Translation of an existing implantable cardiac monitoring device for measurement of gastric electrical slow‐wave activity

Despite evidence that slow‐wave dysrhythmia in the stomach is associated with clinical conditions such as gastroparesis and functional dyspepsia, there is still no widely available device for long‐term monitoring of gastric electrical signals. Actionable biomarkers of gastrointestinal health are critically needed, and an implantable slow‐wave monitoring device could aid in the establishment of causal relationships between symptoms and gastric electrophysiology. Recent developments in the area of wireless implantable gastric monitors demonstrate potential, but additional work and validation are required before this potential can be realized.


| INTRODUC TI ON
Much like the heart, the stomach is an electrically active organ, with its coordinated activity the result of periodic bioelectrical signals called gastric slow waves. 1 From the D'Arsonval instrument used by Alvarez in 1922 2 to the high-resolution electrical mapping of the modern era, 3,4 technological advancements have driven our understanding of gastric electrophysiology.[7] That said, limitations of existing methods have impeded the establishment of causal relationships between symptoms and GI electrophysiology.Current gold-standard techniques of high-resolution mapping have been restricted to intraoperative delivery in fasted subjects under anesthesia, which limits the recording period to a few hours and does not enable the correlation of slow-wave signals with symptoms. 38][9] However, it is limited to a recording duration of a few hours, where subjects must remain sedentary.
Ambulatory studies using a noninvasive adhesive patch 10,11 present an appealing medium-term monitoring solution, but body-surface patch-based methods are inherently limited to days to weeks of continuous monitoring.Longer-term (>1 year) ambulatory slow-wave recording devices would help to expand the potential applications of gastric bioelectrical slow waves as an actionable biomarker of gastric health. 12veral studies have attempted to develop solutions that address this methodological gap using implantable devices.4][15] Subsequent studies proposed minimally-invasive endoscopic devices with wired electrodes implanted in the mucosa.One such device monitored gastric myoelectrical activity for 5 days in ambulatory humans. 16However, as part of its design, wires were passed through the nasal passage of subjects, and although they reported no discomfort, it would not be a feasible solution for long-term use.Several recent studies in wireless implantable gastric monitoring (IGM) have introduced promising devices but have not yet translated to in vivo applications [17][18][19][20][21][22] or validated their measurements using established techniques. 23There remains a need for a gastric monitoring device suitable for human implantation.
Existing implantable bioelectrical monitoring devices used in other organs present an attractive translational solution for the stomach.For example, cardiac monitoring devices can automatically detect irregularities of the electrocardiogram and, in some cases, enable users to record when they feel discomfort. 24These devices contain internal memory, do not require the subject to wear an inductive power pack, and can be left in situ for over a year.
In this study, we hypothesized that an existing cardiac monitoring device (Reveal LINQ™, Medtronic) could be translated to operate as an IGM device.For this exploratory study, the Reveal LINQ was placed directly on the serosal surface of the stomach, adjacent to a high-resolution mapping array, to evaluate its ability to accurately resolve gastric slow waves.

Ethical approval was granted by the University of Auckland Animal
Ethics Committee (AEC3090).Experiments were performed in vivo in seven female crossbreed weaner pigs, which were fasted overnight prior to the study.General anesthesia was induced with zolazepam and tiletamine (0.1 mL/kg, Zoletil, Virbac) and maintained with isoflurane (2%-3% with an oxygen flow of 400 mL/ min in a closed-loop circuit).Blood pressure, heart rate, and core body temperature were continuously monitored and maintained within normal physiological ranges.A midline laparotomy was performed, and the gastric serosal surface was exposed with minimal gastric handling, enabling the serosal placement of the electrode devices.After the experiments, the animals were euthanized with a lethal bolus injection of sodium pentobarbital while still under anesthesia.

| Electrophysiological recordings
The Medtronic Reveal LINQ Model LNQ11 Insertable Cardiac Monitor is a small, leadless programmable device that is inserted under the skin in the chest, serving as an automatically-activated and patient-activated monitoring system that records subcutaneous ECG.For clinical practice, the Reveal LINQ is indicated for use in patients with clinical conditions that place them at an increased risk of cardiac arrhythmias and patients who experience transient symptoms that may suggest a cardiac arrhythmia.The device uses two electrodes on the body of the device to continuously monitor the patient's subcutaneous ECG.The device memory can store up to 27 min of ECG recordings from automatically detected arrhythmias

Key points
• There are currently no widely available solutions for long-term monitoring of the electrical activity of the stomach.
• An existing implantable cardiac monitoring device captured slow waves with a period and signal-to-noise ratio comparable to a validated flexible-printed-circuit electrical mapping array.
• Cardiac monitors may offer an efficient solution for the long-term monitoring of gastric slow-wave frequency.and up to 30 min of ECG recordings from patient-activated episodes.
The system provides three storage options for patient-activated episodes: up to four 7.5-min recordings, up to three 10-min recordings, or up to two 15-min recordings.Arrhythmia detection parameters can be automatically or manually programmed.
In this study, the Medtronic Reveal LINQ was used as a pre- The initialized Reveal LINQ was then placed on the serosal surface of the distal corpus (Figure 1C).A validated flexible-printed-circuit (FPC) high-resolution electrode array (256 electrodes, 16 × 16 array, 4 mm spacing; FlexiMap) 1,3,4 was placed immediately adjacent to the Reveal LINQ device to provide electrical control data.The Reveal LINQ and FPC arrays were covered with warm (39°C) saline-soaked gauze packs to ensure good contact between the electrodes and the serosa and to help maintain the temperature and moisture of the in vivo environment. 3The wound edges were then approximated with surgical clamps to further limit the cooling or drying of the abdominal cavity during electrical recording periods.

| Electrophysiological signal acquisition
Signals from the Reveal LINQ were acquired at 128 Hz and stored within the internal storage of the device.Following the experiment, these data were wirelessly transferred to a tablet computer using the Patient Connector.Signals from the FPC electrodes were acquired at 512 Hz using an ActiveTwo system (BioSemi) modified for passive electrode recordings.The ActiveTwo system was connected to a laptop computer (Asus) running custom acquisition software written in LabVIEW (National Instruments).

| Data preprocessing
FPC data were first visualized in validated analysis software (GEMS; FlexiMap). 25This step enabled the identification of the electrode most proximal to the Reveal LINQ that had captured visible slow waves.The signal from the identified electrode was isolated and processed using a previously described pipeline 26,27 using custom MATLAB code (Version R2021b; MathWorks).In brief, the signal was downsampled to 30 Hz, baseline drift was removed using a 20-s moving-median window, and ventilation artifacts were suppressed using a validated algorithm. 27First, the power spectrum of the signal was computed by performing a fast Fourier transform (FFT).Ventilation frequency was identified by locating the maximum power within the frequency band of 0.2 Hz-0.4 Hz (12-24 cpm) of the FFT result.The ventilation signal was estimated by taking the median of 11 signals shifted forward and backward by integer factors of the ventilation period (between −5 and 5).After subtracting the estimated ventilation signal, a Savitsky-Golay filter was applied (polynomial order: 9, window size: 51).The same pipeline was used to process the Reveal LINQ signals.Automated methods were used to mark slow-wave activation times (AT; i.e., the time when the derivative of the downstroke voltage was most negative), 27,28 followed by a subsequent manual review to ensure accuracy.

| Signal alignment
An initial estimate of alignment between the Reveal LINQ and FPC signals was achieved using the relative offset between the trigger time stored on the Reveal LINQ and the manually recorded trigger time relative to the start of the FPC signal.This alignment was optimized using a correlation technique that leveraged the AT markers extracted above.In brief, a dummy signal was generated for each measurement system, where a series of narrow Gaussians (σ = 0.067 s) represented each AT.The autocorrelation of the dummy signals of the relevant Reveal LINQ and FPC datasets was computed within a 60-s window centered around the preliminary alignment.
Peak autocorrelation within this window determined the relative time shift between datasets.

| Parameter calculation
Once the data were aligned, several parameters (Figure 1E) were calculated for each slow wave captured during the Reveal LINQ recording period.SNR was calculated using a previously reported method 29 To determine downstroke width and amplitude, the peak and trough of each downstroke were estimated using a zero-crossing technique described previously. 30The first derivative of the signal was calculated using a two-point central difference scheme, and the closest zero-crossing positions on either side of the AT were taken as the index of the peak and trough.
The time difference between each peak and trough was defined as the downstroke width and the voltage difference as the amplitude (Figure 1E).The slow-wave period was calculated using the time difference between adjacent ATs.The mean of each slowwave parameter was calculated on a device-and subject-wise basis.Data were excluded from analysis if slow waves were not observed in either the Reveal LINQ or FPC or if the signals could not be time aligned.

| Statistical analysis
Shapiro-Wilk tests were performed to assess the normality of device-wise data.If a parameter associated with either device was significantly non-normal (p < 0.05), a paired Mann-Whitney U test was used to test for a difference between the two devices; otherwise, a paired sample t-test was used.Similarly, Spearman rank correlation coefficients were calculated to assess whether there was a correlation between the non-normal Reveal LINQ and FPC slow-wave parameters.A Pearson correlation coefficient was calculated to evaluate the device-wise correlation for the remaining parameters.Statistical analysis was performed using R Statistical Software. 31Tests were performed with a two-tailed null hypothesis, and significance was set at p < 0.05.

| RE SULTS
Gastric slow waves were measured simultaneously with an FPC array and a Reveal LINQ in seven pigs (Figure 1).For most subjects (n = 6/7), the Reveal LINQ device successfully acquired slow-wave signals that were time aligned with the concomitant FPC signal (Figure 2).Slow waves detected by the FPC array were not captured by the Reveal LINQ in the remaining subject possibly because of intermittent or poor contact with the serosal surface (data not shown).

Quantitative slow-wave characteristics are summarized in
Table 1, and correlation analysis results are summarized in Table 2.
The period of the LINQ-measured slow waves was not significantly different from those measured by the FPC electrodes (Figure 3A; p = 0.69), and they were significantly correlated (p < 0.001), evidencing similar measurements of slow-wave period by both devices.The amplitude of the signal measured by the Reveal LINQ was less than that of the FPC electrode (Figure 3B; p = 0.024).The downstroke width was not significantly different between the measurement systems (Figure 3C; p = 0.98), and these measurements were not correlated (p = 0.13).The SNR of both systems was comparable (Figure 3D; p = 0.58).
There were some standard morphological features of the slow waves captured by each device (Figures 4 and 5).The FPC electrode signal consistently exhibited an initial upstroke, followed by a sudden downstroke, and a two-stage repolarization, characterized by a rapid partial repolarization followed by a slower secondary return to baseline (Figure 5).By contrast, the slow waves recorded by the Reveal LINQ presented only the initial upstroke, sudden downstroke, and rapid recovery to baseline.Essentially, the slow secondary recovery phase to the baseline of the FPC signal was not present in the concomitant Reveal LINQ signal.

| DISCUSS ION
In this study, we measured bioelectrical slow-wave activity in the in vivo stomach using an existing implantable cardiac monitoring device, the Reveal LINQ, produced by Medtronic.When in direct contact with the serosa, the Reveal LINQ device was capable of recording slow waves with consistent quality compared to the control recordings from high-resolution FPC electrode arrays (Figure 3D), which have been validated as the gold-standard approach for extracellular gastric slow-wave recordings. 1,32Altogether, the results of this study demonstrate that the commercially available Reveal LINQ cardiac monitoring device has the potential to function as an IGM.Such an application may provide an efficient solution for future slow-wave research where implanted electrode monitoring is required.
While electrical dysfunction occurs in a range of pathological GI conditions, 5,6,33 there are currently no readily accessible implantable devices that can monitor the electrical activity of the F I G U R E 2 Aligned Reveal LINQ signals (black) overlain with a simultaneously recorded flexible-printed-circuit (FPC) electrode signal (red).For visualization purposes, different scales were used for each device (left: FPC signal, red; right: Reveal LINQ signal, black).
stomach on an ongoing chronic basis.Such a device would be helpful in monitoring pathologies (e.g., gastroparesis, 6 chronic nausea and vomiting, 5 and functional dyspepsia 7 ) and postsurgical recovery periods (e.g., surgical anastomoses 34 and gastric ablation 35,36 ).Outside of a clinical setting, it might be feasible to use an IGM as a monitoring solution during a recovery period between surgical time points when intra-operative high-resolution mapping is used to provide spatiotemporal propagation detail. 35vices such as "gastric seeds" 22,23 show great potential for this application but require further development and validation before they will be readily available in clinical and broader research environments.The Reveal LINQ, on the other hand, offers a commercially available option with extensive clinical use for monitoring cardiac arrhythmias, which could be built on and customized for IGM applications.
Despite it being designed and validated for monitoring cardiac electrophysiological signals, our study confirms that the Reveal LINQ device can record slow-wave data in an acute setting when in contact with the serosal surface (Figure 2).Further aspects of the Reveal LINQ suggest it would be well-suited for future IGM applications.Crucially, it can transfer data wirelessly, is self-contained, and is already approved for subcutaneous implantation for cardiac applications. 24fundamental limitation of the Reveal LINQ is that it provides only a single channel of data and, therefore, lacks the spatiotemporal resolution to classify propagation patterns provided by modern high-resolution mapping techniques. 3,4A modified version or new generation of Reveal LINQ devices could integrate more electrodes for spatial mapping.This modification would enable postsurgical connectivity monitoring, for example, following a surgical anastomosis or ablation lesion, to assess if propagation is re-established across the intervention. 34,358][39] That said, even the current Reveal LINQ system, with a single implantable electrode, may yield clinically significant spectral content of gastric motility.Electrogastrography research 40,41    the Reveal LINQ and the selected FPC electrode was minimized, so it is unlikely that regional variability of gastric electrophysiology was the cause of these discrepancies.Instead, they are likely attributable to hardware filtering in the Reveal LINQ.The Reveal LINQ is optimized for detecting myocardial electrical events and contains filters to minimize the contribution of noncardiac signals to the recorded signal. 42Slow waves occur at a very different frequency to cardiac electrical events (0.05 Hz for slow waves vs. 1 Hz electrocardiogram).Additionally, the healthy cardiac QRS complex occurs on a timescale of less than 120 ms, 43 compared to the 3-4 s associated with the activation recovery index of gastric slow waves. 44The internal filters of the LINQ are likely suppressing the slow repolarization phase observed in the FPC signals (Figure 5).
Future IGM applications would require the Reveal LINQ to be chronically implanted, where this study was limited to the acute setting.Reveal LINQ devices were used to record slow waves for less than 10 min and were held in place with damp gauze (Figure 1D).While these data represent an essential foundational step toward translating the Reveal LINQ from cardiac to gastric bioelectrical monitoring, long-term implanted studies are now needed to realize the potential of this device for GI applications.One of the critical hurdles to overcome is determining how best to securely implant the Reveal LINQ, which is nontrivial.
Traditionally, the Reveal LINQ device is subcutaneously implanted to monitor the electrical activity of the heart.It is possible that, rather than attaching to the stomach directly, the Reveal LINQ could be implanted subcutaneously.Such usage presents challenges, but a logical progression would be to investigate whether slow-wave information is resolvable when the Reveal LINQ is implanted subcutaneously.
To conclude, this study translated the Reveal LINQ cardiac monitoring device to the in vivo stomach, demonstrating that it can clinical research tool to record gastric slow-wave signals for comparison to FPC data.Device programming and data transfer were conducted using a tablet (iPad; Apple) running a non-production, development version of the LINQ Mobile Manager (LMM) application.Communication between the LMM app and the LINQ was enabled via the 24967 Medtronic Patient Connector, and recordings were manually triggered using the record button on the PA96000 Medtronic Patient Activator.

A 7 . 5 -
min recording was obtained from the Reveal LINQ device, with the electrodes in direct contact with the serosa, and a simultaneous recording was obtained from the FPC electrode array.The recording of the FPC data began 30 s before the start time and stopped 30 s after the end time of the Reveal LINQ recording.Timealignment of the Reveal LINQ and FPC recordings is detailed in the 'Signal Alignment' section.F I G U R E 1 Gastric slow-wave measurement methods.(A) Reveal LINQ device.(B) Flexible-printed-circuit (FPC) electrode array.(C) Schematic showing the relative placements of the Reveal LINQ and FPC arrays during experiments.(D) Gross surgical images of the devices in situ.(E) Labeled diagram of slow waves captured by the Reveal LINQ (Sub.4) illustrating each parameter studied.Slowwave activation time (AT) is indicated by red dots.
and recent advances in spectral analysis techniques on body surface mapping data 8,9 have demonstrated that frequency information alone has clinical significance.Hence, despite the lack of spatial information, data generated by the Reveal LINQ could provide valuable insights.There were some morphological differences between the averaged FPC and Reveal LINQ signals.The physical distance between TA B L E 1 Comparison of parameters associated with the signals recorded by a Reveal LINQ with simultaneously recorded signals by gold-standard flexible-printed-circuit (FPC) electrodes.

F I G U R E 4 F I G U R E 5
Morphological comparison of subject-specific Reveal LINQ and flexible-printed-circuit (FPC) electrode signals.Averaged signals are indicated with thicker lines (black, Reveal LINQ; red, FPC electrode), and the shaded regions indicate the standard deviation.Each panel indicates the number of slow waves used to calculate the averaged signals.Morphological comparison of averaged Reveal LINQ and flexible-printed-circuit (FPC) electrode signals.Averaged signals are indicated with thick lines (black, Reveal LINQ; red, FPC electrode), and the shaded regions indicate the standard deviation.Each panel indicates the number of slow waves used to calculate the averaged signals.