Computational Quantification of Peristalsis in Preclinical Mouse Models Using Smartphone Videography

ABSTRACT

Background

A number of diseases and medical interventions affect gastrointestinal motility. However, quantitative methods for measuring effects on peristalsis in live animals are uncommon, cumbersome, and lack standardization.

Methods

Here we present a new computational method for quantifying gut peristalsis in preclinical mouse models that requires only a video taken using a mobile phone camera. Our analytical pipeline processes the videographic data to track motion as it correlates to gastrointestinal transit.

Results

This method is compatible with different mouse models (such as conventionally housed and germ‐free), gut microbiome compositions, and can further be used to screen the impact of drugs on gut motility, which we demonstrate using methacholine as an example. Quantitative peristalsis measurements after methacholine treatment of conventionally housed and germ‐free mice with or without reconstitution of their gut microbiomes demonstrate that our analytical method can detect and quantify changes in peristalsis driven not only by endogenous acetylcholine signaling but also by colonic anatomy and the microbiome.

Conclusion

We envision that our computational tool will find uses in preclinical drug or mouse model screenings and forms the cornerstone to expand this methodology to other non‐gastrointestinal applications.

Key Points

• A new computational method for quantifying gut peristalsis in live animals.

• Requires videos taken using a smartphone camera and our analytical pipeline.

• Our methodology can be expanded to multiple applications including peristaltic assessments of rodent gastrointestinal disease models, screening novel compounds for off and on target gastrointestinal effects, and quantifying body wall movements of other muscle groups.

Plain Language Summary

Many diseases and medical interventions affect gut motility. But quantitative methods for measuring gut motility in live animals are uncommon, cumbersome, and lack standardization. To address this unmet need, we present a new computational method for quantifying gut motility in preclinical mouse models that requires only a video taken using a mobile phone camera. This method is compatible with different mouse models and across distinct gut microbiome compositions (we demonstrate its use for measuring gut motility in conventionally housed mice, germ‐free mice, and germ‐free mice whose microbiomes were subsequently reconstituted). It can further be applied to different preclinical scenarios such as screening of drugs and probiotics, studying the effects of different microbiome compositions, and screening preclinical mouse models. Additionally, our method is amenable for use beyond its application in gastrointestinal motility and could be further enhanced by integrating advances in AI like video processing neural networks.

A new computational method and analytical pipeline for quantifying gut peristalsis in live mouse models that involves taking a video using a smartphone camera and then processing the videographic data using our custom analytical pipeline to quantify gastrointestinal motility.

1. Introduction

Peristalsis forms the basis of the involuntary physiological gastrointestinal (GI) muscle contraction and relaxation that propels chyme through the digestive tract, representing a major mechanism of gut motility. The peristaltic contractions in the alimentary canal are initiated by a number of interrelated signals, many of which are mediated by the interstitial cells of Cajal (ICC), pacemaker cells in the GI tract [1]. The ICC interface with smooth muscle, parasympathetic neurons, and enteric neurons [2], which can stimulate gut movement through release of acetylcholine in the postprandial stage, or motilin to stimulate migratory motility complexes in the fasting or inter‐digestive phases [3]. Acetylcholine acting on M₃ muscarinic receptors on smooth muscle cells [4] and ICC [5] facilitates calcium‐dependent [6, 7] muscle contractions [8] to propel chyme in the anterograde direction.

Dysregulation of GI motility can result from a number of diseases and is also a typical side effect of many commonly used drugs like NSAIDS, antidepressants, and colchicine [9]. Gastroparesis, chronic intestinal pseudo‐obstruction, or neuromuscular junction disorders like myasthenia gravis are characterized by reduced rates of peristaltic contractions and can result in constipation or malnutrition [10, 11]. On the other hand, pathologies such as hyperthyroidism may increase gut motility and cause diarrhea [12]. While some of these disorders arise in adulthood, congenital diseases like Hirschsprung disease result in abnormal or absent peristaltic motility from birth [13]. Irritable bowel syndrome (IBS) may result in diarrhea or constipation, due to fast or slow peristalsis, respectively. Furthermore, numerous classes of drugs have well‐characterized positive or negative effects on GI motility. For example, prokinetic drugs, a class including acetylcholinesterase inhibitors, dopamine antagonists, and serotonergic agonists, increase peristalsis and gastric emptying, while opioids, selective serotonin reuptake inhibitors, antihistamines, and many other classes of drugs inhibit peristalsis and lead to constipation.

Due to these common drug actions on gastrointestinal motility, many methodologies for visualizing, measuring, and quantifying gut motility are currently in clinical use. These include procedures like antroduodenal manometry (which requires invasive catheterization [14]) as well as non‐invasive techniques such as gadolinium‐ or manganese‐contrast magnetic resonance imaging, which offer high resolution of soft tissue but are costly, time consuming, and affected by positioning and breathing [15, 16]. Many related procedures have availability limited to specific tertiary settings and suffer from a lack of standardized protocols [17]. Recently, body surface gastric mapping (BSGM) and ultrasound have been offered as inexpensive, non‐invasive methods of measuring gastrointestinal motility in point‐of‐care settings. BSGM is a promising new technology that utilizes an array of electrodes placed onto the body [18] with recent FDA approval; however, it is currently limited to gastric mapping and lacks multiple key standardization studies [18, 19]. Intestinal ultrasound shows similar promise, allowing high‐resolution visualization of rhythmic intestinal movements in real time in the small intestine [20] and colon [21]. The drawbacks to ultrasound, however, include difficulty visualizing movement when gastrointestinal fluid volume varies amongst patients [22], sometimes necessitating polyethylene glycol laxative ingestion [23], and challenges in standardizing the method to generate reproducible results across modalities [24].

To address these shortcomings, many new methodologies are currently being evaluated using preclinical animal models. For example, ultrasonography on anesthetized live mice shows promise, as it does in clinical settings [25], and trans‐illumination near‐infrared technology has been used to generate detailed spatiotemporal mapping of the gastrointestinal tracts of freely moving live mice [26]. Novel dyes for use with near‐infrared imaging have shown further promise in enhancing gastrointestinal resolution for visualization of peristalsis [27]. While encouraging, these technologies require specialized equipment to generate images. Videography, on the other hand, has been in development for the capture and analysis of uterine movement in mice [28], and represents a methodological advancement primarily due to its low barrier to entry. Recent developments in artificial intelligence (AI) and computational analysis have shown promise to enhance analysis of video data; however, there remains an unmet need for quantitative, standardized tools which can measure intestinal motility via videography. Such a tool could find application at the preclinical stage to characterize effects of drugs or diseases on gastrointestinal motility in animal models, and eventually in point‐of‐care clinical settings for diagnostic or prognostic usage.

In this study, we present a novel computational tool for visualizing and evaluating gut peristalsis in live animals. Our approach involves recording gut motility using a smartphone camera and quantifying motility with a customized video analytical pipeline. We demonstrate the promise of our tool in both the quantification of peristaltic movement as well as for screening the effects of drug concentrations on gastrointestinal motility in conjunction with the FlexiVent ventilator system [29], which allows rapid and controlled administration of nebulized drugs.

2. Methods

2.1. Maintenance of Germ‐Free Animals and Animal Welfare

Mice were maintained in germ‐free conditions in the CWRU gnotobiotic core using the nested isolation system [30] and were regularly tested for contamination by performing standard testing including 16S rRNA sequencing, fecal cultures, and gram staining. All procedures were carried out in accordance with the National Institutes of Health guidelines for care and use of laboratory animals and were approved (4/22/19–4/30/27, protocols #20140158 and #20140035) by the Case Western Reserve University Institutional Animal Care and Use Committee (Assurance Number A‐3145‐01).

2.2. Gavage of Germ‐Free Animals

Each of the 4–8‐week‐old germ‐free mice were orally gavaged with 1 × 10⁸ E. coli strain BW25113 obtained from the KEIO collection from the Coli Genetic Stock Center (CGSC) at Yale University that had been grown overnight in SOC media at 37°C and then resuspended in 300 μL of sterile PBS. This gavage was repeated after 1 week. Limiting dilution assay of fecal samples from gavaged mice demonstrated 3.37 ± 0.39 × 10⁷ colony‐forming units on LB medium. Similarly, 4–8‐week‐old GF mice were orally gavaged with 300 μL PBS that contained normal microbiome that was prepared by resuspending and vortexing fecal material collected from conventionally housed mice in sterile PBS, followed by centrifugation at 3000×g for 1 min to obtain supernatant for oral gavage. This gavage was also repeated after 1 week. Limiting dilution assay of fecal samples from gavaged mice demonstrated 1.64 ± 1.1 × 10⁵ colony‐forming units on LB medium. Mice were used for experiments 4 weeks after the first gavage, during which these mice continued to be housed under germ‐free conditions. DNA was extracted from their fecal samples using the QIAamp PowerFecal Pro DNA kit (Cat# 51804) and was analyzed by 16S rRNA sequencing (Figure S1).

2.3. Mechanical Ventilation and Methacholine Inhalation

Animals were weighed and anesthetized with i.p. ketamine/xylazine (70–100 mg/kg, 5–10 mg/kg body weight; Pfizer, St. Joseph, MO; Lloyd Laboratories, Shenandoah, IA, respectively). Animals were placed supine on a heated surgical bed and tracheostomized via a midline neck incision with a 18G blunt cannula. They were then connected to the small animal ventilator (FlexiVent, SCIREQ Inc., Montreal, QC, CAN) and a muscle relaxant was administered to suppress spontaneous breathing efforts (i.p. 70 μg pancuronium bromide; Sigma‐Aldrich). Mice were mechanically ventilated at default settings: tidal volume of 10 mL/kg, a positive end expiratory pressure of 3 cm H₂O, a rate of 150 breaths/min, and an FiO₂ of 50%. Following two recruitment breaths of deep inspiration up to 30 cm H₂O for 3 s, sequential doses of an inhaled methacholine solution in saline (0 to 200 mg/mL in saline; Sigma‐Aldrich) were aerosolized over 10 s using an ultrasonic nebulizer (Aeroneb, SCIREQ) and diverted into the ventilator's inspiratory flow. Animals were monitored via continuous electrocardiogram during mechanical ventilation. 5‐min video recordings were analyzed at baseline and after methacholine inhalation.

2.4. Video Recording and Post‐Production Editing

Video recordings of the gut peristalsis movements in the ventilated germ‐free mice were made using the camera of an iPhone 15 Pro Max that was mounted to a phone clamp suspended above the supine animal. Videos were recorded at 4 K resolution and 60 frames per second. Video clips were time‐limited to a standard 150 s across all mice and also cropped rostrally at the level of the xiphoid and caudally at the level of the pelvis. The cropped video frames were normalized to a range of [0,1], and denoised. Sudden camera movements were corrected by applying an affine transformation using the OpenCV method “warpAffine” for motion compensation [31].

2.5. Software Design and Algorithm for Quantifying Gut Motility

To quantify peristaltic motion, a corner detection method was applied to extract up to 200 high‐quality feature points using the Shi‐Tomasi corner detection algorithm [32], while excluding a 50‐pixel margin around the edges to minimize noise from fur movement. These features were tracked across frames using the Lucas–Kanade optical flow algorithm [33] and the displacement of each feature point was obtained as the difference between its positions between successive frames to obtain a two‐dimensional displacement vector. This vector was then reduced to a one‐dimensional representation using principal component analysis [34] to capture the direction of maximum motion. The obtained motion signal was subsequently filtered using a low‐pass filter to reduce noise, and peaks were detected to represent motion events. The number of peaks was counted within a sliding frame window (empirically set to 100 frames per window), and the resulting signal was smoothed with a moving average filter. Finally, the mode and standard deviation of peak counts within each window were calculated to characterize the dominant frequency and variability of peristaltic motion (reported as motion peaks per frame window, or mppfw). The mode statistic was selected here to quantify the most frequently occurring number of peaks per frame window, where peaks corresponded to peristaltic motion and increased peristaltic activity corresponded to more frequent peaks within the frame window.

3. Results

To establish our analytical pipeline, we first used germ‐free (GF) mice, which feature a markedly enlarged cecum 4–8× larger than conventional mice [35, 36]. Despite generally decreased peristaltic movement at baseline in GF mice [37], previous research has demonstrated a strong response to cholinergic stimulation in GF mice, such as hyperresponsiveness to forskolin (an adenylate cyclase activator that stimulates acetylcholine release from myenteric neurons [38]) and no reduction in the response to bethanechol (a direct muscarinic agonist [39]). Furthermore, due to the compression of the peritoneal cavity by the enlarged cecum as well as increased smooth muscle mass [36], we hypothesized that cholinergic stimulation could generate more visible gastrointestinal waves.

GF mice were anesthetized with a ketamine and xylazine solution and a tracheostomy was performed to intubate supine mice with the FlexiVent ventilator. With spontaneous breathing suppressed and mechanical ventilation regulated by the FlexiVent, video recordings of the abdomen to capture peristaltic movements were made from above the mouse. These were captured at baseline as well as following inhaled doses of methacholine (0–200 mg/mL), a non‐selective muscarinic receptor agonist that activates acetylcholine‐mediated peristaltic contractions at the neuromuscular junction and stimulates gastrointestinal motility [40, 41]. Methacholine is routinely administered by inhalation as it is poorly absorbed by the gut [42]. The captured videos were then used to measure peristalsis using a custom computational pipeline designed for the purpose of quantifying abdominal movement.

In brief, our analytical pipeline (Figure 1, details provided in Methods) first extracted frames from the recorded videos. After applying masks to the abdomen to select regions of interest, frames were normalized and basal motion was compensated. 200 abdominal points were detected and tracked across frames. Motion analysis involved calculating principal components of motion vectors after noise filtering. Lastly, peaks representing motion events of each point were identified in order to calculate standard deviations and modes as measures of peristaltic activity. This final output represents a quantitative measurement of the extent of abdominal movement (motion peaks per frame window, or mppfw) as well as the variability of the movement, both at baseline and after administering drugs.

FIGURE 1.

A flowchart of our analytical pipeline to quantify peristaltic movement from videos of mice. (A) Data was acquired by recording videos of mice ventilated on the FlexiVent system using a mobile phone after which a region of interest was drawn over the videos of the abdomen. (B) Frames from the videos were analyzed using this pipeline by tracking the movement of abdominal features over time. (C) Final output of our computational tool is the extent and variability of gastrointestinal movement.

Using this algorithm, we assessed intestinal motility of GF mice via video capture, first at baseline and following introduction of methacholine at two doses (100 and 200 mg/mL) to induce peristalsis. As shown in Figure 2 (n = 11 mice), our tool quantified a marked increase in motility upon methacholine treatment. Motility was highest with the 100 mg/mL dosage (10.10 mppfw vs. 3.75 mppfw at baseline, p = 0.06) and lowered slightly upon treatment with the 200 mg/mL dose (6.78 mppfw, p = 0.06 vs. baseline). Muscarinic agonism administered by inhalation is known to increase gut motility, and our measurements suggest that our tool can quantify this peristaltic movement. The low standard deviation in measurements suggests our quantification works with high reproducibility.

FIGURE 2.

Quantitative measurement of peristaltic activity via videographic analysis demonstrates methacholine‐induced gastrointestinal motility and microbiome‐augmented peristalsis. Measurements of peristaltic activity can be compared between GF mice (green), conventionally housed mice (red), and GF mice whose microbiomes were reestablished via oral gavage (blue; subgroups in light blue include E. coli monoculture and fecal microbiota transplant). Output from the analysis is provided as average peak motion (top), mode (middle), and standard deviation (bottom), quantified as motion peaks per frame window (mppfw) for different doses of methacholine (MCh): Baseline, 100 mg/mL (MCh100), 200 mg/mL (MCh200).

We next assessed whether our computational tool was able to detect gastrointestinal motility in conventional non‐GF mice, under the hypothesis that the lack of an enlarged cecum may hinder the ability to detect peristalsis by videography. With a similar experimental setup for video capture, peristaltic movement in conventionally housed mice (n = 12) demonstrated a pattern that was very similar to the GF mice: low but detectable peristalsis at baseline (4.07 mppfw), a significant increase with methacholine administration peaking at the 100 mg/mL dose (11.07 mppfw, p = 0.03 vs. baseline), as well as a relative decrease at the 200 mg/mL dose (7.45 mppfw, p = 0.03 vs. baseline). Our tool may thus be useful for studies across different mouse models as the peristaltic activity does not appear to rely on an artifact of peritoneal development (enlarged cecum) to detect and quantify signal.

In our last experiment, we sought to assess peristalsis in GF mice with a reconstituted microbiome comprised of either E. coli monoculture or normal mouse microbiome (fecal microbiota transplant) under the hypothesis that microbial‐derived metabolites may exert a significant effect on gut motility. Previous research has demonstrated that metabolically active compounds from gut microbiota can influence peristalsis via acetylcholine‐mediated parasympathetic neurons [43, 44] or via a diverse set of other compounds [45, 46]. Specifically, E. coli are responsible for a large proportion of luminal acetylcholine production [47], perhaps even affecting the gut‐brain axis via transport of the neurotransmitter into the central nervous system [48]. We first reconstituted the gut of mice that had been raised GF via an oral gavage of either E. coli or normal microbiome. After allowing 4 weeks for the establishment of their reconstituted gastrointestinal microbiomes, confirming colonization with fecal sampling and 16S sequencing (Figure S1), we quantified peristaltic activity on video captures, as described previously. Altogether, reconstituted GF mice (n = 31) demonstrated the highest overall intestinal motility at baseline (13.53 mppfw, p < 0.01 vs. conventional as well as GF mice), as well as a significantly elevated peristaltic response at both doses of methacholine (19.55 & 19.59 mppfw, p < 0.01 vs. baseline). Separated out by microbiome subgroup, we determined that—although quite similar to one another—GF mice reconstituted with E. coli alone had slightly higher intestinal motility compared to GF mice reconstituted with the full microbiome at 200 mg/mL methacholine (20.01 vs. 19.03 mppfw, p < 0.05). Overall, however, reconstituted GF mice did not demonstrate significantly different peristaltic activity between 100 and 200 mg/mL of methacholine pooled or as individual subgroups, suggesting these conditions may have reached a ceiling of motility. Our data therefore imply that the synergy between the anatomy of mice raised GF (e.g., enlarged cecum) and bioactive compounds released by microbes in the gut, especially E. coli , greatly increase motility in comparison to mice that either lack altered GF anatomy or lack microbial contributions to acetylcholine signaling. Alternatively stated, on their own, neither the enlarged cecum of the GF mouse nor the contributions of the microbiome in conventional mice are enough to substantially raise motility at baseline or in response to methacholine. Together, however, their contributions produce substantial movement even at baseline.

4. Discussion

Changes in gastrointestinal motility reflect key physiological effects of chronic diseases as well as responses to drugs. Here we have presented a new method to study and quantify peristalsis in the preclinical setting in live animals, using video capture via a smartphone camera. Our method utilizes a customized computational tool to automatically quantify overall peristaltic activity in different mouse models. In this case, we also used the FlexiVent ventilator for methodological standardization to control administration of nebulized methacholine and greatly mitigate motion artifacts from erratic breathing. While we have demonstrated the utility of this pipeline using anesthetized, ventilated mice, the analytical computational tool described herein is non‐invasive per se and requires only videography to analyze motion.

Computational measurements from our pipeline were found to distinguish peristaltic motion differences upon drug treatment in both GF and conventionally housed C57BL/6 mice. The GF mice have well‐characterized increases in transit time compared to conventionally housed mice [37, 49]. But since these differences are on the order of hours, not minutes, we were unable to distinguish between GF and conventional mice at baseline despite demonstrating robust and equivalent peristaltic responses in both groups to acute methacholine treatment. It remains possible, however, that tracking abdominal movement over longer periods of time with this method could demonstrate measurable differences between the GF and conventionally housed mice at baseline even without the administration of a drug. Nevertheless, this tool fills an unmet need for standardized quantification of peristalsis using external videography via widely available smartphones and could be used to further assess the impact of drugs on gut motility, such as during a screen for gastrointestinal side effects, by quantifying the change in peristaltic movements (Figure 3A). In conjunction with the FlexiVent, the impact of these drugs on lung function can be simultaneously assessed. This pipeline can thus be used for assessment of agents administered by inhalation, but also through other routes including diet, oral gavage, intravenous, intraperitoneal, or intramuscular. We envision this approach having particular utility on screening of novel opioid derivatives with lower addictive potential, which are both in development pharmacologically and have major impacts on gut motility as well as respiratory function [50].

FIGURE 3.

Potential applications of this method. (A) To study the effect of different drugs on gut motility. (B) To test the impact of microbiome compositions on gut motility. (C) To compare peristalsis in different mouse models (MM1 and MM2) as a response to experimental treatments. (D) To analyze the movement of different tissues, like the example of muscle shown here, by taking a video and using our analytical pipeline to quantify the motion. Illustrations made using BioRender.com.

Our methodology may find particular use in studies on microbiome composition. We demonstrate that GF mice with reconstituted microbiomes showed a significantly higher degree of peristaltic activity, which was further increased when these mice were administered different doses of methacholine. This suggests a clear effect of the microbiome on peristalsis in accordance with a large body of literature [43, 44, 45, 46, 47]. In the future, such a system could be utilized to evaluate the impact of different microbiome compositions on gut motility by gavaging GF mice with different microbes (Figure 3B), an active area of study given the strong relationship between microbiome and gut transit time [51]. Similarly, our method also holds promise for screening the impact of probiotic cocktails or antibiotics on peristalsis.

The presented tool could also enable more detailed comparisons of different animal models of disease by allowing researchers to quantify the effects of specific experimental treatments on gut motility using an objective and well‐controlled system (Figure 3C). In all three mouse models tested here, our tool quantified increased levels of peristaltic activity at a methacholine dose of 100 mg/mL, as well as discerning between changes in gut motility upon administration of different doses of methacholine. A combination of these approaches could help identify which mouse models exhibit the strongest or the weakest response to different drugs, microbial compositions, or probiotic/antibiotic treatments, thereby aiding the selection of appropriate mouse models for downstream applications and further studies.

We do note some limitations of our study. We tested a limited number of mice from different models and a specific number of methacholine doses. The goal here was to demonstrate the utility and proof of concept for our tool that is designed and optimized for processing of videos of peristaltic movement. By integrating newer advances in AI such as video processing neural networks, we envision this tool could find application to quantifying movement in other videographic applications (Figure 3D). These could include measuring muscle contractions in many contexts: rate, regularity, and labor of breathing; or even quantification of motility of macroscopic or microscopic live organisms across a variety of conditions. The intuitive methodology and set of features derived here are intended to serve as a platform for future developments and integration of more advanced AI techniques which could enable a more detailed and comprehensive assessment of motility and motion in mouse models.

5. Conclusion

We have presented a new computational method and analytical pipeline for quantifying gut peristalsis in live preclinical mouse models. Our methodology involves taking a video using a smartphone camera and then processing the videographic data using our custom analytical pipeline to quantify gastrointestinal motility. Our method can find different uses and applications within the laboratory setting, as has been demonstrated and outlined in this manuscript, providing the gastrointestinal research community with a valuable and highly accessible resource for measuring and quantifying gut motility in mouse models under different experimental conditions. Additionally, by integrating advances in AI like video processing neural networks, this method is amenable to be expanded further beyond its applications in gastrointestinal motility.

Author Contributions

J.C.S., P.S., and T.M.R. designed the study. J.C.S., P.S., and T.M.R. performed the experiments and made the video recordings. G.V.P. and S.E.V. performed the analysis and developed the computational quantification pipeline. J.C.S., S.E.V., and P.S. wrote the manuscript with input from all authors.

Funding

SEV and GVP were partially supported by the National Cancer Institute (1R01CA280981‐01A1,1U01CA294415‐01A1), the National Institute of Nursing Research (1R01NR019585‐01A1), the National Institute of Biomedical Imaging and Bioengineering (1R01EB037526‐01), the Veterans Affairs Biomedical Laboratory Research and Development Service (1I01BX006439‐01), and the Northeast Ohio Renal Research Innovation Award. This work was also partially supported by NIH grants F30‐ES03527, T32‐GM007250, and T32‐GM152319 to JCS; R01‐HL172894 to TMR; and R01‐DK128347 to PS. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, the U.S. Department of Veterans Affairs, the Department of Defense, or the United States Government.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: nmo70353‐sup‐0001‐FigureS1.pdf.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.