Introduction

Advances in endoscopic techniques have increased the detection of pharyngeal and esophageal squamous cell carcinoma (SCC) during upper gastrointestinal endoscopy. Iodine staining is a useful method for the early detection of pharyngeal and esophageal SCC. However, due to the irritant nature of Lugol’s solution, its use in routine screening examinations is limited. Moreover, detailed pharyngeal observation is still challenging due to the pharyngeal reflex and anatomical factors. In recent years, the application of artificial intelligence (AI) in the diagnosis and identification of cancer through endoscopy has progressed, but its use in cancer risk prediction remains limited. In this study, we developed an AI model to predict high-risk cases of pharyngeal and esophageal SCC by identifying multiple Lugol-voiding lesions (LVLs) in the esophagus using pharyngeal background mucosa images and evaluated its diagnostic performance.

Aims & Methods

Among cases that underwent esophageal iodine staining at Cancer Institute Hospital between April 2015 and March 2021, we excluded images of patients with a history of pharyngeal surgery and those with pharyngeal SCC. After this exclusion, we obtained a dataset of 22,069 pharyngeal images from 905 cases. A total of 10,074 images from 425 cases, including white light imaging (WLI) and narrow-band imaging (NBI), were used to train the AI model. Model accuracy was validated using 4,029 images from 182 cases. We applied a Triplet Network to learn feature space and suppress overfitting. Using pre-trained features, we constructed an LVL classification model based on Efficient Net. LVLs in the esophagus were graded as follows: A, no lesions; B, 1 to 10 lesions; and C, ≥10 lesions per endoscopic view. The AI model was then evaluated using 7,966 images from 298 test cases to assess its diagnostic performance for LVL-C. Furthermore, we calculated AI-predicted scores for LVL-C among the LVL-A/B/C groups, and performed a Cox proportional hazards analysis to assess the association between pharyngeal and esophageal SCC development and background factors, including AI-based prediction of LVL-C.

Results

Among 298 test cases, the median age was 60 years, and 201 cases (67%) were male. The distribution of cases among the LVL grade was as follows: LVL-A: 144 cases, LVL-B: 41 cases, and LVL-C: 113 cases. For the test cases, the accuracy/sensitivity/specificity of LVL-C prediction using WLI images was 71.5% / 62.5% / 78.3%, respectively. Using NBI images, these were 77.7% / 91.2% / 69.1%. When both WLI and NBI images were combined, accuracy was 78.2%, sensitivity was 89.4%, and specificity was 71.4%. The mean AI-based prediction score of LVL-C was 0.27 for LVL-A, 0.57 for LVL-B, and 0.61 for LVL-C. The median follow-up period for the test cases was 80 months, during which 38 cases developed pharyngeal SCC and 83 cases developed esophageal SCC. Cox proportional hazards analysis revealed that AI-predicted LVL-C was significantly associated with the risk of pharyngeal SCC (hazard ratio :12.00, 95% CI: 1.56-92.50, p=0.02). Similarly, AI-predicted LVL-C and current alcohol intake were significantly associated with the risk of esophageal SCC (HR: 4.20, 95% CI: 1.92-9.16, p<0.01 and HR ;1.62, 95%CI 1.00-2.63, p=0.04 respectively).

Conclusion

The AI-based prediction of LVL-C showed good performance with a sensitivity of 89.4%, and it may be useful for predicting the development of both pharyngeal and esophageal SCC.

Introduction

The presence of bubbles during esophagogastroduodenoscopy (EGD) is common and may impair adequate mucosal visualization (MV).

Aims & Methods

The primary objective was to evaluate the effect of pre-endoscopic simethicone administration on mucosal visibility, using the GRACE (Gastroscopy Rate of Cleanliness Evaluation) scale. Secondary objectives included assessment of endoscopy duration, endoscopist satisfaction, and the detection rate of endoscopic findings.
This was a prospective, single-center, randomized, blinded study including patients undergoing elective EGD between December 2024 and March 2025. Patients were randomized into three groups: A (100 mg simethicone ≤10 minutes before EGD), B (100 mg simethicone >10 minutes before EGD), and C (50 mL water, placebo). All received a total volume of 50 mL. Patients with gastroparesis, achalasia, or prior gastrointestinal surgery were excluded. MV was assessed using the GRACE scale (validated in 2024), which scores each segment (esophagus, stomach, duodenum) from 0 to 3 based on bubble quantity, resulting in a total score from 0 (inadequate) to 9 (excellent). Endoscopist satisfaction was subjectively assessed on a scale from 0 (not satisfied) to 10 (highly satisfied).

Results

A total of 82 patients were included (18 in group A, 34 in group B, and 30 in group C). Fifty-two patients received 100 mg of simethicone and 30 received placebo. Median simethicone administration time was 8 minutes in group A and 19 minutes in group B. Simethicone significantly improved MV compared to placebo (mean GRACE score: 7.62±1.7 vs 5.83±1.7; p<0.001), with significant improvements in all segments (esophagus: 2.65 vs 2.03, p<0.001; stomach: 2.21 vs 1.63, p=0.002; duodenum: 2.69 vs 2.17, p=0.001). A very weak negative correlation was found between GRACE score and simethicone timing (r = -0.058; p = 0.685). Median endoscopist satisfaction was higher in the simethicone group (8 vs 7; p=0.001), and strongly correlated with total GRACE score (r = 0.855; p<0.001). No significant differences were observed in endoscopy duration (7 vs 6 minutes; p=0.800) or diagnostic yield (p=0.416).

Conclusion

Pre-endoscopic administration of simethicone significantly improved mucosal visibility across all upper GI segments, supporting its routine use in clinical practice. The GRACE scale is a useful and objective tool that correlates strongly with subjective endoscopist satisfaction.

References

Esposito G, Dilaghi E, Costa-Santos C, Ligato I, Annibale B, Dinis-Ribeiro M, Areia M; GRACE Investigators. The Gastroscopy RAte of Cleanliness Evaluation (GRACE) Scale: an international reliability and validation study. Endoscopy. 2025 Apr;57(4):312-320. doi: 10.1055/a-2422-0856. Epub 2024 Sep 25. Erratum in: Endoscopy. 2025 Apr;57(4):C2. doi: 10.1055/a-2545-0228. PMID: 39321961
Manfredi G, Bertè R, Iiritano E, Alicante S, Londoni C, Brambilla G, Romeo S, Menozzi F, Griffanti P, Brandi G, Moreschi O, Pezzilli R, Zullo A, Buscarini E. Premedication with simethicone and N-acetylcysteine for improving mucosal visibility during upper gastrointestinal endoscopy in a Western population. Endosc Int Open. 2021 Feb;9(2):E190-E194. doi: 10.1055/a-1315-0114. Epub 2021 Jan 25. PMID: 33532557; PMCID: PMC7834924.

Introduction

Gastric cancer is one of the most common and lethal malignancies worldwide¹. A recognized independent risk factor for the development of this neoplasm is the presence of moderate-to-severe chronic atrophic gastritis (OLGA stage III-IV)^2,3.
In Italy regions with varying epidemiology for gastric cancer coexist. Notably, the Po Valley area, encompassing the provinces of Cremona and Mantua, is a high-risk region, with a standardized incidence rate of 32.2 cases per 100,000 men and 15.9 cases per 100,000 women⁴.

Aims & Methods

The present single-center retrospective observational study aims to evaluate the correlation between the incidence, over a period of approximately 10 years, of dysplastic and neoplastic lesions amenable to endoscopic resection and OLGA staging and other recognized risk factors for gastric cancer in this high-risk population.
We retrospectively analyzed all mucosal resections (EMR), submucosal dissections (ESD), and hybrid resections (Hybrid EMR/ESD) performed at our hospital between December 2014 and September 2024 for visible gastric lesions identified as low-grade dysplasia (LGD) or high-grade dysplasia (HGD) in pre-resection biopsies. For each patient, we collected demographic data, information on risk factors for gastric neoplasia, pre- and post-resection histology and details related to the endoscopic resection procedure. High Definition Chromo-Endoscopy (HD-CE) with at least 5 guided biopsies from antrum, incisura and corpus were performed concurrently with biopsy sampling of visible lesions and all specimens underwent to a second-look pathological examination.

Results

A total of 66 patients (36 males and 30 females) with a mean age of 74.7 years were included in the study. Pre-resection biopsies revealed low-grade dysplasia (LGD) in 62% of cases and high-grade dysplasia (HGD) in the remaining 38%. Forty-two patients underwent ESD, 14 EMR and 10 Hybrid EMR/ESD. However, in 3 patients the ESD procedure was discontinued due to massive intra-procedural bleeding, perforation and evidence of deep infiltration involving the muscularis propria. Examination of the resected specimens identified 29 lesions as LGD (44%), 17 as HGD (26%), 2 as intramucosal adenocarcinomas (3%), 14 as pT1a adenocarcinomas (21%), and 4 as pT1b adenocarcinomas (6%). Regarding risk factors: 9 patients (14%) had a first-degree family history of gastric neoplasia, while 16 patients (24%) were affected by Helicobacter pylori at the time the lesion was identified. The more interesting finding comes from the pre-resection OLGA staging: 4 patients had no gastric mucosal atrophy (OLGA 0), and 35 patients had early-stage chronic atrophic gastritis (15 OLGA I and 20 OLGA II). Among the remaining patients, 23 presented with OLGA stage III, and 4 with OLGA stage IV. Therefore, 59% of patients had an OLGA staging of II or lower at the time of the onset of the dysplastic or neoplastic lesion.

Conclusion

Our experience confirms what has already been observed in Asian countries: traditional risk factors for gastric neoplasia are insufficient in high-risk areas to identify the population that might benefit from short-term endoscopic follow-up.
These findings further support the statement included in the recently published update of the ESGE guidelines on the management of epithelial precancerous conditions and early neoplasia of the stomach (MAPS III), highlighting the need for periodic screening in high-risk populations regardless of additional risk factors⁵.

References

1. Smyth EC, Nilsson M, Grabsch HI, van Grieken NC, Lordick F. Gastric cancer. Lancet. 2020;396(10251):635-648. doi:10.1016/S0140-6736(20)31288-5
2. Rugge M, de Boni M, Pennelli G, et al. Gastritis OLGA-staging and gastric cancer risk: a twelve-year clinico-pathological follow-up study. Aliment Pharmacol Ther. 2010;31(10):1104-1111. doi:10.1111/j.1365-2036.2010.04277.x
3. Yue H, Shan L, Bin L. The significance of OLGA and OLGIM staging systems in the risk assessment of gastric cancer: a systematic review and meta-analysis. Gastric Cancer. 2018;21(4):579-587. doi:10.1007/s10120-018-0812-3
4. Val Padana Cancer Registry, ATS (Health Protection Agency) Val Padana, https://www.ats-valpadana.it/documents/1654672/64558781/Stomaco.pdf/3a88e15f-a32e-9344-a748-fdfa71f138ea
5. Dinis-Ribeiro M, Libânio D, Uchima H, et al. Management of epithelial precancerous conditions and early neoplasia of the stomach (MAPS III): European Society of Gastrointestinal Endoscopy (ESGE), European Helicobacter and Microbiota Study Group (EHMSG) and European Society of Pathology (ESP) Guideline update 2025. Endoscopy. 2025;57(5):504-554. doi:10.1055/a-2529-5025

Introduction

Inspection accuracy on colonoscopy is problematic because of the colon’s complex structure and the need for skillful operators. In particular, lesions in the ascending colon (AC) may be difficult to detect because of its high folds. Additionally, it has recently been reported that post-colonoscopy colorectal cancer is often located in the proximal colon. There are only a few reports on the effectiveness of retroflex inspection after forward view inspection in the AC during colonoscopy. While endoscopes and colonoscopy using artificial intelligence or image enhancement systems to detect lesions have been developed, it is important to reconsider basic endoscopic techniques, such as retroflex inspection, during colonoscopy.

Aims & Methods

We performed a prospective study to investigate the usefulness of retroflex inspection after forward view inspection in the AC during colonoscopy. Between August 2022 and December 2024, a single endoscopist enrolled and performed screening colonoscopies for 335 consecutive patients aged ≥40 years. Patients with inflammatory bowel disease, gastrointestinal polyposis, post-colorectal surgery, or poor preparation (≤1 point in at least one of three areas on the Boston Bowel Preparation Scale) were excluded from this study. The AC was inspected in this sequence: forward direction (Inspection 1), retroflex direction (Inspection 2), and forward direction (Inspection 3). The primary endpoint was the adenoma detection rate (ADR) in the AC. Secondary endpoints were the adenoma miss rate (AMR) in the AC per colonoscopy in Inspection 1 and Inspections 1+2, percentage of cases in which lesions were detected in Inspection 2, average observation time in the AC for each inspection, success rate of retroflexion in the AC, average pain score (0 points: no pain, 5 points: most severe pain), and complication rate during Inspection 2.

Results

We evaluated 300 patients with a median age of 66 years. The ADR in the AC for Inspection 1 vs. 1+2 was 18.7% vs. 28.7%, respectively (odds ratio: 1.749; 95% CI: 1.174–2.622; p = 0.005); the values for Inspection 1+2 vs. 1+2+3 were 28.7% vs. 30.0%, respectively (odds ratio: 1.066; 95% CI: 0.739–1.540; p = 0.788) (Table 1). The average AMR in the AC per colonoscopy for Inspection 1 vs. 1+2 was 12.6% (95% CI: 9.0%–16.2%) vs. 2.6% (95% CI: 0.8%–3.8%) (p < 0.001) (Table 1). The percentage of cases in which lesions were detected in Inspection 2 was 13.8% (41/298; total number of lesions: 47). The average diameter of lesions discovered and resected endoscopically during Inspection 2 was 6.1 mm (range: 3–20 mm). The locations of these lesions were as follows: proximal AC: 27.7% (13/47); mid-AC: 38.3% (18/47); and distal AC: 34.0% (16/47). The average observation times in the AC (Inspection 1 vs. 2 vs. 3) were 80.7 (95% CI: 77.8–83.6) vs. 59.1 (95% CI: 56.9–61.3) vs. 82.1 (95% CI: 79.4–84.8) seconds, respectively. The success rate of retroflexion in the AC was 99.3% (298/300), and the average pain score and complication rate during Inspection 2 were 0.5 points (95% CI: 0.4–0.6 points) and 0%, respectively.

Conclusion

Retroflex inspection following forward inspection in the AC during colonoscopy detects additional adenomas and is minimally invasive and safe. This technique contributes to improving the accuracy of colonoscopy.