Abstract
Background Electrogastrography (EGG) non-invasively evaluates gastric function but has not achieved common clinical adoption due to several technical limitations. Body Surface Gastric Mapping (BSGM) has been introduced to overcome these limitations, but pitfalls in traditional metrics used to analyze spectral data remain unaddressed. This study critically evaluates five traditional EGG metrics and introduces improved BSGM spectral metrics, with validation in a large cohort.
Methods Pitfalls in five EGG metrics were assessed (dominant frequency, percentage time normogastria, amplitude, power ratio, and instability coefficient), leading to four revised BSGM spectral metrics. Traditional and revised metrics were compared to validate performance using a 100 BSGM subject database (30 min baseline; 4-hrs postprandial), recorded using Gastric Alimetry (Alimetry, New Zealand).
Key Results BMI and amplitude were highly correlated (r=-0.57, p<0.001). We applied a conservative BMI correction to obtain a BMI-adjusted amplitude metric (r=-0.21, p=0.037). Instability coefficient was highly correlated with both dominant frequency (r=-0.44, p<0.001), and percent bradygastria (r=0.85, p<0.001), in part due to conflation of low frequency transients with gastric activity. This was corrected by introducing distinct gastric frequency and stability metrics (Principal Gastric Frequency and Gastric Alimetry Rhythm Index (GA-RI)) that were uncorrelated (r=0.14, p=0.314). Only 28% of subjects showed a maximal averaged amplitude within the first postprandial hour. Calculating Fed:Fasted Amplitude Ratio over a 4-hr postprandial window yielded a median increase of 0.31 (IQR 0-0.64) above the traditional ratio.
Conclusions & Inferences The revised metrics resolve critical pitfalls impairing the performance of traditional EGG, and should be applied in future BSGM spectral analyses.
Introduction
Chronic gastroduodenal symptoms affect >8% of the population and impart a substantial quality of life and economic burden.1,2 Existing diagnostic approaches for assessing the contribution of gastric dysmotility are controversial, and new tests are needed to aid diagnosis.3,4
Historically, gastric electrophysiological abnormalities were clinically assessed using electrogastrography (EGG) with sparse epigastric electrodes. Although EGG identified group-level differences between patients and controls in multiple adult and pediatric disorders,5–8 it did not achieve common clinical adoption due to several technical limitations, including inability to account for gastric anatomical variability, poor spatial resolution, and pitfalls in handling artifacts.9–12
Body surface gastric mapping (BSGM) overcomes these limitations, with recent studies showing capability to define novel disease subgroups with symptom correlations.9,13,14 BSGM is performed using a recently-developed medical device (Gastric Alimetry®, Alimetry, New Zealand) comprising a conformable high-resolution array, wearable Reader, and validated symptom-logging App.13,15,16 Using Gastric Alimetry, we have compiled a large database of high-quality recordings encompassing 4.5 hours (30 minutes preprandial, 4 hours postprandial).13,16 Compiling this database enabled us to critically re-evaluate spectral metrics previously applied in EGG, generating fresh insights into additional pitfalls that contributed to EGG’s failure to become a routine diagnostic tool.
In this study, we address these pitfalls and introduce revised spectral metrics for BSGM, demonstrating utility in a large dataset. While BSGM provides both spectral metrics and novel spatial metrics (wave patterns), 13,16,17 the current analysis focuses on spectral analytics.
Methods
Traditional EGG metrics are concerned with frequency, power (or amplitude), and variability, derived from spectrograms calculated using a fast Fourier transform with a sliding window.18 Supplementary Table 1 defines five key EGG metrics: dominant frequency, percentage normal / tachygastria / bradygastria, amplitude, power ratio, and instability coefficient of the dominant frequency.
The pitfalls of each metric are discussed below together with the justification of revised metrics, which are discussed in greater detail in Supplementary Table 2.
Revised BSGM Spectral Metrics
Pitfall 1: Effect of BMI on Amplitude
Adipose tissue reduces extracellular potential transmission; hence higher BMIs attenuate EGG amplitudes.19 While this relationship is known, it has not been robustly quantified, preventing amplitude from being compared across subjects or studies. To resolve this issue, we introduce BMI-adjusted amplitude, employing a multiplicative regression model to correct amplitude attenuation.
Pitfall 2: Conflation of low frequency transients with stable gastric activity
Dominant frequency is typically calculated as the frequency associated with the highest power in the spectrum for a given period/window. This approach can be problematic, because normal cutaneous gastric myoelectrical recordings commonly show transient high-amplitude bursts of low frequency noise (<2 cpm) occurring concurrently with stable gastric activity of normal frequency,12 likely representing colonic activity,20,21 or artifacts.10,12 Dominant frequency calculations may therefore conflate non-gastric low frequency transients with stable gastric activity. Note that while patients with gastric neuromuscular dysfunction also show substantial spectral instability within this same frequency range, this typically occurs when coordinated gastric activity is intermittent or absent.16
To resolve this issue, we developed robustly distinct frequency and stability metrics. Principal Gastric Frequency identifies only the sustained frequency associated with the most stable spectral oscillations, as determined by the separate stability metric, Gastric Alimetry Rhythm Index (GA-RI) (discussed below).
Pitfall 3: Misuse of coefficient of variation
The formula for the instability coefficient (coefficient of variation) serves as a normalized measure of dispersion. However, this metric is inappropriate for gastric rhythm because it implicitly and incorrectly assumes that a dominant frequency varying around a lower frequency is less stable than one varying around a higher frequency (see Supplementary Table 2).
To resolve this issue, we introduce Gastric Alimetry Rhythm Index (GA-RI)™, a proprietary measure of rhythmic stability quantifying the extent to which activity is concentrated within a narrow frequency band over time relative to the residual spectrum. We observed that BMI confounds stability metrics (see Results), hence GA-RI also includes a BMI adjustment.
Pitfall 4: Variability in meal response timing
EGG power normally increases after a meal,13,22 however, we observe substantial variability in the timing of the onset and duration of this response.13,16,23 Therefore, if meal response metrics employ only a brief fixed window (e.g. 45 minute recordings), meal responses may be underestimated.
We therefore introduce Fed:Fasted Amplitude Ratio, a ratio of the maximum amplitude in any single 1-hour over a 4-hr postprandial period, to the preprandial period.
Experimental Methods
The relative performance of the revised metrics was evaluated in a large existing BSGM database (100 participants; 4.5-hr recordings), comprising healthy controls. For full experimental methodology, refer.13,16,24 Ethical approval was granted by the Auckland Health Research Ethics Committee (AH1130) and all subjects provided informed consent.
Results
BMI-Adjusted Amplitude
Regression analyses showed significant associations between BMI and average amplitude (r=-0.57, p<0.001; Fig. 1A). Fig. 1B. reports the relationship between BMI and the revised BMI-adjusted amplitude metric. A conservative adjustment was achieved with a small residual effect (r=-0.21, p=0.037), intentionally designed to avoid over-adjustment in excessively high BMIs.
Relevant associations between traditional EGG metrics, revised BSGM metrics, and/or patient characteristics. (A) BMI and log average amplitude (B) BMI and log BMI-Adjusted Amplitude (C) Instability coefficient and dominant frequency (D) Instability coefficient and percent bradygastria (E) Gastric Alimetry Rhythm Index and Principal Gastric Frequency (F) BMI and instability coefficient (G) BMI and Gastric Alimetry Rhythm Index
Principal Gastric Frequency and GA-RI
Dominant frequency was negatively correlated with instability coefficient (r=-0.44, p<0.001; Fig. 1C), with incorrect identification of dominant frequency as <1.5 cpm contributing in a subset due to misclassification of low frequency transients as bradygastric activity. Fig. 1D reveals a strong correlation between instability coefficient and percent bradygastria (r=0.85, p<0.001); therefore, the extent to which percent bradygastria indicates abnormally slow but coordinated gastric activity, as opposed to high variability in dominant frequency, is unclear.
By contrast, the ability of the revised GA-RI and Principal Gastric Frequency metrics to independently capture frequency and stability is reported in Fig. 1E (r=0.10, p=0.314).
The relationship between BMI and instability coefficient is shown in Fig. 1F (r=0.28, p=0.004). As such, the GA-RI metric includes a conservative BMI adjustment (Fig. 1G; r=-0.20, p=0.044).
Fig. 2A shows an example problematic control spectrogram with both sustained activity at a normal frequency and a large volume of transient low frequency noise at <2 cpm. Here, the traditional dominant frequency metric identifies the gastric frequency as 1.1 cpm, whereas the revised Principal Gastric Frequency metric identifies the gastric frequency as 2.9 cpm. The Principal Gastric Frequency metric therefore correctly indicates that this subject has a normal frequency, while the low frequency activity is separately captured by assigning a GA-RI of 0.18, indicating low background spectral stability.
Representative spectrograms from healthy control subjects, recorded using the Gastric Alimetry system, for (A) Transient low frequency activity resulting in misclassified Dominant Frequency. (B) Inactive periods misclassified as unstable/bradygastric by traditional EGG metrics. (C, D) Examples of early vs late onset meal responses respectively; in these cases, using the entire postprandial average amplitude/power captures significant portions of relative inactivity, whereas a fixed window average (i.e., <1-hour postprandial) would miss either the early or delayed response.
GA-RI also improved on the problem of percent bradygastria and instability coefficient metrics being prone to indicating abnormal gastric activity during times where the stomach was simply inactive. An example is shown in Fig. 2B, depicting a spectrogram with normal periods of relative inactivity preprandially, and after the meal response. In this case, the GA-RI is 0.98 (indicating highly stable rhythmic activity), whereas percent bradygastria and instability coefficient metrics were 25% and 0.23 (above the median instability coefficient), respectively.
Meal Response
Across the whole dataset, median peak amplitude occurred 1.6 hours postprandially (IQR 0.7-2.6). Only 28% of healthy controls had a maximal average amplitude in the first postprandial hour. When moving from a fixed amplitude ratio using the first hour postprandially to the Fed:Fasted Amplitude Ratio over a 4-hr period, the median increase was 0.31 (IQR 0-0.64). Figs. 2C and 2D show examples of spectrograms with early and delayed-onset meal responses.
Discussion
Despite extensive research, EGG did not achieve common clinical adoption owing to technical limitations that impacted clinical utility.9,11 BSGM using Gastric Alimetry has overcome many of these limitations,13,14,16 but pitfalls in traditional metrics have remained unaddressed. Using an extensive BSGM dataset, we critically evaluated past spectral metrics and present a revised metric set that resolves these pitfalls.
The revised metrics, together with the other technical and processing advances of BSGM,10,13,15 offer a substantially more rigorous foundation for advancing the clinical utility of cutaneous gastric electrophysiology, and will be incorporated into the Gastric Alimetry test. However, their clinical application also requires a comprehensive reference range, which is currently being compiled.24
Concerns regarding ‘percent bradygastria’ were addressed here, but problems with the interpretation of ‘percent tachygastria’ also warrant discussion. While many EGG studies report frequencies in the tachygastric range (e.g. >4 cpm),5–8 direct recordings and pacing studies show that human gastric frequencies rarely exceed ∼4.5 cpm.25 Higher frequencies yielded by EGG could therefore also potentially reflect non-gastric sources.26 Gastric Alimetry employs a tight spectral window of 1-6 cpm, capturing the vast majority of relevant gastric activity, while eliminating physiologically implausible data <1 cpm and >6 cpm.12,26
Our study also revealed the importance of correcting stability metrics for BMI. Previous studies have found relationships between obesity and increased percentage of EGG activity beyond the normal frequency range,19 however such data should be interpreted cautiously without BMI-correction.
Significant post-processing research remains before the full potential of BSGM is realized. Spatial analytics encompassing wave patterns have demonstrated clear potential to further improve diagnostic yield, but still require refinement, optimization and standardization.13,14,16,17 In addition, analyses of the shape and dynamics of postprandial BSGM amplitude curves could offer new insights into gastric dysfunction.13 Nevertheless, spectral metrics alone offer fundamental value in characterizing disease states,16,23 and revising these to avoid pitfalls constitutes one more step toward realizing the clinical potential of non-invasive gastric diagnostics.
Data Availability
Data are available upon reasonable request at the discretion of the corresponding author.
Footnotes
Conflicts of interest: GO hold grants and intellectual property in the field of GI electrophysiology and are members of University of Auckland spin-out companies: The Insides Company (GO), FlexiMap (PD), and Alimetry (GS, SC, SW, GO, and AG). CV and JE have no relevant conflicts to declare.
References
