Endoscopic submucosal dissection (ESD) is the standard treatment for early malignant colonic lesions and advanced colorectal neoplasms.¹ Owing to a higher en bloc resection rate, the local recurrence rate in ESD is lower than in endoscopic mucosal resection. Colonic ESD is technically demanding, and is thought to be a higher risk procedure than gastric ESD.^2,3 The lack of ESD specialists and the challenges in learning ESD has limited the use of colonic ESD in Western countries.⁴ A crucial step is exposing the dissection plane to make the ESD procedure safe and efficient.⁵ If resection is performed blindly, complications such as perforation and bleeding can arise.⁶ There have been unsuccessful attempts to make ESD easier and safer by providing countertraction of the mucosal flap.^7-9 With the widespread use of natural orifice transluminal endoscopic surgery, various robotic multitask platforms have been developed to assist colonic ESD.^10-12 Robotic multitask platform enables faster, more precise, and safer ESD procedures, but has the disadvantage of being bulky and incompatible with conventional endoscopes.¹³ Therefore, we developed a retractable modular robotic device for colonic ESD that can be attached to a conventional endoscope (e.g., an endoscopic cap). This study aimed to evaluate the feasibility and safety of our novel assistive robotic device for colorectal ESD. A staged study was performed, beginning with an ex vivo comparative study of the porcine colon, comparing robot-assisted ESD (RESD) and conventional ESD (CESD), and testing the in vivo feasibility in live pigs.

1. Robotic device

Our device comprises a robotic manipulator, user interface (UI), and actuation console (Fig. 1). The manipulator consists of an endoscope cap-like case with a built-in arm and grasping forceps and can be attached to a standard colonoscope. The cap-like case of the traction device has an elliptical shape when viewed from the front. The vertical axis (long axis) includes the robot arm and has a length of 23 mm. The horizontal axis (short axis) does not include the robot arm and has a length of 16 mm. The distal end outer diameter of the colonoscope (CF-HQ290; Olympus, Tokyo, Japan) is 13.2 mm, and the inner diameter of the traction device case is also 13.2 mm. The robotic arm protrudes forward from the cap-like case while assisting with the ESD procedure (Fig. 2). The endoscopist controls the UI to operate the robotic arm that has five degrees of freedom. The robotic arm can be angled upward and downward and to the left and right. It can rotate 180° circumferentially along the distal end of the endoscope. When the robot arm is maximally protruded, the length is 20 mm. The length of the protruding robot arm can be adjusted freely within the range, allowing traction back and forth. The console monitor with a graphic simulator displayed the position of the robotic arm (Fig. 1).

Figure 1. Overview of the concealable robotic add-on assistive device system. Our device consists of an actuation console with a graphical simulation monitor (A), an intuitive robot control user interface (B), and a robotic manipulator that can be attached to an endoscope (C).

Figure 2. Operation details of the concealable robotic add-on assistive device system. (A) The robotic arm is concealed inside the endoscopic cap-like case. (B) It can be extruded forward when assisting the endoscopic submucosal dissection procedure. (C, D) The robotic arm can bend up and down, left and right, and rotate 180° along the circumference of the distal end of the endoscope. The length of the protruding robotic arm can be adjusted freely, allowing traction back and forth.

2. Experimental preparation

We designed two studies to evaluate the efficacy of the robotic device.

1) Study 1

Colorectal ESD procedures were performed, by a single ESD expert, using the robotic device in 18 ex vivo porcine colons. CESD was performed in 18 additional cases for comparative purposes, using the same colonoscope with a standard cap. Resected porcine colon models were used in the ex vivo study. The ex vivo study of porcine colon was conducted within 3 hours on the colons of a pig sacrificed on the morning of the same day. To maintain the vitality, all colons were flushed with normal saline solution and were not refrigerated until the beginning of the study. The ex vivo colon was inverted inside-out to expose the mucosa. After lavage, lesions with 25 mm diameters were marked with a pen on the posterior (6 o’clock position) and anterior colonic walls (12 o’clock position) of different parts of the colon model, using a standard circular template. The colon was divided into three zones: ascending, transverse, and descending (Supplementary Fig. 1A). A total of 36 lesions were created in six resected colons. The number of procedures was equally distributed among the three zones in both the RESD and CESD groups. This corresponded to six procedures at each location for each method. The colon was then everted to its normal anatomical configuration and affixed to a colonic simulator platform (Supplementary Fig. 1A). A customized overtube was inserted through the distal end of the colon model to facilitate entry.

2) Study 2

An in vivo porcine study was conducted to evaluate the feasibility of the proposed robotic device. Two Yorkshire pigs were used for the RESD experiments (Supplementary Fig. 1B). The animals just received boiled rice 2 days before study. Food was withheld, but water was allowed during the preparation for colonoscopy. Each pig received split dose of 2L-PEG (Coolprep; Taejoon Pharm, Seoul, Korea) solution through orogastric intubation. The split doses were divided into two portions, with half of the dose (1L-PEG) given in the evening (6:00 PM) on the day before the examination and the second dose given the next morning (8:00 AM) before the examination (12:00 noon). With the robotic manipulator mounted, the colonoscope was inserted into the most proximal colon. During withdrawal, the colonic wall was carefully examined for mucosal damage. While retracting the endoscope, ESDs were performed on three separate locations (60 cm, 30 cm, and 10 cm from anal verge) in each pig. The participant made markings of a virtual lesion that has diameter of 2 to 3 cm with an ESD dual knife. Six RESDs were performed, two at each of the three locations mentioned above. This study was approved by the Animal Test Center of the College of Medicine, Korea University (IACUC number: KOREA-2021-0057).

All procedures were conducted by an experienced endoscopist who has performed more than 200 gastric ESD and 100 colon ESD procedures without any direct technical assistance. The CESDs and RESDs were performed as follows: a transparent cap (FM-EC0007; Finemedix, Seoul, Korea) or robotic manipulator was attached to the distal end of a standard colonoscope (CF-HQ290; Olympus, Tokyo, Japan) and advanced to the target lesion. The lesion was marked using a dual knife (KD-650Q; Olympus). Submucosal injections were administered using a mixed solution of saline and indigo carmine. The mucosal layer surrounding the lesion was circumferentially cut using a dual knife. We used the ERBE VIO 300 D (Erbe, Tubingen, Germany). The output setting for ESD procedures is described in Supplementary Table 1. In the conventional group, submucosal dissection was performed in a similar manner.¹⁴ In the robotic group, conventional submucosal dissection was performed until the mucosa had sufficient space for grasping. At this stage, the robotic arm was advanced from the cap and the dissected mucosal flap was grasped. The mucosal flap was then lifted and the submucosal plane was exposed by applying countertraction of the flap (Fig. 3). The endoscopist manipulated the robotic arm by using a UI attached to the endoscope (Supplementary Fig. 2). The resected mucosa was removed by grasping by the robot, or by suction into the cap in the conventional group. All experiments were video-recorded.

Figure 3. Graphical illustration of endoscopic submucosal dissection procedure using our robotic device. Multidirectional traction with the robotic arm provides clear visualization of the submucosal plane, minimizing disruption of the current clinical workflow.

3. Outcome evaluation

The following parameters were recorded for all ESD procedures: location, total procedure time, operative time for specific ESD steps (e.g., dissection, incision, and grasping), size of the resected tissue, en bloc or piecemeal resection, complete resection and adverse events. The primary outcome measure was the total procedure time, which was calculated as the time from the initial mucosal marking to the removal of the resected lesions. The secondary outcome measures included submucosal dissection speed (mm²/min), blind dissection rate, complete resection, and complications. The submucosal dissection speed was calculated as the resected area of the specimen divided by dissection time. A blind dissection was defined as a submucosal dissection in which the dissection plane was not clearly visible. The blind dissection rate refers to the ratio of blind dissection time to total dissection time. Grasping time was defined as the time from which the robotic arm stretched out to the time the submucosal dissection began after appropriately holding the flap. To obtain accurate values, the recorded video was analyzed using a stopwatch.

4. Sample size calculation

The sample size of ex vivo study was estimated based on the primary endpoint (comparison of total procedure time). According to our preliminary study on porcine colons, the mean and standard deviation of the total procedure time for CESD were 30 and 10 minutes, respectively. We hypothesized that RESD would lead to a 10-minute reduction in the total procedure time. With a standard deviation of 10 minutes, a two-sided significance level of 5%, and a power of 80%, 18 subjects were required in each group with a 1:1 allocation.

5. Statistical analysis

Continuous variables are reported as mean±standard deviation, and categorical variables are reported as numbers. Ordinal qualitative and quantitative variables were compared using the Wilcoxon or Kruskal-Wallis rank-sum test. Paired comparisons of qualitative variables were performed using the Fisher exact test or the chi-square test. Statistical significance was set at p<0.05, and the results were considered statistically significant. SPSS software (version 25.0; IBM Corp., Armonk, NY, USA) was used for all the analyses.

1. Study 1

Thirty-six target lesions were created in the anterior and posterior walls of the ascending, transverse, and descending colons. ESD was successfully performed in a total of 36 cases using either RESD (n=18) or a conventional endoscope with a cap (n=18). Complete resection was achieved in all procedures. Two CESD cases had minor perforations. Table 1 summarizes the outcomes of RESD compared to those of CESD. There was no difference in the specimen size (RESD vs CESD: 518.4±27.2 mm² vs 525.5±23.8 mm², p=0.40). Total procedure times were significantly reduced in the RESD group compared to those in the CESD group (20.1±4.1 minutes vs 34.3±8.3 minutes, p<0.05) (Table 1). The total procedure time was subdivided into the incision, grasping, and submucosal dissection time. The submucosal dissection speed was significantly higher in the RESD group than in the CESD group (36.8±9.2 mm²/min vs 18.1±4.7 mm²/min, p<0.05). There was no significant difference in incision time between the two groups. In the conventional method, the average blind dissection rate was 35.1% compared to 12.8% for RESD (p<0.05). This value includes the blind dissection that occurred when conventional submucosal dissection was performed in the RESD group to create a submucosal flap. The blind dissection rate in the RESD group after grasping was 8.5%. In RESD, it took an average of 1 minute and 36 seconds to grasp the flap using the robotic arm after making the flap, without a significant difference in the lesion site.

Table 1 . Outcomes of RESD and CESD.

Outcome	RESD (n=18)	CESD (n=18)	p-value
Dissection area, mm2	518.4±27.2	525.5±23.8	0.403
Complete resection rate	100	100	-
Total procedure time, min	20.1±4.1	34.3±8.3	<0.05
Dissection time, min	14.9±4.1	30.9±8.3	<0.05
Dissection speed, mm2/min	36.8±9.2	18.1±4.7	<0.05
Grasping time, min	1.6±0.6	NA	-
Incision time, min	3.4±0.6	3.3±0.5	0.812
Perforations during the procedure	0	2 (11)	0.153
Blind dissection rate	12.8±3.4	35.1±3.9	<0.05

Data are presented as mean±SD or number (%)..

ESD, endoscopic submucosal dissection; RESD, robot-assisted ESD; CESD, conventional ESD; NA, not applicable..

1) Subgroup analysis

The subgroup analysis demonstrated that the total procedure time spent for the posterior wall was significantly longer than that for the anterior wall in the CESD group (28.6±4.2 minutes vs 39.2±6.9 minutes, p<0.05) (Table 2). The dissection speed of CESD was greatly reduced when the lesion was placed on the posterior wall. In the RESD group, no significant differences in the procedure time and dissection speed between the two walls (Table 2). The safety profile was improved when RESD was used for posterior wall lesions, as two posterior wall perforations occurred in the CESD group. We compared the outcomes of ESD performed in three different zones: ascending, transverse, and descending colon. No significant difference in the total procedure time based on locations between the two groups was observed (Table 2).

Table 2 . Subgroup Analysis of the Total Procedure Time Based on the Location of Lesions.

	RESD			CESD
	Total procedure time, min	p-value		Total procedure time, min	p-value
Zone		0.910*			0.670*
Ascending (n=6, each)	19.6±4.5	0.720 (ascending vs transverse)		34.3±5.4	0.460 (ascending vs transverse)
Transverse (n=6, each)	20.7±5.5	0.870 (ascending vs descending)		36.3±3.5	0.830 (ascending vs descending)
Descending (n=6, each)	19.9±2.5	0.780 (transverse vs descending)		34.8±2.7	0.430 (transverse vs descending)
Lesion		0.610			<0.05
Anterior wall (n=9, each)	20.6±4.6			28.6±4.2
Posterior wall (n=9, each)	19.6±3.8			39.2±6.9

Data are presented as mean±SD..

ESD, endoscopic submucosal dissection; RESD, robot-assisted ESD; CESD, conventional ESD..

*p-value for the Kruskal-Wallis test..

2. Study 2

We tested feasibility of our robotic device in the proximal colon above rectum of live pigs. Two Yorkshire pigs were used in robot-assisted in vivo ESD experiments. It was possible to insert up to 70 cm from the anal verge in each experiment, with a robotic manipulator attached to the distal end of a standard colonoscope. The range was limited as the porcine proximal colon is spiral. No mucosal injuries caused by the robotic device during insertion and withdrawal were observed. The robotic arm operated well without significant hysteresis, providing optimal traction. Six RESDs were successfully performed at anal verges of 60 cm, 30 cm, and 10 cm in the pigs without significant adverse events (Table 3, Fig. 4).

Figure 4. Robot-assisted endoscopic submucosal dissection of a live porcine colon. (A-F) The robot arm protrudes from the cap, grasps the dissected mucosal flap, and exposes the dissection plane by countertraction of the flap. The submucosal layer was exposed through countertraction in various directions while making dissections. Arrow indicates the direction of traction applied.

Table 3 . Outcomes of Colorectal Endoscopic Submucosal Dissection Using a Retractable Robotic Device on a Live Porcine Model.

Animal No.	Location	Dissection area, mm2	Total procedure time, min	Mean dissection speed, mm2/min	Grasping time, min	Blind dissection rate, %	Complete resection	Adverse events
1	AV 60 cm	495.8	15.2	46.7	1.2	13.5	Yes	-
1	AV 30 cm	519.7	17.5	38.1	1.5	12.6	Yes	-
1	AV 10 cm	527.7	16.2	41.8	1.2	15.5	Yes	-
2	AV 60 cm	569.1	18.3	42.4	0.8	11.0	Yes	-
2	AV 30 cm	502.7	17.0	39.5	1.8	14.4	Yes	-
2	AV 10 cm	529.2	18.2	41.8	2.3	12.6	Yes	-
Mean		524.0	17.1	41.7	1.5	13.2	Yes	-

AV, anal verge..

Colorectal ESD has clear benefits of high en bloc resection rates and low local recurrence rates. However, it is not a widely used endoscopic technique, except in some East Asian countries, such as Japan and Korea.¹² Its limited adoption is partially due to the procedure being time-consuming, technically demanding, and carrying a high risk of adverse events. A key factor in solving these issues is providing optimal traction to the mucosal flap.¹⁵ Various tissue traction devices can reduce blind dissection. However, few are actually used in practice.¹⁶ Many traction devices have problems, such as difficulty in changing the tension applied to the lesions and in adjusting the point of traction as well as cumbersome preparation, and application on a clinical basis. Robotic machines designed to enable more intuitive bimanual tissue manipulation are currently being developed. Among these, the most active areas of development are robotic endoscopic multitasking platforms. These the potential, such as MASTER¹⁰ and EASE¹¹ systems, have the potential to provide better visualization, improved dexterity as well as improved precision of surgical manipulation. These devices follow the master-slave robotic concept and are designed for telemanipulation. The basic concept is to hold the ESD flap with the robotic arm and proceed with dissection using the other end effector. However, this system has the following disadvantages: the large master console size, learning of the new method, and high cost. Above all, these systems have a robotic end effector at the distal end of the endoscope. Therefore, in colonic ESDs, owing to their size, entering the proximal colon above the rectosigmoid colon will be challenging.

In this study, we designed a robotic device that attaches to the distal end of the colonoscope for colorectal ESD. Colonic ESD is technically difficult, especially in the proximal colon, such as the cecum and the ascending colon. More ESD complications occur in these lesions than those in the distal colon.¹⁷ For this reason, assistance from ESD assistive devices is particularly needed for such cases. However, it may be technically difficult to insert larger assistive devices into the proximal colon. Likewise, robotic colonic ESD studies published thus far have not been conducted above 35 cm from the anal verge.^10-12 The advantage of our robot is that the robot arm can be retracted into a thin endoscope cap so that it can safely enter the proximal colon, without causing mucosal damage. We were able to demonstrate this feature in our live animal study.

An attachable manual traction device called EndoLifter (Olympus Medical Systems, Co., Tokyo, Japan), which is similar to this device, has been developed. It consists of retractable grasping forceps that is attached to a transparent cap by a hinge and allows simultaneous grasping, retracting, and lifting of the mucosa.¹⁸ This product has been commercialized; however, it has not been widely adopted. A limitation of this product is that it can only apply traction along one axis, back and forth, making it challenging to expose the dissection plane precisely. The greatest strength of our robot is that it can accurately provide tension at the discretion of the endoscopist. The robotic manipulator can be used by controlling the UI attached to the control body of the endoscope. It can manipulate tissues up and down, left and right, back and forth and circumferentially. The direction, angle, and power of lifting the grasped tissue with the robot arm can be easily adjusted. It is also easy to repeatedly grab the tissue and hold it in place. This eases the dexterity of the endoscopist to manipulate the endoscope and knife. Other traction platforms often require two or more operators, such as endoscope and console controllers. Our robot has the advantage of being able to be operated by a single endoscopist, and it is relatively inexpensive. The robotic cap is made of stainless steel. We have also developed a polyethylene terephthalate robotic cap that can be easily manufactured using a three dimensional printer. Both are disposable with no risk of cross-contamination. The operation of the robotic arm is based on a simple tendon-and-sheath mechanism. This robotic device is an excellent economical option, especially in low-volume centers with low ESD case rates.

The coaxial motion was a major limitation of these traction devices. When the robotic arm holds the tissue at a fixed point, the lateral movement of the endoscope is limited. This also causes the dissection plane to constantly change, making stable submucosal dissection very difficult. Such shortcomings could be improved by making the connection part between the robot arm and the cap-like case more flexible. Our experiment showed improvement in the lateral movement of the endoscope. We are designing an upgraded flexible robotic arm that can automatically apply more tension in a direction opposite to the dissection direction.

This study demonstrates that this robot allows for a significantly shorter procedure time, faster dissection speed, and fewer submucosal injuries than the CESD technique in an ex vivo porcine model for colonic ESD. It provides simple, effective, and safe multidirectional traction and thus, adequate visibility of the dissection plane (Fig. 4). ESD was successfully performed by entering the proximal colon in both in vivo and ex vivo studies. In the ascending colon, the cable connected to the robot manipulator was bent along the colonic wall. However, the robot arm functioned well without significant hysteresis or malfunction. In addition, although there were differences between the operator and the study environment, the obtained results were not inferior to those reported in ex vivo model studies using other multitasking platforms in terms of dissection speed or blind dissection rate.^10-12 Despite the grasp time, the total procedure time was shorter in the RESD group than in the conventional method group. RESD is particularly effective for posterior wall lesions, where submucosal visualization is difficult owing to gravity (Fig. 5). Furthermore, the dissection speed was faster, and the grasp time was shorter in the in vivo study than in the ex vivo study. In the ex vivo model, the colon was not completely fixed to the simulator floor surface. Therefore, in some cases, the tension was not properly applied, although the flap was held with a robot arm. In the in vivo tissues, countertraction tended to be better applied, resulting in faster dissection speed. This method may be an effective treatment option, especially when dissecting fibrotic lesions such as early colon cancer lesions or colorectal adenomas with severe adhesion.

Figure 5. Using the robotic traction device, submucosal dissection can be effectively performed. (A) Anterior wall lesions of the colon with good visualization of the submucosal layer with the aid of gravity. (B) Visualizing the submucosal dissection plane is difficult if the lesion is on the posterior wall of the colon.

This study has several limitations. The technical difficulties associated with proximal colonic ESD are not well reflected in the ex vivo study. Situations such as endoscope looping or unstable positioning of endoscope did not occur in these experiments. No significant differences were observed among the ascending, transverse, and descending colons in the subgroup analysis of our study (Table 2). In the ex vivo colon, the submucosal solution was not well injected into the tissues; therefore, the inferior ESD outcomes in the CESD group may have been overestimated. It is also a limitation of this study that no statistical comparison was performed between the robotic and conventional groups in the in vivo study.

Many endoscopists do not initially perform a 360° circumferential mucosal incision in colon ESD because it is difficult to maintain a sufficient submucosal cushion.¹⁹ The initial circumferential mucosal incision strategy in the robotic and conventional groups may cause longer procedure time and decrease the dissection speed in CESD in this study. While it was necessary to standardize the process for making step-by-step comparisons, this could potentially hinder the accurate comparison between the conventional method and RESD.

In conclusion, the RESD device is effective for ESD as demonstrated by ex vivo and in vivo studies. This study demonstrated the superiority of RESD over CESD. RESD provided improved visualization and tissue traction, which significantly reduced blind dissection and dissection times.

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (grant no. NRF-2020R1A2C4002621). This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2022R1A2C2006986).

J.H.H. is an editorial board member of the journal but was not involved in the peer reviewer selection, evaluation, or decision process of this article. No other potential conflicts of interest relevant to this article were reported.

Study concept and design: S.H.K., C.K., B.K. Data acquisition: J.I., S.W., B.G.K. Data analysis and interpretation: K.K., D.H., Y.T.J. Drafting of the manuscript: S.H.K., C.K. Critical revision of the manuscript for important intellectual content: T.K., B.K., E.S.K. Statistical analysis: B.G.K., K.K., D.H. Administrative, technical, or material support: B.G.K., K.K., D.H., H.J.J. Study supervision: H.S.C., H.J.C., J.H.H. Approval of final manuscript: all authors.

Supplementary materials can be accessed at https://doi.org/10.5009/gnl230280.