Diagnostic accuracy and interobserver agreement for prediction of gastric preneoplastic lesions with fourth-generation endocytoscopy: Pilot study

Abstract

Background and study aims

Early detection of gastric preneoplastic lesions (GPL) is essential to prevent gastric cancer. Endocytoscopy may improve optical diagnosis through in vivo histologic assessment of mucosa. We assessed diagnostic performance of endocytoscopy and interobserver agreement for GPL diagnosis compared with reference histopathology.

Patients and methods

Eighty gastric areas (36 antrum, 44 corpus) from 25 prospectively recruited patients were assessed by endocytoscopy and classified as non-preneoplastic, preneoplastic, or dysplastic/neoplastic by an expert endoscopist. Targeted biopsies from each area served for histopathological assessment. High-quality images/videos were recorded for independent review by three additional observers (1 expert, 2 non-experts), blinded to endoscopic and histological findings.

Results

Histologically, 25 gastric areas (31.3%) showed GPL, 14 (17.5%) atrophic gastritis, and 11 (13.8%) intestinal metaplasia. Gastric cancer was found in three cases (3.8%). Combined sensitivity, specificity, and accuracy for endocytoscopic diagnosis of GPL or dysplasia were 89.3%, 94.2%, and 92.5% among expert endocytoscopists and 85.7%, 68.3%, and 74.4%, among non-experts, respectively. Interobserver agreement was substantial (κ-value 0.79) between experts and fair (κ-value 0.30) between non-experts.

Conclusions

New-generation endocytoscopy has potential to identify GPL with high diagnostic accuracy, reliability, and reproducibility. Training and clinical experience with endocytoscopy are required to maximize diagnostic performance.

 

Scroll to top

 

Scroll to top

Keywords

Endoscopy Upper GI Tract - Precancerous conditions & cancerous lesions (displasia and cancer) stomach - Diagnosis and imaging (inc chromoendoscopy, NBI, iSCAN, FICE, CLE) - Ulcers (peptic and other)

Introduction

Patients with chronic atrophic gastritis with or without intestinal metaplasia have an increased risk of developing gastric adenocarcinoma. Therefore, early detection of these gastric preneoplastic lesions (GPL) is essential to prevent progression and reduce cancer-related mortality [1]. Current guidelines recommend integrating conventional high-definition white-light imaging (WLI) with other image-enhanced endoscopy (IEE) techniques, such as virtual chromoendoscopy and/or magnification endoscopy, whenever available. This approach improves optical diagnosis of GPL and early gastric neoplasia and facilitates targeting biopsy sampling more effectively [2].

Endocytoscopy is an advanced imaging modality used in gastrointestinal endoscopy for detailed in vivo examination of the mucosal surface. It allows for real-time visualization of cellular and subcellular structures at ultra-high magnification, providing information about tissue morphology at the microscopic level. Endocytoscopy has been shown to improve the detection rate for various precancerous lesions throughout the gastrointestinal tract, such as dysplasia in Barrett’s esophagus [3], in patients with ulcerative colitis [4], and in colorectal polyps [5]. Previous studies evaluating endocytoscopy for characterization of stomach lesions have primarily focused on gastric neoplasia, consistently demonstrating high accuracy in discriminating between gastric cancer and non-neoplastic tissue [6]. However, there is currently a lack of data regarding diagnostic utility of endocytoscopy for detecting GPL in both the antrum and corpus. Recently, we characterized endocytoscopic alterations related to GPL at different stages of gastric carcinogenesis using a modern endocytoscopy system and image enhancement technologies, such as texture and color enhancement (TXI) [7]. The aims of this study were to assess diagnostic accuracy of endocytoscopy for detecting GPL compared with standard histopathology and to assess interobserver agreement among endoscopists with different levels of experience with this technique.

 

Scroll to top

Patients and methods

Study population

Patients undergoing upper gastrointestinal endoscopy (UGE) with dyspeptic symptoms or because of alarming symptoms (weight loss, anemia, vomiting, or dysphagia) were prospectively recruited at the LMU University Hospital in Munich from November 2020 to May 2021.

Subjects were recruited within the ERANET Bavaria and Helicopredict projects (German clinical trials register, DRKS-ID: DRKS00028629), large-scale prospective studies focused on studying different aspects of Helicobacter pylori infection, including assessment of endoscopic alterations of the gastric mucosa occurring during gastric carcinogenesis. This study was approved by the local ethics committee and government authorities (Project Number 20–875) and was conducted in accordance with current Good Clinical Practice guidelines and the Declaration of Helsinki [8]. Written informed consent was obtained from each participant. Exclusion criteria included a history of prior gastric surgery and use of anticoagulants or antibiotics within the 4 weeks preceding endoscopy. Regular use of proton pump inhibitors or previous H. pylori eradication therapy did not represent exclusion criteria.

 

Scroll to top

Endocytoscopic procedure and histological assessment

Enrolled patients underwent diagnostic UGE using standard video gastroscopes, as well as a fourth-generation endocytoscope (GIF-HQ190, GIF-H290EC) on a EVIS X1 system (Olympus, Tokyo, Japan). All examinations were performed by one endoscopist (A) with expertise in performing endocytoscopy.

To optimize staining quality for endocytoscopy, a 50-mL water-based solution containing a mucolytic agent (600 mg acetylcysteine, Hexal, Holzkirchen, Germany) and 5 mL dimethicone (Sab simplex, Pfizer, Berlin, Germany) was administered to all patients 20 minutes before endoscopy. Endoscopic procedures were performed under intravenous sedation with propofol and/or midazolam. After initial assessment with WLI and narrow-band imaging (NBI), suitable areas (or lesions) were stained using a NaCl-based solution containing 0.05 % crystal violet and 0.1 % methylene blue, as previously described [7]. Endocytoscopic assessment started approximately 2 minutes after dye absorption and was carried out using different grades of magnification (up to a maximum of 520x) as well as the TXI modus. After exhaustive documentation with images and video recording for each area, targeted biopsies were obtained for histopathological diagnosis, which followed the updated Sydney system criteria [9]. Biopsies were obtained immediately after endocytoscopic assessment at the highest magnification (520x). At maximum magnification, endocytoscopy requires direct contact between the endoscope tip and the gastric mucosa, which is facilitated by controlled modulation of air suction through the endoscope.

Biopsy forceps were introduced through the working channel of the endoscope while the tip of the instrument maintained contact with the mucosa. This approach ensured that the biopsy was taken precisely from the site visualized under maximal magnification, minimizing discrepancies between the observed and sampled areas.

Endocytoscopic diagnosis relied on an existing classification system [10] [11], which was specifically adjusted for evaluating gastric lesions, considering: 1) foveolar architecture; 2) vascular structure; 3) presence/absence of goblet cells; and 4) configuration of cell nuclei in gastric areas. Based on these endocytoscopic features, gastric areas were categorized as follows: non-preneoplastic (EC1A), preneoplastic (EC1B), adenoma (EC2), and cancer (EC3). Within the EC1B category, lesions were further subdivided into EC1B (atrophy) if they exhibited features indicative of gastric atrophy, EC1B (IM) if they displayed characteristics typical of intestinal metaplasia, and EC1B (atrophy+IM) if there was a combination of both gastric atrophy and intestinal metaplasia. Exemplary images demonstrating characteristic changes in endocytoscopy for each of these gastric conditions are shown in [Fig. 1]. [Fig. 2] highlights the benefit of using the TXI mode over WLI for endocytoscopic evaluation. An exemplary video showing the diagnostic procedure leading to the diagnosis of GPL is provided elsewhere [7].

Fig. 1 Exemplary endocytoscopic images using texture and color enhancement imaging (TXI) modus and ultra-high magnification (350×-520×) showing characteristic findings in different conditions of the gastric mucosa. a Normal gastric mucosa in antrum and b corpus, with regular arrangement of foveolar structures/gastric pits, small and poorly-stained nuclei, and regular honeycomb-like subepithelial capillary network. c, d Chronic atrophic gastritis, with patchy distribution of atrophic areas, deformation and loss of the glandular architecture, reduced dye uptake and translucent vessels. e Intestinal metaplasia, with compactly arranged glandular structures, narrowing of the lumens (slit-like lumens), increased dye uptake, and multiple goblet cells. f Gastric neoplasia, with complete disarrangement of gastric foveolae and nuclear atypia (hyperchromatic and pleomorphic nuclei, increased nuclear/cytoplasmic ratio). These images were classified as a, b EC1A, c, d EC1B (atrophy), e EC1B (IM), and f EC3.

Fig. 2 Comparison of images from the same gastric areas obtained using conventional white-light imaging (WLI) (left) and texture and color enhancement imaging (TXI) (right) modes during endocytoscopy. a, b Corpus, 150×. c, d Corpus, 520×. e, f Antrum, 250×. g, h Antrum, 520×.

 

Scroll to top

Statistical analyses and interobserver agreement

Following independent endocytoscopic diagnosis, high-quality images and video sequences representing each area were selected from the database by endoscopist A. These were then organized into evaluation sets for post hoc analysis. On average, four images per lesion at different magnification levels, with at least one image at the highest magnification (> 450x), were used for evaluation. Video sequences also were used for lesions with a multifocal pattern that could not be adequately represented by a single still image. Each evaluation set contained the same collection of data, but the lesions were arranged in different randomly reassigned orders using Microsoft Excel (Microsoft Corporation, Redmond, Washington, United States). Three endoscopists (B, C, and D) reviewed the evaluation sets, assessing the above-mentioned features and providing a final endocytoscopic diagnosis. Reviewers were blinded to patient data, as well as to previous endoscopic and histopathologic findings. Endoscopists A and B, with experience in more than 50 endocytoscopy procedures, were considered experts. Endoscopists C and D had no experience in performing endocytoscopy but were trained for image assessment. All endoscopists had more than 5 years of experience in screening endoscopy, advanced endoscopic procedures, chromoendoscopy, and conventional magnifying endoscopy. Training for assessment of endocytoscopy images was provided to reviewers C and D through a 1-hour lecture by an expert endoscopist (A). The training covered principles of the endocytoscopy technique, diagnostic criteria, and correlation with histopathologic findings, as well as lesion classification based on published literature. The training involved evaluation of exemplary images for each lesion category, which were not included in the study data set. Furthermore, all reviewers were regularly involved in joint multidisciplinary meetings with a pathologist (JN), during which endoscopic findings of GPL were correlated with histopathologic results, thereby providing them with a basic understanding of pathology relevant to the study.

Descriptive statistical analysis was performed using IBM SPSS Statistics 29.0.0 (IBM Corporation, New York, United States). Sensitivity, specificity, accuracy, and positive and negative predictive value of endocytoscopy for detection of GPL were calculated using histology as the gold standard. Diagnostic performances were estimated for each observer separately and for expert (A, B) or non-expert (C, D) observers combined. The McNemar test was used to compare sensitivities and specificities between two observers. Interobserver agreement was assessed by estimating the pairwise concordance rates as well as the Cohen’s κ-value among expert and non-expert observers. The Fleiss’ κ-value was used for assessment of interrater reliability among all observers.

 

Scroll to top

 

Scroll to top

Results

Twenty-five patients (12/13 male/female; mean age 58.1±14.6 years) scheduled for elective UGE were enrolled. A total of 80 gastric areas (36 in antrum, 44 in corpus) were assessed by endocytoscopy. Histopathological assessment of paired targeted biopsies identified GPL in 25 of 80 gastric areas (31.3%), including 14 cases (17.5%) with chronic atrophic gastritis and intestinal metaplasia and 11 cases (13.8%) with intestinal metaplasia alone. Gastric neoplasia was found in three cases (3.8%).

Diagnostic performance of endocytoscopy compared with histological assessment is shown in [Table 1]. Overall, discrimination of GPL or gastric dysplasia from non-preneoplastic lesions showed combined sensitivity, specificity, and accuracy of 89.3%, 94.2%, and 92.5%, respectively, among endoscopists with experience in performing endocytoscopy. These diagnostic measures were significantly lower when evaluation of endocytoscopic images was performed by endoscopists with no experience in performing endocytoscopy, with combined sensitivity, specificity, and accuracy of 85.7%, 68.3% and 74.4%, respectively.

Table 1 Diagnostic performance of endocytoscopy for the diagnosis of GPLs according to assessment of 4 reviewers (endoscopists A-D), as compared with histological diagnosis of target biopsies (gold standard).

Absence of GPL (EC1A)	Sensitivity	Specificity	Accuracy	PPV	NPV
95% confidence intervals are indicated in brackets. GPL, gastric preneoplastic lesion; IM, intestinal metaplasia; NPV, negative predictive value; PPV, positive predictive value.
Endoscopist A	98.1% (89.7%–99.9%)	96.4% (81.7%–99.9%)	97.5% (91.3%–99.7%)	98.1% (88.2%–99.7%)	96.4% (79.5%–99.5%)
Endoscopist B	90.4% (79.0%–96.8%)	82.1% (63.1%–93.9%)	87.5% (78.2%–93.8%)	90.4 %(80.9%–95.4%)	82.1% (66.3%–91.5%)
Endoscopist C	61.5% (47.0%–74.7%)	96.4% (81.7%–99.9%)	73.8% (62.7%–83.0%)	97.0% (82.2%–99.6%)	57.5% (48.7%–65.7%)
Endoscopist D	75.0% (61.1%–86.0%)	75.0% (55.1%–89.3%)	75.0% (64.1%–84.0%)	84.8% (74.2%–91.5%)	61.8% (49.1%–73.0%)
Combined (A+B)	94.2% (87.9%–97.9%)	89.3% (78.1%–96.0%)	92.5% (87.3%–96.1%)	94.2% (88.5%–97.2%)	89.3% (79.2%–94.8%)
Combined (C+D)	68.3% (58.4%–77.1%)	85.7% (73.8%–93.6%)	74.4% (66.9%–80.9%)	89.9% (82.2%–94.5%)	59.3% (51.8%–66.3%)
Presence of GPL (EC1B)	Sensitivity	Specificity	Accuracy	PPV	NPV
Endoscopist A	96.0% (79.7%–99.9%)	98.2% (90.1%–99.9%)	97.5% (91.3%–99.7%)	96.0% (77.5%–99.4%)	98.2% (88.9%–99.7%)
Endoscopist B	80.0% (59.3%–93.2%)	90.4% (79.0%–96.8%)	87.0% (77.4%–93.6%)	80.0% (63.0%–90.4%)	90.4% (81.0%–95.4%)
Endoscopist C	84.0% (63.9%–95.5%)	70.9% (57.1%–82.4%)	75.0% (64.1%–84.0%)	56.8% (45.6%–67.2%)	90.7% (79.6%–96.1%)
Endoscopist D	68.0% (46.5%–85.1%)	78.2% (65.0%–88.2%)	75.0% (64.1%–84.0%)	58.6% (44.5%–71.4%)	84.3% (74.9%–90.6%)
Combined (A+B)	88.0% (75.7%–95.5%)	94.6% (88.5%–98.0%)	92.5% (87.3%–96.0%)	88.0% (77.0%–94.1%)	94.6% (89.1%–97.4%)
Combined (C+D)	76.0% (61.8%–86.9%)	74.6% (65.4%–82.4%)	75.0% (67.6%–81.5%)	57.6% (48.7%–66.0%)	87.2% (80.5%–91.9%)
Presence of GPL (EC1B) or adenoma (EC2) or cancer (EC3)	Sensitivity	Specificity	Accuracy	PPV	NPV
Endoscopist A	96.4% (81.7%–99.9%)	98.1% (89.7%–99.9%)	97.5% (91.3%–99.7%)	96.4% (79.5%–99.5%)	98.1% (88.2%–99.7%)
Endoscopist B	82.1% (63.1%–93.9%)	90.4% (79.0%–96.8%)	87.5% (78.2%–93.8%)	82.1% (66.3%–91.5%)	90.4% (80.9%–95.4%)
Endoscopist C	96.4% (81.7%–99.9%)	61.5% (47.0%–74.7%)	73.8% (62.7%–83.0%)	57.5% (48.7%–65.7%)	97.0% (82.2%–99.6%)
Endoscopist D	75.0% (55.1%–89.3%)	75.0% (61.1%–86.0%)	75.0% (64.1%–84.0%)	61.8% (49.0%–73.0%)	84.9% (74.2%–91.5%)
Combined (A+B)	89.3% (78.1%–96.0%)	94.2% (87.9%–97.9%)	92.5% (87.3%–96.1%)	89.3% (79.2%–94.8%)	94.2% (88.5%–97.2%)
Combined (C+D)	85.7% (73.8%–93.6%)	68.3% (58.4%–77.1%)	74.4% (66.9%–80.9%)	59.3% (51.8%–66.3%)	89.9% (82.2%–94.5%)
Gastric atrophy (EC1B atrophy)	Sensitivity	Specificity	Accuracy	PPV	NPV
Endoscopist A	64.3% (35.1%–87.2%)	100 % (94.6%–100%)	93.8% (86.0%–97.9%)	100 % (66.4%–100%)	93.0% (86.7%–96.4%)
Endoscopist B	35.7% (12.8%–64.9%)	98.5% (91.8%–100%)	87.5% (78.2%–93.8%)	83.3% (38.7%–97.5%)	87.8% (83.0%–91.4%)
Endoscopist C	50.0% (23.0%–77.0%)	92.4% (83.2%–97.5%)	85.0% (75.3%–92.0%)	58.3% (34.2%–79.1%)	89.7% (83.7%–93.7%)
Endoscopist D	57.1% (28.9%–82.3%)	80.3% (68.7%–89.1%)	76.3% (65.4%–85.1%)	38.1% (82.7%–94.2%)	89.8% (82.7%–94.2%)
Combined (A+B)	50.0% (30.7%–69.4%)	99.2% (95.9%–100%)	90.6% (85.0%–94.7%)	93.3% (65.7%–99.0%)	90.3% (86.6%–93.1%)
Combined (C+D)	53.6% (33.9%–72.5%)	86.4% (79.3%–91.7%)	80.6% (73.6%–86.4%)	45.5% (32.5%–59.1%)	89.8% (73.6%–86.4%)
Intestinal metaplasia (EC1B IM)	Sensitivity	Specificity	Accuracy	PPV	NPV
Endoscopist A	84.0% (63.9%–95.5%)	98.2% (90.1%–99.9%)	93.8% (86.0%–97.9%)	95.5% (74.9%–99.3%)	93.1% (84.6%–97.1%)
Endoscopist B	68.0% (46.5%–85.1%)	92.7% (82.4%–98.0%)	85.0% (72.3%–92.0%)	81.0% (61.4%–91.9%)	86.4% (78.2%–91.9%)
Endoscopist C	68.0% (46.5%–85.1%)	80.0% (67.0%–89.6%)	76.3% (65.4%–85.1%)	60.7% (46.1%–73.7%)	84.6% (75.4%–90.8%)
Endoscopist D	40.0% (21.1%–61.3%)	100 % (93.5%–100%)	81.3% (71.0%–89.1%)	100% (69.2%–100%)	78.6% (72.7%–83.5%)
Combined (A+B)	76.0% (61.8%–86.9%)	95.5% (89.7%–98.5%)	89.4% (83.5%–93.7%)	88.4% (76.1%–94.8%)	89.7% (84.2%–93.5%)
Combined (C+D)	54.0% (39.3%–68.2%)	90.0% (82.8%–94.9%)	78.8% (71.6%–84.8%)	71.1% (57.0%–82.0%)	81.2% (76.0%–85.4%)

Level of concordance among observers as well as interrater reliability are reported in [Table 2]. Endocytoscopic diagnosis of GPL or gastric neoplasia made by endoscopists A and B showed a substantial agreement with a κ-value of 0.79 and high proportions of pairwise concordance, ranging from 77.5% to 100%, for each morphological feature considered. Level of agreement for diagnosis of GPL or gastric dysplasia among endoscopists C and D was fair (κ-value 0.30). In this group of reviewers, the highest proportion of concordance was observed in detection of goblet cells (90.0%), translucent vessels (81.3%) and nuclear atypia (87.5%). Overall, interrater reliability among all observers for prediction of GPL or gastric neoplasia was moderate (κ-value 0.48).

Table 2 Interobserver agreement among 4 reviewers for both final endocytoscopic diagnosis and each of the endocytoscopic features evaluated.

	Experts (A,B)*				Non-experts (C,D)†				All observers (A,B,C,D)
*A and B had experience in performing endocytoscopy. †C and D had no experience in performing endocytoscopy and were trained for image assessment.
Endocytoscopic features	Pairwise concordance		Cohenʼs Kappa		Pairwise concordance		Cohenʼs Kappa		Fleissʼ Kappa
	n/n	%	value	95% CI	n/n	%	value	95% CI	value	95% CI
Endocytoscopic diagnosis	72/80	90.0	0.79	0.65–0.93	47/80	58.8	0.30	0.12–0.47	0.48	0.41–0.56
Glandular architecture	62/80	77.5	0.63	0.48–0.77	33/80	41.3	0.23	0.10–0.37	0.41	0.36–0.46
Vascular structure	75/80	93.8	0.67	0.40–0.94	65/80	81.3	0.01	-0.21–0.24	0.39	0.30–0.48
Goblet cells	67/80	83.8	0.50	0.27–0.74	72/80	90.0	0.63	0.40–0.87	0.59	0.50–0.68
Nuclear atypia	80/80	100	1.00	1.00–1.00	70/80	87.5	0.31	0.00–0.63	0.45	0.36–0.54

 

Scroll to top

Discussion

High-quality endoscopic evaluation is essential for diagnosis of GPL and early identification of patients at higher risk for gastric cancer, who would benefit from a surveillance strategy [2]. Here, we report on diagnostic accuracy of endocytoscopic assessment of gastric mucosa for optical diagnosis of GPL in comparison with the histological gold standard. The endocytoscopic findings facilitate discrimination with high accuracy between non-preneoplastic mucosa and presence of GPL or gastric dysplasia. However, interobserver accuracy varied substantially according to degree of experience in performing this technique.

Our study is the first to attempt validation of endocytoscopy for detection and characterization of GPL, in both stomach antrum and corpus, using a last-generation integrated-type endocytoscope and the TXI modus. The combination of image enhancement through TXI further improves endocytoscopy assessment, enabling acquisition of high-quality images for structural evaluation. Only a limited number of studies have assessed accuracy of endocytoscopy for diagnostic classification of gastric lesions. All these studies were conducted in Eastern countries (9 in Japan and 1 in China), and the large majority of them focused on diagnosis of gastric cancer [6] [11] [12] [13] [14] [15] [16] [17] [18] [19]. In a prospective pilot study by Sato et al., non-neoplastic changes in the gastric antral mucosal were assessed in a cohort of 64 patients. Endocytoscopic mucosal patterns suggestive of atrophic gastritis and intestinal metaplasia highly correlated with the histological diagnosis of GPL, demonstrating a sensitivity of 87.0% and a specificity of 95.1%, with an almost perfect interobserver agreement (κ-value ranging from 0.86 to 0.93 among reviewers experienced in endocytoscopy) [16]. Our study further confirmed these findings, extending the analysis not only to the antral mucosa but also to corpus areas. Most of the endocytoscopic features previously reported by others were consistently observed in our cohort of patients with GPL, including increased dye uptake and detection of goblet cells for discrimination of intestinal metaplasia [16] [17] and presence of nuclear atypia (i.e. enlarged nuclear sign [11]) in neoplastic lesions. Especially in identifying goblet cells or nuclear atypia, the highest levels of interrater reliability were observed, even among observers without prior experience in endocytoscopy. Furthermore, diagnostic measures of non-expert reviewers (C and D) showed relatively high sensitivity and NP values for detecting GPL or neoplastic patterns (EC1B and EC2/3). In the clinical setting, this enables targeted biopsies for further histological confirmation in cases of uncertainty, thus minimizing risk of missing potential high-risk lesions.

Our findings suggest that the most prominent endocytoscopic changes can be readily recognized even after minimal training in image assessment. However, these results also highlight the importance of appropriate observer training for accurate endocytoscopic assessment. Previous studies have described substantially higher interobserver agreement among observers with experience in observing at least 30 endocytoscopic procedures [16] or when the focus was limited to identifying much more evident alterations occurring in gastric cancer [13]. In our study, endoscopists lacking prior onsite experience with endocytoscopy received instructions on how to evaluate different features in a single session, using a limited set of still images. This approach may have provided insufficient training for reviewers C and D, thus presenting a limitation of our study. To date, no formal learning steps or competence criteria for optical diagnosis by endocytoscopy have been defined. Similar to other advanced imaging endoscopic techniques [20], onsite training with experts, as well as dedicated training sessions alongside gastrointestinal pathologists, may be necessary to achieve proficiency in this technique.

Another potential limitation is that all procedures were performed by a single expert endoscopist (A), who also selected images for the evaluation sets. However, reviewer A was not aware of the histopathological findings at the time of evaluation, because his endocytoscopic diagnosis had to be provided immediately after the procedure. Unlike the other reviewers, his diagnostic assessment may have been influenced (and supported) by prior inspection of lesions with WLI and NBI, as well as by availability of dynamic imaging. Nevertheless, this reflects real-life clinical practice, where multiple imaging modalities are used intraprocedurally to guide endoscopic decision-making. Furthermore, the high concordance of endocytoscopic diagnoses with an independent expert reviewer (B), who was blinded to endoscopic and histological results, supports the validity of the observed endocytoscopic changes.

Future studies are needed to assess the cost-effectiveness, clinical applicability, and potential added benefit of endocytoscopy compared with conventional modalities. Accurate differentiation between normal and preneoplastic mucosa can be achieved in most cases using high-definition WLI combined with virtual chromoendoscopy. A recent meta-analysis demonstrated that NBI, along with similar technologies such as blue laser/light imaging and I-Scan, is the most effective modality for detecting gastric intestinal metaplasia, with pooled data for sensitivity and specificity ranging from 0.79 to 0.84 and 0.91 to 0.95, respectively, whereas detection of atrophic gastritis with these technologies still suffers from an unsatisfactory diagnostic yield [21]. Moreover, current evidence suggests that integrating ultra-high magnification with virtual chromoendoscopy further improves diagnostic performances for intestinal metaplasia [21] [22] and gastric cancer detection [23] [24]. Studies directly comparing diagnostic performance of endocytoscopy and conventional IEE modalities for detecting GPL are currently lacking. Although WLI and NBI are established as gold-standard techniques for GPL detection, endocytoscopy holds significant promise as a complementary tool in the endoscopist armamentarium when making optical diagnoses, especially for cases in which virtual chromoendoscopy alone might not provide definitive results.

Further development of this technology is expected through integration of artificial intelligence (AI) solutions for interpreting endocytoscopic images to improve diagnostic accuracy of GPL. Recent advancements in this field have been reported in upper gastrointestinal endoscopy, in which AI-assisted endocytoscopic diagnosis significantly increased accuracy of identifying early gastric cancer and Barrett’s-associated dysplasia [3] [12].

 

Scroll to top

Conclusions

In conclusion, fourth-generation endocytoscopy has the potential to identify GPL with high diagnostic accuracy, high reliability, and good reproducibility. In vivo histological evaluation and optical diagnosis by endocytoscopy may be powerful tools to target biopsies in patients at high risk for gastric cancer. Diagnostic performance is significantly influenced by the experience of the endoscopist with this procedure, and therefore, dedicated training is required to increase diagnostic benefit.

 

Scroll to top

 

Scroll to top

Conflict of Interest

The authors declare that they have no conflict of interest.

Acknowledgement

The authors would like to thank Ulrich Lang for organizing the data and the endoscopy team from Medical department II for their excellent technical assistance during study procedures.

• References

• 1 Rugge M, Genta RM, Malfertheiner P. et al. RE.GA.IN.: the Real-world Gastritis Initiative-updating the updates. Gut 2024; 73: 407-441
• 2 Pimentel-Nunes P, Libânio D, Marcos-Pinto R. et al. Management of epithelial precancerous conditions and lesions in the stomach (MAPS II): European Society of Gastrointestinal Endoscopy (ESGE), European Helicobacter and Microbiota Study Group (EHMSG), European Society of Pathology (ESP), and Sociedade Portuguesa de Endoscopia Digestiva (SPED) guideline update 2019. Endoscopy 2019; 51: 365-388
• 3 van der Laan JJH, van der Putten JA, Zhao X. et al. Optical biopsy of dysplasia in Barrett's oesophagus assisted by artificial intelligence. Cancers (Basel) 2023; 15
• 4 Kudo S-e, Maeda Y, Ogata N. et al. Combined endocytoscopy with pit pattern diagnosis in ulcerative colitis-associated neoplasia: Pilot study. Dig Endosc 2022; 34: 133-143
• 5 Barua I, Mori Y, Bretthauer M. Colorectal polyp characterization with endocytoscopy: Ready for widespread implementation with artificial intelligence?. Best Pract Res Clin Gastroenterol 2021; 52-53
• 6 Kaise M, Ohkura Y, Iizuka T. et al. Endocytoscopy is a promising modality with high diagnostic accuracy for gastric cancer. Endoscopy 2015; 47: 19-25
• 7 Vasapolli R, Neuhaus L, Schirra J. et al. Microscopic alterations of the gastric mucosa in preneoplastic lesions as assessed by new-generation endocytoscopy. Endoscopy 2023; 55: E998-e1000
• 8 World Medical Association. World Medical Association Declaration of Helsinki: Ethical Principles for Medical Research Involving Human Subjects. JAMA 2013; 310: 2191-2194
• 9 Dixon MF, Genta RM, Yardley JH. et al. Classification and grading of gastritis. The updated Sydney System. International Workshop on the Histopathology of Gastritis, Houston 1994. Am J Surg Pathol 1996; 20: 1161-1181
• 10 Kudo SE, Wakamura K, Ikehara N. et al. Diagnosis of colorectal lesions with a novel endocytoscopic classification - a pilot study. Endoscopy 2011; 43: 869-875
• 11 Abad MRA, Inoue H, Ikeda H. et al. Utilizing fourth-generation endocytoscopy and the 'enlarged nuclear sign' for in vivo diagnosis of early gastric cancer. Endosc Int Open 2019; 7: E1002-e1007
• 12 Noda H, Kaise M, Higuchi K. et al. Convolutional neural network-based system for endocytoscopic diagnosis of early gastric cancer. BMC Gastroenterol 2022; 22: 237
• 13 Kaise M, Kimura R, Nomura K. et al. Accuracy and concordance of endocytoscopic atypia for the diagnosis of gastric cancer. Endoscopy 2014; 46: 827-832
• 14 Tsurudome I, Miyahara R, Funasaka K. et al. In vivo histological diagnosis for gastric cancer using endocytoscopy. World J Gastroenterol 2017; 23: 6894-6901
• 15 Sato H, Inoue H, Ikeda H. et al. In vivo gastric mucosal histopathology using endocytoscopy. World J Gastroenterol 2015; 21: 5002-5008
• 16 Sato H, Inoue H, Hayee B. et al. In vivo histopathology using endocytoscopy for non-neoplastic changes in the gastric mucosa: a prospective pilot study (with video). Gastrointest Endosc 2015; 81: 875-881
• 17 Chiu PW, Ng EK, To KF. et al. Recognition of goblet cells upon endocytoscopy indicates the presence of gastric intestinal metaplasia. Dig Endosc 2014; 26: 52-56
• 18 Horiuchi Y, Hirasawa T, Ishizuka N. et al. Diagnostic performance in gastric cancer is higher using endocytoscopy with narrow-band imaging than using magnifying endoscopy with narrow-band imaging. Gastric Cancer 2021; 24: 417-427
• 19 Horiuchi Y, Hirasawa T, Ishizuka N. et al. Evaluation of microvascular patterns alone using endocytoscopy with narrow-band imaging for diagnosing gastric cancer. Digestion 2022; 103: 159-168
• 20 Dekker E, Houwen B, Puig I. et al. Curriculum for optical diagnosis training in Europe: European Society of Gastrointestinal Endoscopy (ESGE) Position Statement. Endoscopy 2020; 52: 899-923
• 21 Rodríguez-Carrasco M, Esposito G, Libânio D. et al. Image-enhanced endoscopy for gastric preneoplastic conditions and neoplastic lesions: a systematic review and meta-analysis. Endoscopy 2020; 52: 1048-1065
• 22 Iwagami H, Uedo N, Yip H-C. et al. Ultra-magnifying narrow-band imaging for endoscopic diagnosis of gastric intestinal metaplasia: a pilot image analysis study. Endosc Int Open 2021; 9: E522-E529
• 23 Horiuchi Y, Hirasawa T, Ishizuka N. et al. Diagnostic performance in gastric cancer is higher using endocytoscopy with narrow-band imaging than using magnifying endoscopy with narrow-band imaging. Gastric Cancer 2021; 24: 417-427
• 24 Yoo I, Park JC, Lee H. et al. A comparative study of magnifying endoscopy with narrow-band image and endocytoscopy in the diagnosis of gastric neoplasm: a pilot study. Eur J Gastroenterol Hepatol 2023; 35: 530-536

Publication History

Received: 06 November 2024

Accepted after revision: 28 May 2025

Accepted Manuscript online:
23 June 2025

Article published online:
24 July 2025