Artifact Rejection Methodology Enables Continuous, Noninvasive Measurement of Gastric Myoelectric Activity in Ambulatory Subjects

Abstract

The increasing prevalence of functional and motility gastrointestinal (GI) disorders is at odds with bottlenecks in their diagnosis, treatment, and follow-up. Lack of noninvasive approaches means that only specialized centers can perform objective assessment procedures. Abnormal GI muscular activity, which is coordinated by electrical slow-waves, may play a key role in symptoms. As such, the electrogastrogram (EGG), a noninvasive means to continuously monitor gastric electrical activity, can be used to inform diagnoses over broader populations. However, it is seldom used due to technical issues: inconsistent results from single-channel measurements and signal artifacts that make interpretation difficult and limit prolonged monitoring. Here, we overcome these limitations with a wearable multi-channel system and artifact removal signal processing methods. Our approach yields an increase of 0.56 in the mean correlation coefficient between EGG and the clinical "gold standard", gastric manometry, across 11 subjects (p < 0.001). We also demonstrate this system's usage for ambulatory monitoring, which reveals myoelectric dynamics in response to meals akin to gastric emptying patterns and circadian-related oscillations. Our approach is noninvasive, easy to administer, and has promise to widen the scope of populations with GI disorders for which clinicians can screen patients, diagnose disorders, and refine treatments objectively.

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1

(a) Illustration of an ambulatory monitoring system to record gastric electrical activity during unrestricted living outside of the clinic, which includes (b) the recording hardware with skin-mounted electrodes, and (c) a smartphone application for logging events. (d) Workflow includes processing of recordings to remove artifacts, extract gastric activity, and synchronize with event timing such that clinical assessments can be made.

Figure 2

(a) Cartoon of the stomach depicting the placement of the manometry catheter. The red dots indicate the locations where the pressure is measured; five sensors in the antrum with 1 cm spacing. (b) An X-ray image of a subject showing the position of the EGG sensor array relative to the manometry channels.

Figure 3

(a) Example time-series of raw EGG data containing motion artifacts (black arrows). (b) The same time-series following artifact removal using the LMMSE method (black arrows aligned by time with data in A). (c) A frequency domain representation of both time series (colors matched to a, b). The red star indicates the EGG frequency peak from this time series matching the expected frequency peak of gastric contractions, which only becomes apparent after removal of artifacts (green curve).

Figure 4

(a) Spectrogram of the raw EGG signal with motion artifacts visible as red vertical bands, and (b) spectrogram of the same signal with artifacts removed by the LMMSE method for Subject 2. (c) The percentage of normal gastric slow-wave activity across subjects using electrodes at the traditional EGG location (50 ± 10%), the highest SNR location (64 ± 8%), and the highest SNR location after artifact removal (90 ± 4%). Gastric activity is normal if the spectrum exhibits dominant power in the range of 2–4 cpm. In humans, the normal percentage of gastric slow-wave is typically defined as 70% (horizontal gray line).

Figure 5

(a) Overlay of the EGG power between 0.04–0.06 Hz and the manometry motility index (r = 0.74, p < 0.001). (b) Correlation between EGG and manometry motility index across subjects using electrodes at the traditional EGG location (r = 0.01 ± 0.26), the highest SNR location (r = 0.40 ± 0.24), and the highest SNR location after artifact removal (r = 0.57 ± 0.17). Hollow circles indicate correlations that are not statistically significant positive correlations.

Figure 6

(a) Spectrogram representation of EGG over a 24 hour period after the removal of artifacts. The EGG signal can be seen in the 0.04–0.06 Hz frequency band. (b) Extracted EGG power (mean of the 0.04–0.06 Hz frequency band) with event markers. The accelerometer magnitude is plotted in gray.

Figure 7

Ambulatory EGG extracted from multiple recordings (n = 8) on a single healthy subject. (a) Pre- and post-prandial EGG power response to isolated meals (n = 6). The time of meal completion is indicated by the vertical gray line. (b) High resolution mean of the EGG frequency from multiple continuous recordings from a single subject throughout a day (n = 8). Average time with the subject asleep in dark gray at top, +/− one standard deviation (light gray), with average time in wake in white. (c) The same signals averaged by time of sleep onset rather than by time of day. Differences in shape and distribution of variance suggest unique contributions of circadian and sleep regulation to GI activity profiles. The error bars are +/− one standard deviation.