GI Genius increases small and right-sided adenoma and sessile serrated lesion detection rate when used with EndoCuff in a real-world setting: a retrospective United States study

Jeong Hoon Kim; Jade Wang; Colton Pence; Patrick Magahis; Enad Dawod; Felice Schnoll-Sussman; Reem Z. Sharaiha; David Wan

doi:10.5946/ce.2024.271

Articles

Page Path: HOME > Clin Endosc > Ahead-of print articles > Article

Original Article GI Genius increases small and right-sided adenoma and sessile serrated lesion detection rate when used with EndoCuff in a real-world setting: a retrospective United States study: Jeong Hoon Kim^,*, Jade Wang^,*, Colton Pence, Patrick Magahis, Enad Dawod, Felice Schnoll-Sussman, Reem Z. Sharaiha, David Wan; DOI: https://doi.org/10.5946/ce.2024.271
Published online: April 22, 2025

Division of Gastroenterology and Hepatology, Weill Cornell Medicine, New York-Presbyterian Hospital, New York, NY, USA

Correspondence: Jeong Hoon Kim Division of Gastroenterology and Hepatology, Weill Cornell Medicine, New York-Presbyterian Hospital, 1300 York Ave, New York, NY 10065, USA E-mail: jhn.kim99@gmail.com

*Jeong Hoon Kim and Jade Wang contributed equally to this work as co-first authors.

• Received: October 8, 2024 • Revised: November 11, 2024 • Accepted: November 13, 2024

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (https://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

453 Views
48 Download

Full Article

Download PDF

Abstract
Graphical abstract
INTRODUCTION
METHODS
RESULTS
DISCUSSION
NOTES
REFERENCES

See letter "Where technology meets technique: computer-aided detection and mucosal exposure device to improve adenoma detection".

Abstract

Background/Aims
The real-world efficacy of computer-aided detection (CADe) systems, such as GI Genius (Medtronic), is unclear. We examined the colonoscopy metrics using CADe alone and with a mucosal exposure device (EndoCuff; Olympus) in a real-world setting.
Methods
We retrospectively reviewed screening and surveillance colonoscopies before, during, and after CADe use in a large tertiary care center. Outcomes included the adenomas per colonoscopy (APC), sessile serrated lesions per colonoscopy, adenoma detection rate (ADR), sessile serrated lesion detection rate (SSLDR), advanced ADR, total polyp detection rate, and true histology rate. The ADR and SSLDR were further examined according to size, colon location, and EndoCuff use.
Results
A total of 798 colonoscopies were performed, including 386 pre-CADe, 178 CADe, and 234 post-CADe. In cases where CADe was used with the EndoCuff, the 1 to 5 mm ADR increased from 36.3% (pre-CADe) to 52.1% (CADe) (p=0.01). The 1 to 5 mm SSLDR increased from 9.6% (pre-CADe) to 17.1% (CADe) (p=0.02). The right-sided ADR increased from 30.8% (pre-CADe) to 42.7% (CADe) (p=0.03). The right-sided SSLDR increased from 12.3% (pre-CADe) to 24.8% (CADe) (p<0.001). No significant changes were observed when only CADe was used. No differences were found in other outcome measures. Post-CADe metrics returned to pre-CADe levels.
Conclusions
GI Genius is useful for identifying small and right-sided polyps only when used with the EndoCuff.
Keywords: Artificial intelligence; Colonoscopy; Colorectal neoplasms

Graphical abstract

INTRODUCTION

Interest in the computer-aided detection (CADe) of precursor colorectal cancer polyps during colonoscopy has grown in recent years. Artificial intelligence (AI) systems, which utilize machine learning techniques to differentiate polyps from the background mucosa, can assist endoscopists in identifying polyps in real time.¹ A recent meta-analysis found that even endoscopists with a relatively high adenoma detection rate (ADR), i.e., ≥35%, still missed up to 17% of adenomas.² As a significant number of colorectal cancers are detected shortly after colonoscopy, and 50% to 60% of interval colorectal cancers are thought to be due to missed adenomas,³ real-time AI assistance has the potential to benefit patient care.

Several CADe systems based on machine learning principles have been developed recently. In randomized controlled trials, they have shown an increase in the adenomas detected per colonoscopy (APC) and/or ADR.⁴ These systems include the GI Genius (Medtronic), the first United States Food and Drug Administration-approved CADe system.⁵ In validation testing, the GI Genius has been shown to be capable of detecting most polyps observed by endoscopists, with a low false-positive rate.¹^,⁶ Nonetheless, whether the GI Genius can increase the ADR in routine clinical practice is unclear. While some studies have found that various CADe systems, including GI Genius, increase the overall ADR,⁷^-¹⁰ others have found no significant difference or a decrease in the overall ADR between AI-assisted (including GI Genius-assisted) and non-AI colonoscopies.¹¹^-¹³

This study examined the impact of the GI Genius system on the primary outcomes of the APC, sessile serrated lesions per colonoscopy (SSLPC), ADR, and sessile serrated lesion detection rate (SSLDR). We also studied other procedure- and endoscopist-related metrics of screening and surveillance colonoscopies, with the expectation that GI Genius would significantly improve all outcomes of interest. We also examined how these metrics were affected by discontinuation of the system.

METHODS

Study design

This retrospective review of screening and surveillance colonoscopies conducted at a single tertiary care center between January and December 2022. Inclusion criteria included a patient age ≥18 years and indications for screening or surveillance colonoscopy. The exclusion criteria included colonoscopies performed for other indications. Colonoscopies were conducted by 16 experienced faculty endoscopists who had baseline ADRs ≥35%, 14 of whom had baseline ADRs ≥45%.

The study population comprised three groups. Cases performed using GI Genius were considered AI cases. Cases performed by the same endoscopists in the months prior to GI Genius implementation, i.e., January–April 2022, were considered pre-AI controls. Cases performed by the same endoscopists after the GI Genius trial ended, i.e., July–December 2022, were considered post-AI. Endoscopists were included in the study if they had performed at least two AI-assisted procedures. A minimum of 20 pre- and post-AI cases per endoscopist were included, even for endoscopists with <20 AI cases. Endoscopists who performed a higher number of AI cases (i.e., 30–40) were matched with the same number of pre- and post-AI controls.

GI Genius implementation

Two GI Genius modules were provided on a trial basis and used for colonoscopies conducted in May and June 2022. The modules were placed in two separate endoscopy rooms. The use of GI Genius was at the provider’s discretion. If an endoscopist expressed interest in trialing the system, they and their patients were moved to one of the rooms equipped with GI Genius. Subsequently, the endoscopists chose whether to use the system on a case-by-case basis. GI Genius identified polyps in a manner consistent with previous descriptions in the literature.¹^,⁶^,⁸

All colonoscopies were performed using an high-definition colonoscope (Olympus). A device attached to the distal end of the colonoscope was used to manipulate the folds and increase the amount of exposed mucosa. EndoCuff Vision (Olympus) was used at the discretion of the provider. All providers performed at least one second look, either in the forward or retroflexed view, at their discretion. Gastroenterology fellows were present in some cases for training purposes. In all cases, the visualized polyps were removed using a snare or forceps at the discretion of the provider and according to standard practice guidelines. The removed polyps were sent to NYP-Weill Cornell for histopathological analysis. Individual patient-level data were stored in a personal health information-protected database (REDCap; Vanderbilt University). The use of the EndoCuff, Aronchick scale or Boston bowel preparation scale, presence of a GI fellow, and withdrawal time were documented. The Aronchick score was coded numerically as follows: poor, 0; fair, 1; good, 2; and excellent, 3.

Outcomes

The primary outcomes of interest were the APC, SSLPC, ADR, and SSLDR. Secondary outcomes included the advanced ADR, total polyp detection rate (PDR), and true histology rate (THR). According to existing guidelines, APC was defined as the total number of detected adenomas divided by the total number of colonoscopies performed.¹⁴ The SSLPC was defined as the total number of sessile serrated polyps detected divided by the total number of colonoscopies performed. The ADR was defined as the percentage of colonoscopies that detected at least one adenoma. The SSLDR was defined as the percentage of colonoscopies that detected at least one sessile serrated lesion (SSL).¹⁴

Definitions

Advanced adenomas were defined as any adenoma ≥10 mm or with a villous component or high-grade dysplasia. The advanced ADR was defined as the percentage of colonoscopies that detected at least one advanced adenoma. We further classified the ADR and SSLDR by size (1–5, 6–9, and ≥10 mm) and location (right- and left-sided). Right-sided polyps were defined as polyps occurring in the cecum, ascending colon, hepatic flexure, or transverse colon. Left-sided polyps were defined as those occurring in the splenic flexure, descending colon, sigmoid, rectosigmoid, or rectum. The PDR was defined as the percentage of colonoscopies that detected at least one polyp. The THR was defined as the percentage of the total resected polyps confirmed to be adenomatous polyps or SSLs on pathological examination.

Statistical analysis

Excel (Microsoft Corp.) was utilized for statistical analysis. A two-tailed t-test was used to assess the statistical significance of continuous variables between groups. The chi-square test was used to assess the statistical significance of categorical variables between groups. Stratified analysis was performed to control for EndoCuff use during the procedure. Statistical significance was set at p-values <0.05.

Ethical statement

This review was approved by the Institutional Review Board (protocol #2207025058) on January 24, 2023.

RESULTS

A total of 798 colonoscopies were performed: 386 pre-AI, 178 GI Genius-assisted AI, and 234 post-AI. The characteristics and demographics of the three groups are listed in Table 1. The groups were similarly balanced in all characteristics, except for the rate of the use of EndoCuff, which was 65.7% in the AI group and 37.8% in the pre-AI group. The number of AI procedures performed by each endoscopist ranged from 2 to 40.

The primary outcomes of interest, the APC, SSLPC, ADR, and SSLDR, were not significantly different between the three groups (Table 2). The advanced ADR, THR, and PDR were not significantly different between the three groups (Table 3).

Notably, when we examined the ADR by size (1–5, 6–9, and ≥10 mm), the AI group had a significant increase in 1 to 5 mm ADR compared to that in the pre-AI group (45.5% vs. 33.2%, p=0.03). This significant increase in 1 to 5 mm ADR did not persist in the post-AI cases performed in subsequent months; instead, the post-AI ADR for all adenoma sizes was comparable to that of the pre-AI group. In addition, the ADR was not significantly different between the AI and pre-AI groups for larger (6–9 and ≥10 mm) adenomas. The findings are summarized in Table 4 and Figure 1.

Similarly, when the SSLDR was examined by lesion size, a significant increase in the SSLDR was noted between the AI and pre-AI groups (15.2% vs. 8.0%, p=0.01) for small (1–5 mm) SSL. The AI and pre-AI groups had no significant difference in SSLDR for larger (6–9 mm and ≥10 mm) SSL. The findings are summarized in Table 5 and Figure 2.

The ADR and SSLDR were also examined in right- and left-sided locations. There was no difference among the three groups in either right- or left-sided ADR, although the AI and pre-AI groups showed a trend toward an increased right-sided ADR (p=0.05). However, the AI and pre-AI groups had a significant increase in right-sided SSLDR (12.7% vs. 22.5%, p=0.01). The three groups showed no differences in left-sided SSLDR. These findings are summarized in Tables 6 and 7 and Figure 3.

Statistically significant metrics, namely differences between pre-AI and AI cases in 1–5 mm ADR and SSLDR and right-sided SSLDR, were further stratified by EndoCuff usage to address the higher utilization of EndoCuff in the AI group than in the pre-AI group. We also examined the effect of EndoCuff use on the overall ADR, SSLDR, APC, and SSLPC, as these were the primary outcomes, and on the right-sided ADR because of its strong trend toward significance (Table 8). The overall SSLDR, APC, and SSLPC showed no significant change, regardless of EndoCuff usage. However, in cases where the EndoCuff was used, the overall ADR, 1 to 5 mm ADR, 1 to 5 mm SSLDR, right-sided ADR, and right-sided SSLDR were significantly increased in the AI group compared with those in the pre-AI group. These metrics were not significantly different when only AI was used. The 6 to 10 mm ADR and ≥10 mm ADR in the pre-AI and AI groups were calculated to determine their contribution to the increase in overall ADR with EndoCuff use, and no significant differences were found.

DISCUSSION

This study aimed to determine the utility of the AI CADe system, GI Genius, in a real-world setting limited to screening and surveillance colonoscopies. Therefore, we examined the metrics across pre-AI baseline cases, AI-assisted cases, and post-AI cases performed after GI Genius was discontinued. Our results indicate that for the primary outcomes of the APC, SSLPC, ADR, and SSLDR, as well as secondary outcomes including the advanced ADR, THR, and PDR, GI Genius did not provide any advantage over the pre-AI baseline. However, using the combination of AI and EndoCuff Vision significantly increased the 1 to 5 mm ADR, 1 to 5 mm SSLDR, right-sided ADR, and SSLDR. The overall ADR also increased with EndoCuff usage. However given that there were no significant increases in 6 to 10 mm and ≥10 mm ADR, the enhancement of the overall ADR can be solely attributed to the increase in the 1 to 5 mm ADR. Interestingly, when GI Genius was used alone, none of these metrics showed a significant increase. These results suggest a synergistic relationship between EndoCuff and GI Genius.

Our findings may explain the variation in the prior literature examining the effect of CADe systems on colonoscopy metrics. Many investigations have found that different CADe systems, including GI Genius, increase the overall ADR.⁷^-¹⁰ For example, an Italian randomized controlled trial found that GI Genius significantly increased the APC from 0.71 to 1.07, overall ADR from 40.4% to 54.8%, and ADR for ≤5 mm and 6 to 9 mm adenomas compared to controls.⁷ Additionally, a prospective study in Singapore demonstrated that GI Genius improves the ADR.⁸ A recently published study utilizing another CADe system, ENdoscopy as AI-powered Device (ENAD), showed significant improvements in the ADR, SSLDR, and APC with CADe use, particularly for small (<10 mm) and non-polypoid lesions.¹⁵ Other investigators have also studied the effects of CADe systems on the PDR and polyp detection sensitivity. Deep-GI is another CADe device demonstrating enhanced polyp detection sensitivity while decreasing false-positive rates, compared with pre-existing CADe models and the naked eye.¹⁶

Conversely, other studies found either no significant difference or a decrease in the overall ADR between CADe-assisted (including GI Genius-assisted) and non-CADe colonoscopies.¹¹^-¹³ A retrospective study by Levy et al.¹¹ showed a decrease in the ADR between colonoscopies utilizing GI Genius and colonoscopies performed in the previous year without AI (30.3% vs. 35.2%). Another retrospective study found no significant differences in the APC (1.08 vs. 1.04) or ADR (50.4% vs. 53.0%) between the pre-AI and AI groups.¹² This heterogeneity could exist because of the differential use of adjunctive tools, such as the EndoCuff, which have not been examined or controlled for in all prior studies. In our study, the use of GI Genius did not significantly enhance the PDR when used alone; however, there were significant increases in certain metrics when used in combination with the EndoCuff. Prior literature also suggests that AI is most effective for detecting smaller adenomas¹⁵^,¹⁷^-¹⁹ while providing less or no benefit for adenomas larger than 6 to 10 mm; however, not all prior studies conducted such subgroup analyses. This was supported by our own study, which showed that the GI Genius-EndoCuff combination significantly increased the ADR only for the smallest sized (1–5 mm) adenomas, although GI Genius alone did not. A Danish single-center prospective observational study evaluating the real-life efficacy of GI Genius alone showed no statistically significant differences in the ADR with GI Genius use, similar to our own overall ADR results. However, this study did not evaluate the combination of mucosal exposure devices with the EndoCuff and did not differentiate the location or size of the polyps.²⁰

The differences in results between randomized trials and real-world studies may also be explained by behavioral differences outside clinical trials. Levy et al.¹¹ found that both the ADR and PDR decreased during the AI period, as did the procedure time. Endoscopists may replace their typical meticulous withdrawal inspection technique when they rely on AI, perhaps even foregoing a second look. The study did not describe the use of a second look, and more importantly, the use of EndoCuff. Meanwhile, in our study, all endoscopists utilized the second look throughout the pre-AI, AI-assisted, and post-AI periods.

The additive effect of EndoCuff and GI Genius on small (1–5 mm) adenomas and SSL detection was notable. We found a significant increase in the right-sided ADR and SSLDR in cases where the GI Genius was used with the EndoCuff and no difference in cases where AI was used alone. The miss rates for small and right-sided adenomas are as high as 28% and 26%, respectively,² and the increase in the ADR achieved by the GI Genius-EndoCuff combination is promising. This is compelling, given that standard colonoscopies do not provide as much mortality benefit for right-sided as for left-sided colon cancers,²¹^,²² likely due to the unique biological and morphological characteristics of right-sided polyps.²³ In addition, the near-doubling in both the 1 to 5 mm and right-sided SSLDR associated with the combined use of GI Genius and EndoCuff was also notable, as it has been estimated that up to 20% of colon cancers develop from SSL,²⁴ and the overall miss rate of SSL on colonoscopy is as high as 27%.² Our findings are consistent with those of a recent randomized controlled trial, which found that using the GI Genius-EndoCuff combination increased the ADR compared to using GI Genius alone, although the authors did not find any differences in the SSLDR.²⁵ The baseline group of the trial used only AI, whereas our baseline group used only EndoCuff, which may explain the different results. These findings were echoed in a recent randomized controlled trial that compared CADe alone with a CADe-EndoCuff combination. The authors employed Cad Eye (Fujifilm) rather than GI Genius and found that the combination of CADe and EndoCuff significantly increased the overall ADR compared to a standard colonoscopy protocol without the use of either device (41.9% vs. 58.8%, p<0.05). Interestingly, the CADe-EndoCuff combination in this study also resulted in a significant increase in the advanced ADR compared with standard colonoscopy (7.7% vs. 13.6%, p<0.05). Neither CADe nor EndoCuff alone led to significant increases in these metrics.²⁶

These results indicate that a mucosal exposure device, such as EndoCuff, which exposes the colonic folds to increase the endoscopic field of view, has a synergistic effect when used with AI because AI can only draw attention to polyps within a given field. When used alone, both AI and EndoCuff have been shown to increase the ADR.⁴^,⁷^,⁸^,¹⁷^,²⁷^-²⁹ The former appears to provide the greatest benefit for discovering smaller polyps that are more commonly missed. Meanwhile, the latter exposes polyps located behind the colonic folds that are easily missed.²^,¹⁷ Thus, our findings support the idea that these complementary methods enhance beneficial results when used together. In summary, to optimize the benefits of GI Genius, routine employment of EndoCuff should be considered, unless contraindicated.

We found that combining GI Genius and EndoCuff results in a statistically significant increase in the detection of 1 to 5 mm adenomas and SSL; however, whether this increase actually decreases the risk of developing post-colonoscopy colorectal cancer (CRC) is unclear. Two studies found that the overall ADR is inversely proportional to the risk of developing post-colonoscopy CRC³⁰^,³¹; however, in both instances, the investigators did not stratify ADR according to polyp size. Exploring the relationship between the 1 to 5 mm ADR and post-colonoscopy CRC risk represents a promising future direction, especially as the increasing use of AI will likely increase the detection of smaller polyps.

Our study included a post-AI group consisting of non-GI Genius-assisted cases performed in the months after the GI Genius trial was discontinued. This served to examine whether the temporary implementation of GI Genius affected the endoscopists’ subsequent ability to detect polyps. Our findings indicated no difference in post-AI performance in any of the metrics compared with those of pre-AI baseline cases, suggesting that the use of GI Genius did not have any lasting effect on the endoscopists’ subsequent polyp detection ability. However, the duration of GI Genius usage (only two months) should be considered, as it may have been insufficient to impact endoscopists’ long-term polyp detection ability. Finally, some concerns regarding CADe involve the distracting nature of the bounding box; however, the use of GI Genius did not result in a significant increase in the procedure time. Moreover, highlighting more polyps did not cause the endoscopists to remove more hyperplastic polyps.

This study had several limitations. All endoscopists in this study had a high historical ADR of ≥35%; indeed, the vast majority (14/16) had an ADR of ≥45%. With such a high baseline ADR, the additional benefits of an AI or a combined AI-EndoCuff detection system may not be as pronounced. Wang et al.¹⁸ demonstrated that an AI system provides considerable benefits in increasing the overall ADR for endoscopists with a low baseline ADR (approximately 20%). Thus, the results in our cohort of endoscopists may be a relative underestimation of the effect of the GI Genius-EndoCuff combination. Nevertheless, the distinct increases in small- and right-sided ADR and SSLDR that we observed suggest that there is a benefit, even for endoscopists with a high baseline ADR. This effect may be more apparent among providers with a lower baseline ADR. Examining the impact of GI Genius or GI Genius-EndoCuff on providers with a wider range of ADRs, especially those in the lower quartiles of the detection rate, represents another valuable avenue for further study. In addition, our study was performed at a single tertiary center. Future multicenter collaborations that include endoscopists with a lower baseline ADR will enhance the generalizability of the results.

Furthermore, GI Genius was only available for two months, which limited the number of cases and providers (n=16) included and may have impacted the generalizability of our results. In particular, short-term usage of CADe may not show variance in colonoscopy metrics that could become apparent with prolonged usage. Providers may become more proficient in utilizing CADe alone and/or in combination with EndoCuff over time, and differences in colonoscopy metrics between CADe cases and controls may become even more apparent. Therefore, further studies that examine longer periods of CADe use are warranted. In addition, as this was an observational study of the real-life efficacy of CADe, GI Genius and EndoCuff were used at the discretion of the endoscopists. Endoscopists who are more willing to try new technologies, such as CADe, may have traits, behaviors, or tool usage that favor polyp detection compared to those who are less willing, thus introducing the potential for selection bias. Investigator bias may be possible if endoscopists are aware of the potential benefits of the CADe and EndoCuff combination. However, these potential sources of bias were mitigated by the endoscopists serving as their own controls. Providers with a higher baseline ADR may be more willing to use mucosal exposure devices, such as the EndoCuff; however, given the low variability in the ADR among our provider cohort, we believe this played less of a confounding role. A future randomized controlled trial investigating the combined use of CADe and EndoCuff will help reduce this potential bias.

Finally, the use of historical rather than prospectively enrolled controls was another limitation. We ensured that each provider had at least 20 control cases and that those with a larger number of AI cases were matched with a higher number of control cases. However, without randomization, the patient population may have differed between groups. We attempted to mitigate the effects of bias by limiting our inclusion criteria to screening or surveillance colonoscopies. Ultimately, the two groups were demographically similar, and all control and AI cases were conducted in the same endoscopy suite within a 6-month span. This scenario reflects how AI will likely be used in common practice; thus, our results provide valuable insights into AI and AI-EndoCuff efficacy.

Combining the use of GI Genius and a mucosal exposure device such as EndoCuff shows promise for identifying polyps that have historically been difficult for endoscopists to detect, namely small and right-sided lesions. This optimization of polyp detection can be observed even for endoscopists with a very high baseline ADR. Future studies that include more centers in a variety of settings, especially those with proceduralists with a lower baseline ADR, and combining CADe with other colonoscopic aids, may further add to our understanding of the impact of AI on colonoscopy performance.

Conflicts of Interest

Felice Schnoll-Sussman, MD, is a consultant for Medtronic, Sebela, Ethicon, Phathom, and Braintree. Reem Z. Sharaiha, MD, is a consultant for Boston, Cook, Olympus, and Surgical Intuitive. The other authors have no potential conflicts of interest.

Funding

None.

Author Contributions

Conceptualization: ED, FSS, RZS, DW; Data curation: JHK, JW, CP, PM; Formal analysis: JHK, JW; Investigation: all authors; Methodology: all authors; Supervision: ED, FS, RS, DW; Validation: all authors; Writing–original draft: JHK, JW; Writing–review & editing: all authors.

Fig. 1.

Artificial intelligence (AI) cases increased 1 to 5 mm adenoma detection rate (ADR) compared to pre-AI cases. *Statistical significance.

Fig. 2.

Artificial intelligence (AI) cases increased 1 to 5 mm sessile serrated lesion detection rate (SSLDR) compared to pre-AI cases. *Statistical significance.

Fig. 3.

Artificial intelligence (AI) cases increased right-sided sessile serrated lesion detection rate (SSLDR) compared to pre-AI cases. *Statistical significance.

Table 1.

Characteristics of the pre-AI, AI, and post-AI groups

Characteristic	Pre-AI	AI	Post-AI
Total subjects	386 (100.0)	178 (100.0)	234 (100.0)
Age (yr)	57.4±9.6	57.9±11.2	59.8±10.2
Female	231 (59.8)	94 (52.8)	156 (66.7)
Screening	273 (70.7)	114 (64.0)	142 (60.7)
Surveillance	113 (29.3)	64 (36.0)	92 (39.3)
Withdrawal time (min)	17.1±8.5	17.8±8.7	17.7±8.5
Aronchick average total	2.2±0.69	2.3±0.84	2.3±0.71
BBPS average total	8.3±1.25	7.9±1.79	8.1±1.27
Fellow present	124 (32.1)	71 (39.9)	94 (40.2)
EndoCuff used	146 (37.8)	117 (65.7)	125 (53.4)

Values are presented as number (%) or mean±standard deviation.

AI, artificial intelligence; BBPS: Boston bowel preparation scale.

Table 2.

Comparison of adenoma per colonoscopy, sessile serrated lesion per colonoscopy, adenoma detection rate, and sessile serrated lesion detection rate between the pre-AI, AI, and post-AI groups

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
Adenoma per colonoscopy	0.86	1.08	0.92	0.12	0.66
Sessile serrated lesion per colonoscopy	0.21	0.28	0.30	0.20	0.11
Adenoma detection rate (%)	40.3	52.5	42.7	0.05	0.70
Sessile serrated lesion detection rate (%)	15.8	21.2	19.2	0.18	0.40

AI, artificial intelligence.

Table 3.

Comparison of advanced adenoma detection rate, true histology rate, and polyp detection rate between the pre-AI, AI, and post-AI groups

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
Advanced adenoma detection rate (%)	5.2	4.5	6.0	0.76	0.73
True histology rate (%)	52.8	61.8	59.4	0.90	0.93
Total polyp detection rate (%)	73.8	84.3	74.8	0.23	0.91

AI, artificial intelligence.

Table 4.

Comparison of ADR by adenoma size between the pre-AI, AI, and post-AI groups

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
1–5 mm ADR (%)	33.2	45.5	34.6	0.03	0.80
6–9 mm ADR (%)	9.1	9.0	5.6	0.98	0.24
≥10 mm ADR (%)	4.7	3.9	4.7	0.74	0.99

ADR, adenoma detection rate; AI, artificial intelligence.

Table 5.

Comparison of SSLDR by sessile serrated lesion size between the pre-AI, AI, and post-AI groups

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
1–5 mm SSLDR (%)	8.0	15.2	12.4	0.01	0.12
6–9 mm SSLDR (%)	6.0	5.1	6.8	0.71	0.72
≥10 mm SSLDR (%)	3.4	4.5	2.6	0.54	0.66

SSLDR, sessile serrated lesion detection rate; AI, artificial intelligence.

Table 6.

Comparison of ADR by location between the pre-AI, AI, and post-AI groups

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
Right-sided ADR (%)	30.8	41.6	38.5	0.05	0.17
Left-sided ADR (%)	18.4	22.5	15.0	0.34	0.42

ADR, adenoma detection rate; AI, artificial intelligence.

Table 7.

Comparison of SSLDR by location between the pre-AI, AI, and post-AI groups

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
Right-sided SSLDR (%)	12.7	22.5	16.7	0.01	0.27
Left-sided SSLDR (%)	2.6	1.7	3.4	0.57	0.61

SSLDR, sessile serrated lesion detection rate; AI, artificial intelligence.

Table 8.

Stratification of ADR by size category, 1 to 5 mm SSLDR, right-sided SSLDR, right-sided ADR, overall APC, overall SSLPC, overall ADR, and overall SSLDR by EndoCuff use

Metric	Pre-AI	AI	p-value (pre-AI vs. AI)
ADR without EndoCuff (%)	46.2	49.2	0.67
ADR with EndoCuff (%)	48.0	63.2	0.03
SSLDR without EndoCuff (%)	16.2	19.7	0.40
SSLDR with EndoCuff (%)	19.2	27.4	0.06
1–5 mm ADR without EndoCuff (%)	31.2	32.8	0.78
1–5 mm ADR with EndoCuff (%)	36.3	52.1	0.01
6–10 mm ADR without EndoCuff (%)	11.2	13.1	0.58
6–10 mm ADR with EndoCuff (%)	5.5	6.8	0.56
≥10 mm ADR without EndoCuff (%)	3.8	3.3	0.81
≥10 mm ADR with EndoCuff (%)	6.2	4.3	0.45
1–5 mm SSLDR without EndoCuff (%)	7.1	11.5	0.10
1–5 mm SSLDR with EndoCuff (%)	9.6	17.1	0.02
Right-sided SSLDR without EndoCuff (%)	12.9	18.0	0.16
Right-sided SSLDR with EndoCuff (%)	12.3	24.8	<0.001
Right-sided ADR without EndoCuff (%)	30.8	39.3	0.13
Right-sided ADR with EndoCuff (%)	30.8	42.7	0.03
APC without EndoCuff	0.87	0.91	0.87
APC with EndoCuff	0.84	1.17	0.07
SSLPC without EndoCuff	0.20	0.24	0.61
SSLPC with EndoCuff	0.21	0.29	0.27

ADR, adenoma detection rate; SSLDR, sessile serrated lesion detection rate; APC, adenomas per colonoscopy; SSLPC, sessile serrated lesions per colonoscopy; AI, artificial intelligence.

REFERENCES

1. Milluzzo SM, Cesaro P, Grazioli LM, et al. Artificial intelligence in lower gastrointestinal endoscopy: the current status and future perspective. Clin Endosc 2021;54:329–339.Article PubMed PMC PDF
2. Zhao S, Wang S, Pan P, et al. Magnitude, risk factors, and factors associated with adenoma miss rate of tandem colonoscopy: a systematic review and meta-analysis. Gastroenterology 2019;156:1661–1674.Article PubMed
3. Robertson DJ, Lieberman DA, Winawer SJ, et al. Colorectal cancers soon after colonoscopy: a pooled multicohort analysis. Gut 2014;63:949–956.Article PubMed
4. Hassan C, Spadaccini M, Iannone A, et al. Performance of artificial intelligence in colonoscopy for adenoma and polyp detection: a systematic review and meta-analysis. Gastrointest Endosc 2021;93:77–85.Article PubMed
5. Spadaccini M, Marco A, Franchellucci G, et al. Discovering the first US FDA-approved computer-aided polyp detection system. Future Oncol 2022;18:1405–1412.Article PubMed
6. Hassan C, Wallace MB, Sharma P, et al. New artificial intelligence system: first validation study versus experienced endoscopists for colorectal polyp detection. Gut 2020;69:799–800.Article PubMed
7. Repici A, Badalamenti M, Maselli R, et al. Efficacy of real-time computer-aided detection of colorectal neoplasia in a randomized trial. Gastroenterology 2020;159:512–520.Article PubMed
8. Koh FH, Ladlad J; SKH Endoscopy Centre, et al. Real-time artificial intelligence (AI)-aided endoscopy improves adenoma detection rates even in experienced endoscopists: a cohort study in Singapore. Surg Endosc 2023;37:165–171.Article PubMed
9. Shaukat A, Colucci D, Erisson L, et al. Improvement in adenoma detection using a novel artificial intelligence-aided polyp detection device. Endosc Int Open 2021;9:E263–E270.Article PubMed PMC
10. Xu H, Tang RS, Lam TY, et al. Artificial intelligence-assisted colonoscopy for colorectal cancer screening: a multicenter randomized controlled trial. Clin Gastroenterol Hepatol 2023;21:337–346.Article PubMed
11. Levy I, Bruckmayer L, Klang E, et al. Artificial intelligence-aided colonoscopy does not increase adenoma detection rate in routine clinical practice. Am J Gastroenterol 2022;117:1871–1873.Article PubMed
12. Nehme F, Coronel E, Barringer DA, et al. Performance and attitudes toward real-time computer-aided polyp detection during colonoscopy in a large tertiary referral center in the United States. Gastrointest Endosc 2023;98:100–109.Article PubMed
13. Quan SY, Wei MT, Lee J, et al. Clinical evaluation of a real-time artificial intelligence-based polyp detection system: a US multi-center pilot study. Sci Rep 2022;12:6598.Article PubMed PMC PDF
14. Calderwood AH, Jacobson BC. Colonoscopy quality: metrics and implementation. Gastroenterol Clin North Am 2013;42:599–618.Article PubMed PMC
15. Park JB, Bae JH. Effectiveness of a novel artificial intelligence-assisted colonoscopy system for adenoma detection: a prospective, propensity score-matched, non-randomized controlled study in Korea. Clin Endosc 2025;58:112–120.Article
16. Tiankanon K, Karuehardsuwan J, Aniwan S, et al. Performance comparison between two computer-aided detection colonoscopy models by trainees using different false positive thresholds: a cross-sectional study in Thailand. Clin Endosc 2024;57:217–225.Article PubMed PMC PDF
17. Barua I, Vinsard DG, Jodal HC, et al. Artificial intelligence for polyp detection during colonoscopy: a systematic review and meta-analysis. Endoscopy 2021;53:277–284.Article PubMed
18. Wang P, Berzin TM, Glissen Brown JR, et al. Real-time automatic detection system increases colonoscopic polyp and adenoma detection rates: a prospective randomised controlled study. Gut 2019;68:1813–1819.Article PubMed PMC
19. Repici A, Spadaccini M, Antonelli G, et al. Artificial intelligence and colonoscopy experience: lessons from two randomised trials. Gut 2022;71:757–765.Article PubMed
20. Rønborg SN, Ujjal S, Kroijer R, et al. Assessing the potential of artificial intelligence to enhance colonoscopy adenoma detection in clinical practice: a prospective observational trial. Clin Endosc 2024;57:783–789.Article
21. Baxter NN, Goldwasser MA, Paszat LF, et al. Association of colonoscopy and death from colorectal cancer. Ann Intern Med 2009;150:1–8.Article PubMed
22. Singh H, Nugent Z, Demers AA, et al. The reduction in colorectal cancer mortality after colonoscopy varies by site of the cancer. Gastroenterology 2010;139:1128–1137.Article PubMed
23. Baran B, Mert Ozupek N, Yerli Tetik N, et al. Difference between left-sided and right-sided colorectal cancer: a focused review of literature. Gastroenterology Res 2018;11:264–273.Article PubMed PMC
24. Rashtak S, Rego R, Sweetser SR, et al. Sessile serrated polyps and colon cancer prevention. Cancer Prev Res (Phila) 2017;10:270–278.Article PubMed PDF
25. Spadaccini M, Hassan C, Rondonotti E, et al. Combination of mucosa-exposure device and computer-aided detection for adenoma detection during colonoscopy: a randomized trial. Gastroenterology 2023;165:244–251.Article PubMed
26. Aniwan S, Mekritthikrai K, Kerr SJ, et al. Computer-aided detection, mucosal exposure device, their combination, and standard colonoscopy for adenoma detection: a randomized controlled trial. Gastrointest Endosc 2023;97:507–516.Article PubMed
27. Nazarian S, Glover B, Ashrafian H, et al. Diagnostic accuracy of artificial intelligence and computer-aided diagnosis for the detection and characterization of colorectal polyps: systematic review and meta-analysis. J Med Internet Res 2021;23:e27370.Article PubMed PMC
28. Walls M, Houwen BB, Rice S, et al. The effect of the endoscopic device Endocuff™/Endocuff vision™ on quality standards in colonoscopy: a systematic review and meta-analysis of randomized trials. Colorectal Dis 2023;25:573–585.Article PubMed PDF
29. Facciorusso A, Del Prete V, Buccino RV, et al. Comparative efficacy of colonoscope distal attachment devices in increasing rates of adenoma detection: a network meta-analysis. Clin Gastroenterol Hepatol 2018;16:1209–1219.Article PubMed
30. Corley DA, Jensen CD, Marks AR, et al. Adenoma detection rate and risk of colorectal cancer and death. N Engl J Med 2014;370:1298–1306.Article PubMed PMC
31. Schottinger JE, Jensen CD, Ghai NR, et al. Association of physician adenoma detection rates with postcolonoscopy colorectal cancer. JAMA 2022;327:2114–2122.Article PubMed PMC

Figure & Data

REFERENCES

Citations

Citations to this article as recorded by

PubReader
ePub Link

Cite

CITE: export Copy Download Format; Close

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

RIS — For EndNote, ProCite, RefWorks, and most other reference management software
BibTeX — For JabRef, BibDesk, and other BibTeX-specific software

Include:

Citation for the content below
Citation and abstract for the content below

GI Genius increases small and right-sided adenoma and sessile serrated lesion detection rate when used with EndoCuff in a real-world setting: a retrospective United States study

DOI: https://doi.org/10.5946/ce.2024.271

XML Download

Figure

GI Genius increases small and right-sided adenoma and sessile serrated lesion detection rate when used with EndoCuff in a real-world setting: a retrospective United States study

Fig. 1. Artificial intelligence (AI) cases increased 1 to 5 mm adenoma detection rate (ADR) compared to pre-AI cases. *Statistical significance.

Fig. 2. Artificial intelligence (AI) cases increased 1 to 5 mm sessile serrated lesion detection rate (SSLDR) compared to pre-AI cases. *Statistical significance.

Fig. 3. Artificial intelligence (AI) cases increased right-sided sessile serrated lesion detection rate (SSLDR) compared to pre-AI cases. *Statistical significance.

Graphical abstract

Fig. 1.

Fig. 2.

Fig. 3.

Graphical abstract

GI Genius increases small and right-sided adenoma and sessile serrated lesion detection rate when used with EndoCuff in a real-world setting: a retrospective United States study

Characteristic	Pre-AI	AI	Post-AI
Total subjects	386 (100.0)	178 (100.0)	234 (100.0)
Age (yr)	57.4±9.6	57.9±11.2	59.8±10.2
Female	231 (59.8)	94 (52.8)	156 (66.7)
Screening	273 (70.7)	114 (64.0)	142 (60.7)
Surveillance	113 (29.3)	64 (36.0)	92 (39.3)
Withdrawal time (min)	17.1±8.5	17.8±8.7	17.7±8.5
Aronchick average total	2.2±0.69	2.3±0.84	2.3±0.71
BBPS average total	8.3±1.25	7.9±1.79	8.1±1.27
Fellow present	124 (32.1)	71 (39.9)	94 (40.2)
EndoCuff used	146 (37.8)	117 (65.7)	125 (53.4)

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
Adenoma per colonoscopy	0.86	1.08	0.92	0.12	0.66
Sessile serrated lesion per colonoscopy	0.21	0.28	0.30	0.20	0.11
Adenoma detection rate (%)	40.3	52.5	42.7	0.05	0.70
Sessile serrated lesion detection rate (%)	15.8	21.2	19.2	0.18	0.40

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
Advanced adenoma detection rate (%)	5.2	4.5	6.0	0.76	0.73
True histology rate (%)	52.8	61.8	59.4	0.90	0.93
Total polyp detection rate (%)	73.8	84.3	74.8	0.23	0.91

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
1–5 mm ADR (%)	33.2	45.5	34.6	0.03	0.80
6–9 mm ADR (%)	9.1	9.0	5.6	0.98	0.24
≥10 mm ADR (%)	4.7	3.9	4.7	0.74	0.99

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
1–5 mm SSLDR (%)	8.0	15.2	12.4	0.01	0.12
6–9 mm SSLDR (%)	6.0	5.1	6.8	0.71	0.72
≥10 mm SSLDR (%)	3.4	4.5	2.6	0.54	0.66

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
Right-sided ADR (%)	30.8	41.6	38.5	0.05	0.17
Left-sided ADR (%)	18.4	22.5	15.0	0.34	0.42

Metric	Pre-AI	AI	Post-AI	p-value (pre-AI vs. AI)	p-value (pre-AI vs. post-AI)
Right-sided SSLDR (%)	12.7	22.5	16.7	0.01	0.27
Left-sided SSLDR (%)	2.6	1.7	3.4	0.57	0.61

Metric	Pre-AI	AI	p-value (pre-AI vs. AI)
ADR without EndoCuff (%)	46.2	49.2	0.67
ADR with EndoCuff (%)	48.0	63.2	0.03
SSLDR without EndoCuff (%)	16.2	19.7	0.40
SSLDR with EndoCuff (%)	19.2	27.4	0.06
1–5 mm ADR without EndoCuff (%)	31.2	32.8	0.78
1–5 mm ADR with EndoCuff (%)	36.3	52.1	0.01
6–10 mm ADR without EndoCuff (%)	11.2	13.1	0.58
6–10 mm ADR with EndoCuff (%)	5.5	6.8	0.56
≥10 mm ADR without EndoCuff (%)	3.8	3.3	0.81
≥10 mm ADR with EndoCuff (%)	6.2	4.3	0.45
1–5 mm SSLDR without EndoCuff (%)	7.1	11.5	0.10
1–5 mm SSLDR with EndoCuff (%)	9.6	17.1	0.02
Right-sided SSLDR without EndoCuff (%)	12.9	18.0	0.16
Right-sided SSLDR with EndoCuff (%)	12.3	24.8	<0.001
Right-sided ADR without EndoCuff (%)	30.8	39.3	0.13
Right-sided ADR with EndoCuff (%)	30.8	42.7	0.03
APC without EndoCuff	0.87	0.91	0.87
APC with EndoCuff	0.84	1.17	0.07
SSLPC without EndoCuff	0.20	0.24	0.61
SSLPC with EndoCuff	0.21	0.29	0.27

Table 1. Characteristics of the pre-AI, AI, and post-AI groups

Values are presented as number (%) or mean±standard deviation.

AI, artificial intelligence; BBPS: Boston bowel preparation scale.

Table 2. Comparison of adenoma per colonoscopy, sessile serrated lesion per colonoscopy, adenoma detection rate, and sessile serrated lesion detection rate between the pre-AI, AI, and post-AI groups

AI, artificial intelligence.

Table 3. Comparison of advanced adenoma detection rate, true histology rate, and polyp detection rate between the pre-AI, AI, and post-AI groups

AI, artificial intelligence.

Table 4. Comparison of ADR by adenoma size between the pre-AI, AI, and post-AI groups

ADR, adenoma detection rate; AI, artificial intelligence.

Table 5. Comparison of SSLDR by sessile serrated lesion size between the pre-AI, AI, and post-AI groups

SSLDR, sessile serrated lesion detection rate; AI, artificial intelligence.

Table 6. Comparison of ADR by location between the pre-AI, AI, and post-AI groups

ADR, adenoma detection rate; AI, artificial intelligence.

Table 7. Comparison of SSLDR by location between the pre-AI, AI, and post-AI groups

SSLDR, sessile serrated lesion detection rate; AI, artificial intelligence.

Table 8. Stratification of ADR by size category, 1 to 5 mm SSLDR, right-sided SSLDR, right-sided ADR, overall APC, overall SSLPC, overall ADR, and overall SSLDR by EndoCuff use

ADR, adenoma detection rate; SSLDR, sessile serrated lesion detection rate; APC, adenomas per colonoscopy; SSLPC, sessile serrated lesions per colonoscopy; AI, artificial intelligence.