like SDRR (standard deviation of R–R intervals). Results revealed three main points. First, there was an increase in DET, L, LMAX, and TT, with a decrease in ENTR in the POST stage compared to PRE, indicating more regular and predictable patterns. Second, the STIM2 stage enhanced the predictability and stability of cardiac dynamics compared to PRE, as indicated by L, LMAX, and TT. Third, no significant changes were observed in conventional indices, except for an increase in SDRR in the POST stage compared to STIM1. These findings suggest a reduction in complexity and nonlinear behavior of fHRV patterns after musical stimulus. The increase in SDRR during the POST stage appears to coincide with fetal movements, as indicated by the number of fetal accelerations found.

The study of fetal heart rate dynamics can provide valuable insights into the developmental state of the fetal autonomic nervous system, particularly when exposed to external stimuli such as music. Traditional linear methods for analyzing fetal heart rate variability (fHRV) often fail to capture the complex, nonlinear nature of these signals. In this context, we investigated the effects of classical music exposure on the nonlinear behavior of fetal heart rate fluctuations in fetuses between 32 and 40 weeks of gestation. We employed recurrence quantification analysis (RQA) to evaluate changes in fHRV patterns during four distinct stages: Pre-exposure (PRE), exposure to the first musical piece (STIM1), exposure to the second musical piece (STIM2), and post-exposure (POST). Our findings suggest that RQA indices (determinism, average diagonal line length, maximum line length, entropy, and trapping time) effectively characterize changes in the complexity of fHRV, with notable increases in regularity and predictability observed after music exposure. These results align with previous evidence indicating the potential of music to modulate fetal cardiac dynamics, offering new perspectives on the use of music in perinatal care to support fetal well-being.

The intricate relationship between music and physiological responses has long fascinated researchers, particularly its impact on fetal neurodevelopment. The influence of music on the fetal autonomic nervous system (ANS) and fetal heart rate variability (fHRV) has been documented across various gestational ages and conditions.1–3 However, its impact on nonlinear fetal heart rate dynamics remains underexplored, despite emerging evidence suggesting its significant potential.

Music therapy, a non-invasive intervention, has gotten attention for its ability to modulate physiological and psychological states. Brillo et al. (2019) found that fetuses exposed to specific musical pieces showed significant increases in heart rate accelerations and movements during listening sessions, indicating a response to music listened to during previous sessions.4 Relevant research has shown that fetuses at 30–38 weeks of gestational age exhibited less movement to fixed singing activities compared to irregular singing,5 reinforcing the idea that fetuses can discern and respond to different musical patterns. This nuanced modulation of fetal responses by musical exposure underscores its potential influence on fetal autonomic regulation.

Contrasting findings, however, emerged from a meta-analysis involving 1419 pregnant women, which found no significant changes in the fetal heart rate, number of fetal movements, or number of accelerations due to prenatal music therapy.6 That systematic review highlighted the complexity of music's impact, music type, exposure duration, and methodological variations among studies. In contrast, a relevant non-randomized controlled trial by Fathi et al. (2023) investigated the impact of classical music. That study found that while no significant differences were observed in the mean fetal heart rate between the groups, the intervention group exposed to classical music exhibited notable increases in heart rate accelerations and a decrease in the number of decelerations.7 Furthermore, Massimello et al. (2022) also provided evidence of the effect of classical music stimulation on fHRV.3 That study was relevant because it evaluated fetal heart rate variability on a beat-to-beat basis, offering a detailed analysis of the autonomic responses to music. Their findings indicated that although music did not alter the mean fetal heart rate, it significantly increased the total fHRV, possibly through potential parasympathetic activation. Critical insights into the optimal gestational periods for fetal auditory stimulation were provided by Hibiya et al. (2020).8 Their study showed significant increases in fetal heart rate accelerations from 26 weeks of gestation onward, with the most robust responses between 28 and 37 weeks. This period aligns with the development of fetal hearing capabilities, indicating a critical window for auditory interventions.

In addition to the observed effects on autonomic regulation, prenatal auditory stimulation, such as exposure to classical music, may have important implications for early cognitive development. Previous research has indicated that prenatal sensory stimulation can enhance brain plasticity, potentially facilitating the development of cognitive skills like memory and attention during early childhood.9,10 This suggests that the impact of prenatal auditory exposure extends beyond immediate physiological responses, potentially influencing long-term cognitive outcomes.

Previous studies have indicated that fetal heart rate exhibits fundamental nonlinearity, and valuable insights can be derived from analyzing the dynamics of fHRV using nonlinear statistical methods.11,12 Interestingly, recurrence quantification analysis (RQA), which quantifies recurrent patterns in time series, has been successfully applied to fetal cardiovascular time series in preclinical or experimental models of ewes, yielding promising results.13,14

According to the consulted literature, it remains a lack of comprehensive studies examining the nonlinear features of fetal heart rate fluctuations in response to music. Traditional linear methods of fHRV analysis may not fully capture the complex and dynamic changes induced by an auditory stimulus. A deeper understanding of how music influences fetal autonomic regulation may be achieved by complementing the conventional HRV analysis with other methods, such as the RQA.

The aims of this study are twofold: (i) To assess the response of classical music exposure on the nonlinear behavior of fetal heart rate fluctuations in fetuses between 32 and 40 weeks of gestation using RQA and (ii) to propose a physiological interpretation of the potential changes in the RQA indices, complemented by a conventional HRV analysis. We predict that the nonlinear dynamics of fetal heart rate fluctuations, as evaluated by RQA, will show significant modifications when comparing stages of no musical exposure with those having a musical stimulus.

This study was conducted at the Hospital Reina Madre Clínicas de la Mujer in Toluca, State of Mexico, Mexico, from April to July 2024. We enrolled 100 pregnant women in the third trimester, between the 28th and 40th weeks of gestation, during their prenatal follow-up consultations. Inclusion criteria specified that pregnancies be singleton, uncomplicated, and that the mothers were not obese. Exclusion criteria included pregnancies involving more than one fetus, such as twins or triplets, fetal malformations, and drug abuse. Participants were excluded from the study if they experienced intrauterine fetal death, gestational diabetes, gestational hypertension, intrauterine growth restriction, or withdrew consent. This non-probabilistic sampling method ensured a representative sample within the study period. All participants provided informed consent, and the research protocol was approved on March 20, 2024, by the Reina Madre Clínicas de la Mujer hospital's Bioethics Committee, which is registered with the Mexican National Bioethics Commission (CONBIOETICA 15CHB09020190318).

For data collection of transabdominal signals, we utilized the Monica AN24 maternal-fetal monitor (Monica AN24 ®, Monica Healthcare, Nottingham, UK) with a sampling frequency of 900 Hz.15 The AN24 device features three separate bipolar channel leads, which produce three separate electrical channels using a common lead. Additionally, a ground lead is used. Five electrodes were strategically placed on the maternal abdomen: One on the midline above the umbilicus, another to the left of the umbilicus, one on the midline just above the pubic hairline, one to the right of the umbilicus, and the final electrode close to the previous one with the silver studs in proximity (Fig. 1).

FIG. 1.

Bipolar detection channels and electrode placement on the maternal abdomen during the experiment. The Monica AN24 maternal-fetal monitor uses three bipolar channel leads, color-coded as orange, white, and green, which create three separate electrical channels with the yellow lead, serving as the common lead. A ground lead with a black color-coded connector is also utilized. These leads are connected to five electrodes positioned on the maternal abdomen. The electrodes are placed strategically: One on the midline just above the umbilicus, one to the left of the umbilicus, one on the midline above the pubic hairline, and two to the right of the umbilicus. Throughout the monitoring process, headphones are kept on the maternal abdomen to provide musical stimulation, specifically during the 5th to 15th min of the recording.

FIG. 1.

Bipolar detection channels and electrode placement on the maternal abdomen during the experiment. The Monica AN24 maternal-fetal monitor uses three bipolar channel leads, color-coded as orange, white, and green, which create three separate electrical channels with the yellow lead, serving as the common lead. A ground lead with a black color-coded connector is also utilized. These leads are connected to five electrodes positioned on the maternal abdomen. The electrodes are placed strategically: One on the midline just above the umbilicus, one to the left of the umbilicus, one on the midline above the pubic hairline, and two to the right of the umbilicus. Throughout the monitoring process, headphones are kept on the maternal abdomen to provide musical stimulation, specifically during the 5th to 15th min of the recording.

Close modal

The transabdominal signals were captured using disposable electrodes (Ambu ® BlueSensor VL) arranged in a bipolar configuration while the participants were seated. The electrode placement involved cleaning the abdominal area with an alcohol swab and lightly abrading the skin with sandpaper tape to lower skin impedance. This procedure ensured accurate and consistent data collection throughout the study.

Consistent with previous methodologies, including those used by Massimello et al. (2022), we verified the effect of auditory stimuli on fetuses using classical music. To explore different cultural impacts and ensure comprehensive analysis, we selected two musical pieces: “The Swan,” composed by Camille Saint-Saëns from Carnival of the Animals, is an instrumental quiet composition often associated with relaxation and tranquility due to its andantino tempo (80–108 beats per min) and melodic lines. This piece has a duration of 3 min and 5 s, in the key of G major, which has been used in music therapy studies for the rehabilitation of stroke patients.16 The classic Mexican song “Arpa de Oro,” composed by Abundio Martínez, is a traditional Mexican classical song with an instrumental quiet rhythmic structure, featuring an andante tempo (76–108 bpm) and a duration of 4 min and 50 s in the key of D major. Both pieces were played for 5 min each through stereo headphones (MDR-ZX110, Sony Corporation, Tokyo, Japan) placed on the mother's abdomen to minimize maternal acoustic interference (Fig. 1). The sound intensity during the entire musical stimulus averaged 60 dB, ranging from 45 to 74 dB, as verified with a sound level meter (1352H, TES Electrical Electronic Corp, Taipei, Taiwan). The volume was maintained below 85 dB to avoid potential harm to the fetuses.17 This corresponds to approximately 70% of the volume on a personal computer. The whole 20-min auditory stimulus including the 5 min baseline and post-exposure periods is available in an MP4 file, featuring a bitrate of 144 kbps, two stereo channels, and a sampling rate of 44.100 kHz. This file can be downloaded from the institutional repository of UAM-L at http://hdl.handle.net/20.500.12222/434 or https://xogi.ler.uam.mx/items/3c19d1de-7c8e-4765-9d59-b316e2bfd9fd.

The study was structured into four stages to thoroughly evaluate fetal heart rate fluctuations in response to the music. The first stage, PRE, involved 5 min of silence before any stimulus to establish a baseline. The second stage, STIM1, included a 5-min period during which the first musical piece, “The Swan,” was played. The third stage, STIM2, involved another 5-min period during which the second musical piece, “Arpa de Oro,” was played. Finally, the POST stage consisted of 5 min of silence following the musical exposure to observe any lasting effects.

To ensure synchronization between the fetal physiological signal and the musical stimuli, both the Monica AN24 device and the headphones used for musical playback were connected to the same data acquisition PC. The recordings were initiated simultaneously with the start of the musical playback on this unified system, ensuring that both signals were time-locked. This careful and simultaneous initialization allowed for the alignment of physiological data with the periods of musical exposure, facilitating reliable analysis of fetal heart rate responses in relation to the auditory stimuli.

To ensure a controlled environment and minimize acoustic interferences during the experiment, each session was conducted in the same quiet room with participants in a seated position. Mothers were specifically instructed to speak only if absolutely necessary, and all individuals present in the room were asked to remain silent throughout the session. Additionally, the room was free from external noise sources, and no other electronic devices that could produce sound were used during the experiments.

From transabdominal recordings of the 20-min sessions, the continuous fetal beat-to-beat R–R time series were extracted using Monica DK software (Monica Healthcare, Nottingham, UK). This software employs a subtraction algorithm to remove the maternal ECG complex before detecting the fetal ECG complexes.9,18,19 Once extracting the fetal R–R time series with this software, only those series with a global loss of <10% of fetal beats during the 20-min recording period were selected. Additionally, the Monica DK software was also used to estimate periods of acceleration and deceleration in each stage. Large accelerations were defined as periods lasting 15 or more seconds above the fetal heart rate baseline, with a peak increase of 15 or more beats per min. Large decelerations were identified as periods lasting 15 or more seconds below the fetal heart rate baseline, with a significant drop in beats per min.

The fetal R–R signals were then processed using an adaptive filter as proposed by Wessel et al. (2000).20 The filtering algorithm improves heart rate data accuracy through three key steps. First, it removes implausible R–R intervals, such as those of zero length, beat-to-beat intervals shorter than 200 ms (which are below the human refractory period), and significant pauses where the heart does not show a contraction. This step ensures the basic integrity of the data. The second step applies an adaptive percent filter that accounts for heart rate variability by smoothing the data and excluding outliers, based on an adaptive mean and standard deviation calculated from a binomial-7-filtered series. This process maintains the natural variability of the heart rate while effectively removing anomalies. Lastly, an adaptive control filter re-evaluates the data to detect and correct any remaining irregularities, ensuring that the final dataset accurately reflects true cardiac behavior. This systematic approach is crucial for reliable fHRV analysis. A MATLAB script for this implementation is freely available for download at https://tocsy.pik-potsdam.de/ada.php

Finally, the filtered fetal R–R time series were temporally segmented into PRE, STIM1, STIM2, and POST stages (Fig. 2). The initial sample consisted of 100 cases; however, after applying the selection or elimination criteria, only 37 fetal R–R time series remained, resulting in a final sample size of N = 37. The unfiltered fetal R–R time series are freely accessible in the institutional repository of UAM-L at the following link: http://hdl.handle.net/20.500.12222/434 or https://xogi.ler.uam.mx/items/3c19d1de-7c8e-4765-9d59-b316e2bfd9fd

FIG. 2.

Representative continuous fetal R–R interval time series for the entire experiment in a single fetus. The graph illustrates the fluctuation of fetal R–R intervals throughout the whole recording. The segments are labeled as follows: PRE represents the period before the musical stimulation, STIM1 indicates the musical stimulation with the first song, STIM2 represents the musical stimulation with the second song, and POST is the period after the musical stimulation. Subsequently, the 5-min segments of each stage were analyzed and compared.

FIG. 2.

Representative continuous fetal R–R interval time series for the entire experiment in a single fetus. The graph illustrates the fluctuation of fetal R–R intervals throughout the whole recording. The segments are labeled as follows: PRE represents the period before the musical stimulation, STIM1 indicates the musical stimulation with the first song, STIM2 represents the musical stimulation with the second song, and POST is the period after the musical stimulation. Subsequently, the 5-min segments of each stage were analyzed and compared.

Close modal
To visualize recurrences in a time series or dataset, we calculate the N × N recurrence matrix as follows:
(1)

In this equation, R i , j represents the element of the recurrence matrix at the ith row and jth column, indicating whether the state at time point i is recurrent with the state at time point j. The Heaviside step function Θ assigns a value of 1 if the distance between the state vectors x i and x j is less than the predefined threshold εi, and 0 otherwise. Here, εi denotes the threshold distance, which determines the maximum allowable distance for two states to be considered recurrent. The norm | | | |, such as the Euclidean norm, measures the distance between these state vectors.

The vectors x i and x j are points in an m-dimensional phase space R m, where m is the embedding dimension, a critical parameter that defines the number of delayed copies of the time series used to reconstruct the phase space. The embedding time delay τ is the interval between these delayed copies and is essential for capturing the underlying dynamics of the time series. The phase space vectors for the one-dimensional time series ui from observations are reconstructed using the following equation:
(2)

The indices i and j run from 1 to N, representing each time point in the series, with N being the total number of data points.

A multidimensional state space is reconstructed from the one-dimensional fetal R–R intervals in the time series using the time-delay embedding method. Each point in this reconstructed phase space corresponds to the system's state at a specific time and is defined by the m coordinates of a chosen embedding dimension.

The embedding time delay or τ was determined as the first zero crossing point of the averaged autocorrelation function across all recordings.21,22 Figure 3 (upper panels) displays the autocorrelation function for each individual recording (thin lines) and the averaged values per group (thick lines). The averaged autocorrelation function crosses zero between lags 14 and 16. Based on this observation, we chose an embedding time delay τ = 14 for all analyses.

FIG. 3.

Assessment of the autocorrelation function (upper panels) and false nearest neighbors' method (lower panels) for each recording (thin lines) and averaged values per group (thick lines). The groups correspond to different stages of musical stimulation: PRE (before stimulation), STIM1 (during the first phase of stimulation), STIM2 (during the second phase of stimulation), and POST (after stimulation) in fetal R–R intervals.

FIG. 3.

Assessment of the autocorrelation function (upper panels) and false nearest neighbors' method (lower panels) for each recording (thin lines) and averaged values per group (thick lines). The groups correspond to different stages of musical stimulation: PRE (before stimulation), STIM1 (during the first phase of stimulation), STIM2 (during the second phase of stimulation), and POST (after stimulation) in fetal R–R intervals.

Close modal

Figure 3 (lower panels) shows the proportion of false nearest neighbors as a function of the embedding dimension for each recording (thin lines) and the averaged values per group (thick lines). We observed that the proportion of false nearest neighbors decreased sharply with increasing embedding dimension for most recordings. Specifically, at least 50% of the recordings met the criterion of having less than 10% (0.1) false nearest neighbors at an embedding dimension of 5. According to Abarbanel and Kennel (1996), an embedding dimension where the percentage of false nearest neighbors falls below 10% is considered sufficient for accurately reconstructing the system's dynamics without significant projection errors.23 Therefore, we selected m = 5 as the embedding dimension for all recordings. The distances between individual points in the matrix, representing the state of the system at a specific time, were calculated using the fixed amount of nearest neighbors (“fan”) option in the recurrence plots toolbox. A consistent threshold distance (εi) of 7% was applied to all recordings to define the radius of the neighborhood.

The following RQA measures such as determinism (DET), averaged diagonal line length (L), maximum line length (LMAX), laminarity (LAM), entropy (ENTR), trapping time (TT), recurrence times of first type (T1), and recurrence times of second type (T2)24 were derived from the fetal R–R time series of PRE, STIM1, STIM2, and POST stages and computed by using the Cross Recurrence Plot Toolbox for MATLAB,25 Version 5.29 (R38) freely available in https://tocsy.pik-potsdam.de/CRPtoolbox/.

Determinism (DET) is defined as the proportion of recurrence points that form diagonal lines of at least a specified minimum length (lmin) in relation to the total number of recurrence points. In this study, lmin was set to two data points for all analyses, following the default setting commonly used in RQA analysis.26 In Eq. (2), P(l) represents the frequency distribution of these diagonal line lengths,27 
(3)
The average length of diagonal lines (L) quantifies the duration for which the trajectories in the phase space remain within the same regions, indicating the system's predictability over time,27 
(4)
Maximum line length (LMAX) refers to the length of the longest diagonal sequence of recurrent states within a system. This metric offers insights into the stability of the underlying attractor dynamics. Additionally, it establishes a theoretical link to the broader field of dynamical systems: The maximum line length is inversely proportional to the largest Lyapunov exponent, a commonly used measure of attractor stability.28 The formula for LMAX is given in Eq. (4) as follows:
(5)

In this formula, l represents the length of each diagonal sequence of recurrent states and P(l) denotes the frequency distribution of these lengths.

Entropy (ENTR) measures the system's complexity by evaluating the Shannon entropy of the line length distribution. The entropy obtained from a recurrence plot is inversely proportional to the largest Lyapunov exponent. As a result, it decreases when the signal's complexity and chaotic behavior increase,29 
(6)
Laminarity (LAM) measures the proportion of recurrent points that form vertical line structures. These vertical lines indicate patterns where the system remains in a single state, exhibiting the same behavior repeatedly.30 In the following formula, w denotes the length of the vertical lines, P(w) represents the probability of finding a vertical line of length w, and wmin is the minimum length considered. In this study, wmin was set to two data points for all analyses, following the default setting commonly used in RQA analysis,25 
(7)
Trapping time measures the average length of vertical lines in a plot,30 indicating the stability of trapped states. The formula for trapping time (TT) is
(8)
Recurrence time of the first type (T1) quantifies the temporal distance between recurrent points in the ith column of a recurrence plot. It is also viewed as the average duration for a point in the embedding space to be revisited,31 
(9)
Recurrence time of the second type (T2) indicates whether a point belongs to the recurrence vector. It is also seen as the average time for a point in the embedding space to be revisited, excluding one-unit time intervals,31 
(10)

Conventional linear HRV indices were calculated from the filtered fetal R–R time series extracted with the Monica DK using PyBios software.32 The indices included the mean R–R period (RRmean), the mean heart rate (HRmean), the SDRR (standard deviation of all R–R intervals), the root mean square of successive differences between normal-to-normal R–R intervals (RMSSD),33 and the pNN5 (percentage of successive R–R intervals with differences greater than 5 ms).34 Power spectral indices were obtained by first removing a linear trend, resampling at 4 Hz, and applying a non-overlapped Hanning window of 300 data points with 50% overlap. The frequency domain indices included normalized low-frequency power (LFnu) in the 0.05–0.2 Hz range, normalized high-frequency power (HFnu) in the 0.2–1 Hz range, and the LF/HF ratio.35 

Continuous variables are reported as mean ± standard deviation or median (25th percentile–75th percentile), depending on the data distribution. Normality was assessed using the Kolmogorov–Smirnov test. For normally distributed indices, we used the mean and parametric tests (one-way repeated measures ANOVA with Tukey's post-hoc tests), estimating effect sizes with eta squared (η2). For non-normally distributed indices, we used the median and non-parametric tests (Friedman test with Dunn's multiple comparisons), estimating effect sizes with Kendall's W. This a priori approach aligns the choice of central tendency measure, statistical test, and effect size estimation with the characteristics of the data, ensuring accurate and reliable results. All analyses were performed using GraphPad Prism version 8.4.2 (GraphPad Software, La Jolla, CA), with a significance level set at p ≤ 0.05.

From the 100 recorded cases, 57 R–R time series were excluded because these had a global loss greater than 10% of fetal beats during the 20-min recording period. Additionally, six more time series were eliminated because the women were diagnosed with gestational diabetes. Thus, a total of 37 fetal R–R time series were analyzed in the PRE, STIM1, STIM2, and POST segments.

Table I summarizes the relevant clinical characteristics of the mothers, fetuses, and newborns in the study population. The average gestational age was 36.2 ± 3.2 weeks, and the mean maternal age was 29.2 ± 6.0 years. The mean maternal weight and height were 74.6 ± 13.7 kg and 1.64 ± 0.10 m, respectively.

TABLE I.

Clinical characteristics of the study population.

Variable
Sample (N)  37 
Weeks of gestation (weeks)  36.2 ± 3.2 
Maternal age (years)  29.2 ± 6.0 
Maternal weight (kg)  74.6 ± 13.7 
Maternal height (m)  1.64 ± 0.10 
Birth weight (g)  3149 ± 394 
APGAR at 5 min (points)  8.8 ± 0.5 
Biological gender (Masculine, %)  49 
Variable
Sample (N)  37 
Weeks of gestation (weeks)  36.2 ± 3.2 
Maternal age (years)  29.2 ± 6.0 
Maternal weight (kg)  74.6 ± 13.7 
Maternal height (m)  1.64 ± 0.10 
Birth weight (g)  3149 ± 394 
APGAR at 5 min (points)  8.8 ± 0.5 
Biological gender (Masculine, %)  49 

Notably, the number of cumulative accelerations of the fetal heart rate showed variations across the different stages. The raw number of cumulative accelerations for all cases in the PRE stage was 24, in the STIM1 stage was 24, in the STIM2 stage was 21, and in the POST stage was 31. It is important to highlight that a higher number of cumulative long accelerations in the fetal heart rate was observed in the POST stage. Additionally, no decelerations in the fetal heart rate were observed in any of the stages according to the criteria described above.

Table II shows the RQA indices of fetal R–R time series in 37 fetuses across four stages of musical stimulation: PRE, STIM1, STIM2, and POST. The metrics were first calculated individually for each fetus in each condition. Subsequently, group-level means and medians were obtained by averaging or taking the median of the individual values across all fetuses in each condition. The results are presented as mean ± standard deviation (SD) or median (25th percentile–75th percentile). All calculations were performed with a time delay of 14 and an embedding dimension of 5.

TABLE II.

Recurrence quantification analysis (RQA) indices of fetal heart rate fluctuations of 37 fetuses across four stages of musical stimulation: PRE (before stimulation), STIM1 (during the first phase of stimulation with the first piece of music), STIM2 (during the second phase of stimulation with the second piece of music), and POST (after stimulation). Data are shown as mean ± standard deviation (SD) or median (25th percentile–75th percentile). All calculations were performed with a time delay of 14 and an embedding dimension of 5.

PRE STIM1 STIM2 POST
DET  0.455 ± 0.135  0.490 ± 0.140  0.484 ± 0.134  0.514 ± 0.138a 
2.52 (2.32–2.82)  2.58 (2.38–3.00)  2.58 (2.38–2.96)b  2.67 (2.44–3.13)c 
LMAX  10.0 (8.5–15.0)  14.0 (8.50–25.50)  17.0 (9.0–33.5)d  13.0 (9.0–27.0) 
LAM  0.600 ± 0.114  0.628 ± 0.113  0.620 ± 0.107  0.644 ± 0.109 
ENTR  1.00 ± 0.31  1.08 ± 0.34  1.08 ± 0.36  1.14 ± 0.33a 
TT  3.01 (2.65–3.35)  3.04 (2.69–3.59)  3.05 (2.74–3.56)c  3.20 (2.81–3.63)b 
VMAX  14.0 (11.5–19.0)  17.0 (13.0–23.0)  16.0 (13.5–21.0)  17.0 (13.5–22.05) 
T1  14.51 (13.27–15.62)  14.97 (13.93–15.80)  14.73 (13.75–15.52)  14.18 (13.55–15.83) 
T2  24.67 (21.65–27.97)  25.56 (22.04–32.10)  25.23 (22.32–30.91)  23.32 (13.28–32.05) 
PRE STIM1 STIM2 POST
DET  0.455 ± 0.135  0.490 ± 0.140  0.484 ± 0.134  0.514 ± 0.138a 
2.52 (2.32–2.82)  2.58 (2.38–3.00)  2.58 (2.38–2.96)b  2.67 (2.44–3.13)c 
LMAX  10.0 (8.5–15.0)  14.0 (8.50–25.50)  17.0 (9.0–33.5)d  13.0 (9.0–27.0) 
LAM  0.600 ± 0.114  0.628 ± 0.113  0.620 ± 0.107  0.644 ± 0.109 
ENTR  1.00 ± 0.31  1.08 ± 0.34  1.08 ± 0.36  1.14 ± 0.33a 
TT  3.01 (2.65–3.35)  3.04 (2.69–3.59)  3.05 (2.74–3.56)c  3.20 (2.81–3.63)b 
VMAX  14.0 (11.5–19.0)  17.0 (13.0–23.0)  16.0 (13.5–21.0)  17.0 (13.5–22.05) 
T1  14.51 (13.27–15.62)  14.97 (13.93–15.80)  14.73 (13.75–15.52)  14.18 (13.55–15.83) 
T2  24.67 (21.65–27.97)  25.56 (22.04–32.10)  25.23 (22.32–30.91)  23.32 (13.28–32.05) 
a

p < 0.05, PRE vs POST by Tukey's multiple comparisons test.

b

p < 0.05, PRE vs STIM2 by Dunn's multiple comparisons test.

c

p < 0.05, PRE vs POST by Dunn's multiple comparisons test.

d

p < 0.01, PRE vs STIM2 by Dunn's multiple comparisons test.

The mean DET values significantly increased from PRE (0.455 ± 0.135) to POST (0.514 ± 0.138) (ANOVA: F = 3.321, p = 0.0226, η2 = 0.024, indicating a small effect size; Tukey's multiple comparisons test, p = 0.0117).

Median values of L significantly increased from PRE (2.52 [2.32–2.82]) to POST (2.67 [2.44–3.13]), p = 0.0414 (Dunn's multiple comparisons test; Friedman: Fr = 10.56, p = 0.0144, Kendall's W = 0.095, indicating a small effect size) and from PRE (2.52 [2.32–2.82]) to STIM2 (2.58 [2.38–2.96]), p = 0.0238.

Median LMAX values significantly increased from PRE (10.0 [8.5–15.0]) to STIM2 (17.0 [9.0–33.5]) (Friedman: Fr = 12.73, p = 0.0053, Kendall's W = 0.115, indicating a small effect size; Dunn's multiple comparisons test, p = 0.0037).

The mean ENTR values significantly increased from PRE (1.00 ± 0.31) to POST (1.14 ± 0.33) (ANOVA: F = 3.251, p = 0.0246, η2 = 0.023, indicating a small effect size; Tukey's multiple comparisons test, p = 0.0131).

TT median values significantly increased from PRE (3.01 [2.65–3.35]) to POST (3.20 [2.81–3.63]), with significant differences between PRE and POST (Dunn's multiple comparisons test, p = 0.0238; Friedman: Fr = 11.85, p = 0.0079, Kendall's W = 0.107, indicating a small effect size) and from PRE (3.01 [2.65–3.35]) to STIM2 (3.05 [2.74–3.56]), p = 0.0132. The other RQA indices did not show significant differences between the different stages.

Figure 4 presents a representative response of fHRV to musical stimulation across the study stages in a single fetus. The top two rows show the underlying time series for R–R intervals during each stage. It is apparent that these time series exhibit different patterns of variability and complexity across the stages of the study. The lower row shows the recurrence plots for each stage, where changes in the density and structure of recurrence points reflect the variations observed in the RQA indices. Specifically, in this example, some relevant RQA values are DET rising from 0.6221 (PRE) to 0.8183 (STIM1), then slightly decreasing to 0.8030 (STIM2) and 0.7878 (POST); L increasing from 3.14 (PRE) to 4.47 (STIM1), then decreasing to 4.29 (STIM2) and 4.07 (POST); and LMAX increasing from 17 (PRE) to 71 (STIM1), then decreasing to 61 (STIM2) and 49 (POST). Additionally, some linear fHRV values are HRmean changing from 147 bpm (PRE) to 156 bpm (STIM1), then to 147 bpm (STIM2) and 145 bpm (POST); SDRR increasing from 27.9 ms (PRE) to 29.6 ms (STIM1), 35.4 ms (STIM2), and 35.7 ms (POST).

FIG. 4.

Representative example of the response of fetal heart rate fluctuations (or fetal heart rate variability) to musical stimulation at different stages in a single fetus. The stages are PRE (before stimulation), STIM1 (during the first phase of stimulation), STIM2 (during the second phase of stimulation), and POST (after stimulation). The figure shows the underlying time series (upper panels) and recurrence plots (lower panels). Reconstruction parameters are delay = 14 and embedding dimension = 5 for all stages.

FIG. 4.

Representative example of the response of fetal heart rate fluctuations (or fetal heart rate variability) to musical stimulation at different stages in a single fetus. The stages are PRE (before stimulation), STIM1 (during the first phase of stimulation), STIM2 (during the second phase of stimulation), and POST (after stimulation). The figure shows the underlying time series (upper panels) and recurrence plots (lower panels). Reconstruction parameters are delay = 14 and embedding dimension = 5 for all stages.

Close modal

Table III reports the results of the conventional linear HRV indices for the study, as described previously. Interestingly, significant differences were found only in the SDRR index between the STIM1 and POST phases. The SDRR values were 18.81 ± 6.84 ms during STIM1 and 21.94 ± 7.93 ms in the POST phase (ANOVA: F = 2.824, p = 0.0422, η2 = 0.027, indicating a small effect size; Tukey's multiple comparisons test, p = 0.0464), indicating an increase in the fetal heart rate variability following the musical stimulus.

TABLE III.

Conventional indices of heart rate fluctuations of 37 fetuses across four stages of musical stimulation: PRE (before stimulation), STIM1 (during the first phase of stimulation with the first piece of music), STIM2 (during the second phase of stimulation with the second piece of music), and POST (after stimulation). Data are shown as mean ± standard deviation (SD) or median (25th percentile–75th percentile).

PRE STIM1 STIM2 POST
RRmean (ms)  435.0 ± 25.99  435.7 ± 25.45  435.2 ± 24.71  434.4 ± 26.21 
HRmean (bpm)  138.4 ± 8.2  138.2 ± 8.1  138.3 ± 7.7  138.6 ± 7.9 
SDRR (ms)  19.22 ± 6.97  18.81 ± 6.84  19.49 ± 7.99  21.94 ± 7.93a 
RMSSD (ms)  10.89 ± 4.35  9.83 ± 2.81  10.14 ± 3.21  10.94 ± 3.72 
pNN5 (%)  39.55 ± 12.40  38.73 ± 9.41  39.05 ± 9.32  40.37 ± 11.00 
HFnu  43.1 (30.5–57.0)  34.7 (22.0–51.9)  40.1 (24.5–50.2)  39.6 (24.9–56.8) 
LFnu  56.9 (42.9–69.5)  65.3 (48.1–77.9)  59.9 (49.8–75.5)  60.4 (43.1–75.1) 
LF/HF  1.32 (0.75–2.27)  1.87 (0.92–3.53)  1.49 (0.99–3.09)  1.52 (0.75–3.02) 
PRE STIM1 STIM2 POST
RRmean (ms)  435.0 ± 25.99  435.7 ± 25.45  435.2 ± 24.71  434.4 ± 26.21 
HRmean (bpm)  138.4 ± 8.2  138.2 ± 8.1  138.3 ± 7.7  138.6 ± 7.9 
SDRR (ms)  19.22 ± 6.97  18.81 ± 6.84  19.49 ± 7.99  21.94 ± 7.93a 
RMSSD (ms)  10.89 ± 4.35  9.83 ± 2.81  10.14 ± 3.21  10.94 ± 3.72 
pNN5 (%)  39.55 ± 12.40  38.73 ± 9.41  39.05 ± 9.32  40.37 ± 11.00 
HFnu  43.1 (30.5–57.0)  34.7 (22.0–51.9)  40.1 (24.5–50.2)  39.6 (24.9–56.8) 
LFnu  56.9 (42.9–69.5)  65.3 (48.1–77.9)  59.9 (49.8–75.5)  60.4 (43.1–75.1) 
LF/HF  1.32 (0.75–2.27)  1.87 (0.92–3.53)  1.49 (0.99–3.09)  1.52 (0.75–3.02) 
a

p < 0.05, STIM1 vs POST by Tukey's multiple comparisons test.

The spectral indices, including HFnu, LFnu, and the LF/HF ratio, showed a Friedman statistic of 7.476 with a p-value of 0.0582 (Kendall's W = 0.0168, indicating a very small effect). For individual comparisons using Dunn's multiple comparisons test, the comparison between PRE and STIM1 indicated near-significant differences (p = 0.0541). Specifically, HFnu decreased from 43.1 (30.5–57.0) in PRE to 34.7 (22.0–51.9) during STIM1, while LFnu increased from 56.9 (42.9–69.5) to 65.3 (48.1–77.9). Similarly, the LF/HF ratio increased from 1.32 (0.75–2.27) in PRE to 1.87 (0.92–3.53) in STIM1. These changes reflect a consistent pattern among the indices, with LFnu and LF/HF ratio increasing and HFnu decreasing during STIM1, although these changes did not reach statistical significance.

This study explored the effects of two musical stimuli on fetal heart rate fluctuations in women during the third trimester of pregnancy. Notably, this is one of the first studies using RQA to evaluate some non-linear properties of fetal cardiac dynamics resulting from musical exposure at the perinatal stage using RQA. Our findings can be summarized into three main points. First, we observed an increase in determinism, more recurrent behaviors, lower complexity, and greater stability of trapped states in the recurrence plots generated from the fetal R–R time series in the POST compared to the PRE stage, as indicated by DET, L, ENTR, and TT, respectively. Second, the STIM2 stage uniquely enhanced the predictability and stability of cardiac dynamics compared to the PRE stage, as indicated by L, LMAX, and TT. However, no significant changes were observed during the STIM1 stage when compared to the PRE stage across all RQA measures and conventional HRV indices. This lack of significant findings with STIM1 suggests that the first musical stimulus did not elicit measurable changes in fetal cardiac dynamics, in contrast to the effects observed with STIM2 and POST stages. Third, no significant changes were observed in the conventional HRV indices, including HRmean, except for an increase in total heart rate variability measured by SDRR in the POST stage compared to STIM1.

An increase in DET from the PRE to POST stage may indicate a shift toward more regular and predictable patterns in the fetal R–R intervals following the musical exposure. This enhanced predictability suggests that the musical auditory stimuli elicited a response in the fetal cardiac autonomic system, leading to more deterministic heart rate behaviors. In a related study conducted by Dimitriev et al. in 2023,36 it was found that DET levels were notably elevated in adults exposed to stimulating music compared to silence, implying a shift from stochastic to deterministic patterns in heart rate behavior. Additionally, recent evidence by Wright and Palmer indicates that nonlinear cardiac dynamics in adults were also more deterministic during rhythm perception than synchronization.37 

The increase in L, LMAX, and TT from PRE to POST and from PRE to STIM2 indicates more prolonged and regular fHRV behaviors. The longer diagonal lines and increased trapping states in the recurrence plots for STIM2 and POST suggest sustained regularity and stability in fHRV patterns. This may reflect a robust autonomic response to the musical stimuli, particularly during the second musical piece, “Arpa de Oro,” which induced extended periods of regularity and stability in the fHRV. This suggests that musical exposure, especially the specific characteristics of “Arpa de Oro,” may modulate the fetal cardiac autonomic system. A study found that in healthy male volunteers, the average R–R interval variability and mean blood pressure remained stable (as a result of musical stimulus). However, the LMAX in the RQA was significantly lower in a silence condition compared to noise levels at 70, 80, and 90 dB, indicating that RQA indices effectively assess low-frequency noise exposure in short-term HRV time series.38 Interestingly, previous findings have shown that the electrocardiogram analysis of 18 volunteers before and after listening to a motivational song showed cardiac activity changes. An artificial neural network (ANN) classification confirmed this with over 85% accuracy, highlighting key RQA features TT, LMAX, and DET.39 Other evidence suggests that cardiac measures at early testing times indicated increased heart rates and more predictable cardiac dynamics in pianists during music performance than at baseline rest.40 

In our results, the decrease in ENTR from PRE to POST suggests a reduction in the complexity and nonlinear-like behavior of the fHRV patterns post-musical stimulus. Lower entropy values indicate more regular and predictable patterns, aligning with the observed increases in DET, L, TT, and LMAX. Contrarily, some studies have revealed an increase in cardio-vagal modulation and higher complexity, assessed by short-term HRV indices, suggesting that music, especially of higher frequency, has a positive relaxing effect on the human adult organism.41 Therefore, we speculate that the observed reduction in complexity and increased predictability in our fetal study might reflect the developed physiological state of the fetal autonomic nervous system at the third trimester. It is possible that fetuses respond differently to auditory musical stimuli compared to adults, perhaps because such stimuli either evoke discomfort or engage the evolving sensory perception of the fetus in a distinct manner.

Accordingly, the changes observed here in the recurrence measures of fHRV dynamics might mirror the increase in constraints on heartbeat variability during music listening. Specifically, an increase in TT, L, and LMAX from PRE to POST and from PRE to STIM2 can be interpreted as a stronger attractor for heartbeat fluctuations. This might suggest that the fetal cardiac autonomic system adapts to the auditory stimuli, potentially increasing its regularity and stability in response to music. Using RQA on electroencephalographic time series from 20 subjects, significant differences in brain activity were found when listening to happy vs sad North Indian Classical Music, indicating that RQA parameters can detect emotional changes due to music. Specifically, during happy music, the average entropy in the left frontal region (F3–F4) was approximately 0.64, while during sad music, it was around 0.51. In the fronto-temporal region (F7–T8), the entropy values were about 0.70 for happy music and 0.50 for sad music. Although our study did not evaluate emotional aspects, we found that entropy appears to discriminate between the PRE and POST stages, suggesting that entropy is an RQA parameter sensitive to musical stimuli at both cardiac and cerebral levels.42 

The role of the intrauterine and placental environment in modulating fetal responses to music was further emphasized by Pino (2023),43 showing that music stimulation can influence neurobehavioral development and parental interaction. Additionally, a systematic review of the benefits of music therapy on prenatal and delivery experiences concluded that music therapy reduces pain during childbirth, lowers maternal anxiety and stress, improves sleep quality, and increases fetal movements and basal fetal heart rate.44 Specifically, during the POST period, a higher number of raw cumulative accelerations of the fetal heart rate were observed. These accelerations might be associated with fetal movements that manifest in coincidence with the increase in the SDRR index at the final stage of the study. This could also explain the trends observed in spectral indices. In a cohort primarily consisting of term pregnancies, Van Laar et al. found that both LF and LFnu were higher in the active state compared to the quiet state. Although HF also showed a tendency toward an increase in the active state compared to the resting state, this was not significant, and HFnu decreased in the active state compared to the resting state.45 

Studies like those by Tristao (2020) further support these findings as they have demonstrated the value of combining behavioral scales and fHRV measures to understand fetal responses to different tempos and sources of music, indicating a sophisticated level of fetal auditory processing and memory.46 The review by Movalled (2023) suggests that prenatal sound stimulation can form specific memory traces during the fetal period and affect the neonatal neural system, highlighting the potential long-term benefits of music exposure in utero.47 

Furthermore, in addition to the observed effects on autonomic regulation, prenatal music exposure might also play a role in shaping the emotional development of the fetus. Repeated exposure to certain types of music could facilitate early familiarization and emotional responsiveness, potentially influencing mood regulation and emotional development postnatally. These early influences suggest that music might contribute to the formation of postnatal emotional responses, underscoring the need for further research into how prenatal experiences, such as auditory stimulation, contribute to early emotional development and mood regulation.

Several limitations should be acknowledged in this study. Based on evidence from studies on adults, factors such as the genre, tempo, volume, and duration of the music samples can influence HRV.48 In our study, we selected two instrumental classical pieces: “The Swan” by Camille Saint-Saëns, representing European classical music, and “Arpa de Oro” by Abundio Martínez, representing Mexican classical music. Both pieces featured tempos of andantino and andante, average volumes of 60 dB, and durations of 5 min each. These pieces were chosen for their typical musical characteristics and cultural diversity, providing a robust framework to study fetal HRV responses to music. Future research may build upon these findings by exploring the effects of other musical styles or characteristics.

Despite initially enrolling 100 patients, only 37 R–R time series were included in the final analysis due to the challenges in acquiring and recording continuous fetal ECG signals that meet stringent quality selection criteria. Moreover, the small effect sizes observed in our study may limit the strength and generalizability of our findings. In future studies, we aim to increase the sample size to enhance the statistical power and robustness of our results, which may help to detect subtle changes in fetal heart rate variability in response to musical stimuli.

The absence of a direct assessment of fetal behavioral states (fBS) during the POST stage limits our ability to definitively attribute the changes in fHRV to specific fetal movements or behavioral states. The gold standard for fBS classification involves expert visual inspection of fHRV patterns and actograms, often complemented by additional monitoring techniques. Future studies incorporating comprehensive fBS evaluations could provide more detailed insights into the fetal responses observed, particularly in understanding whether the increase in cumulative accelerations post-music application indicates a shift to a more active behavioral state—perhaps suggesting that the fetuses became more alert after the music stopped. It is important to note that fetal movements should be considered alongside fHRV both scientifically and clinically to achieve a more detailed interpretation of results and to minimize the risk of misinterpreting fHRV outcomes.49 According to Brändle et al. (2015), an active state (state 4F) is characterized by accelerations greater than 15 bpm accompanied by fetal movements.50 Their study demonstrated that fHRV parameters could help differentiate between fetal behavioral states and provide insights into neurovegetative development in utero. For instance, relevant evidence suggests that parameters like the SDRR were higher in active states compared to passive ones.51 Although we did not conduct a formal fBS assessment, the observed increase in SDRR during the POST stage and the rise in cumulative accelerations may suggest increased fetal movement and a possible transition to a more active behavioral state. Additionally, studies like that of Frank et al. (2006) have shown that nonlinear indices can improve fBS classification based on fetal heart rate analysis.52 Future research should consider employing RQA and novel classification methods to effectively evaluate changes in fBS.

We acknowledge that the non-randomized order of musical stimulus presentation constitutes a significant limitation of our study. Specifically, the fixed sequence of PRE → STIM1 → STIM2 → POST may introduce a systematic exposure bias, wherein the progression from the first to the second stimulation could inherently influence fHRV independent of the specific musical pieces administered. This potential bias might affect the consistency and comparability of the responses observed between STIM1 and STIM2, thereby impacting the overall interpretation of our results. Consequently, our findings should be interpreted with caution, recognizing that order effects may confound the relationship between musical exposure and fetal autonomic regulation. To address this limitation, future research should employ a randomized or counterbalanced design for the presentation of musical stimuli.

Finally, another limitation of our study is the potential influence of varying numbers of data points in the 5-min time series across different stages. Since the number of data points depends on the fetal heart rate, which may slightly vary between stages, this variability could potentially affect the RQA metrics. Although we did not explicitly examine whether the number of data points influenced the RQA metrics, we observed no significant differences in mean heart rate or mean R–R intervals across the four stages (Table III). Therefore, we do not expect that the varying number of data points systematically affected our results. Future research should consider investigating the potential impact of time series length and data point variability on RQA metrics in fetal heart rate variability analysis to better understand these potential effects.

This study assessed the effects of two musical stimuli on fetal heart rate fluctuations in women during the third trimester of pregnancy using recurrence quantification analysis (RQA). Our findings highlight three main points. First, there was an increase in determinism, more recurrent behavior, lower complexity, and greater stability of trapped states in the POST stage compared to PRE, indicated by DET, L, ENTR, and TT, respectively. Second, the STIM2 stage uniquely enhanced the predictability and stability of cardiac dynamics compared to PRE, as indicated by L, LMAX, and TT. Third, no significant changes were observed in conventional HRV indices, except for an increase in total heart rate variability measured by SDRR in the POST stage compared to STIM1.

The decrease in ENTR from PRE to POST suggests a reduction in the complexity and nonlinear behavior of fHRV patterns post-stimulus, indicating more regular and predictable patterns. This may reflect the developed physiological state of the fetal autonomic nervous system at the third trimester. The observed increase in SDRR during the POST phase could be coincident with the manifestation of fetal movements, emphasizing the need for concomitant ultrasound evaluations in future studies. Overall, this study underscores the potential of music to enhance the predictability and stability of fetal cardiac dynamics, offering new insights into the effect of music in perinatal care.

We thank all the staff of Hospital Reina Madre Clínicas de la Mujer. Additionally, we appreciate the participation of Selina Magaly Vicente-Reyes for their support in data collection.

The authors have no conflicts to disclose.

All participants provided informed consent, and the research protocol was approved by the Reina Madre Clínicas de la Mujer hospital's Bioethics Committee, which is registered with the Mexican National Bioethics Commission (CONBIOETICA 15CHB09020190318). Participants got voluntarily involved in the research project, full information was given before the study, and they gave signed consent of their participation. Participants' integrity was respected at all times.

 José Javier Reyes-Lagos and Hugo Mendieta-Zerón contributed equally to this paper.

José Javier Reyes-Lagos: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Supervision (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Hugo Mendieta-Zerón: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Methodology (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Migdania Martínez-Madrigal: Data curation (equal); Formal analysis (supporting)l; Methodology (supporting); Writing – review & editing (supporting). Juan Carlos Santiago-Nuñez: Data curation (supporting); Formal analysis (supporting); Visualization (supporting). Luis Emilio Reyes-Mendoza: Data curation (supporting); Methodology (supporting); Writing – review & editing (supporting). Ximena Gonzalez-Reyes: Data curation (supporting); Supervision (supporting). Juan Carlos Echeverría: Conceptualization (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal). Eric Alonso Abarca-Castro: Conceptualization (equal); Formal analysis (equal); Supervision (equal); Writing – review & editing (equal). Ana Karen Talavera-Peña: Data curation (equal); Formal analysis (supporting); Methodology (supporting); Writing – review & editing (supporting). Sara Avilés-Hernández: Data curation (equal); Methodology (equal); Resources (equal). Claudia Lerma: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Methodology (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).

The musical stimulus data and fetal R–R interval signals that support the findings of this study will be freely accessible in the institutional repository of UAM Lerma (https://xogi.ler.uam.mx/items/3c19d1de-7c8e-4765-9d59-b316e2bfd9fd or http://hdl.handle.net/20.500.12222/434). For any additional data requests, please contact the corresponding author.

1
B. S.
Kisilevsky
,
S. M. J.
Hains
,
A.-Y.
Jacquet
,
C.
Granier-Deferre
, and
J. P.
Lecanuet
, “
Maturation of fetal responses to music
,”
Dev. Sci.
7
(
5
),
550
559
(
2004
).
2
N. H.
Al-Qahtani
, “
Foetal response to music and voice
,”
Aust. New Zealand J. Obstet. Gynaecol.
45
(
5
),
414
417
(
2005
).
3
F.
Massimello
,
L.
Billeci
,
A.
Canu
,
M. M.
Montt-Guevara
,
G.
Impastato
,
M.
Varanini
,
A.
Giannini
,
T.
Simoncini
, and
P.
Mannella
, “
Music modulates autonomic nervous system activity in human fetuses
,”
Front. Med.
9
,
857591
(
2022
).
4
E.
Brillo
,
V.
Tosto
,
A.
Ceccagnoli
,
N.
Nikolova
,
V.
Pinzaglia
,
F.
Bordoni
,
F.
Spano
,
V.
Bini
,
I.
Giardina
, and
G. C.
Renzo
, “
The effect of prenatal exposure to music on fetal movements and fetal heart rate: A pilot study
,”
J. Matern. Fetal Neonatal Med.
34
(
14
),
2274
2282
(
2021
).
5
L.
Lee
,
Y.-H.
Chang
,
W.-J.
Liang
, and
Y.-C.
Huang
, “
The effect of music intervention on fetal education via Doppler fetal monitor
,”
Children
9
(
6
),
918
(
2022
).
6
H.
He
,
J.
Huang
,
X.
Zhao
, and
Z.
Li
, “
The effect of prenatal music therapy on fetal and neonatal status: A systematic review and meta-analysis
,”
Complement. Ther. Med.
60
,
102756
(
2021
).
7
L.
Fathi
,
A.
Shakarami
,
K.
Amraei
,
F.
Yari
, and
A.
Behzadvand
, “
Effects of music therapy on the fetal outcomes of non-stress test and maternal anxiety
,”
Neuropsychiatr. Enfance Adolesc.
71
(
6
),
316
319
(
2023
).
8
R.
Hibiya-Motegi
,
M.
Nakayama
,
R.
Matsuoka
,
J.
Takeda
,
S.
Nojiri
,
A.
Itakura
,
T.
Koike
, and
K.
Ikeda
, “
Use of sound-elicited fetal heart rate accelerations to assess fetal hearing in the second and third trimester
,”
Int. J. Pediatr. Otorhinolaryngol.
133
,
110001
(
2020
).
9
H.
Narayan
,
E.
Vignola
,
W.
Fifer
, and
I.
Williams
, “
Assessment of cardiac rate and rhythm in fetuses with arrhythmia via maternal abdominal fetal electrocardiography
,”
Am. J. Perinatol. Rep.
05
(
02
),
e176
e182
(
2015
).
10
E.
Inguaggiato
,
G.
Sgandurra
, and
G.
Cioni
, “
Brain plasticity and early development: Implications for early intervention in neurodevelopmental disorders
,”
Neuropsychiatr. Enfance Adolesc.
65
(
5
),
299
306
(
2017
).
11
A.
Kikuchi
,
T.
Shimizu
,
A.
Hayashi
,
T.
Horikoshi
,
N.
Unno
,
S.
Kozuma
, and
Y.
Taketani
, “
Nonlinear analyses of heart rate variability in normal and growth-restricted fetuses
,”
Early Hum. Dev.
82
(
4
),
217
226
(
2006
).
12
D.
Cysarz
,
P.
Van Leeuwen
, and
H.
Bettermann
, “
Irregularities and nonlinearities in fetal heart period time series in the course of pregnancy
,”
Herzschr. Elektrophys.
11
(
11
),
179
183
(
2000
).
13
C.
Del Gaudio
,
A.
Carotti
,
M.
Grigioni
, and
U.
Morbiducci
, “
Nonlinear analysis of heart rate variability to assess the reaction of ewe fetuses undergoing fetal cardiac surgery
,”
Int. J. Artif. Organs
35
(
5
),
376
384
(
2012
).
14
M. G.
Frasch
,
C. L.
Herry
,
Y.
Niu
, and
D. A.
Giussani
, “
First evidence that intrinsic fetal heart rate variability exists and is affected by hypoxic pregnancy
,”
J. Physiol.
598
(
2
),
249
263
(
2020
).
15
Z.
Rauf
and
Z.
Alfirevic
, “
666: Continuous remote fetal monitoring with MONICA AN24 during home induction of labor
,”
Am. J. Obstet. Gynecol.
204
(
1
),
S263
(
2011
).
16
E. Z.
Yakupov
,
A. V.
Nalbat
,
M. V.
Semenova
, and
K. A.
Tlegenova
, “
Efficacy of music therapy in the rehabilitation of stroke patients
,”
Neurosci. Behav. Physiol.
49
(
1
),
121
128
(
2019
).
17
N. M.
Lalande
,
R.
Hétu
, and
J.
Lambert
, “
Is occupational noise exposure during pregnancy a risk factor of damage to the auditory system of the fetus?
,”
Am. J. Ind. Med.
10
(
4
),
427
435
(
1986
).
18
J.
Reinhard
,
B.
Hayes-Gill
,
S.
Schiermeier
,
W.
Hatzmann
,
T.
Heinrich
,
H.
Hüsken-Janßen
,
E.
Herrmann
, and
F.
Louwen
, “
Change of spectral analysis of fetal heart rate during clinical hypnosis: A prospective randomised trial from the 20th week of gestation till term
,”
Geburtshilfe Frauenheilkd.
72
(
4
),
316
321
(
2012
).
19
A.
Karpov
,
A.
Simakova
,
O.
Frolova
,
G.
Shiferson
, and
I.
Yemelianov
, “
Differential diagnosis of monotonous fetal heart rate
,” in
Obstetrics
, edited by
H.
Abduljabbar
(
InTech
,
2017
).
20
N.
Wessel
,
A.
Voss
,
H.
Malberg
,
C.
Ziehmann
,
H. U.
Voss
,
A.
Schirdewan
,
U.
Meyerfeldt
, and
J.
Kurths
, “
Nonlinear analysis of complex phenomena in cardiological data
,”
Herzschr. Elektrophys.
11
(
3
),
159
173
(
2000
).
21
M.
Calderón-Juárez
,
G. H.
González-Gómez
,
J. C.
Echeverría
,
H.
Pérez-Grovas
, and
C.
Lerma
, “
Association between mean heart rate and recurrence quantification analysis of heart rate variability in end-stage renal disease
,”
Entropy
22
(
1
),
114
(
2020
).
22
M.
Javorka
,
Z.
Trunkvalterova
,
I.
Tonhajzerova
,
Z.
Lazarova
,
J.
Javorkova
, and
K.
Javorka
, “
Recurrences in heart rate dynamics are changed in patients with diabetes mellitus
,”
Clin. Physiol. Funct. Imaging
28
(
5
),
326
331
(
2008
).
23
H. D. I.
Abarbanel
and
M. B.
Kennel
, “
Local false nearest neighbors and dynamical dimensions from observed chaotic data
,”
Phys. Rev. E
47
(
5
),
3057
3068
(
1993
).
24
N.
Marwan
,
M.
Carmen Romano
,
M.
Thiel
, and
J.
Kurths
, “
Recurrence plots for the analysis of complex systems
,”
Phys. Rep.
438
(
5–6
), 237–329 (
2007
).
25
N.
Marwan
, “Cross recurrence plot toolbox for MATLAB, version 5.29 (R38),” see https://tocsy.pik-potsdam.de/CRPtoolbox/; accessed 17 December 2024.
26
B.
Babaei
,
R.
Zarghami
,
H.
Sedighikamal
,
R.
Sotudeh-Gharebagh
, and
N.
Mostoufi
, “
Selection of minimal length of line in recurrence quantification analysis
,”
Physica A
395
,
112
120
(
2014
).
27
L. I. T.
de Santana
,
I. D. C.
Barreto
,
L. S.
Araújo
, and
T.
Stosic
, “
Recurrence quantification analysis of São Francisco river flow: Hydrological alterations caused by the construction of Sobradinho dam
,”
Res. Soc. Dev.
9
(
11
),
e87491110467
(
2020
).
28
A. D.
Likens
,
K. S.
McCarthy
,
L. K.
Allen
, and
D. S.
McNamara
, “
Recurrence quantification analysis as a method for studying text comprehension dynamics
,” in
ACM International Conference Proceeding Series
(
Association for Computing Machinery
,
Sidney
,
2018
), pp.
111
120
.
29
C.
Letellier
, “
Estimating the Shannon entropy: Recurrence plots versus symbolic dynamics
,”
Phys. Rev. Lett.
96
(
25
),
254102
(
2006
).
30
N.
Marwan
,
N.
Wessel
,
U.
Meyerfeldt
,
A.
Schirdewan
, and
J.
Kurths
, “
Recurrence-plot-based measures of complexity and their application to heart-rate-variability data
,”
Phys. Rev. E
66
(
2
),
026702
(
2002
).
31
D. T.
Mewett
,
K. J.
Reynolds
, and
H.
Nazeran
, “
Principal components of recurrence quantification analysis of EMG
,” in
Conference Proceedings of the 23rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society
(
IEEE
,
Istanbul
,
2001
), pp.
1592
1595
.
32
L. E. V.
Silva
,
R.
Fazan
, and
J. A.
Marin-Neto
, “
Pybios: A freeware computer software for analysis of cardiovascular signals
,”
Comput. Methods Programs Biomed.
197
,
105718
(
2020
).
33
M.
Malik
, “
Time-domain measurement of heart rate variability
,”
Card. Electrophysiol. Rev.
1
(
3
),
329
334
(
1997
).
34
C. I.
Montalvo-Jaramillo
,
A. C.
Pliego-Carrillo
,
M. Á.
Peña-Castillo
,
J. C.
Echeverría
,
E.
Becerril-Villanueva
,
L.
Pavón
,
R.
Ayala-Yáñez
,
R.
González-Camarena
,
K.
Berg
,
N.
Wessel
,
G.
Pacheco-López
, and
J. J.
Reyes-Lagos
, “
Comparison of fetal heart rate variability by symbolic dynamics at the third trimester of pregnancy and low-risk parturition
,”
Heliyon
6
(
3
),
e03485
(
2020
).
35
M.
Romano
,
L.
Iuppariello
,
A. M.
Ponsiglione
,
G.
Improta
,
P.
Bifulco
, and
M.
Cesarelli
, “
Frequency and time domain analysis of foetal heart rate variability with traditional indexes: A critical survey
,”
Comput. Math. Methods Med.
2016
,
1
12
.
36
D.
Dimitriev
,
O.
Indeykina
, and
A.
Dimitriev
, “
The effect of auditory stimulation on the nonlinear dynamics of heart rate
,”
Noise Health
25
(
118
),
165
175
(
2023
).
37
S. E.
Wright
and
C.
Palmer
, “
Auditory rhythm complexity affects cardiac dynamics in perception and synchronization
,”
PLoS One
18
(
11
),
e0293882
(
2023
).
38
S. T.
Chen
,
C. Y.
Chou
, and
L. H.
Tseng
, “
Recurrence plot analysis of HRV for exposure to low-frequency noise
,”
Adv. Mater. Res.
1044–1045
,
1251
1257
(
2014
).
39
S.
Paul
,
G.
Yadu
,
S. K.
Nayak
,
A.
Dey
, and
K.
Pal
, “
Recurrence quantification analysis of electrocardiogram signals to recognize the effect of a motivational song on the cardiac electrophysiology
,” in
Computational Advancement in Communication Circuits and Systems
, Lecture Notes in Electrical Engineering Vol. 575, edited by K. Maharatna, M. R. Kanjilal, S. C. Konar, S. Nandi, and K. Das (Springer, Singapore, 2020), pp. 165–172.
40
S. E.
Wright
and
C.
Palmer
, “
Physiological and behavioral factors in musicians’ performance tempo
,”
Front. Hum. Neurosci.
14
,
311
(
2020
).
41
D.
Parizek
,
N.
Visnovcova
,
K. H.
Sladicekova
,
M.
Veternik
,
J.
Jakus
,
J.
Jakusova
,
Z.
Visnovcova
,
N.
Ferencova
, and
I.
Tonhajzerova
, “
Effect of selected music soundtracks on cardiac vagal control and complexity assessed by heart rate variability
,”
Physiol. Res.
72
(
5
),
587
596
(
2023
).
42
M.
Sushrutha Bharadwaj
,
S.
Hegde
,
D.
Narayana Dutt
, and
A. P.
Rajan
, “
Nonlinear signal processing method detects emotional changes induced by Indian classical music
,”
Int. J. Eng. Adv. Technol.
9
(
1
),
6200
6206
(
2019
).
43
O.
Pino
,
S.
Di Pietro
, and
D.
Poli
, “
Effect of musical stimulation on placental programming and neurodevelopment outcome of preterm infants: A systematic review
,”
Int. J. Environ. Res. Public Health
20
(
3
),
2718
(
2023
).
44
C.
Ji
,
J.
Zhao
,
Q.
Nie
, and
S.
Wang
, “
The role and outcomes of music therapy during pregnancy: A systematic review of randomized controlled trials
,”
J. Psychosomatic Obstet. Gynecol.
45
(
1
),
2291635
(
2024
).
45
J. O. E. H.
van Laar
,
G. J. J.
Warmerdam
,
K. M. J.
Verdurmen
,
R.
Vullings
,
C. H. L.
Peters
,
S.
Houterman
,
P. F. F.
Wijn
,
P.
Andriessen
,
C.
van Pul
, and
S.
Guid Oei
, “
Fetal heart rate variability during pregnancy, obtained from non-invasive electrocardiogram recordings
,”
Acta Obstet. Gynecol. Scand.
93
(
1
),
93
101
(
2014
).
46
R. M.
Tristao
,
J. A.
Lacerda De Jesus
,
M.
De Lima Lemos
, and
R. D.
Freire
, “
Foetal music perception: A comparison study between heart rate and motor responses assessed by APIB scale in ultrasound exam
,”
Preprints (Basel)
arXiv:2007.0345.v1 (
2020
).
47
K.
Movalled
,
A.
Sani
,
L.
Nikniaz
, and
M.
Ghojazadeh
, “
The impact of sound stimulations during pregnancy on fetal learning: A systematic review
,”
BMC Pediatr.
23
(
1
),
183
(
2023
).
48
S.
Koelsch
and
L.
Jancke
, “
Music and the heart
,”
Eur. Heart J.
36
(
44
),
3043
3049
(
2015
).
49
A. R.
Zizzo
,
I.
Kirkegaard
,
J.
Hansen
,
N.
Uldbjerg
, and
H.
Mølgaard
, “
Fetal heart rate variability is affected by fetal movements: A systematic review
,”
Front. Physiol.
11
,
578898
(
2020
).
50
J.
Brändle
,
H.
Preissl
,
R.
Draganova
,
E.
Ortiz
,
K. O.
Kagan
,
H.
Abele
,
S. Y.
Brucker
, and
I.
Kiefer-Schmidt
, “
Heart rate variability parameters and fetal movement complement fetal behavioral states detection via magnetography to monitor neurovegetative development
,”
Front. Hum. Neurosci.
9
,
147
(
2015
).
51
L.
Semeia
,
K.
Sippel
,
J.
Moser
, and
H.
Preissl
, “
Evaluation of parameters for fetal behavioural state classification
,”
Sci. Rep.
12
(
1
),
3410
(
2022
).
52
B.
Frank
,
B.
Pompe
,
U.
Schneider
, and
D.
Hoyer
, “
Permutation entropy improves fetal behavioural state classification based on heart rate analysis from biomagnetic recordings in near term fetuses
,”
Med. Biol. Eng. Comput.
44
(
3
),
179
187
(
2006
).