Reproducibility of the full gut spectral signature over extended periods

Figure 1 shows GutPrints for four patients, each taken at multiple timepoints over 2 to 3 years. Although each patient has their own unique response pattern revealing the relative level of motor activity of their stomach, small intestine and colon, from one test to the next their patterns remain quite stable. One can argue that there is something fundamental about these patterns as there was no effort to curate their diets. Patients ate whatever they preferred at the time and went about their normal daily business during the tests. We think of this as a new Vital Sign for the gut.

Figure 1: Repeat spectra for four different patients recorded at multiple time points over 2 or 3 years.

Pre-clinical validation of external recording fidelity

Myoelectric signals generated in the muscles layers in the walls of the stomach and intestines are strongest at the organs and get progressively weaker traveling to the skin surface. Only those rhythmic contractions strong enough to be detected and distinguished from background noise are included in subsequent analysis. A study was carried out in Yucatan mini-pigs to learn what fraction of internally recognizable events were also detected at the skin surface, using internal strip electrodes placed on the serosal surface of the stomach, small intestine and colon and standard GutTracker patches on the pig abdomens. Pigs were free to move about their pens during the study, which lasted for up to 6 days.
A representative sample of data comparing peak events recorded internally and externally over a 3-day period is shown in Figure 2A. Blue crosses represent the frequency of peaks detected with the internal electrodes while red circles stand for the peaks detected externally. The vertical axis is frequency and the horizontal axis is time. The agreement between internal and external peaks in frequency and time is obvious, so much so that it’s difficult to see most of the red crosses. Figure 2B shows just those peaks where the internal and external peaks were consistent in frequency and occurred at essentially the same time. An analysis of the level of agreement between internal and external peaks across the seven pigs studied yielded concordances ranging from a low of 55% to as high as 81%, as shown in the table in Figure 4.

Figure 2: (A) Peak frequency plots for internal electrodes (blue crosses) and external patches (red circles) superimposed. (B) Frequency plot of all peaks for which the external patch frequencies matched that of the internal electrodes.

Figure 3: Table of concordance values between internal electrodes and GutTracker patch readings

Validation of colon and small intestine peaks by comparison to SmartPill pressure measurements

In a longitudinal study led by the Parkinson’s Institute, patients were tested at time 0, 3, 12, 24 and 36 months using SmartPill, anorectal manometry and G-Tech’s prototype GutTracker system. Over 150 total tests were completed. The SmartPill records pressure reading as it traverses the GI tract, and uses a pH reading to determine which organ it is currently in. Pressure readings are only sampled once per second after the first day, and the data is subject to high noise, large drifts and gaps in recording This may one reason why one sees very little about SmartPill pressure readings in the literature; the primary parameters commonly used clinically are traversal time of the stomach, small intestine and colon.

Using our algorithms to clean up the pressure readings and perform spectral analysis we are able to make direct comparisons of the internally recorded pressure readings and our external myoelectrical signals.

Figure 4 shows three examples of spectra recorded simultaneously with GutTracker and the SmartPill while it was in the stomach, small intestine and colon. Not only the frequencies of the major peaks but also subtleties like peak asymmetries and side peaks are also reproduced.

Figure 4: Spectra from GutTracker myoelectric recordings and SmartPill internal pressure measurements at the simultaneously for approximately one hour duration. In (a), (b) and (c) the SmartPill resided in the stomach, small intestine and colon respectively.

Peak analysis allows for a more comprehensive comparison than the compelling but limited examples in
Figure 2-?. Frequency vs time dot plots for a 12 hour data stream are shown in Figure 3-?. Noting that the GutTracker system records signals from all organs and so will have more peaks at different frequencies than the SmartPill, it can be seen that there is a strong concordance in the 14 to 16 cpm frequency range that is the most active for the SmartPill in this example.

Figure 5: Frequency vs. time dot plots for SmartPill and GutTracker readings while SmartPill was in the colon,

Histograms of the peak frequencies shown in the dot plots of Figure 5, weighted by peak area, are shown in Figure 6. There is strong agreement of the dominant 14 to 15 cpm frequency in both plots.

Figure 6: Histograms of the peaks detected in SmartPill and GutTracker measurements

Comments on Myoelectric recordings vs. Pressure measurements

A few caveats to mention when in the comparison of GutTracker and SmartPill or any other pressure measurement approach (e.g. manometry). GutTracker patches sample all myoelectric signals reaching the skin surface at the patch’s location. Any given pair of electrodes (there are four per patch) may sense signals from more than a single organ, or from a particular location in an organ depending on source location, intervening patch conductivity, and signal strength. In contrast, the SmartPill records the local pressure at whatever its location is at that moment in time, as do the individual sensors in a manometry device.

The physics of using pressure measurement as a proxy for rhythmic contractions in the stomach and intestines is not simple. Unless the device is in a volume that is closed at both ends (occluded) to allow pressure to change locally, it will not be able to detect contractions. There may be clinically significant peristaltic actions that do not cause in pressure changes at the location of the device. The stomach is a good example, where the clearest 3 cpm signals detected by SmartPill are in the last moments before it passes into the antrum and through the pylorus. During most of the time in the stomach it does not record the 3 cpm contractions even though they would have been required to move the pill to the antrum and pylorus.

Another important point about internal pressure measurements as opposed to myoelectric measurements is the frequency response curve of the device in situ. In a linear device the response is equal at all frequencies. However at the higher colonic frequencies of 15 to 20 cpm, a half cycle from low to high is two seconds or less. While the sensing element may not have a problem at this frequency the intestine may not have enough time to fully occlude. This will result in an effective frequency response curve that drops at higher frequencies.

Separately but compounding the challenge of detecting higher frequency activity, is that spectral peaks at higher frequencies have proportionally greater peak widths. The total energy represented by a peak is proportional to its area, so a wider peak will be shorter, such that a 15 cpm peak at the same energy level as one from the stomach at 3 cpm will be 5 times broader and 5 times shorter. This will make it much harder to detect above the background noise. This may be why to date there is only a relatively small amount of literature confirming the existence of frequencies above 10 cpm for the colon.

Existence of strong nighttime colon activity

Traditional thinking has asserted that the gut goes to sleep at night. The esteemed researcher Dr. Satish Rao and collaborators published a paper in 2001 presenting ambulatory manometric readings over 24 hours on 22 healthy volunteers that used 6 strain sensors inserted into the colon. They reported lower activity at night as measured by pressure changes that appeared as waves on a visually examined tracing on a computer monitor, with supplemental analysis by a computer program. In a widely read 2010 book Colonic Motility – From Bench Side to Bedside the noted pioneer in the field Sushil Sarna quoted Rao’s results which no doubt has influenced many gastroenterologists.

With the advent of modern technology such as multi-day wearable electrodes, wireless data transfer, gigabyte memory modules and terahertz processing capability, we’ve been able to assess motility for up to 6 days continuously for the entire GI tract, in both healthy volunteers and patients of many types. Spectral analysis using fast processors and advanced algorithms to remove artifacts and noise provide significant improvements in the ability to extract true rhythmic signals and characterize them by frequency, width, total area and likelihood of being real vs. random.

Our results in many hundreds of tests show that the colon is actually quite busy at night in most cases, and that there is a pattern in the frequency from day to night. Figure 7 shows frequency vs time dot plots for a 3 day period as measured by a SmartPill while in the colon, and by GutTracker. Clusters of dots between 12 and 14 cpm can be clearly seen in both plots, all of which begin right around midnight and last for 6 to 8 hours. During the daytime more scattered dots at higher frequencies, and can be seen on trajectories, jumping away from and then gradually moving back to the nighttime clusters. In this example the difference of about 6 cpm between day and night colon frequency is unusually large, more often is one or two cpm, or not particularly visible due to few daytime colon contractions.

Figure 7: Frequency vs time dot plots for 12 hour period as recorded by SmartPill and GutTracker

Figure 8 shows the same data as in Figure 7, but presented with the three days wrapped upon one another – so that the x-axis is not absolute date and time but time of day. With the additional statistical power of three days of data the patterns described become even more clear.

Figure 8: The same data as in Figure 7 plotted as a function of time of day rather than absolute date and time

Frequencies above 12 cpm from the colon have previously been described in the literature, although it is more common to see frequencies below 9 cpm and as low as 1 or 2 cpm mentioned, especially in earlier work. We don’t dispute that some of the activity at low frequencies that we measure may be from the colon, particularly those frequencies below the narrow 3 cpm stomach peak and between 3 cpm and the small intestine signals from 8 or 9 up to 12 cpm. The SmartPill data in the figures do show peaks at frequencies below 12 cpm so some of what we measure in this region must be colonic. However anything above 12 cpm is almost certainly from the colon.

We have considered alternative sources of the nighttime activity in depth. The most obvious one is breathing artifacts, which one might expect would be lower and easier to detect in the quiet of the night. To examine this we ran tests with controlled breathing at a constant frequency, at a normal level and while taking heavy, and then very heavy deep breaths. Only with the deepest breathing were we able to induce noticeable rhythmic activity in the patch readings. There is a reason this is so. The design of the GutTracker patches uses four sets of bipolar electrode pairs on each patch, that are relatively closely located and by being part of a single patch do not move relative to one another during breathing, or for that matter during most ambulatory movements. Using separate patches across the abdomen that can move freely relative to one another without inducing a signal rather than a larger unit is an important reason why the GutTracker system allows patients to go about their normal daily lives during the test.

In patients who are monitored immediately after having abdominal surgery we generally see an increase in activity in each of the organs over several days, depending on the procedure. In such cases there is no sign of nighttime colon clusters, except in the uncommon cases where a patient has had a longer than usual hospital stay. It is likely that the return of the nighttime clusters is a good sign that the colon has returned to health. But the absence in the first few days is another piece of evidence that the clusters are not simply experimental artifacts.

In presentations at the 2024 ANMS and ACG conferences we had posters that showed that patients in the Mayo clinic chronic nausea and vomiting study, who reported constipation effects via the G-Tech app, had a statistically significant lack of the nighttime clusters. With further verification and larger N this could become a biomarker for at least some forms of constipation that could be used to objectively assess efficacy of interventions.

A gastrectomy patient provides validation that the sharp signal at 3 cpm is from the stomach

A patient who had previously had their stomach removed and was being fed through a jejunal feeding tube was tested to understand how her remaining organs were functioning. Comparing their GutPrint to a typical healthy control (Figure 9) there is similar activity at the 9 to 10 cpm small intestine frequencies, and strong colon activity above 15 cpm. However in the patient there is a complete lack of activity where one normally sees a sharp peak from the stomach at 3 cpm. Interestingly there is essentially no activity in the 1 to 5 cpm region in this patient despite literature referring to colon activity at those frequencies and our own SmartPill data analysis showing the same. This anecdotal result suggests that when we see broad activity between 1 and 5 cpm it may be from the stomach.

Another less obvious difference between this patient and the control is a higher ratio of small intestine to colon activity. This could be due to higher than normal activity induced by the jejunal feeding tube.

Figure 9

Response of the gut to motility drugs

A volunteer subject who was experiencing long term constipation performed a study guided by a gastroenterologist that sequentially evaluated a selection of constipation treatments while wearing a single patch. The subject was one for whom a single 4 channel patch captured signals from each of her GI organs. The patches used were an earlier version with three day battery life. The protocol involved beginning a drug one day before the 3-day test and continuing it through the end of the test. Several weeks were allowed between trials, with a baseline test at the beginning and again later. Baseline spectra were very similar to one another and are shown in blue in the plots in Figure 10.

The orange traces in each plot represent the spectrum for the drug listed in the legend, namely erythromycin, linaclotide, a probiotic, metoclopramide, bisacodyl, and polyethylene glycol in a, b, c, d, e and f. Erythromycin lowered overall activity and led to nausea, linaclotide increased activity at 11 cpm (likely small intestine) but did not relieve constipation, the probiotic and metoclopramide had little effect on the signal or symptoms. Bisacodyl caused a large increase from 1 to5 cpm, a more modest effect form 5 to 10 cpm, and a large increase at 12 cpm. However the subject diary did not indicate any change in stool habit. In contrast to all the previous drugs, polyethylene glycol caused a large increase at 11 cpm and uniquely created a new peak where there had been only background before at 15 to 16 cpm. This relieved the subject’s constipation. On daily basis the first day the 16 cpm peak was small and the constipation had not abated, but on days 2 and 3 the peak emerged and so did the stool.

Figure 10: Spectra from tests in which a subject sequentially tried 6 different constipation remedies

Colonoscopy prep validation of propulsive colon frequencies

The same volunteer subject that did the response to drugs study also wore a single patch during preparation for a scheduled colonoscopy. The test was run beginning a day before fasting and colon prep right through the procedure. Figure 11 shows four spectra, a) at baseline before fasting; b) near the end of fasting before colon prep; c) during the colon prep; and d) during the colonoscopy.

The baseline period recorded from the single patch shows a strong stomach peak at 3 cpm and some small peaks and enhancements at small intestine and colon frequencies. Toward the end of the fasting period the entire spectrum has become near quiescent, an important result in itself as it indicates that what might be considered the background between peaks is actually affiliated with gastrointestinal motor activity and not just from motion artifact, skeletal muscle or electrical noise sources. During bowel prep and evacuation a very strong peak from the colonic activity emerges at 14 to 15 cpm. Finally during the actual procedure the reaction of the colon to insufflation is robust, with very sharp peaks at 12.5, 15 and 18 cpm. Not surprisingly these frequencies are very similar to the 16 cpm peak induced by polyethylene glycol in the same subject.

Figure 11

Variability in day-to-day gastric activity for patients with Gastroparesis Syndrome

A patient’s GutPrint® is a representation of the level of activity in their stomach, small intestine and colon for a given time period, as constructed by creating a histogram of the frequencies of spectral peaks detected within 10 minute time intervals. Hour to hour through the day and night activity varies and so will the appearance of the GutPrint, but over a 24 hour period it becomes reproducible – the next 24 hour period will look similar just as one full heartbeat on an EKG tracing looks like the next., but any interval shorter than a full heartbeat, say half a second, will look quite different from the next.

Over a 6 day test there is substantial averaging such that if a patient is tested months – or even years – later and their GI health hasn’t changed, they will have a GutPrint that is recognizably theirs. Much like a fingerprint. But in a single “GutBeat” of 24 hours, there are day to day differences. Figure 12 shows 6 daily GutPrints for two subjects in a study of patients with chronic nausea and vomiting. In this case patients were tested at the same time they had a gastric emptying scintigraphy test, so there was special interest in the activity of the stomach, which is at 3 cpm.

It’s quite easy to see that from one day to the next the height of the peak at 3 cpm – which is a proxy for gastric activity – varies a great deal. If one were to run a test such as gastric emptying on day 1 for both patients they would likely have normal emptying. On day two patient 1 might have rapid emptying but patient 2 would likely show slow emptying. And so on day after day.

Not all patients have such day to day variability. There seems to be a higher degree with patients that have lower activity on average. In Figure 13 the percent variation from day to day was studied for a total of 37 patients, with Weak, Moderate or Strong gastric activity. It didn’t quite meet the usual standard of P < 0.05 but there is definitely a trend toward higher variability in the patients with lower average activity. Exactly the group that has issues the tests are hoping to uncover.

This shows how important it is to monitor for multiple full days, to assess the true performance of the stomach, not to mention the rest of the digestive tract.

Figure 12

Figure 13

Daytime vs. nighttime and intraday GI activity

Figure 14

#####