Abstract
Background/Aims:
The integration of artificial intelligence (AI)-powered image analysis and mucosal exposure devices, such as distal attachment caps, has been demonstrated to significantly improve the adenoma detection rate (ADR) during colonoscopy. This study aimed to compare AI-assisted colonoscopy (AIC) with cap-assisted colonoscopy (CAC).
Methods:
This retrospective propensity score-matched cohort study was performed at a tertiary care hospital between January 2022 and May 2022. Data were extracted from the electronic health record system and colonoscopy video recordings. Adult patients aged 40 years who underwent screening or surveillance colonoscopies were included. The primary outcome was the ADR, whereas the secondary outcome was the polyp detection rate (PDR).
Results:
A 1:1 propensity score-matched analysis was performed, resulting in 49 well-matched patient pairs. One patient from each pair was assigned to the CAC group, whereas the other was assigned to the AIC group. No significant difference in ADR was observed between the CAC and AIC groups (47% vs. 51%, p=0.69). Similarly, PDR did not significantly differ between the two groups (80% vs. 71%, p=0.35).
Conclusions:
Both CAC and AIC have the potential to increase ADR and PDR. However, neither modality offers a significant advantage.
Keywords: Adenoma, Artificial intelligence, Colonoscopy, Polyps
Graphical abstract
INTRODUCTION
In 2020, more than 1.9 million new cases of colorectal cancer (CRC) were reported. In the Philippines, CRC is the 4th leading cause of cancer-related deaths.1 It is widely known that the majority of CRCs originate from adenomas; hence, early detection and complete resection of these premalignant lesions can decrease the incidence and mortality associated with CRC.2,3 One of the most important indicators of high-quality colonoscopy is the adenoma detection rate (ADR), defined as the proportion of screening colonoscopies that detect at least one colorectal adenoma or adenocarcinoma confirmed through histopathology.4,5 The importance of ADR cannot be overemphasized because a 1% increase in ADR is associated with an estimated 3% decrease in interval CRC.6 Colorectal adenomas can be easily missed once they are located behind the colonic flexures and folds. Adequate mucosal exposure behind the folds, bowel distention, and withdrawal time are important for detecting these adenomas.7 New adjunctive devices and techniques for improving the ADR have been reported in the literature, including image-enhanced endoscopy, chromoendoscopy, and distal attachment devices, with the objective of improving visualization of hard to detect polyps.8 A distal attachment cap is a simple, low-cost, and easy-to-use device to improve the quality of colonoscopy. Depression of the colonic folds created by the distal cap can allow for better endoscopic views of colonic adenomas, and little evidence has shown improved ADR and cecal intubation rate compared with standard colonoscopy.9-11 Artificial intelligence (AI) is increasingly being incorporated into gastrointestinal endoscopy, particularly in the detection and characterization of colorectal polyps.12-14 Recent studies have compared the outcomes of AI and mucosal exposure devices, specifically Endocuff Vision, in improving ADR.15-17 Nonetheless, few studies have directly compared the use of cap-assisted colonoscopy (CAC) and AI-assisted colonoscopy (AIC). This study aimed to evaluate the comparative efficacy of the distal attachment cap and AI system in improving the ADR during colonoscopic procedures.
METHODS
Study design and population
This retrospective, propensity score-matched study was conducted at the Chinese General Hospital and Medical Center, Philippines, between January and May 2022. Data were extracted from the electronic health record system and endoscopy unit database of the institution. All colonoscopy video recordings were independently reviewed by two separate teams. Each team comprised a gastroenterology fellow and two consultant gastroenterologists who were tasked with evaluating the accuracy of the polyp count, Paris classification, and Japan NBI Expert Team (JNET) classification. The primary cause of inter-observer variability was polyp type classification. To resolve inter-observer variability among endoscopists, a majority vote of the video reviewers was performed. This study included adult patients aged ≥40 years who underwent elective colonoscopy for the following indications: colon cancer screening, polyp surveillance, positive fecal occult blood test or fecal immunohistochemical test, or evaluation of abdominal symptoms. The inclusion criteria were expanded to patients aged ≥40 years, instead of ≥50 years, because of the rising incidence of advanced adenomas in the 40–49-year age range.18 Patients were excluded from the study if they had a history of CRC, colon resection, inflammatory bowel disease, hereditary polyposis syndrome and pregnant patients.
Colonoscopy
Colonoscopies were performed by 13 consultant gastroenterologists and two gastroenterology fellows-in-training. All consultant gastroenterologists were board-certified specialists with a minimum of ten years of experience in endoscopic procedures and a documented history of performing more than 1,000 colonoscopies. Gastroenterology fellows-in-training had less than two years of experience in the field and underwent direct supervision by consultant gastroenterologists during all procedures. Both trainees and consultants participated in colonoscopies in the AIC and CAC study groups. All examinations were performed using a high-definition colonoscope (EC-760P-V/L; Fujifilm) or CF-HQ290I (Olympus Medical Systems). In the CAC group, a transparent cap was attached to the colonoscope tip. The CAD EYE AI system (Fujifilm) was used in the AIC group. CAD EYE (Fujifilm) is an endoscopic AI that has been available in the Philippines since 2021. This slim box-type device is similar to a processor or light source device. In addition to polyp recognition, this modality provides a supplementary examination mode that enables real-time optical characterization of polyps during colonoscopy.19 Following successful cecal intubation within the AIC group, endoscopists initiated the computer-aided detection (CADe) algorithm, specifically the CAD EYE system, via a pre-programmed control interface integrated into the colonoscope's button. Subsequently, withdrawal was executed, during which the real-time image analysis and polyp detection capabilities of the CAD EYE system were actively utilized to augment the visual inspection.
For each procedure, the endoscopist was responsible for choosing the colonoscope and using the distal cap or AI system. Bowel preparation was performed using either a standard dose of polyethylene glycol 3350 or sodium picosulfate solution. The Boston bowel preparation scale (BBPS) was used to grade the cleanliness of bowel preparation. Endoscopic procedures were performed under intravenous sedation using midazolam, fentanyl, or propofol, depending on the endoscopist. Cecal intubation was photographically documented. Intubation and withdrawal times were measured using a timer, excluding the periods of therapeutic intervention and mucosal cleansing.
Adenoma detection
All detected polyps were evaluated for their location, size, and morphology. Polyp size was endoscopically measured using biopsy forceps as a guide. The morphology of the polyps was classified according to the Paris classification, while the JNET classification was used for histological type classification.20,21 All endoscopically resected polyps were sent to the laboratory for histopathological confirmation and classified as hyperplastic polyps, adenoma, or cancer.
Outcomes
The primary outcome measured was the ADR between the CAC and AIC groups. In this study, the ADR was defined as the proportion of patients with at least one adenoma confirmed by histopathology. The secondary outcome was the polyp detection rate (PDR), which was defined as the proportion of patients with at least one polyp identified during colonoscopy.
Statistical analysis
The use of CAC in our institution was lower than that of AIC owing to endoscopist preferences. Propensity score matching was used to balance the assignment of the included patients, adjust for the procedure type (CAC or AIC), and limit confounding variables that could affect the outcome of the study. Propensity score matching was performed using known factors such as age, sex, indication for colonoscopy, and BBPS score, which may affect the ADR as matching variables. Calipers were set to 0.2 standard deviation (SD) of the logit of the propensity score. Continuous variables are presented as mean±SD or median (interquartile range), depending on data distribution. Categorical variables are expressed as frequencies and percentages. Continuous data were compared using the independent t-test or Mann-Whitney U-test. Categorical data were compared using the chi-squared test or Fisher exact test. Statistical significance was set at p<0.05.
Ethical statements
This study was approved by the Research Ethics Review Board of the Chinese General Hospital and Medical Center (CGHMC RERB 2022-C-80). The requirement for informed consent was waived by the Institutional Ethics Review Panel.
RESULTS
A total of 971 patients underwent colonoscopy between January and May 2022. A substantial proportion of these procedures (n=672) were conducted without the use of a distal attachment cap or AI technology. This limitation was primarily due to the scarcity of AI-equipped rooms (only 1 of 4 rooms) and the preference of several senior endoscopists for performing colonoscopies without using distal attachment devices. After applying the exclusion criteria, 201 patients were included in the study. Propensity score matching was used to balance the distribution of significant confounding variables. Using a 1:1 matching process, 49 well-matched patient pairs were identified and included in the analysis. One patient from each pair was assigned to the CAC group, and the other was assigned to the AIC group (Fig. 1). Patient characteristics in each group are summarized in Table 1. After propensity score matching, no statistically significant differences were observed between groups in terms of age, sex, indications for colonoscopy, or bowel preparation quality. All colonoscopies involved cecal intubation. No significant difference in the median colonoscopy withdrawal time was observed (11 vs. 11 minutes, p=0.83). The ADR was 47% (95% confidence interval [CI], 33%–62%) and 51% (95% CI, 36%–66%) in the CAC and AIC groups, respectively; no statistically significant difference was found between the two groups (p=0.69). The PDR was 80% (95% CI, 66%–90%) and 71% (95% CI, 57%–83%) in the CAC and AIC groups, respectively; no statistically significant difference was noted between the two groups (p=0.35) (Table 2).
Fig. 1.
Flowchart showing patient selection. AI, artificial intelligence; AIC, artificial intelligence-assisted colonoscopy; CAC, cap-assisted colonoscopy.
Table 1.
Demographic characteristics of patients in the full cohort and propensity score-matched cohort
| Characteristic | Full (unmatched) cohort (n=201) |
Matched cohort (n=98) |
||||||
|---|---|---|---|---|---|---|---|---|
| All patients (n=201) | With the use of distal cap (n=54) | With the use of AI (n=147) | p-value | All patients (n=98) | With the use of distal cap (n=49) | With the use of AI (n=49) | p-value | |
| Age (yr) | 59 (53–68) | 55 (50–67) | 61 (55–68) | 0.0150a),* | 58 (51–67) | 55 (52–68) | 58 (50–66) | 0.7839a) |
| Sex | ||||||||
| Male | 102 (50.7) | 28 (51.9) | 74 (50.3) | 0.849b) | 48 (49.0) | 23 (46.9) | 25 (51.0) | 0.686b) |
| Female | 99 (49.3) | 26 (48.1) | 73 (49.7) | 50 (51.0) | 26 (53.1) | 24 (49.0) | ||
| BMI (kg/m2) | 22.4 (20.8–24.4) | 22.7 (20.8–24.7) | 22.3 (20.5–24.4) | 0.7128a) | 23 (20.8–25) | 22.7 (20.8–24.4) | 23.4 (20.5–25.3) | 0.4670a) |
| Underweight | 15 (7.5) | 3 (5.6) | 12 (8.2) | 0.667b) | 7 (7.1) | 3 (6.1) | 4 (8.2) | 0.591b) |
| Normal | 143 (71.1) | 38 (70.4) | 105 (71.4) | 66 (67.3) | 35 (71.4) | 31 (63.3) | ||
| Overweight | 40 (19.9) | 13 (24.1) | 27 (18.4) | 23 (23.5) | 11 (22.4) | 12 (24.5) | ||
| Obese | 3 (1.5) | 0 (0) | 3 (2.0) | 2 (2.0) | 0 | 2 (4.1) | ||
| Indication | ||||||||
| Screening | 195 (97.0) | 51 (94.4) | 144 (98.0) | 0.346c) | 93 (94.9) | 47 (95.9) | 46 (93.9) | 1.000c |
| Surveillance | 6 (3.0) | 3 (5.6) | 3 (2.0) | 5 (5.1) | 2 (4.1) | 3 (6.1) | ||
| Boston bowel preparation scale score | ||||||||
| 2 | 63 (31.3) | 6 (11.1) | 57 (38.8) | <0.000c),* | 11 (11.2) | 6 (12.2) | 5 (10.2) | 0.749c) |
| 3 | 138 (68.7) | 48 (88.9) | 90 (61.2) | 87 (88.8) | 43 (87.8) | 44 (89.8) | ||
| Sedation | ||||||||
| Propofol | 166 (82.6) | 54 (100.0) | 112 (76.2) | <0.0001c),* | 91 (92.9) | 49 (100.0) | 42 (85.7) | 0.012c),* |
| Mid/Fen | 35 (17.4) | 0 (0) | 35 (23.8) | 7 (7.1) | 0 (0) | 7 (14.3) | ||
Values are presented as median (interquartile range) or number (%).
AI, artificial intelligence; BMI, body mass index; Mid, midazolam; Fen, fentanyl.
Mann-Whitney U-test,
chi-squared test,
Fisher exact test.
*p-value is statistically significant.
Table 2.
Summary of the per-patient analysis
| Characteristic | All patients (n=98) | With the use of distal cap (n=49) | With the use of AI (n=49) | p-value |
|---|---|---|---|---|
| Cecal intubation rate | 98 (100) | 49 (100) | 49 (100) | - |
| Colonoscopy withdrawal time (min) | 11 (8–14) | 11 (8–13) | 11 (9–14) | 0.8299a) |
| Polyp detection rate | 74 (75.5) | 39 (79.6) | 35 (71.4) | 0.347b) |
| Polyps detected per patient | 2 (1–3) | 2 (1–4) | 2 (0–3) | 0.3121a) |
| Adenoma detection rate | 48 (49.0) | 23 (46.9) | 25 (51.0) | 0.686b) |
| Adenoma detected per patient | 0 (0–1) | 0 (0–1) | 1 (0–1) | 0.7395a) |
| Adverse events (yes) | 0 | 0 | 0 | - |
Values are presented as number (%) or median (interquartile range).
AI, artificial intelligence; -, not applicable.
Mann-Whitney U-test,
chi-squared test.
The per-lesion analysis results are presented in Table 3. Overall, 247 colorectal lesions were identified: 132 in the CAC group and 115 in the AIC group. Of these, 43% were adenomas. No statistically significant difference in the proportion of adenomas was observed between the two groups (p=0.77). Most adenomas were identified in the rectum and sigmoid colon. However, a significant difference in lesion distribution was observed between the groups. The CAC group had more lesions in the ascending colon (21% vs. 7%, p=0.001) and sigmoid colon (30% vs. 18%, p=0.001). In contrast, the AIC group demonstrated a higher ADR in the transverse colon (16% vs. 6%, p=0.001) and descending colon (22% vs. 13 %, p=0.001). Most polyps were classified as Paris classification type I. Statistically significant intergroup differences were observed in the Paris classification. The CAC group had more Paris type IIa polyps than the AIC group (7% vs. 0%, p=0.007). The two groups exhibited no significant differences with respect to polyp size, JNET classification, or pathological diagnosis.
Table 3.
Clinicopathologic features of the resected lesions
| Characteristic | All patients (n=247) | With the use of distal cap (n=132) | With the use of AI (n=115) | p-value |
|---|---|---|---|---|
| No. of adenomatous polyps (yes) | 105 (42.5) | 55 (41.7) | 50 (43.5) | 0.774a) |
| Location (n=223) | ||||
| Ascending | 33 (14.8) | 26 (21.3) | 7 (6.9) | 0.001a),* |
| Cecum | 14 (6.3) | 7 (5.7) | 7 (6.9) | |
| Descending | 38 (17.0) | 16 (13.1) | 22 (21.8) | |
| Rectum | 61 (27.4) | 30 (24.6) | 31 (30.7) | |
| Sigmoid | 54 (24.2) | 36 (29.6) | 18 (17.8) | |
| Transverse | 23 (10.3) | 7 (5.7) | 16 (15.9) | |
| Paris classification (n=223) | ||||
| Is | 194 (87.0) | 100 (81.9) | 94 (93.1) | 0.007b),* |
| Ip | 10 (4.5) | 8 (6.6) | 2 (2.0) | |
| Isp | 10 (4.5) | 5 (4.1) | 5 (4.9) | |
| IIa | 9 (4.0) | 9 (7.4) | 0 (0) | |
| Size (mm) (n=223) | 3 (3–5) | 3 (3–4) | 3 (3–5) | 0.7188c) |
| ≤5 | 200 (90) | 112 (91.8) | 88 (87.1) | 0.392b) |
| 6-9 | 16 (7) | 6 (4.9) | 10 (9.9) | |
| ≥10 | 7 (3) | 4 (3.3) | 3 (3.0) | |
| JNET classification (n=223) | ||||
| 1 | 117 (52.5) | 66 (54.1) | 51 (50.5) | 0.832c) |
| 2A | 103 (46.2) | 54 (44.2) | 49 (48.5) | |
| 2B | 3 (1.3) | 2 (1.7) | 1 (1.0) | |
| Histopathology (n=167) | ||||
| Non-neoplastic | 88 (52.7) | 56 (54.4) | 32 (50.0) | 0.582a) |
| Neoplastic | 79 (47.3) | 47 (45.6) | 32 (50.0) | |
| Specific non-neoplastic (n=88) | ||||
| Hyperplastic polyp | 86 (97.7) | 55 (98.2) | 31 (96.9) | 0.598c) |
| Inflammatory polyp | 1 (1.1) | 1 (1.8) | 0 (0) | |
| Normal mucosa | 1 (1.1) | 0 (0) | 1 (3.1) | |
| Specific neoplastic (n=79) | ||||
| Tubular adenoma | 73 (92.4) | 41 (87.3) | 32 (100.0) | 0.105c) |
| Tubulovillous adenoma | 5 (6.3) | 5 (10.6) | 0 (0) | |
| Sessile serrated | 1 (1.3) | 1 (2.1) | 0 (0) |
Values are presented as number (%) or median (interquartile range).
AI, artificial intelligence; JNET, Japan NBI Expert Team.
Chi-squared test,
Fisher exact test,
Mann-Whitney U-test.
*p-value is statistically significant.
DISCUSSION
In this single-center, propensity score-matched, retrospective study, the comparison of AIC and CAC revealed no significant differences in ADR or PDR. Recent randomized controlled trials (RCTs) have demonstrated that both AIC and Endocuff Vision, individually and in combination, can enhance ADR rates relative to standard high-definition colonoscopy.15-17 This study used a standard transparent distal attachment cap instead of an Endocuff device. In a resource-limited setting, Endocuff devices, which have an average cost of 30 United States dollars, may impose an additional financial burden on patients.22 Additionally, Endocuff devices are not widely available in the Philippines. A recent meta-analysis indicated that CAC and Endocuff-assisted colonoscopy had comparable ADR rates. However, the utilization of Endocuff devices significantly enhanced the detection rate of diminutive adenomas or polyps relative to CAC.23 The ADR for the CAC group in our study was 47%, which was higher than the ADR of the CAC group in Li et al.’s study (45.1%),23 but comparable to the ADR of the Endocuff group (47%). The ADR of second-generation distal attachment cuffs, specifically Endocuff Vision, was even much higher, ranging from 49.6% to 54%.16,17,24 This discrepancy in ADR rates may be attributed to the fact that the study of Li et al.23 primarily utilized first-generation Endocuff devices, whereas Endocuff Vision, a second-generation device, was employed in only one study. Endocuff Vision, with its single row of softer, 2-mm longer flexible arms, may be associated with a higher ADR compared to first-generation devices.24
All CACs were performed under propofol sedation, a procedural choice dictated by endoscopist preference, owing to the perceived patient discomfort associated with the distal cap. Although there was a significant difference in anesthetic administration, the authors determined that this factor did not significantly affect ADR. Bowel preparation score and colonoscopy withdrawal time are the primary determinants of ADR,4,5 and there were no significant intergroup differences were observed in this study.
Our lesion size analysis revealed no clinically significant differences in the ADR between CAC and AIC, regardless of lesion size. Our findings contradict those of Aniwan et al.,16 who reported a statistically significant difference in ADR for adenomas <10 mm between the AIC and mucosal exposure devices (43.9% vs. 45.7%). However, no significant difference in ADR was observed for adenomas ≥10 mm between the two modalities.16 The discrepancy between our findings and those of Aniwan et al.16 may be attributed to the use of Endocuff Vision in their study. This advanced mucosal exposure device has demonstrated superior performance compared to earlier generations, potentially leading to improved detection rates, especially for smaller lesions.23,24
In terms of the location of the detected adenomas, the CAC group showed a higher detection rate in the ascending colon (21% vs. 7%) and sigmoid colon (30% vs. 18%), while the AIC group exhibited enhanced ADR in the transverse (16% vs. 6%) and descending colon (22% vs. 13%). Our study is consistent with the findings of Aniwan et al.,16 who demonstrated a higher ADR in the proximal colon using a mucosal exposure device (34.9%) than AIC (33.3%). This result was observed even though the study used a different mucosal exposure device (Endocuff Vision) compared to our study.17 A recent meta-analysis by Desai et al.25 demonstrated that CAC significantly outperformed standard colonoscopy in detecting right-sided or proximal adenomas. Right-sided colon adenomas are frequently overlooked for several reasons. Lesions situated near the haustral folds pose a significant challenge to colonoscopic detection. Additionally, the right colon has haustral folds that exhibit distinct morphological characteristics, being thin and fragile, in contrast to the complex, truncated, and bulbous folds of the left colon. In another study by Ishiyama et al.,26 the authors hypothesized that the CADe system was less effective for lesions situated within anatomical blind spots. Conversely, the relatively shallow folds of the transverse and descending colons facilitate lesion detection using CADe. Although the ascending colon has a more complex anatomical structure with deeper folds, its larger spatial dimensions and increased lens-to-mucosal distance can reduce the apparent lesion size. Consequently, the potential for CADe-mediated improvements in adenoma detection is likely to be greater in this region. The impact of CADe on adenoma detection in the distal colon, particularly in the sigmoid colon, may be limited by the unique anatomical characteristics of this region. The narrow lumen, pronounced flexure, and numerous blind spots inherent to the sigmoid colon can limit the effectiveness of this system.
Regarding polyp morphology, the CAC group had significantly more flat-elevated polyps (Paris type IIa) than the AIC group (7% vs. 0%, p=0.007). Our results were again similar to the findings of Aniwan et al.,16 indicating a superior detection rate of flat lesions with a mucosal exposure device relative to the AIC (25.1% vs. 21.5%; p<0.001).
Another study by Kim et al.27 showed that CADe demonstrated enhanced utility for identifying diminutive (1 to 5 mm) and right-sided polyps, specifically when employed in conjunction with a mucosal exposure device such as Endocuff. Furthermore, a study by Soleymanjahi et al.28 observed only a marginal increase of 0.02 incidence rate difference in sessile serrated lesion detection with CADe, suggesting limited clinical improvement, particularly in the proximal colon. Insufficient training data likely leads to a high proportion of false negatives in CADe-assisted procedures, particularly for these lesions, as shown in recent studies.29,30
CADe systems are limited by the extent of mucosal visualization. Therefore, maximizing mucosal exposure during colonoscopy is essential for the effective detection of adenomas. The efficacy of CADe relies on the operator's skill, as it cannot overcome the limitations caused by poor procedural quality or inadequate visualization.31 Recent studies on the combined application of CADe and mucosal exposure devices, specifically Endocuff Vision, indicate a significant improvement in adenoma detection performance. This improvement, achieved through optimized mucosal visualization, exceeded that of CADe alone, even for proficient endoscopists.15-17
The limitations of this study include its retrospective, non-blinded, single-center design. While propensity score matching can reduce selection bias, it cannot eliminate it entirely, particularly in the presence of unmeasured confounders. Endoscopist proficiency was not incorporated as a matching variable in this study because the same gastroenterology fellows-in-training were present in both the CAC and AIC groups and all were under direct consultant supervision. A significant limitation of this study is the single-room implementation of the AI systems. This may have introduced an unintended selection bias, potentially affecting the generalizability of our findings. Another limitation of the study was the absence of a standard colonoscopy group for direct comparison, and the limited sample size of the combined CAC and AIC group (n=6), which precluded the inclusion of this modality. The primary focus of this study was to compare the ADR of CAC and AIC, limiting the scope of the investigation. This study was conducted in a non-academic, private tertiary hospital setting; hence, it is subject to limitations inherent in real-world clinical practice. Specifically, the heterogeneity of endoscopist procedural techniques and the restricted utilization of AI and distal attachment cap technology may have affected the external validity of our findings. Consequently, the results should be interpreted with caution, acknowledging the potential limitations of the generalizability to diverse endoscopic settings. To mitigate these limitations, future multicenter prospective RCTs are warranted. These RCTs should incorporate a standard colonoscopy control arm to enable a more rigorous comparative analysis and quantify the incremental benefits conferred by CAC and AIC.
A notable strength of this study was the use of a distal attachment cap, which is a readily available and cost-effective device. This approach offers a practical and accessible solution for enhancing ADR, particularly in resource-limited settings where advanced mucosal exposure devices such as Endocuff Vision may not be widely available. Furthermore, in endoscopy units where AIC is not feasible, the use of a distal attachment cap may offer a potential strategy to improve ADR, similar to that observed with AI. In conclusion, both CAC and AIC demonstrated the potential to enhance the ADR and PDR. However, no significant advantage was observed when one technique was used. While the CAC exhibited a superior ability to detect lesions in the ascending and sigmoid colon, the AIC demonstrated a stronger performance in the transverse and descending colon. Additionally, CAC showed a higher detection rate for flat lesions than AIC. The complementary nature of these two modalities suggests that a combined approach may optimize adenoma detection and mitigate the limitations of each technique.
Footnotes
Conflicts of Interest
The authors have no potential conflicts of interest.
Funding
This work was supported by a grant from the Korean Gastrointestinal Endoscopy Research Foundation through the KSGE/IDEN Fine Crystal Award 2022.
Author Contributions
Conceptualization: JRLT; Data curation: JRLT, KBTE, KVTA; Formal analysis: JRLT; Funding acquisition: JRLT; Investigation: JRLT, KBTE, KVTA, MAGG; Methodology: JRLT, MAGG; Project administration: JRLT; Resources: KBTE, KVTA; Software: KBTE, KVTA; Supervision: MAGG, MLOL, JTC; Validation: all authors; Visualization: MAGG, MLOL, JTC; Writing–original draft: JRLT; Writing–review & editing: all authors.
REFERENCES
- 1.Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71:209–249. doi: 10.3322/caac.21660. [DOI] [PubMed] [Google Scholar]
- 2.Winawer SJ, Zauber AG, Ho MN, et al. Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup. N Engl J Med. 1993;329:1977–1981. doi: 10.1056/nejm199312303292701. [DOI] [PubMed] [Google Scholar]
- 3.Nishihara R, Wu K, Lochhead P, et al. Long-term colorectal-cancer incidence and mortality after lower endoscopy. N Engl J Med. 2013;369:1095–1105. doi: 10.1056/nejmoa1301969. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Rex DK, Schoenfeld PS, Cohen J, et al. Quality indicators for colonoscopy. Gastrointest Endosc. 2015;81:31–53. doi: 10.1016/j.gie.2014.07.058. [DOI] [PubMed] [Google Scholar]
- 5.Kaminski MF, Regula J, Kraszewska E, et al. Quality indicators for colonoscopy and the risk of interval cancer. N Engl J Med. 2010;362:1795–1803. doi: 10.1056/nejmoa0907667. [DOI] [PubMed] [Google Scholar]
- 6.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. doi: 10.1056/nejmoa1309086. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Rex DK, Repici A, Gross SA, et al. High-definition colonoscopy versus Endocuff versus EndoRings versus full-spectrum endoscopy for adenoma detection at colonoscopy: a multicenter randomized trial. Gastrointest Endosc. 2018;88:335–344. doi: 10.1016/j.gie.2018.02.043. [DOI] [PubMed] [Google Scholar]
- 8.Gkolfakis P, Tziatzios G, Facciorusso A, et al. Meta-analysis indicates that add-on devices and new endoscopes reduce colonoscopy adenoma miss rate. Eur J Gastroenterol Hepatol. 2018;30:1482–1490. doi: 10.1097/meg.0000000000001245. [DOI] [PubMed] [Google Scholar]
- 9.Frieling T, Neuhaus F, Heise J, et al. Cap-assisted colonoscopy (CAC) significantly extends visualization in the right colon. Z Gastroenterol. 2012;50:279–284. doi: 10.1055/s-0031-1281707. [DOI] [PubMed] [Google Scholar]
- 10.Kondo S, Yamaji Y, Watabe H, et al. A randomized controlled trial evaluating the usefulness of a transparent hood attached to the tip of the colonoscope. Am J Gastroenterol. 2007;102:75–81. doi: 10.1111/j.1572-0241.2006.00897.x. [DOI] [PubMed] [Google Scholar]
- 11.Rastogi A, Bansal A, Rao DS, et al. Higher adenoma detection rates with cap-assisted colonoscopy: a randomised controlled trial. Gut. 2012;61:402–408. doi: 10.1136/gutjnl-2011-300187. [DOI] [PubMed] [Google Scholar]
- 12.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. doi: 10.5946/ce.2023.145. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.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. doi: 10.5946/ce.2024.038. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.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. doi: 10.5946/ce.2024.168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lui TK, Lam CP, To EW, et al. Endocuff with or without artificial intelligence-assisted colonoscopy in detection of colorectal adenoma: a randomized colonoscopy trial. Am J Gastroenterol. 2024;119:1318–1325. doi: 10.14309/ajg.0000000000002684. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.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. doi: 10.1016/j.gie.2022.09.023. [DOI] [PubMed] [Google Scholar]
- 17.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. doi: 10.1055/s-0043-1765359. [DOI] [PubMed] [Google Scholar]
- 18.Tseng HH, Su CW, Chang WC, et al. Impact of time period and birth cohort on the trend of advanced neoplasm prevalence in the 40-49 average-risk screening population. Clin Gastroenterol Hepatol. 2025;23:644–652. doi: 10.1016/j.cgh.2024.07.044. [DOI] [PubMed] [Google Scholar]
- 19.Yoshida N, Inoue K, Tomita Y, et al. An analysis about the function of a new artificial intelligence, CAD EYE with the lesion recognition and diagnosis for colorectal polyps in clinical practice. Int J Colorectal Dis. 2021;36:2237–2245. doi: 10.1007/s00384-021-04006-5. [DOI] [PubMed] [Google Scholar]
- 20.The Paris endoscopic classification of superficial neoplastic lesions: esophagus, stomach, and colon: November 30 to December 1, 2002. Gastrointest Endosc. 2003;58(6 Suppl):S3–S43. doi: 10.1016/s0016-5107(03)02159-x. [DOI] [PubMed] [Google Scholar]
- 21.Sano Y, Tanaka S, Kudo SE, et al. Narrow-band imaging (NBI) magnifying endoscopic classification of colorectal tumors proposed by the Japan NBI Expert Team. Dig Endosc. 2016;28:526–533. doi: 10.1111/den.12644. [DOI] [PubMed] [Google Scholar]
- 22.Yu TM, Tradonsky A, Tang J, et al. Cost-effectiveness of adding Endocuff® to standard colonoscopies for interval colorectal cancer screening. Clinicoecon Outcomes Res. 2019;11:487–504. doi: 10.2147/ceor.s201328. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Li A, Yang JZ, Yang XX, et al. Endocuff-assisted colonoscopy versus cap-assisted colonoscopy for adenoma detection rate: a meta-analysis of randomized controlled trials. J Gastroenterol Hepatol. 2020;35:2066–2073. doi: 10.1111/jgh.15155. [DOI] [PubMed] [Google Scholar]
- 24.Patel HK, Chandrasekar VT, Srinivasan S, et al. Second-generation distal attachment cuff improves adenoma detection rate: meta-analysis of randomized controlled trials. Gastrointest Endosc. 2021;93:544–553. doi: 10.1016/j.gie.2020.09.045. [DOI] [PubMed] [Google Scholar]
- 25.Desai M, Sanchez-Yague A, Choudhary A, et al. Impact of cap-assisted colonoscopy on detection of proximal colon adenomas: systematic review and meta-analysis. Gastrointest Endosc. 2017;86:274–281. doi: 10.1016/j.gie.2017.03.1524. [DOI] [PubMed] [Google Scholar]
- 26.Ishiyama M, Kudo SE, Misawa M, et al. Impact of the clinical use of artificial intelligence-assisted neoplasia detection for colonoscopy: a large-scale prospective, propensity score-matched study (with video) Gastrointest Endosc. 2022;95:155–163. doi: 10.1016/j.gie.2021.07.022. [DOI] [PubMed] [Google Scholar]
- 27.Kim JH, Wang J, Pence C, et al. 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. Clin Endosc. 2025;58:438–447. doi: 10.5946/ce.2024.271. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Soleymanjahi S, Huebner J, Elmansy L, et al. Artificial intelligence-assisted colonoscopy for polyp detection: a systematic review and meta-analysis. Ann Intern Med. 2024;177:1652–1663. doi: 10.7326/annals-24-00981. [DOI] [PubMed] [Google Scholar]
- 29.Ali S, Jha D, Ghatwary N, et al. A multi-centre polyp detection and segmentation dataset for generalisability assessment. Sci Data. 2023;10:75. doi: 10.1038/s41597-023-01981-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Rex DK, Mori Y, Sharma P, et al. Strengths and weaknesses of an artificial intelligence polyp detection program as assessed by a high-detecting endoscopist. Gastroenterology. 2022;163:354–358. doi: 10.1053/j.gastro.2022.03.055. [DOI] [PubMed] [Google Scholar]
- 31.Spadaccini M, Menini M, Massimi D, et al. AI and polyp detection during colonoscopy. Cancers (Basel) 2025;17:797. doi: 10.3390/cancers17050797. [DOI] [PMC free article] [PubMed] [Google Scholar]