The assessment of gastric myoelectrical activity between different stomach regions based on three-channel electrogastrogram signals: a pilot study

Previous studies have widely used surface electrogastrogram (EGG) signals to non-invasively evaluate gastric motility. However, it focused only on the assessment of electrical activity in a certain stomach region and lacked the analysis of interactions between different gastric regions. In the study, we aimed to assess the correlations of myoelectrical activity between distinct stomach areas using functional connectivity (FC) methods by three-channel EGG for the first time. The study contained 20 healthy subjects and the dataset for each subject consisted of two three-channel EGG signals (labeled as CH1, CH2 and CH3; CH1 projected onto the region of lesser curvature, CH3 projected onto the region of greater curvature, and CH2 positioned the region between CH1 and CH3) with a sampling rate of 2 Hz acquired from three locations of stomach during a 20-minute period of fasting and postprandial states respectively. Firstly, the EGG signals were band-pass filtered into three slow wave (SW) bands by finite impulse response (FIR) filter, namely bradygastria (0.017-0.033 Hz), normogastria (0.033-0.067 Hz) and tachygastria (0.067-0.15 Hz). Secondly, we employed four FC algorithms (i.e. Spearman correlation coefficient (SCC), phase lag index (PLI), mutual information (MI) and coherence (Coh)) to calculate the connection strength of each pair of gastric areas. Finally, the FC strength between fasting and postprandial conditions for all subjects or the normal weight (NW) or overweight (OW) people and between the NW and OW groups for both fasting and postprandial states were compared, respectively. The results indicated that compared to the fasting state, lower SCC of CH1 and CH3 in bradygastria for the NW group and significantly increased SCC of CH2 and CH3 in normogastria for all subjects were observed under the postprandial state. The NW group had greater MI of CH1-CH3 in bradygastria and normogastria during the fasting condition relative to the OW group. It was also found that the connections between CH1 and CH3 for Coh in normogastria and tachygastria and for MI in tachygastria and for SCC in normogastria were markedly increased in the normal individuals than those in the OW group under the postprandial state. Moreover, the connections of CH1-CH3 for SCC, MI in normogastria and for Coh in normogastria and tachygastria showed significantly negative correlations with body mass index during the postprandial condition. The findings of the study demonstrated that the FC techniques can be used as an exploratory analytical approach to assess gastric functional activity, and the FC strength in the NW group were remarkably increased in comparison with the OW group under the postprandial state, suggesting that it should be considered to separately analyze the EGG signals of individuals with normal weight and those who are overweight when conducting the gastric electrical analysis in scientific research or practical clinical practice.

Surface electrogastrography (EGG) has been utilized as a non-invasive electrophysiological technique to record gastric myoelectrical activity by positioning cutaneous electrodes on the abdomen over the stomach (Yin and Chen, 2013), which is used to non-invasively assess gastric motility function. As early as in 1922, EGG was first applied and introduced by Alvarez (Alvarez, 1922), but it has achieved remarkable progress in scientific and clinical applications in the late 20 years due to the great improvement of recording instruments as well as its increasing measurement precision (Oczka et al., 2024). Just as electroencephalograghy (EEG) contains much important information about neural electrical activity (Wang et al., 2016), EGG records electrical activity from the stomach directly reflecting the state of stomach smooth muscle motility (Chaudhari et al., 2022). Previous studies have demonstrated that abnormalities in EGG are closely associated with many gastroduodenal diseases including gastroparesis, functional dyspepsia and nausea, etc (Varghese et al., 2021; Carson et al., 2022; Somarajan et al., 2022). Although EGG has great potential as an easy-to-use diagnostic tool for detecting the symptoms of digestive system diseases or other illnesses such as Parkinson’s disease (Sakakibara et al., 2009; Araki et al., 2021) and schizophrenia (Peupelmann et al., 2009), its widespread adoption is limited by many technical limitations (e.g. the lack of a standard for electrode placement), which have prevented it from being extensively applied in clinic. However, the non-invasive nature of EGG makes it one of the most powerful tools for assessing gastric functional activity.

Studies have shown that EGG is a combination of basic gastric electric rhythms generated by a network of specialized cells in the stomach wall (Wolpert et al., 2020), known as the interstitial Cells of Cajal (ICCs). ICCs as pacemakers in the gastrointestinal tract continuously and intrinsically generate and propagate slow currents, determining that the compositions of EGG signal are mainly slow waves (SW). Generally, the dominant frequency of EGG in human is at around three circles per minute (cpm) or 0.05 Hz (Chang, 2005; Li et al., 2022). Furthermore, there are three common sub-bands in EGG, which includes 0.017-0.033 Hz (1-2 cpm; bradygastria), 0.033-0.067 Hz (2-4 cpm; normogastria) and 0.067-0.15 Hz (4-9 cpm; tachygastria) (Chaudhari et al., 2022; Araki et al., 2021; Popović et al., 2019). So far, numerous previous researches have only focused on the evaluation of the characteristic parameters of a single-channel EGG signal in these bands such as dominant frequency (DF) and power (DP), amplitude and percentage of each of gastric slow wave (Oczka et al., 2024; Popović et al., 2019; Lim et al., 2012; Schamberg et al., 2023). For example, the mean dominant frequencies of gastric electricity in 20 healthy participants were calculated and compared for both fasting and postprandial states (Popović et al., 2019). The results showed a significant increase of DFs after meal intake. Moreover, abnormalities in gastric slow waves were also observed in patients with Parkinson’s disease (Li et al., 2022), suggesting gastric dysrhythmia may be a reliable method for the diagnosis of clinical diseases. While the characteristics extracted from the single-channel EGG remain common in clinical settings due to its simplicity, it can not capture the information about spatial propagation or connectivity. Recent advances using high-resolution EGG (Gharibans et al., 2017) or invasive recordings (O’Grady et al., 2012) have demonstrated that gastric slow waves exhibit distinct spatial propagation vectors and phase gradients across the gastric wall. These high-dimensional approaches have fundamentally advanced our understanding of gastric electrophysiology by establishing the directional nature of slow-wave conduction. However, the clinical translation of these techniques is constrained by the need for the placement of high-density electrode arrays, accurate anatomical coregistration, and the complexity of invasive surgical procedures. Moreover, from a physiological perspective, the quantification of propagation direction does not necessarily equate to the quantification of functional coupling strength between distal gastric regions. Additionally, the functional connectivity (FC) method, phase locking value, was utilized to investigate the synchronization between stomach and brain, revealing a novel resting-state network based on the delayed-connectivity of the brain and the normal gastric rhythm (Rebollo et al., 2018). It should be acknowledged that recent advances in high-resolution body surface gastric mapping (BSGM) have significantly improved the ability to capture detailed spatial and propagation information of gastric slow-wave. However, it is also worth noting that most existing EGG studies have focused primarily on characterizing gastric electrical features within local regions (Chaudhari et al., 2022; Araki et al., 2021; Schamberg et al., 2023; Varghese et al., 2025; Humph et al., 2026). In contrast, relatively little attention has been paid to the regional relationships of gastric electrical activity between different areas of the stomach. In other words, existing research has mainly described EGG signals within individual gastric regions while overlooking inter-regional functional interactions, the exploration of which could provide new insights into the assessment of the motility of gastric activity.

In this study, various FC methods were employed to calculate the connection strength of gastric myoelectrical activity among different gastric regions using three-channel EGG signals. The main purpose of our study was to compare gastric activity differences between the fasting and postprandial states as well as between normal weight and overweight individuals based on amplitude, phase, information theory, and frequency of EGG, offering alternative approaches for evaluating gastric motility function across different stomach regions from the perspective of functional connectivity. Besides, although body mass index (BMI) is not related to gastric physiology or does not directly reflect the gastric electrical activity, previous study has suggested potential association between body mass index (BMI) and dominant frequency of EGG (Popović et al., 2019). Thus, as an exploratory study, we further analyzed the correlations between BMI and the indicators of FC to determine whether EGG characteristic parameters differ across BMI levels, which can provide a theoretical basis for further research on the physiological mechanism.

Participants and EGG data acquisition

The EGG data used in this study comes from an open access database provided by Popović et al. (freely available at https://zenodo.org/records/3878435). The datasets include 20 healthy participants (12 males and 8 females; mean years = 25.0 ± 2.7 years; mean heights = 179.3 ± 11.5 cm; mean weight = 78.1 ± 16.5 kg; mean BMI = 24.1 ± 3.1 kg/cm 2). According to the BMI values, all subjects were further classified into two subgroups, that is the normal weight (NW) group with BMI <25 kg/cm 2 and the overweight (OW) group with BMI >25 kg/cm 2, and there were 10 volunteers in each group. Please refer to the previous research by Popović et al. for the details of demographic information (Popović et al., 2019). Exclusion criteria for this study included symptoms of digestive disorders (nausea, vomiting or stomach pain), and the intake of drugs in the past 6 months. Before participation, all subjects gave their written informed consent. The study was approved by the Ethics Committee of University of Belgrade and the whole experimental process conformed to the standards outlined in the Declaration of Helsinki.

Three channels (labeled as CH1, CH2 and CH3) EGG signals were recorded using standard surface, self-adhesive, Ag/AgCl electrodes H135SG (Kendall/Covidien, Dublin, Ireland) placed on the skin above the different stomach areas, shown in Figure 1 (please review the study of Popović et al. (Popović et al., 2019)). The sampling rate of EGG signal was set to 2 Hz. The CH1 and CH3 were placed based on surface anatomical landmarks over the upper abdomen corresponding to the projected regions of the lesser and greater curvatures, respectively. And the location for the CH2 was between the CH1 and CH3, in which the angle between line CH1—common electrode and line CH2—common electrode equaled the angle between line CH2—common electrode and line CH3—common electrode. Twenty-minute EGG signals were collected from each subject under the fasting and postprandial states with 10 minutes break for meal intake, respectively. It should be noted that no imaging verification (e.g., ultrasound or CT) was performed to confirm the underlying gastric anatomy, therefore any regional designations in this study were based on surface projections rather than direct visualization.

According to previous experience, the obtained EGG data was band-pass filtered into three commonly used slow waves by a zero-phase finite impulse response (FIR) filter with a Hamming window (implemented using the function pop_eegfiltnew.m in MATLAB software equipped with the EEGLAB Toolbox): bradygastria (0.017-0.033 Hz), normogastria (0.033-0.067 Hz) and tachygastria (0.067-0.15 Hz). An example of 4-min EGG signal with the three slow waves for CH1 during the fasting and postprandial conditions is shown in Figure 2. The pictures in the left column display the gastric waves with bradygastria, normogastria and tachygastria from top to bottom under the fasting condition and the pictures in the right column exhibit respectively the gastric waves with bradygastria, normogastria and tachygastria under the postprandial condition. It is not difficult to see from the pictures that the wave forms under the condition of the same frequency band are very similar for fasting and postprandial states.

Gastric myoelectrical activity analysis

As far as we know, there are currently few relevant literature reports that have evaluated the connection strength between the gastric myoelectrical activity in different regions of the stomach using FC methods. So, in the current study, we employed four FC methods commonly used in the brain research due to the first exploration, which includes Spearman correlation coefficient (SCC), phase lag index (PLI) (Mao et al., 2022), mutual information (MI) (Al-Angari et al., 2016) and coherence (Coh) (Li et al., 2015), to comprehensively investigate the interaction of myoelectrical activity between different gastric regions from multiple perspectives of time domain, signal phase, information theory and frequency domain. A brief description of these algorithms is given below.

Spearman correlation coefficient (SCC) is a linear method to measure the correlation between two variables. Compared to Pearson correlation coefficient, the SCC is less strict on the analyzed variables, which is a measurement of the relationship of the ranks of elements in two variables sorted in ascending order. Given two signals X and Y, its formula is defined as follow:

Where N denotes the total elements for X or Y, and X k r a n k Y k r a n k represents the rank of the ascended elements in X (Y). Therefore, SCC is also called rank correlation coefficient with the value ranging from −1 to 1.

Phase lag index (PLI) can evaluate the degree of coupling between the phase information of two signals. Given a signal X (t), its Hilbert transform X H (t) is defined as:

The corresponding analytic signal Z (t) of X (t) is then expressed as: where j is the imaginary unit, A (t) and θ (t) denote the instantaneous amplitude and instantaneous phase of the signal at time t respectively, where the instantaneous phase is calculated as follows:

Δ φ t So, the mathematical expression for PLI is defined as follows: where sign indicates a symbolic function, Δ φ t is the phase difference between the signals X and Y at time t and T is the total number of signal sampling points.

Mutual information (MI) can be used to assess the mutual independence of two signals. In the study, MI was computed via the histogram-based approach. It is assumed that X and Y represent EGG signals in two different regions of the abdomen. Each signal was partitioned into the same equally-spaced bins to obtain the marginal and joint probability histograms. The formula of MI is expressed as: where P X x and P Y y denote the marginal probability of signals X and Y respectively, and P X Y x, y is their joint probability. To eliminate the influence of signal entropy and enable cross-channel comparison, normalized MI was applied:

Coherence (Coh) refers to the degree of synchronization for two signals at a particular frequency. Coh was estimated using the Welch’s method in the study. Given two signals X (t) and Y (t), the signals were first divided into overlapping segments (a 50% overlap between adjacent segments was applied in the study). Second, each segment was multiplied by a Hamming window to suppress spectral leakage. Then, the discrete Fourier transform was computed for each windowed segment, and self- and cross-power spectral densities S XX (f), S YY (f) and S XY (f) were averaged over all segments at the frequency f. In practice, the spectrum coherence is normalized and the amplitude squared coherence is obtained, which is defined as follows: where S XY (f) is the cross-spectral density between the signals X and Y, and S XX (f) and S YY (f) are respectively the self-spectral densities of X and Y.

Besides, it can be seen from the definitions that the values of PLI, MI, and Coh range from 0 to 1. The larger the value is, the stronger the connection between the two signals is.

To obtain the FC strength, we first divided the whole three-channel signals into five 4-min epochs in this study. Then, the connection between gastric electrical activities for any pair of channels for each epoch was calculated via the above-mentioned techniques. After that, the FC results of five epochs were averaged and used for further comparisons of the fasting and postprandial states for all subjects or the NW or OW groups, and between the normal weight and overweight groups for the fasting or postprandial state respectively.

Statistical analysis

The statistical analysis was conducted by SPSS software (version 25, IBM Inc., New York, United States). The paired t -test was used to examine the connection difference in gastric myoelectrical activity of two different stomach regions between the fasting and postprandial conditions. As to the connection comparisons of gastric activity between the normal-weight and overweight groups under either the fasting or postprandial condition, the Kolmogorov-Smirnov test was applied to assess the normality of the data distribution in each group. If both groups met the normality assumption, an independent-sample t -test was then used to examine the difference between the two groups of data; otherwise, using Mann-Whitney U -test. For each frequency band (i.e. bradygastria, normogastria or tachygastria) and each FC metric, multiple comparisons including three channel pairs (CH1-CH2, CH1-CH3 and CH2-CH3) between the fasting and postprandial conditions or between the OW and NW groups under the fasting or postprandial condition were corrected by the Benjamini-Hochberg method to control false discovery rate (FDR). Results were considered to be statistically significant at the level of p < 0.05 after FDR adjustment. In addition, the correlations between the metric values and BMI were assessed when the compared result of FC showed a (marginally) significant difference. For the multiple comparisons under each condition, only one set of channel pairs at most showed a significant difference, therefore the p values of the correlation analysis did not require correction in this study.

Functional connectivity comparisons of gastric activity between the fasting and postprandial states

The comparison results of the connections based on SCC between fasting state and postprandial state across three slow wave bands for all subjects are presented in Figure 3. It was found that individuals had significantly increased SCC between CH2 and CH3 in tachygastria during the postprandial condition in comparison with the fasting condition (t = −2.787, p / p corr = 0.012/0.037). Figure 4 shows the compared results of SCC of the fasting and postprandial conditions for the normal weight group. The results indicated that compared to fasting state, decreased SCC of CH1-CH3 in bradygastria (t = −2.385, p / p corr = 0.041/0.123) but increased SCC for CH2-CH3 in tachygastria (t = 2.870, p / p corr = 0.018/0.056) were observed in the NW group for postprandial state. As shown in Figure 5, the fasting state exhibited notably reduced SCC of CH2-CH3 in normogastria (t = −3.391, p / p corr = 0.012/0.035) and a decrease of SCC of CH2-CH3 in bradygastria (t = −2.534, p / p corr = 0.039/0.112) than postprandial condition for the overweight individuals. As for the metric of PLI, no significant differences were observed between the fasting and postprandial states for three SW bands. Furthermore, higher MI between CH1 and CH3 in tachygastria was observed in the fasting state for all subjects (t = 2.466, p / p corr = 0.024/0.072; please see Supplementary Figure S1). In addition, there was no any significant difference between fasting state and postprandial state for Coh of electrical activities in different stomach regions in all SW bands.

Functional connectivity comparisons of gastric activity between the normal weight and overweight individuals

For the fasting condition, as indicated in Figure 6, it was observed that the connection for CH2-CH3 with respect to SCC in normogastria in NW group was higher than that in OW group (t = −2.299, p / p corr = 0.035/0.105). Also, the normal weight subjects had greater MI for CH1-CH3 in both bradygastria and normogastria relative to the OW group (bradygastria: t = −2.188, p / p corr = 0.043/0.129; normogastria: t = −2.593, p / p corr = 0.018/0.054; please see Supplementary Figure S2). For the postprandial condition, it is displayed in Figure 7 that as compared to the overweight subjects, dramatically increased SCC between CH1 and CH3 in normogastria was found in the NW group (t = −4.024, p / p corr = 0.001/0.003). Furthermore, the results also demonstrated remarkable differences between the OW and NW groups in different SW bands in terms of PLI, MI and Coh during the postprandial condition (please see Supplementary Figures S3–S5).

Correlations between functional connectivities of gastric myoelectrical activity and body mass index

For the functional connectivities between the normal weight and overweight group with (marginally) significant differences, the relationships between FC metrics and BMI by means of the linear regression method were further investigated to find promising electrophysiological biomarkers for the assessment of gastric activity of different stomach areas. Figure 8 displays that BMI had a strongly negative correlation with the connection of CH1-CH3 for MI in normogastria (r = −0.528, 95% CI = [-0.79, −0.11], p = 0.017). Significantly moderate negative correlations between BMI and the connection of CH1-CH3 (i.e. SCC in normogastria (r = −0.556, 95% CI = [-0.81, −0.12], p = 0.017), MI in normogastria (r = −0.469, 95% CI = [-0.75, −0.03], p = 0.037), Coh in normogastria and tachygastria (normogastria: r = −0.510, 95% CI = [-0.78, −0.07], p = 0.026; tachygastria: r = −0.494, 95% CI = [-0.77, 0.05], p = 0.031)) were found under the postprandial state (Figures 9A,D,F,G). Furthermore, the results also showed that there was a highly positive association between PLI of CH1-CH2 in bradygastria with BMI (r = 0.583, 95% CI = [0.16, 0.83], p = 0.011; Figure 9B). In addition, the correlations between PLI of CH2-CH3 in bradygastria with BMI (Figure 9C, r = −0.280, 95% CI = [-0.65, 0.20], p = 0.245), and between MI of CH1-CH3 in tachygastria with BMI (Figure 9E, r = −0.441, 95% CI = [-0.76, 0.05], p = 0.077) were not significant.

Due to the fact that previous studies lacked the analysis of electrical activity between different gastric regions, the purpose of this study aim was to employ functional connectivity methods for assessing the interplay of gastric myoelectrical activity of different stomach regions between the fasting and postprandial conditions, and between the normal weight and overweight subjects. Three-channel EGG signals were recorded from different regions of the stomach (the CH1, CH2 and CH3 channels; CH1 projected onto the region of lesser curvature, CH3 projected onto the region of greater curvature, and CH2 positioned the region between CH1 and CH3). It was found that compared to the fasting state, the NW group exhibited decreased SCC of CH1-CH3 in bradygastria and the overweight people had significantly greater SCC values between CH2 and CH3 in normogastria during the postprandial condition. The results also showed that higher MI of CH1 and CH3 in normogastria was observed in the normal weight individuals for the fasting state in comparison with the overweight group. Furthermore, the OW group had dramatically lower connections values between CH1 and CH3 for Coh in normogastria and tachygastria for MI in tachygastria and for SCC in normogastria than those in the normal weight group under the postprandial state. In addition, the results also demonstrated that there existed significantly negative correlations between BMI and the connections of CH1-CH3 for SCC, MI in normogastria and for Coh in normogastria and tachygastria under the postprandial state. Together, the findings suggested that the interactions of myoelectrical activity between different gastric areas can be evaluated by the FC approaches and the assessment of gastric motility for the normal weight and overweight individuals need to be analyzed separately.

Most previous studies have used a single-channel EGG signal to analyze the gastric activity in a certain area of the stomach. In time domain, the features extracted from the EGG signal were mainly the average amplitude over a period of time (Schamberg et al., 2023; Sukasem et al., 2022; Hendry et al., 2025). On the other hand, for frequency domain, the spectral analysis methods were widely applied to calculate characteristic parameters, such as dominant frequency (defined as the frequency corresponding to the highest power in the overall average spectrum) (Schamberg et al., 2023; Miljkovic et al., 2024), dominant power (defined as the power associated with the dominant frequency) (Schamberg et al., 2023; Friesen et al., 2007), percentage of basic rhythms of gastric myoelectrical activity including bradygastria, normogastria and tachygastria (denoted as the percentage of time windows where the dominant frequency belonged to the analyzed slow wave across the whole signal period) (Li et al., 2022; Lim et al., 2012; Schamberg et al., 2023; Noar and Khan, 2023). As it is well known, the ICCs are located within the muscular layers or at the middle of the circular and longitudinal muscular layer in the stomach wall and generate the gastric basic rhythms (Wolpert et al., 2020; Rebollo et al., 2018). Under the healthy state, the stomach has only one pacemaker, so there is only one dominant frequency for EGG controlling the gastric electrical rhythm at any given moment. In spite of this, the EGG typically is weighted summation of three sub-frequency bands of slow waves presented in the stomach, namely bradygastria, normogastria and tachygastria. Moreover, another study performed time-frequency analysis on multi-channel EGG signals by using the continuous wavelet transform based on the fast Fourier transform (Komorowski and Pietraszek, 2016), and the results have demonstrated that the power proportions of different gastric electrical rhythms in EGG signals vary at different moments. So we suspected that there might be some relationships between the gastric activity collected from different stomach regions within the range of specific frequency band. However, the related studies on the interactions of stomach activity are still scare until now. Therefore, a preliminary exploration of the evaluation of for the calculation of connection strength in gastric myoelectrical activity between different regions of the stomach was conducted by using different FC methods (including SCC, PLI, MI and Coh) for the first time in the current study. Each metric captures different aspects of connectivity and has distinct sensitivities to potential artifacts. Notably, PLI is specifically designed to ignore zero-lag phase coupling, thereby effectively excluding spurious connectivity arising from volume conduction (Stam et al., 2007). This property is particularly valuable in gastric electrophysiology, where signals recorded from adjacent regions may be contaminated by passive electrical spread. In contrast, SCC reflects monotonic correlation of signal amplitudes, MI measures both linear and non-linear statistical dependence, and Coh quantifies linear frequency-domain synchronization. All these three indicators are affected by the volume conductor, which can cause deviations in the estimation of the connection due to signal mixing. In short, the joint use of the four methods enables a more robust and comprehensive functional connectivity analysis for assessing gastric motility function.

Compared to the fasting state, we found prominently increased SCC of CH2 (located between the CH3 (i.e. the corresponding projection area was greater curvature) and CH1 (i.e. the corresponding projection area was lesser curvature) channels) and CH3 in tachygastria in all subjects during the postprandial state (see Figure 3) and in normogastria for the overweight individuals (see Figure 5), but a moderate increase in tachygastria for the normal individuals (see Figure 4). However, we also observed that a reduction of SCC between CH1 and CH3 in bradygastria for the normal weight group (see Figure 4) and a decrease of MI between CH1 and CH3 in tachygastria for all people (see Supplementary Figure S1) under the postprandial condition. Compared to the fasting state, the changes of the indicators SCC and MI were more obvious than those of the indicators PLI and Coh during the postprandial state, indicating that the changes in gastric myoelectrical activity in the amplitude/information dimensions were more pronounced. If the effect were simply driven by enhanced global gastric contraction postprandially due to the propagating activity from the interstitial cells of cajal in the stomach wall, then the results would be uniform changes across all channel pairs, metrics and frequency bands, rather than the results observed in the study. Therefore, we speculated that the gastric motility activity may be regulated by neural activity (Wolpert et al., 2020; Williams et al., 2016; Bellusci et al., 2022; Jiang et al., 2024). Furthermore, the modulation effect of the neurons on ICCs is different for the normal weight and overweight individuals, and these differentiated ICC networks causes them to be respectively associated with different gastric myoelectrical activity and then lead to different results for the FC comparisons of gastric myoelectrical activity between the fasting and postprandial states. Although the significant differences in the comparisons of the fasting and postprandial states were mainly concentrated in the indicator SCC, the other three indicators PLI, MI and Coh as well as SCC showed remarkable differences between the NW and OW groups under the postprandial condition. Specifically, relative to the overweight subjects, the NW group had greater MI and Coh between CH1 and CH3 in both normogastria and tachygastria (Supplementary Figures S4, S5) and SCC for normogastria (Figure 7) under the postprandial condition. These consistent results suggested that overweight individuals exhibited altered connectivity patterns in gastric myoelectrical activity between different regions of the stomach. Previous studies have demonstrated that obesity subjects are characterized by altered gastric functions (Cardoso-Júnio et al., 2007; Cifuentes et al., 2021), which further supports our findings to some extent. Additionally, we also found increased PLI of CH2-CH3 but significantly decreased PLI of CH1-CH2 in bradygastria in the NW group during the postprandial condition when comparing to the OW group (Supplementary Figure S3). This may be related to the fact that the projected greater curvature located in the mid/upper corpus as a dominant pacemaker area generates the gastric electrical rhythm and actively propagates the slow waves with a gradient pattern along the stomach via the ICC network as well as electrically coupled smooth muscle cells (Wolpert et al., 2020). Accordingly, the direction and speed of gastric slow waves vary for the projection areas (i.e. lesser and greater curvatures) within the proximal and distal stomach (Gharibans et al., 2017; Hosseini et al., 2023). Due to the different phase information extracted from CH1 and CH3, the above results were thus produced. Besides, compared with high-resolution electrical mapping studies (Gharibans et al., 2017; O’Grady et al., 2012), the findings of the study revealed a complementary approach rather than a repetitive manner in gastric electrophysiology. Previous studies have established that gastric slow waves follow a clear spatial propagation trajectory, which implies phase based connectivity and shared information. In contrast, our analysis of the connectivity between channels did not attempt to reconstruct these spatial vectors, but rather quantified the strength of pairwise interactions between recording sites without considering the propagation direction. Even in the absence of the propagation wave fronts, our FC approaches remain effective. Therefore, the FC methods of the study do not compete with the high-density electrode array methodology. Instead, it is a practical alternative that places functional coupling above the spatial mapping using surface EGG with a minimal electrode array, and is particularly suitable for scenarios where array recording is not feasible. Together, all these findings indicated that the functional connectivity approaches were able to characterize the interaction between EGG signals and mine the useful information that were not be obtained by a single-channel signal. Furthermore, the connection strength between the gastric electrical activity acquired from different stomach regions in the NW group was higher than that in the OW group. Nevertheless, there is no standard for the proper placement of EGG electrodes up to now. Hence, the future work should further investigate the effect of the selection of different stomach regions on the results.

However, it is important to acknowledge that the differences in functional connectivity metrics between the overweight and normal weight groups might be caused by the signal quality of the two groups. As is well-known, the varying levels of muscle and adipose pose barriers to the transmission of biological electrical signals, thereby resulting in the attenuation of the signals. Particularly, a study that used Bayesian inverse problems to estimate gastric surface electrical activity from cutaneous recordings has also pointed out that the different conductivities of tissues (e.g. fat and muscle) between the stomach and abdomen surface needed to be taken into account for the impact on signal transmission (Allegra et al., 2019). So, individuals with higher BMI typically exhibit lower signal-to-noise ratio (SNR) in surface EGG recordings. This will directly affect the joint distribution of two signals and tend to decrease the measure of statistical dependency. Therefore, the potential confounding effect of BMI-related signal quality on connectivity metrics should be considered when interpreting the findings between different BMI groups in the study. In future work, the differences of FC observed in individuals with different BMI individuals in the study need to be further verified based on serosal or mucosal electrode recordings that minimize the attenuation of the muscle and adipose tissues.

Since food intake can cause changes in the amplitudes of gastric myoelectrical signals, we examined whether the changes in the root mean square (RMS) of EGG signals could explain the changes in FC. Thus, we have computed the RMS values of EGG signals for each channel for both fasting and postprandial states (the results are presented in Supplementary Table S1). It can be seen from the table that compared to fasting state, the RMS values increased slightly for all frequency bands for all subjects during postprandial state. This might have altered the rank-order relationships between different channels, resulting in a significant change in SCC. PLI is completely insensitive to the amplitude of the signal, but it depends on the distribution of phase differences between the signals. The changes in the amplitudes of gastric electrical signals do not necessarily significantly alter the phases of the signals during the postprandial state compared to the fasting state. Therefore, PLI may not have significantly changed. MI is sensitive to both linear and non-linear dependencies between the signals, but its estimation variance is relatively high. If the amplitude of the signal merely increases or decreases as a whole, the distribution pattern remains largely unchanged. So, MI would also remain basically unchanged. Coherence is based on the self-spectral densities of signals and the cross-spectral density between the signals, of which normalization step largely removes amplitude information, making it also insensitive to pure amplitude changes. Furthermore, we observed a slight increase in RMS value in each frequency band in the NW group, but a very small decrease of RMS was found in the OW group under postprandial state as comparison with fasting state. Nevertheless, there was no any significant difference between fasting and postprandial states for each frequency band (p / p corr > 0.05). Therefore, the results indicated that it may be difficult to detect a significant difference in the amplitude fluctuation of EGG signals between fasting and postprandial states in the time domain. We further compared the results of RMS and FC methods and found that the increase or decrease of the results did not showed synchronization or opposite trends. That is to say, the observed changes in FC can not be explained by the changes in the amplitudes of the EGG signals. Furthermore, the results of our study also varied among different FC methods. In fact, the RMS and FC methods analyze gastric electrical activity from the two perspectives of local areas and the interactions between different regions of the stomach, thus the two types of measures are not directly equivalent in a physiological sense. Hence, RMS and functional connectivity represent two distinct dimensional characteristics of EGG signals, and there is no obvious corresponding relationship between them.

In the current study, we utilized the linear regression approach to explore the relation between the connection of different gastric activity and body mass index. It was found that the functional connectivity strength of the CH1 and CH3 channels based on SCC, MI in normogastria and Coh in normogastria and tachygastria had statistically significantly negative relationships with BMI for the postprandial condition, suggesting the higher the BMI, the lower the connection of CH1-CH3 (Figures 8, 9). On the contrary, there was a positive correlation between BMI and the PLI of CH1 and CH2 in bradygastria under the postprandial state (Figure 9). Consistent with the previous study (Popović et al., 2019), the results on the correlation analysis again demonstrated that the analysis of EGG signals using FC should also take into account the subject’s body mass index. However, the BMI correlation analyses were exploratory and should be interpreted as hypothesis-generating rather than confirmatory, especially given the small sample size as well as the relatively wide confidence intervals for the obtained correlation coefficients. Here, it is important to note that this study mainly employed functional connectivity methods to analyze the electrical interactions between different regions of the stomach, rather than being used to distinguish between individuals with normal weight and those who are overweight.

However, there are some limitations in the study. First, the data of the study was obtained from a publicly available dataset collected in accordance with a strict experimental protocols, but the sample size was relatively small. So, more participants are needed to be recruited to confirm the findings of the study in our future work. Second, there were only three channels available EGG signals for each subject, which is physically incapable of capturing the propagation patterns or complex regional interactions (e.g. the gastric network) that characterize gastric motility in this study. In the future, the collected gastric electrical signals should cover the entire stomach as much as possible. Third, it should be acknowledged that the standard anthropometric methods (e.g. body fat percentage and waist circumference) still have a greater advantage in the assessment of obesity. BMI is not a direct measure of gastric physiology, so any observed correlation between BMI and FC parameters extracted EGG signals should be interpreted as exploratory analysis rather than causality. Fourth, the findings of the study are limited to the general adult population (not including these specific groups such as athletes, bodybuilders, and elderly individuals), so the interpretation of these findings should be done with caution. Future work incorporating more precise body composition measures is warranted to better understand the relationship between BMI and gastric myoelectrical activity. Fifth, the quality of the data in this study is reliable (Jovanovi et al., 2020), but it should also be acknowledged that the potential factors of measurement conditions (attenuation, diffusion and electrode position geometry) have the impact on the correlation between EGG and BMI, which can not be completely eliminated. Finally, the EGG data used in this study were obtained from a public database with an original sampling rate of 2 Hz. While this sampling rate satisfies the Nyquist criterion for the frequency range of gastric slow waves, higher sampling rates (e.g., ≥10 Hz) would be preferable for more effective artifact rejection and higher-fidelity functional connectivity analysis. Future studies with higher sampling rates and longer recordings are warranted to validate and extend our findings. Furthermore, we acknowledge that a 20-min EGG recording cannot capture the entire meal response trajectory, but it is methodologically sufficient and clinically relevant for the specific question of postprandial functional connectivity of gastric myoelectrical rhythms. In addition, imaging verification (e.g., ultrasound or CT) of electrode placement was not available due to the use of a publicly available dataset. Consequently, our regional characterizations (e.g., distinctions between putative gastric regions) are based on surface anatomical landmarks and relative electrode positions, not direct stomach visualization.

In this study, based on three-channel EGG signals, the relationships of electrical activity between different stomach regions were firstly evaluated by different functional connectivity methods. Our findings mainly demonstrated that there were significant differences between the fasting and postprandial conditions. Furthermore, compared to the overweight group, the normal weight subjects had significantly higher connection strength of electrical activity of different gastric areas during the postprandial condition for multiple FC methods and slow wave bands. So, the results of the study suggested that the functional connectivity analysis approaches may serve as a potential auxiliary tool for evaluating the interactions of gastric myoelectrical activity in stomach areas and the EGG signals of individuals with normal weight and those who are overweight should be analyzed separately both in scientific research or clinical practice.