Introduction

Third space endoscopic procedures like endoscopic submucosal dissection (ESD) and peroral endoscopic myotomy (POEM) are technically challenging and bear the risk of operator dependant complications such as bleeding and perforation. We aimed at developping and evaluating a deep learning algorithm for the real time detection and delineation of submucosal blood vessels during third space endoscopy.

Aims & Methods

We used 5.470 annotated endoscopic images from 59 full length ESD and POEM videos and 179.681 unlabeled images from the same procedures for the semi-supervised training of a DeepLabv3+ based neural network. The model detects and delineates submucosal blood vessels on endoscopic images in real time. The algorithm was tested on 101 standardized video clips with 200 predefined blood vessels derived from 15 separate third space endoscopy videos. Vessel detection rate, time and duration, defined as the percentage of single frames, on which the algorithm detected a defined vessel in relation to the gold standard were determined. Furthermore, the performance of 8 endoscopists plus or minus AI support with regard to blood vessel detection was evaluated.

Results

The mean Dice score for blood vessels was 68%. The overall vessel detection rate (VDR) of the algorithm was 94% (ESD 96%, POEM 74%). The median overall vessel detection time (VDT) of the AI algorithm was 0.32 sec (ESD 0.3 sec, POEM 0.62 sec). The mean overall vessel detection duration (VDD) was 59.1% (ESD 60.6%; POEM 44.8%) of the gold standard. All eight endoscopists had a higher VDR with AI than without. Mean VDR with AI was 71.2% and 56.4% without (p<0.001).

Conclusion

Regarding blood vessel detection, all endoscopists performed better with AI support than without. A vessel detection time of well under one second and a vessel detection duration of over 50% was considered adequate for clinical application. AI support was associated with a significantly higher vessel detection rate but in many cases a lower rate than AI alone. This was attributed to the subtle and in some situations not clearly visible delineation by the AI algorithm on the endoscopic image. In future prospective trials the AI algorithm should be evaluated for clinical relevance, i.e. lower complication rate and faster procedure time.

Disclosure

LHSL: GenieBiome Ltd. (research product support, lecture fees), Olympus Co. Ltd. (Lecture fees), Boston Scientific Co. Ltd. (Lecture fees)
PWYC: Cornerstone Robotics (Board Member), Endomaster Pte Ltd (Scientific Advisor), Boston Scientific Co Ltd (Research Collaboration), Olympus Co Ltd (Research Grant)

Introduction

Gastroesophageal reflux disease (GERD) is common after per-oral endoscopic myotomy (POEM). Selective sparing of oblique fibers (OFS) during POEM may reduce the incidence of reflux esophagitis after POEM.

Aims & Methods

Aim: In this study, we aim to compare the incidence of GERD between conventional (CM) versus oblique fiber sparing (OFS) techniques of endoscopic myotomy in cases with idiopathic achalasia.
Methods: Eligible patients with type I and II achalasia who underwent POEM from Jan 2020 to October 2020 were randomized into two groups (CM and OFS). Exclusion criteria were: type III achalasia, sigmoid esophagus and history of Heller’s myotomy with fundoplication. The primary outcome of the study was comparison of reflux esophagitis in the two groups. The secondary aims included reflux symptoms, esophageal acid exposure, and clinical success.

Results

115 patients were randomized in CM (58) and OFS (57) groups. Both the groups were similar with regards to the symptom scores, sub-types of achalasia and history of previous treatment. The mean operation time was similar in both the groups (CM 43.7±23.8 mins versus OFS 46.4±24.7 mins). Mean length of total myotomies were comparable in the two groups (9.9±3.0cm vs 9.5 ±3.1cm). Clinical success (Eckardt≤3) was recorded in all the patients. Overall, reflux esophagitis was found in 56 (48.7%) patients. Grade B or higher esophagitis was similar in the two groups (25.9% vs 31.6%, p=0.541). The mean number of reflux episodes (48.2±36.6 vs 48.9±40.3, p=0.933), proportion of patients with increased esophageal acid exposure >6% (45.5% vs 31.7%, p=0.266) and high DeMeester scores (>14.7) (38.6% vs 41.5%, p=0.827) were similar in both the groups. Symptomatic GERD (GERDQ>7) and use of PPIs at 1-year was similar in the two groups (GERDQ>7: 29.8% vs 26.5%, 0.822)(PPIs: 34% vs 28.6%, p=0.661).

Conclusion

Sparing of sling fibers has no significant impact on the incidence of GERD after POEM. Novel strategies need to be explored to prevent GERD after POEM. (NCT04229342)

Disclosure

No disclosure

Introduction

There is no unanimity regarding the most appropriate needle to use for endoscopic ultrasound-guided fine needle biopsy (EUS-FNB). To date, new types of FNB needles have been designed, including the Fork-tip and Franseen needles. The Fork-tip is characterized by two sharp tips on opposite sides of the lumen, and the Franseen needles are characterized by three symmetric cutting tips. This study primarily aimed to compare the diagnostic utility and histological quality between the use of the Fork-tip and Franseen neesles in EUS-FNB for solid pancreatic lesions.

Aims & Methods

We retrospectively analyzed 187 patients at our center for solid pancreatic lesions, 75 of whom underwent EUS-FNB using a 22 G Franseen needle and 112 using a 22 G Fork-tip needle, from December 2019 to October 2022.

Results

After propensity-matched analysis, each group was adjusted to 66 cases. The results revealed that sex, age, route for puncture, and tumor size, site, and tumor type were not different between the Franseen and Fork-tip groups. Both groups were not significantly different in terms of number of punctures, procedure time, core tissue score, blood score, and technical success rate. From the histological findings, the most common final diagnosis was adenocarcinoma (107 patients), which was followed by a benign lesion (19 patients), and finally, the NET (6 patients). In this study, no adverse events during the procedure were encountered. The diagnostic accuracy of the Fork-tip group (92.4%; 61/66) was similar to that of the Franseen group (90.9%; 60/66) (P >0.99). The outcomes of the two groups were similar for the distinction between benign and malignant lesions. The rate of changes of operators from a trainee was significantly less in the Fork-tip group (9.1%, 6/66) than in the Franseen group (24.2%, 16/66) (P = 0.03). The reasons for operator changes were as follows: In the Franseen group, there were ten cases of difficulty in adjusting the puncture route from the second part of the duodenum, six cases of difficulty in penetrating the gastric wall from the stomach. In contrast, in the Fork-tip group, there were two cases of difficulty in adjusting the puncture route from the second part of the duodenum, two cases of difficulty in penetrating the gastric wall from the stomach, two cases of difficulty in avoiding vessels. All cases were successfully punctured after an expert took over.

Conclusion

Although in both groups, the diagnostic accuracy and other outcomes of procedures conducted, were not significantly different, the rate of changes of operator from a trainee to an expert was significantly less in the Fork-tip group than in the Franseen group. Thus, the Fork-tip needle was observed as good choice for trainees and the appropriate needle must be selected according to the situation.

Introduction

Endoscopic full thickness resection (EFTR) is emerging as an effective modality for mucosal and submucosal lesions in colorectum.

Aims & Methods

In this systematic review and meta-analysis, we aimed to analyze the success and safety of device assisted EFTR in colon and rectum. A literature search was performed in Embase, PubMed, Medline databases for studies evaluating device assisted EFTR between inception to October-2022. The primary outcome of the study was clinical success (R0 resection) with EFTR. Secondary outcomes included technical success and adverse events.

Results

29 studies with 3467 patients [2045(59%) males] with 3492 lesions were included in the analysis. The lesions were located in right colon in 1660 (47.5%), left colon in 990 (28.6%) and rectum 840 (24.3%). EFTR was performed for sub-epithelial lesions in 252 (7.2%) patients. Pooled mean size of the lesions was 16.6 mm (95% CI 14.9 – 18.2, I ² 98%). Technical success was achieved in 87.1% (95% CI 85.1%-88.9%, I²39%) procedures. The pooled rate of en bloc resection was 87.9% (95% CI 85.1%-88.9%, I²50%) and R0 resection was 81.6% (95% CI 78.8% to 84.2%, I²57%). In sub-epithelial lesions, pooled rate of R0 resection was 94.3% (95% CI 89.7% to 96.9%, I² 0%). Pooled rate of adverse events was 12.1%(95% CI 10.3%-14.1%, I²44%) and major adverse events requiring surgery was 2.5% (95% CI 2.0%-3.1%, I²0%).

Conclusion

Device assisted EFTR is a safe and effective treatment modality in cases with adenomatous and sub-epithelial colorectal lesions. Comparative studies are required with conventional resection techniques including endoscopic mucosal resection and submucosal dissection.

Introduction

Rigid surgical instruments used for conventional laparoscopic surgery provide limited movement, and it is difficult to perform skillful surgery due to their long, jointless straight features. Articulated instruments could have the same advantages seen in robotic surgery, such as excellent control in smaller spaces, restoration of more intuitive and ergonomic movements, and the ability to overcome obstacles on a direct visual line. The aim of this study was to evaluate the surgical performance and clinical outcomes of patients following laparoscopic colorectal surgery with multijoint laparoscopic instruments, ArtiSential® (LivsMed, Seongnam, Korea), compared with those using conventional instruments.

Aims & Methods

Patients aged 19–80 years undergoing elective laparoscopic surgery for colorectal disease between December 2021 and November 2022, were consecutively enrolled. The first 20 cases underwent conventional laparoscopic surgery, and the latter 50 cases underwent laparoscopic surgery using ArtiSential®. Unedited operative videos were submitted to the Crowd-Sourced Assessment of Technical Skills (C-SATS) platform for evaluation using the Global Operative Assessment of Laparoscopic Skills (GOALS), which is a validated tool for technical proficiency. Each of GOALS’ four domains (bimanual dexterity, depth perception, efficiency, and tissue handling) was scored on a Likert scale between 1 and 5, scoring anchor, with a total score ranging from 4 to 20. The learning curve analysis was performed using the cumulative sum control chart (CUSUM) for GOALS grades (Registration number: KCT0006798).

Results

There was no significant difference between group with regards to type of operation, hospital stay, and 30-day postoperative complication. The GOALS’ grades were similar between the two groups (15.7 ± 0.7 vs. 15.9 ± 0.5, P = 0.190), but the scores in two domains (depth perception and tissue handling) were significantly higher in the ArtiSential® group (3.9 ± 0.3 vs. 4.0 ± 0.2, P = 0.014 and 3.9 ± 0.2 vs. 4.1 ± 0.2, P = 0.048, respectively) compared with those in the conventional group. In addition, patients in the ArtiSential® group experienced a shorter operation time (122.3 ± 58.9 vs. 88.1 ± 35.1, P = 0.022) and less intraoperative blood loss (56.5 ± 69.9 vs. 29.8 ± 1.4, P = 0.002) than those in the conventional group. According to the CUSUM learning curve analysis, the first 10 cases using ArtiSential® were in the inexperienced phase, and the remaining 40 cases were in the experienced phase. The GOALS’ score in the experienced phase of ArtiSential® was improved compared with that in the conventional group (16.0 ± 0.5 vs. 15.7 ± 0.7, P = 0.037).

Conclusion

A plateau in the learning curve of laparoscopic colorectal surgery performed with an articulated instrument was observed in approximately 10 cases, and it can be expected that surgical proficiency in the experienced phase increases without deterioration of clinical outcome as compared to that in conventional surgery.

References

.

Disclosure

This research was supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grantnumber: HI20C2092).

Introduction

Endoscopic full thickness resection (EFTR) of the colon has been described as an effective practicable method for resection of recurrent /advanced neoplastic lesions, but its performance and success may be difficult or different according to location of the lesion in the corresponding colonic segment.
The aim was to investigate the segment-specific efficacy and safety of EFTR in the colon using a full thickness resection device (FTRD) from real-world data.

Aims & Methods

EFTR was conducted in patients being referred to either of the 4 participating centers after previous colonoscopy leading to a diagnosis suitable for EFTR. The primary efficacy measure of EFTR was histologically complete resection (R0) of the lesion. Safety was investigated considering the frequency of adverse events (AE) to EFTR occurring during the procedure or afterwards.

Results

105 patients (66m/39f; median 70yrs) were included in the analysis. FTRD was technically successful in 101 patients (96.2%) with a median lesion size of 20 x 19mm (15-24).
Histologically R0 resection was obtained in 77.1% of the patients with a segment-specific range from 90% (rectum), 70.6% sigmoid, 78.6% descending colon to 62.5% (transverse colon), while in ascending colon and cecum resection rates of 76.9 and 83.3% could be achieved.
Overall, adverse event rate (AE, any) were documented in 33 patients (31.4%), with only 1 perforation in sigmoid (0.95%), while the others were minor, abdominal postsurgical pain (18.1%), hematochezia (9.5%), and haemoglobin decline (7.6%). Topographical analysis revealed that AE were most attributable to transverse colon (12.5%), while in the other segments AE were each lower or equal than 8.2%; the lowest rate of AE was found in rectum with 1.4%.

Conclusion

EFTR is efficacious for the endoscopic treatment of colorectal lesions with best outcome parameters in rectum (high R0 rate, low AE), followed by cecum. The other segments appeared to be more difficult for complete resection and showed a higher AE rate. Transverse colon was identified with lowest R0 resection rates and highest frequency of AE. While only one major complication occurred, minor complications in all other segments were below 8.2% and short-lived.