2.1.

Tethered Capsule Endomicroscopy Design Requirements

The design of the TCE device was derived through an iterative process based on clinical, engineering, and regulatory requirements (Fig. 2). From a clinical use perspective, the imaging device should be minimally invasive, allowing a well-tolerated and short procedure requiring no sedation or special patient preparation. Its resolution should comprehensively, and without sampling error, enable microscopic visualization of structural changes at the level required for diagnosing the conditions of interest. Its diagnostic sensitivity should be high so as not to miss BE cases, yet also with moderate-high specificity to avoid unnecessary follow up EGDs. Its cost should be low without the specialized requirements of endoscopy suites or physicians to administer the device. Many of the engineering requirements for the device were also dictated by the nature of the digestive system (e.g., resistance to gastric acid and capability to be effectively disinfected) and were common with the design of the standard gastrointestinal (GI) endoscopes. In addition to obtaining the highest safety profile, a large effort was put into improving the ease of manufacturing the device.

Fig. 2

Schematic representation of the device design process.

2.2.

Tethered Capsule Endomicroscopy Device

Figure 1 shows a photograph of the TCE device that was used in the primary care office. The capsule has an 11 mm outer diameter and a weight of 6 g to support natural peristalsis via gravity and improve esophageal transit. A custom multilayered sheath (microlumen) ensured flexibility while maintaining its shape, pullability, and resistance to kinking. The connection between the capsule and the sheath was covered with a flexible, tapered strain relief to improve reusability of the device, and maintain the integrity of the capsule-tether connection. A total of four catheters were manufactured for use in the PCP study; one was reused for 11 different examinations.

2.3.

Tethered Capsule Endomicroscopy Pilot Study in the Primary Care Physician’s Office

The PCP pilot study was focused on testing the safety and feasibility of TCE in primary care practice (IRB P002014-001519). The study was approved in July 2014 for imaging 20 adult patients at the Internal Medicine Associates primary care clinic at Massachusetts General Hospital (MGH) [Fig. 3(a)]. Subjects scheduled for nonurgent visits to the clinic were eligible for this study. The inclusion criteria were an age over 18 and informed consent to the study. The exclusion criteria were history of intestinal strictures, current symptoms of fever, nausea, dysphagia or sore throat, and pregnancy.

Fig. 3

(a) Photograph of the OCT imaging console in the primary care examination room and (b) schematic representation of the TCE procedure.

The PCP study approval was followed by a meeting of all investigators and personnel where the study design was reviewed and each participant was assigned a role in the study. The technical staff responsible for the conduct of the study introduced the device to the staff at the clinic, including the nurses and medical assistants. The PCP (L.H.S.) prescreened subjects based on the date of their scheduled appointment and their potential interest and eligibility for the study. After confirmation that all inclusion and exclusion criteria were met, subjects received a consent form and description of the study by mail and then were contacted by the research nurse over the phone. Subjects were asked to fast for 4 h prior to the scheduled appointment. On the day of the procedure, subjects met with the research nurse and a prequestionnaire was filled out to ensure that subjects did not have any exclusion criteria before meeting with the PCP. Fasting status was confirmed, and written consent was obtained by PCP (L.H.S.).

2.4.

Tethered Capsule Endomicroscopy Imaging Procedure

The TCE procedure was approved to be performed before or after the scheduled PCP visit depending on the arrival of the patient and the availability of the consenting physician. After the subject was given informed consent, clinical characteristics such as age, sex, BMI, current medication regimen, and GI history were recorded. Prior to the procedure, the subjects were invited to watch a short presentation that included a recording of prior TCE procedures and a description of the study.

Prior to study initialization, the research nurse (C.N.G.) was trained to perform all TCE procedures. Operator training comprised: reviewing recordings of TCE procedures captured during previous TCE studies performed by an experienced research nurse and a demonstration of the optimal speed for catheter navigation in the esophagus. In the next training step, the nurse performed two TCE procedures in an ongoing TCE study (IRB P0011-2619): the first one under supervision of the endoscopist with previous TCE experience and the second one without supervision. A simplified diagram of the procedure [Fig. 3(b)] was available to the nurse at all times. The nurse was accompanied by an experienced TCE imaging system operator (M.J.G.) and an additional study staff responsible for filling out case report forms (A.T.).

As shown in Fig. 3(b), the TCE clinical procedure was divided into three phases. The first swallowing phase was similar to the administration of video capsule endoscopy devices: the nonsedated subject swallowed the capsule voluntarily while sipping water. The subjects were given a maximum of five swallowing attempts. They were also encouraged to sip water to help the swallowing process. Subjects were offered the option of coating the capsule with Pill Glide^™, a flavored over-the-counter lubricant spray that previously was successfully used for enhancing subjects’ ability to swallow.

After the capsule was swallowed, the imaging phase began and the TCE operator navigated the device through the upper GI tract similarly to a transnasal endoscope,²¹ except that the capsule’s transit was governed by gravity and peristalsis until the capsule reached the stomach. Another difference was the way the device passed anatomical sphincters, such as the lower esophageal sphincter or upper esophageal sphincter during withdrawal. The natural response of these sphincters was to constrict and prevent the capsule from moving proximally from one portion of the GI tract to the other. During the pullback of the capsule using the tether, the sphincters were relaxed with a dry swallow or small sips of water taken by the subject. The study was approved for imaging the esophagus and proximal stomach up to four times. Throughout the capsule’s transit, the research nurse read the capsule’s position from the incisors using 5-cm marks on the tether. These readings were recorded in case report forms with corresponding TCE image frame numbers.

In order to collect OCT images of the highest quality, it was critical to ensure good contact between the lumen and the capsule wall during transit.²² If, during the imaging phase, loss of contact due to peristaltic relaxation was observed in the OCT image displayed in real-time, the catheter operator was instructed to stop the capsule by holding the tether at the incisors. The subject was then asked to dry swallow or sip water to reengage peristalsis. The dynamic response of the esophagus could be seen in the OCT data in real-time with first the expansion of the lumen, then passage of the bolus of water, and finally a constriction of the lumen around the capsule, at which point the capsule’s motion was reinitiated.

The last phase of the procedure involved the removal of the capsule through the upper esophageal sphincter and the mouth. Even though sipping water efficiently relaxed the upper esophageal sphincter, it was rather cumbersome for the patient and the operator to pull the capsule out while swallowing liquid. The recommended method was to use breathing to open the sphincter. Once the capsule was out of the mouth, the total time of the procedure was recorded.

After the TCE examination, subjects were asked a set of questions:

According to the sterilization method in place at the MGH GI unit for endoscopes and manometry probes, the TCE device underwent a high level disinfection before each use at the MGH Sterile Processing Department by submersion in Cidex OPA for 12 min. After use, the TCE device was disinfected, sealed, and stored.

3.1.

Subject Enrollment in the Primary Care Physician Setting

Twenty subjects were enrolled from September 2014 to March 2015 [Fig. 4(a)]. Figure 4(b) represents the enrollment process for the study. Approximately 240 subjects were scheduled for nonurgent visits during enrollment hours, 217 were prescreened by the PCP, and 195 received information about the study by mail. Thirty-eight out of 69 subjects reached by phone were interested in participating in the study; 20 subjects were enrolled based on personnel and equipment availability. On the day of each study, the TCE system was transferred to the PCP unit and set up for the procedure in a designated outpatient examination room.

Fig. 4

(a) Patient enrollment process and (b) timeline of the PCP study. Blue circles represent successful PCP imaging procedures. Stars represent investigator meetings. Triangles correspond to procedures in which the capsule was not swallowed and diamonds correspond to training procedures.

3.2.

Primary Care Physician Study Population and Study Outcomes

Out of 20 enrolled subjects 11 were men, the average BMI was $27\pm 4.7\mathrm{kg}/{\mathrm{m}}^2$, and the average age was $52\pm 13$ years old (Table 1). After reviewing the medical files, eight patients (40%) were found to have a history of GERD and three had already undergone an EGD for conditions other than BE (15%).

Table 1

Demographic characteristics and procedural tolerance in the study population.

Subjects were, on average, a little anxious before the procedure with preswallowing anxiety of $2.1\pm 0.8$. The average number of swallow attempts was $2\pm 1.5$. Three patients (15%) were unable to swallow the capsule. Among the subjects who successfully swallowed the capsule, 16 (94%) declared they would definitely recommend the procedure and 1 (6%) said they would probably recommend it. The total TCE procedure time was $4.8\pm 2.6\min$ with an average of $3.7\pm 0.7$ passes in the esophagus before removal. Subjects experienced only a little discomfort ($1.9\pm 0.9$) during the procedures.

During the study, the research nurse introduced a new method for capsule removal based on a double breath technique, where the capsule was pulled out during the subject’s second deep breath. With this method, the percentage of subjects who identified capsule removal as the most difficult part of the procedure was reduced from 42% found in a prior study^12,13 to 29% in the current study.

3.3.

Tethered Capsule Endomicroscopy Image Analysis

The start and end positions of the capsule in the esophagus were recorded for all imaging passes based on marks on the tether. For 56 out of 61 passes, the capsule positions versus frame numbers were recorded every 5 cm. Based on these recordings, the average length of the esophagus imaged during all studies was $15\pm 6\mathrm{cm}$ with the average starting point in the proximal esophagus at $30\pm 6\mathrm{cm}$. The average velocity of ascending imaging, determined by manual pullback of the tether, was $33.7\mathrm{mm}/\mathrm{s}$ with a standard deviation of $16.8\mathrm{mm}/\mathrm{s}$. This variability in the length of the imaged esophagus was observed even though the same research nurse performed all the procedures. The optimal pullback velocity is a tradeoff between minimal frame separation (equal to $30\mu \mathrm{m}$ spot size), short procedure time, and ease of pulling the capsule. Based on the feedback from the catheter operator, it was difficult to perform a very slow pullback. In the interest of maintaining a short procedure duration, the optimal pullback speed was set to be $20\mathrm{mm}/\mathrm{s}$ corresponding to a $100\text{-}\mu \mathrm{m}$ longitudinal separation between cross-sectional TCE frames. In a majority of cases, the capsule was pulled out faster than the optimal pullback speed, providing an average frame separation for the collected 64,489 frames in 61 imaging runs of $\sim 140\mu \mathrm{m}$.

The image signal-to-noise ratio (iSNR) was calculated for each frame by using the region in the image between the reflectance from the capsule inner and outer surfaces as the noise floor. The average iSNR of all datasets was $24.46\pm 3.44\mathrm{dB}$, which indicates an adequate and reliable iSNR throughout every imaging run. Even the lowest iSNR from all 17 patients (17.23 dB) provided a usable dataset, as determined by expert reviewers (M.G. and G.T.). As a result, image quality was considered to be adequate for all subjects in this study.

For all subjects who swallowed the capsule, the device passed the lower esophageal sphincter and entered the stomach, allowing imaging of the gastroesophageal junction (GEJ), where BE can develop. A side-by-side comparison of these anatomical landmarks in two selected subjects is shown in Fig. 5. Normal features of the esophageal wall, such as a layered mucosa and a good tissue penetration depth, are displayed in Figs. 5(a) and 5(d). The proximal stomach or cardia [Figs. 5(c) and 5(f)], on the other hand, has a very poor penetration into the tissue and a typical “pit pattern” (blue arrowheads) that is visible on the surface of the tissue. The transition between these two tissues types at the GEJ is shown with a presumably normal appearance in a 36-year-old female [Fig. 5(b)] that can be compared to another transition with somewhat atypical glandular architecture in a 66-year-old female subject [Fig. 5(e)]. Similarly, irregularly shaped glands were also seen in the gastric cardia [Fig. 5(f)] of this subject. All three tissue types (squamous, GEJ, and cardia) were present in 57 out of the total of 61 passes. In three passes, only images of the esophagus were captured and in one pass, only images of the stomach were captured.

Fig. 5

Examples of TCE cross-sectional frames selected from three-dimensional data obtained in (a, b, c) a 36-year-old female—subject A and (d, e, f) a 66-year-old female—subject B. These images depict typical features and landmarks found in the OCT images of this study. The top row shows characteristic layered architecture of the esophageal wall with submucosal glands (blue arrowheads) in both subjects. In the example on the right (d), the submucosal gland empties into a duct that emanates at the lumen (green asterisk). The middle row shows a side-by-side comparison of the GEJ, demonstrating the transition of tissue from the squamous mucosa to the stomach’s cardia, depicting a common appearance in subject A (b) and an uncommon appearance characterized by the presence of glandular architecture (blue arrowhead) in subject B (d). The bottom row shows a typical pit pattern (blue arrowheads) on the surface of the gastric cardia in both subjects. In addition, subsurface glandular structures (green asterisks) can be seen below the surface of the cardia in subject B (f).

Visualization of more than 50% of the full circumference per frame was achieved in 92.3% of all 57,679 frames acquired in the esophagus and the GEJ. In every subject who swallowed the capsule, we obtained at least one imaging run where the entire circumference of the esophagus and the GEJ were clearly visualized in more than 85% of the frames. The last ascending imaging run typically had the best tissue visualization, where the full circumference of the esophagus and GEJ were clearly seen in 97 ± 5% of the frames, and the first descending run had the lowest tissue visualization, where the full circumference of the esophagus and GEJ were clearly visible in $86\pm 9\%$ of the frames.