Magnifying Endoscopy With Narrow-band Imaging Research Articles

A deep learning model based on magnifying endoscopy with narrow-band imaging to evaluate intestinal metaplasia grading and OLGIM staging: A multicenter study

Background and PurposePatients with stage III or IV of operative link for gastric intestinal metaplasia assessment (OLGIM) are at a higher risk of gastric cancer (GC). We aimed to construct a deep learning (DL) model based on magnifying endoscopy with narrow-band imaging (ME-NBI) to evaluate OLGIM staging. MethodsThis study included 4473 ME-NBI images obtained from 803 patients at three endoscopy centres. The endoscopic expert marked intestinal metaplasia (IM) regions on endoscopic images of the target biopsy sites. Faster Region-Convolutional Neural Network model was used to grade IM lesions and predict OLGIM staging. ResultsThe diagnostic performance of the model for IM grading in internal and external validation sets, as measured by the area under the curve (AUC), was 0.872 and 0.803, respectively. The accuracy of this model in predicting the high-risk stage of OLGIM was 84.0%, which was not statistically different from that of three junior (71.3%, p = 0.148) and three senior endoscopists (75.3%, p = 0.317) specially trained in endoscopic images corresponding to pathological IM grade, but higher than that of three untrained junior endoscopists (64.0%, p = 0.023). ConclusionThis DL model can assist endoscopists in predicting OLGIM staging using ME-NBI without biopsy, thereby facilitating screening high-risk patients for GC.

P643 Magnifying Endoscopy with Narrow Band Imaging for Graft Failure and Disease Recurrence in Patients with Crohn’s Disease after Intestinal Transplantation

Abstract Background Crohn’s disease (CD) is one of the common causes of short bowel syndrome (SBS) and intestinal failure (IF). Intestinal transplantation (IT) is sometimes needed for patients with CD who develop IF after multiple intestinal resections due to CD-related complications, such as uncontrollable bleeding and penetrating diseases. However, there have been limited case reports concerning the endoscopic surveillance of CD patients after IT. Methods We retrospectively investigated two patients with CD who underwent IT because of SBS-IF after multiple intestinal resection. We administered post-transplantation immunosuppressants and conducted regular magnifying endoscopy with narrow-band imaging (ME-NBI) via ileostomy chimney for surveillance of grafte rejection. Endoscopic prediction of rejection were based on our previously published "VENCH" score which is composed of five parameters: "V" (villi appearance), "E" (erythema), "N" (capillary network), "C" (crypt widening), and "H" (heterogeneity). Endoscopic biopsy were taken for histologic confirmation of graft rejection severity. Results Case 1 was a 60-year-old male patient with CD who received three times of intestinal resection for small bowel stricture and perforation, followed by IT due to SBS-IF. He lived uneventful until seven years thereafter with bloody diarrhea presented. Endoscopic surveillance revealed normal VENCH score of background mucosa but deep longitudinal ulcers which was indicative of recurrence of CD rather than graft rejection. Anti-tumor necrosis factor alpha with adalimumab was prescribed and his diarrhea and graft ulcers improved. Case 2 was a 39-year-old male patient with CD who received extensive intestinal resection due to multiple perforations, followed by IT because of SBS-IF. Endoscopic surveillance revealed normal VENCH score without histologic graft rejection two years thereafter. Both enrolled cases demonstrated favorable outcomes after endoscopic surveillance of intestinal graft with ME-NBI and management accordingly. Conclusion The post IT follow-up of CD patients remains challenging. We highlight the usefulness of the ME-NBI using VENCH score for early detection of graft rejection and its value in monitoring CD patients after IT. Additionally, ulcerations without changes in morphology of surrounding villi should take into consideration the disease recurrence of underlying CD. Further research is needed in endoscopic follow-up and molecular genetics to better understand CD after IT.

Gastric Juvenile Polyposis with Intramucosal Cancer Diagnosed by Magnifying Endoscopy with Narrow-band Imaging: A Case Report.

Although gastric juvenile polyposis (GJP) often coexists with gastric cancer, a preoperative accurate diagnosis is still difficult to obtain. A 70-year-old woman was referred for epigastralgia and anemia. Esophagogastroduodenoscopy with a conventional endoscope showed numerous gastric polyps with no cancerous findings. Magnifying endoscopy with narrow-band imaging (M-NBI) showed cancerous findings, and a target biopsy revealed adenocarcinoma. Histopathological findings after endoscopic resection confirmed a diagnosis of juvenile polyposis with intramucosal adenocarcinoma. Genetic analyses revealed a germline pathogenic variant of SMAD4. A target biopsy using M-NBI and endoscopic resection proved useful for confirming the preoperative diagnosis of coexisting cancerous lesions in GJP.

341. THE PREDICTION STRATEGY OF SUBMUCOSAL INVASION IN SUPERFICIAL BARRETT’S ADENOCARCINOMA BY THE COMBINATION OF HISTOLOGY IN BIOPSY AND MAGNIFYING ENDOSCOPY

Abstract Background Endoscopically predicting the depth of invasion of superficial Barrett’s esophageal adenocarcinoma (s-BEA) is important for choosing an appropriate treatment. This study aimed to evaluate the endoscopic and histopathological characteristics to diagnose s-BEA with submucosal invasion. Methods We retrospectively reviewed 105 tumors from 97 patients endoscopically and pathologically diagnosed with s-BEA in short-segment Barrett’s esophagus (SSBE) (60 lesions in 60 patients) and long-segment Barrett’s esophagus (LSBE) (45 lesions in 37 patients). One hundred five s-BEA lesions were classified into 60 intramucosal (pT1a) tumors and 45 submucosal (pT1b) tumors. Clinicopathological and endoscopic characteristics, including the histology in the biopsy specimens, were analyzed with comparison in pT1a vs pT1b tumors. The histology in the biopsy specimens was classified into two types such as the presence or absence of poorly differentiated component (PDC). Risk factors for submucosal invasion identified using univariate analysis were entered into a multivariate logistic regression analysis. Subsequently, the findings of magnifying endoscopy with narrow-band imaging (ME-NBI) were assessed to determine whether the poorly differentiated component (PDC) in the resected specimens can be identified or not. Results pT1b tumors showed significantly higher frequency than pT1a tumors in the following features: PDC in the biopsy specimens (21/45, 46.7%; vs 1/60, 1.7%; OR = 24.8), complex macroscopic type (33/45, 73.3%; vs 15/60, 25.0%; OR = 4.47), and tumor diameter > 20 mm (31/45, 68.9%; vs 13/60, 21.7%; OR = 4.40). In 87 tumors which were able to be evaluated for endoscopic findings with ME-NBI, 9 out of 11 tumors with corkscrew pattern or wavy micro-vessels showed PDC in the resected specimens. The sensitivity, specificity, and positive predicted value of these irregular vessels for predicting PDC in the resected specimens were 40.9%, 96.9%, and 81.8%, respectively. Conclusion PDC in the biopsy specimens, tumor diameter > 20 mm, and complex macroscopic type were independent predictors for submucosal invasion of BEA. The presence of PDC in the biopsy specimens was an especially useful predictor of T1b cancer with the highest odds. Further, ME-NBI can be useful to diagnose the presence of PDC in BEA.

Lightweight deep learning model incorporating an attention mechanism and feature fusion for automatic classification of gastric lesions in gastroscopic images.

Accurate diagnosis of various lesions in the formation stage of gastric cancer is an important problem for doctors. Automatic diagnosis tools based on deep learning can help doctors improve the accuracy of gastric lesion diagnosis. Most of the existing deep learning-based methods have been used to detect a limited number of lesions in the formation stage of gastric cancer, and the classification accuracy needs to be improved. To this end, this study proposed an attention mechanism feature fusion deep learning model with only 14 million (M) parameters. Based on that model, the automatic classification of a wide range of lesions covering the stage of gastric cancer formation was investigated, including non-neoplasm(including gastritis and intestinal metaplasia), low-grade intraepithelial neoplasia, and early gastric cancer (including high-grade intraepithelial neoplasia and early gastric cancer). 4455 magnification endoscopy with narrow-band imaging(ME-NBI) images from 1188 patients were collected to train and test the proposed method. The results of the test dataset showed that compared with the advanced gastric lesions classification method with the best performance (overall accuracy = 94.3%, parameters = 23.9 M), the proposed method achieved both higher overall accuracy and a relatively lightweight model (overall accuracy =95.6%, parameter = 14 M). The accuracy, sensitivity, and specificity of low-grade intraepithelial neoplasia were 94.5%, 93.0%, and 96.5%, respectively, achieving state-of-the-art classification performance. In conclusion, our method has demonstrated its potential in diagnosing various lesions at the stage of gastric cancer formation.

The Pink Zone Pattern (PP) sign: A novel simple marker for early gastric cancer.

We previously found a pink-colored change in early gastric cancer (EGC) lesions under magnifying endoscopy with narrow-band imaging (ME-NBI) and named it the "Pink Zoon Pattern" (PP) sign, which appeared independent of microvascular and microstructural changes. The aim of this study was to further investigate the characteristics of the PP sign in EGC. The consecutive patients with suspicious gastric lesions detected by ME-NBI and confirmed by pathology at Zhejiang Cancer Hospital between November 2020 and December 2021 were enrolled in the study. The suspicious lesions were observed and assessed by the "VS" system and the PP sign respectively. We found that in the PP-positive group, 238 lesions (96.0%) were diagnosed as malignant. The overall accuracy, sensitivity, and specificity were 84.7%, 85.3%, and 81.8%. Among 164 EGC lesions diagnosed with low confidence (Grades 2, 3, and 4) using the VS system, the overall accuracy of PP to discriminate tumor from normal was 82.3%. The sensitivity and specificity were 82.7% and 81.5% respectively. The PP sign could be a new simple sign for the diagnosis of EGC and as an effective supplement to VS system when using ME-NBI.

Type B2 vessels and infiltrative growth patterns b and c are associated with lymphatic invasion in pT1a-lamina propria mucosa esophageal squamous cell carcinoma.

Tumor growth pattern correlates with outcomes in patients with esophageal squamous cell carcinoma (ESCC), however, the clinical significance of the tumor growth pattern in pT1a-lamina propria mucosa (LPM) type of ESCC was unclear. This study was conducted to clarify clinicopathological features of tumor growth patterns in pT1a-LPM type ESCC and the relationship between tumor growth patterns and magnifying endoscopic findings. Eighty-seven lesions diagnosed as pT1a-LPM ESCC were included. Clinicopathological findings including tumor growth pattern and narrow band imaging with magnifying endoscopy (NBI-ME) in the LPM area were investigated. Eighty-seven lesions were classified as infiltrative growth pattern-a (INF-a): expansive growth (n = 81), INF-b: intermediate growth (n = 4) and INF-c: infiltrative growth pattern (n = 2). Lymphatic invasion was shown in one INF-b and one INF-c lesion. NBI-ME and histopathological images were matched for 30 lesions. The microvascular pattern was classified into types B1 (n = 23) and B2 (n = 7) using the JES classification. All 23 type B1 lesions were classified as INF-a without lymphatic invasion. Type B2 lesions were classified as INF-a (n = 2), INF-b (n = 4) and INF-c (n = 1), and lymphatic invasion was present in two lesions (INF-b and INF-c). The rate of lymphatic invasion was significantly higher in type B2 than type B1 (p = 0.048). The tumor growth pattern of pT1a-LPM ESCC was mostly INF-a in type B1 patterns. Type B2 patterns are rarely present in pT1a-LPM ESCC, however lymphatic invasion with INF-b or INF-c was frequently observed. Careful observation before endoscopic resection with NBI-ME is important to identify B2 patterns to predict histopathology.

Magnifying Endoscopy with Narrow-Band Imaging for Duodenal Neuroendocrine Tumors.

Duodenal neuroendocrine tumors (NETs) are rare subepithelial tumors that arise from the neuroendocrine cells beneath the epithelial layer. However, an accurate histopathological diagnosis is difficult when tissue samples are obtained using conventional endoscopic forceps biopsy alone. This study aimed to evaluate the magnifying endoscopy with narrow-band imaging (ME-NBI) findings of duodenal NETs. We retrospectively analyzed a database of 22 duodenal NETs from 21 patients who underwent ME-NBI between January 2011 and June 2022. The ME-NBI, endosonographic, and histopathologic findings of duodenal NETs were analyzed. Nineteen lesions were located in the bulb, two were located in the superior duodenal angle, and one was located in the second portion of the duodenum. Eighteen lesions (82%) had IIa morphology, and nine (41%) had central depression on the surface. On endoscopic ultrasonography, almost all lesions (20/22, 91%) were located in the second and/or third layers, and the median tumor size was 6 mm. During ME-NBI, the microsurface pattern was regular in 18 lesions (82%) and absent in 4 (18%). The microvascular pattern was regular in 17 lesions (77%), irregular in 4 (18%), and absent in 1 (5%). Thickened subepithelial vessels were observed in 15 (68%) lesions. There was no difference in tumor size according to the presence or absence of thickened subepithelial vessels (6.1 ± 1.8 mm vs. 5.9 ± 3.8 mm, p = 0.860). In conclusion, the characteristic ME-NBI findings of duodenal NETs were regular microsurface and microvascular patterns and the presence of thickened subepithelial vessels. These ME-NBI features may be useful for differentiating duodenal NETs from other duodenal subepithelial lesions.

Utility of image-enhanced magnifying endoscopy and traction wire for a successful endoscopic submucosal dissection in early gastric cancer.

Characterizing tumor extent in early gastric cancer (EGC) is difficult and magnifying endoscopy with narrow-band imaging (M-NBI) is often necessary [1] [2]. Submucosal dissection of large lesions via endoscopic submucosal dissection (ESD) is a time-consuming task. Here, we report the benefit of the traction wire (ERD-TW35; Medtronic, Minneapolis, Minnesota, USA) [3] [4] [5] as a unique technique for ESD.

A comparative study of magnifying endoscopy with narrow-band image and endocytoscopy in the diagnosis of gastric neoplasm: a pilot study.

Endoscopic technologies have recently advanced to optimize the detection and diagnosis of gastric lesions. Endocytoscopy aids in the virtual realization of histology. Herein, we aimed to investigate gastric lesions using single-stain endocytoscopy and compare them using magnifying endoscopy with narrow-band imaging (ME-NBI) in terms of diagnostic yield in vivo. In the present prospective study, we registered 24 patients with gastric neoplasms and retrospectively reviewed their images. Three endoscopists reviewed the images of gastric neoplasms using white light, ME-NBI, and endocytoscopy. The diagnostic yield of endocytoscopy in early gastric cancer (EGC) was assessed using histopathology as the gold standard. Endocytoscopy was performed in 24 patients with gastric neoplasms. Of these, 15 patients had adenocarcinomas, while nine patients had low-grade dysplasia. The sensitivity, specificity, and accuracy of endocytoscopy for EGC detection were reported as 80.0% [95% confidence interval (CI), 51.9-95.7], 66.7% (95% CI, 58.4-91.9), and 75.0% (95% CI, 53.3-90.2) by endoscopist A; 80.0% (95% CI, 51.9-95.7), 44.4% (95% CI, 13.7-78.8), and 66.7% (95% CI, 44.7-84.4) by endoscopist B; and 93.3% (95% CI, 68.1-99.8), 55.6% (95% CI, 21.2-86.3), and 79.2% (95% CI, 57.9-92.8) by endoscopist C; these findings were not inferior to NBI. The inter-observer agreement, κ statistic = 0.67 (95% CI, 0.43-0.90) was favorable. Endocytoscopy aid in the diagnosis of EGC because of its better sensitivity and accuracy compared to NBI or white-light imaging. However, further large-scale studies are required to confirm our findings.

Comparison of clinical and pathological features between early-stage gastric-type and intestinal-type differentiated adenocarcinoma: a retrospective study

BackgroundThe clinicopathological features and endoscopic characteristics under magnifying endoscopy with narrow band imaging (ME-NBI) between early-stage gastric-type differentiated adenocarcinoma (GDA) and intestinal-type differentiated adenocarcinoma (IDA) remain controversial.MethodsEarly gastric adenocarcinomas that underwent endoscopic submucosal dissection (ESD) in Nanjing Drum Tower Hospital between August 2017 and August 2021 were included in the present study. GDA cases and IDA cases were selected based on morphology and immunohistochemistry staining of CD10, MUC2, MUC5AC, and MUC6. Clinicopathological data and endoscopic findings in ME-NBI were compared between GDAs and IDAs.ResultsThe mucin phenotypes of 657 gastric cancers were gastric (n = 307), intestinal (n = 109), mixed (n = 181) and unclassified (n = 60). No significant difference was observed in terms of gender, age, tumor size, gross type, tumor location, background mucosa, lymphatic invasion, and vascular invasion between patients with GDA and IDA. GDA cases were associated with deeper invasion than IDA cases (p = 0.007). In ME-NBI, GDAs were more likely to exhibit an intralobular loop patten, whereas IDAs were more likely to exhibit a fine network pattern. In addition, the proportion of none-curative resection in GDAs was significantly higher than that in IDAs (p = 0.007).ConclusionThe mucin phenotype of differentiated early gastric adenocarcinoma has clinical significance. GDA was associated with less endoscopically resectability than IDA.

AI-assisted identification of intrapapillary capillary loops in magnification endoscopy for diagnosing early-stage esophageal squamous cell carcinoma: a preliminary study.

Esophageal squamous cell carcinoma (ESCC) is one of the most common histological types of esophageal cancers. It can seriously affect public health, particularly in Eastern Asia. Early diagnosis and effective therapy of ESCC can significantly help improve patient prognoses. The visualization of intrapapillary capillary loops (IPCLs) under magnification endoscopy (ME) can greatly support the identification of ESCC occurrences by endoscopists. This paper proposes an artificial-intelligence-assisted endoscopic diagnosis approach using deep learning for localizing and identifying IPCLs to diagnose early-stage ESCC. An improved Faster region-based convolutional network (R-CNN) with a polarized self-attention (PSA)-HRNetV2p backbone was employed to automatically detect IPCLs in ME images. In our study, 2887 ME with blue laser imaging (ME-BLI) images of 246 patients and 493 ME with narrow-band imaging (ME-NBI) images of 81 patients were collected from multiple hospitals and used to train and test our detection model. The ME-NBI images were used as the external testing set to verify the generalizability of the model. The experimental evaluation revealed that the proposed method achieved a recall of 79.25%, precision of 75.54%, F1-score of 0.764 and mean average precision (mAP) of 74.95%. Our method outperformed other existing approaches in our evaluation. It can effectively improve the accuracy of ESCC detection and provide a useful adjunct to the assessment of early-stage ESCC for endoscopists.

The usefulness of image-enhanced endoscopy to distinguish gastric carcinoma in tumors initially diagnosed as adenomas by endoscopic biopsy: A retrospective study.

Superficial epithelial gastric neoplasms can be divided into adenomas and early carcinomas. Histological diagnosis by endoscopic forceps biopsy is crucial for the diagnosis and management of gastric neoplasms. It is difficult to distinguish features of gastric neoplasms in small biopsy specimens; hence, gastric carcinomas can be underdiagnosed as adenomas. Recent developments in image-enhanced endoscopy have improved the ability to differentiate between carcinomatous and non-carcinomatous lesions. To investigate the prevalence of gastric carcinoma in lesions initially diagnosed as adenomas by forceps biopsy and assess the usefulness of image-enhanced endoscopy in distinguishing carcinomas. A total of 142 lesions of gastric adenomas, diagnosed by biopsy and resected endoscopically between January 2010 and May 2020, were retrospectively evaluated. Images were captured by white-light endoscopy (WLE), magnifying endoscopy with narrow-band imaging (M-NBI), and magnifying endoscopy with acetic acid and narrow-band imaging (M-AANBI); they were analyzed and compared with histopathological results. The diagnostic performance of M-AANBI was compared with that of M-NBI. Of the 142 lesions, 58 (40.8%) were pathologically diagnosed as adenocarcinomas. On WLE images, a depressed macroscopic type and size ≥20 mm were significant predictors of carcinoma (P < .001); however, they displayed low sensitivities (32.8% and 41.4%, respectively). M-AANBI displayed significantly higher sensitivity, specificity, and accuracy for distinguishing carcinomas than M-NBI (94.8% vs 74.1%, 81.0% vs 72.6%, and 86.6% vs 73.2%, P < .05). In conclusion, carcinoma was prevalent in 40.8% of gastric lesions initially diagnosed as adenomas by forceps biopsy. M-AANBI may be more useful than M-NBI and WLE in distinguishing gastric carcinomas from adenomas.

Diagnosis and segmentation effect of the ME-NBI-based deep learning model on gastric neoplasms in patients with suspected superficial lesions - a multicenter study.

Endoscopically visible gastric neoplastic lesions (GNLs), including early gastric cancer and intraepithelial neoplasia, should be accurately diagnosed and promptly treated. However, a high rate of missed diagnosis of GNLs contributes to the potential risk of the progression of gastric cancer. The aim of this study was to develop a deep learning-based computer-aided diagnosis (CAD) system for the diagnosis and segmentation of GNLs under magnifying endoscopy with narrow-band imaging (ME-NBI) in patients with suspected superficial lesions. ME-NBI images of patients with GNLs in two centers were retrospectively analysed. Two convolutional neural network (CNN) modules were developed and trained on these images. CNN1 was trained to diagnose GNLs, and CNN2 was trained for segmentation. An additional internal test set and an external test set from another center were used to evaluate the diagnosis and segmentation performance. CNN1 showed a diagnostic performance with an accuracy, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of 90.8%, 92.5%, 89.0%, 89.4% and 92.2%, respectively, and an area under the curve (AUC) of 0.928 in the internal test set. With CNN1 assistance, all endoscopists had a higher accuracy than for an independent diagnosis. The average intersection over union (IOU) between CNN2 and the ground truth was 0.5837, with a precision, recall and the Dice coefficient of 0.776, 0.983 and 0.867, respectively. This CAD system can be used as an auxiliary tool to diagnose and segment GNLs, assisting endoscopists in more accurately diagnosing GNLs and delineating their extent to improve the positive rate of lesion biopsy and ensure the integrity of endoscopic resection.

Effect of adding acetic acid when performing magnifying endoscopy with narrow band imaging for diagnosis of Barrett's esophageal adenocarcinoma.

Background and study aims  Magnifying endoscopy with narrow band imaging (M-NBI) was developed to diagnose Barrett's esophageal adenocarcinoma (BEA); however, this method remains challenging for inexperienced endoscopists. We aimed to evaluate a modified M-NBI technique that included spraying acetic acid (M-AANBI). Patients and methods  Eight endoscopists retrospectively examined 456 endoscopic images obtained from 28 patients with 29 endoscopically resected BEA lesions using three validation schemes: Validation 1 (260 images), wherein the diagnostic performances of M-NBI and M-AANBI were compared - the dataset included 65 images each of BEA and non-neoplastic Barrett's esophagus (NNBE) obtained using each modality; validation 2 (112 images), wherein 56 pairs of M-NBI and M-AANBI images were prepared from the same BEA and NNBE lesions, and diagnoses derived using M-NBI alone were compared to those obtained using both M-NBI and M-AANBI; and validation 3 (84 images), wherein the ease of identifying the BEA demarcation line (DL) was scored via a visual analog scale in 28 patients using magnifying endoscopy with white-light imaging (M-WLI), M-NBI, and M-AANBI. Results  For validation 1, M-AANBI was superior to M-NBI in terms of sensitivity (90.8 % vs. 64.6 %), specificity (98.5 % vs. 76.9 %), and accuracy (94.6 % vs. 70.4 %) (all P  < 0.05). For validation 2, the accuracy of M-NBI alone was significantly improved when combined with M-AANBI (from 70.5 % to 89.3 %; P  < 0.05). For validation 3, M-AANBI had the highest mean score for ease of DL recognition (8.75) compared to M-WLI (3.63) and M-NBI (6.25) (all P <  0.001). Conclusions  Using M-AANBI might improve the accuracy of BEA diagnosis.

Determining Invasion Depth in Superficial Pharyngeal Carcinoma by Transoral Ultrasonography.

To examine the clinical usefulness of transoral ultrasonography (US) in determining the invasion depth of superficial pharyngeal carcinoma (SPC). Determining the invasion depth of SPC is crucial for transoral surgery including determining treatment strategy. This study aimed to examine the usefulness of transoral US in determining the invasion depth of SPC. Forty-six patients with 51 lesions who underwent both magnifying endoscopy with narrow-band imaging (ME-NBI) and transoral US were included. The primary outcomes were the sensitivity, specificity, positive (PPV), and negative predictive values (NPV) of ME-NBI and transoral US findings for pathological tumor depth in SPCs. The accuracy (82.4%), sensitivity (85.2%), PPV (82.1%), and NPV (82.6%) rates of US for subepithelial propria (SEP) were higher than those of ME-NBI and macroscopic classification, indicating that transoral US is superior to ME-NBI in determining the invasion depth. All cases where the SEP was clearly invaded (SEP deep) could be diagnosed as SEP by transoral US. Transoral US may be useful in determining the invasion depth of SPCs. 3 Laryngoscope, 133:2192-2197, 2023.

Magnifying endoscopy with narrow-band imaging for diagnosis of subtype of gastric intestinal metaplasia.

Patients with incomplete gastric intestinal metaplasia (GIM) have a higher risk of gastric cancer (GC) than those with complete GIM. We aimed to clarify whether micromucosal patterns of GIM in magnifying endoscopy with narrow-band imaging (M-NBI) were useful for diagnosis of incomplete GIM. We enrolled patients with a history of endoscopic resection of GC or detailed inspection for suspicious or definite GC. The antrum greater curvature and corpus lesser curvature were regions of interest. Areas with endoscopic findings of light blue crest and/or white opaque substance (WOS) were defined as endoscopic GIM, and subsequent M-NBI was applied. Micromucosal patterns were classified into Foveola and Groove types, and targeted biopsies were performed on GIM with each pattern. GIM was classified into complete and incomplete types using mucin (MUC)2, MUC5AC, MUC6, and CD10 immunohistochemical staining. The primary endpoint was the association between micromucosal pattern and histological subtype. The secondary endpoint was endoscopic findings associated with incomplete GIM. We analyzed 98 patients with 156 GIMs. Univariate analysis (odds ratio [OR] 3.4, P=0.004), but not multivariate analysis (OR 0.87, P=0.822), demonstrated a significant association between micromucosal pattern and subtype. The antrum (OR 3.7, P=0.006) and WOS (OR 43, P=0.002) were independent predictors for incomplete GIM. The WOS had 69% sensitivity and 93% specificity. The M-NBI micromucosal pattern is not useful for diagnosis of GIM subtype. WOS is a promising endoscopic indicator for diagnosis of incomplete GIM. (UMIN-CTR000041119).

Endoscopic Features of Gastric Epithelial Neoplasm of Fundic Gland Mucosa Lineage

The endoscopic features of gastric epithelial neoplasms of fundic gland mucosa lineage (GEN-FGML) have not been well investigated. We aimed to clarify the endoscopic features of GEN-FGML and differences between gastric adenocarcinoma of the fundic gland type (GA-FG) and fundic gland mucosa type (GA-FGM). A total of 62 GEN-FGML lesions, including 52 GA-FG and 10 GA-FGM, were retrospectively analyzed using endoscopic and clinicopathological findings to provide information of diagnostic value using white light imaging (WLI) and magnifying endoscopy with narrow-band imaging (M-NBI). GA-FG frequently presented with a whitish, submucosal tumor (SMT) shape with dilated vessels with branching architecture and background mucosa without atrophic change in WLI, an indistinct demarcation line (DL), dilatation of the crypt opening and intervening part (IP), and microvessels without distinct irregularity in M-NBI. GA-FGM frequently presented as a reddish, elevated lesion in WLI, with a distinct DL, dilatation of the IP, and an irregular microvascular pattern in M-NBI. As for an M-NBI diagnosis, five GA-FGM lesions met the diagnostic criteria for cancer, whereas none of the GA-FG lesions met the same criteria. We highlight the endoscopic features of GEN-FGML, and the differentiation between GA-FG and GA-FGM might be possible by combination of lesion color and morphology in WLI and M-NBI diagnoses.

ClusterNet: a clustering distributed prior embedded detection network for early-stage esophageal squamous cell carcinoma diagnosis.

Early and accurate diagnosis of esophageal squamous cell carcinoma (ESCC) is important for reducing mortality. Analyzing intrapapillary capillary loops' (IPCLs) patterns on magnification endoscopy with narrow band imaging (ME-NBI) has been demonstrated effective in the diagnosis of early-stage ESCC. However, even experienced endoscopists may face difficulty in finding and classifying countless IPCLs onME-NBI. We propose a novel clustering prior embedded detection network: ClusterNet. ClusterNet is capable of analyzing the distribution of IPCLs on ME-NBI automatically and enables endoscopists to overview multiple types of visualization. With ClusterNet assisting, endoscopists may observe ME-NBI images more efficiently, thus they may also predict the pathology and make medical decisions more easily. We propose the first large-scale ME-NBI dataset with fine-grained annotations by consensus of expert endoscopists. The dataset is splitted into a training set and an independent testing set based on patients. With two strategies for embedding, ClusterNet can automatically take the clustering effect into consideration. Prior to this work, none of the existing approaches take the clustering effect, which is rather important in classifying the IPCLs, into account. ClusterNet achieves an average precision of 81.2% and an average recall of 90.0% for the detection of IPCLs patterns on each patient of the independent testing set. We also compare ClusterNet with other state-of-the-art detection approaches. The performance of ClusterNet with embedding strategies is consistently superior to that of other approaches in terms of average precision, recall andF2-Score. Experiments demonstrate that our proposed method is able to detect almost all the IPCLs patterns on ME-NBI and classify them according to the Japanese Endoscopic Society (JES) classificationaccurately.

Automatic classification of gastric lesions in gastroscopic images using a lightweight deep learning model with attention mechanism and cost-sensitive learning

Magnification endoscopy with narrow-band imaging (ME-NBI) technology is widely used in the early diagnosis of precancerous lesions and gastric cancer, which is critical to reducing the incidence of gastric cancer and improving the survival rate of gastric cancer patients. The diagnosis based on ME-NBI image is mostly in manual way in clinics. In order to improve its objectivity and efficiency, here we proposed a lightweight attention mechanism deep learning model to automatically classify ME-NBI images for artificial intelligence (AI) diagnosis of early gastric cancer, low-grade intraepithelial neoplasia, and non-neoplasm. We collected 4,098 images from 990 patients for model training and validation and evaluated the performance of our model by comparisons with that of other models, such as the benchmark model. An additional 587 images from 296 patients were collected as an independent test set to further evaluate our method’s performance. The validation set showed that the overall accuracy, recall, precision, F1 score, and the area under the curve of our method were higher than those of other methods. The independent test set showed that our method achieved state-of-the-art classification for low-grade intraepithelial neoplasia (accuracy = 93.9%, sensitivity = 92.6%). Our method displayed the advantages of lightweight and high effectiveness in classifying effectiveness, which is the potential for AI diagnosis of early gastric cancer, low-grade intraepithelial neoplasia, and non-neoplasm.