Introduction

The quality of colonic preparation remains a critical benchmark in ensuring the efficacy of colonoscopy procedures (1). Despite concerted efforts and elaborate protocols, inadequate preparation persists in up to 20% of colonoscopies (2). To address this challenge, there is a need for innovative solutions to assess bowel preparation providing more reliable and standardized results before the procedure.

Aims & Methods

The aim of this study is to assess and compare the predictive accuracy of models predicated on artificial neural network frameworks with those established via traditional logistic regression analysis, to predict the efficacy of the bowel preparation using bowl images characteristics. This prospective study included 512 bowl images from 100 patients taken after the second dose of a split-dose preparation (using high volume while adhering to a three day low fiber diet), between April 2023 and October 2023 in Hotel Dieu de France, a tertiary hospital center in Beirut. The Boston bowel preparation scale (BBPS) was divided into two categories: Adequate (≥ 8) and Inadequate (< 8). Manual cropping techniques were employed to refine accuracy and extract images’ features using snipping tool. Then, we specified the corresponding mean scores of the image in RGB (Red, Green, Blue) and HSV (Hue, Saturation, Value). Two statistical analysis strategies were considered: the artificial neural network analysis, and the logistic regression analysis.

Results

The artificial neural network for predicting the BBPS from the bowl image characteristics showed an AUROC= 0.797, and an accuracy of 90.4% for the training group (forming 68.2% of the sample), and 92.8% for the testing group (representing 31.8% of the sample). For the traditional statistical analysis, a threshold was defined for each image feature using the area under the ROC curve (AUC). Each feature was categorized into four quartiles. The binary logistic regression showed a significant correlation between blue (threshold= 67) and green colors (threshold= 98), the saturation (threshold= 0.63064), the first quartile of the hue (threshold ≤ 0.073), and the last quartile of the value (threshold ≥ 0.527). Then, a binary logistic regression formula was established: -1.23 * First quartile of the hue (threshold ≤0.073) - 1.7 * Last quartile of the Value (threshold ≥ 0.527) - 1.321 * Saturation (threshold= 0.63064) + 0.688 * Blue (threshold= 67) + 1.08 * Green (threshold=98) + 2.932. The AUROC of the new formula was equal to 0.787 and revealed a threshold= 1.55. The classification based on the new formula (BBPS as dependent factor) showed an accuracy of 88.1%.

Conclusion

The characteristics of the bowl images serve as predictive indicators of colonic preparation quality. A model utilizing artificial neural network architecture yields higher accuracy in prediction than one derived from conventional statistical methods. This method may allow the adjustment of the preparation strategy before undergoing colonoscopy to improve the outcome of the procedure.

References

1. Rex DK. Key quality indicators in colonoscopy. Gastroenterol Rep (Oxf). 2023 Mar 10;11:goad009. doi: 10.1093/gastro/goad009. PMID: 36911141; PMCID: PMC10005623.
2. van Riswijk MLM, van Keulen KE, Neumann H, Siersema PD. An intraprocedural bowel cleansing system for difficult-to-prepare patients-A multicenter prospective feasibility study. United European Gastroenterol J. 2024 Feb;12(1):56-65. doi: 10.1002/ueg2.12501. Epub 2024 Jan 11. PMID: 38213079; PMCID: PMC10859716.

Introduction

Gastric cancer ranks as the fifth most common cancer and the second leading cause of cancer-related deaths. Multifocal atrophic gastritis and gastric intestinal metaplasia (GIM) often coexist and are both associated with the risk of developing gastric cancer, with GIM being the first irreversible mucosal transformation to cancer. Endoscopic surveillance was associated with earlier cancer detection and improved survival. This study aims to evaluate the prevalence, epidemiological characteristics, and clinical features of patients with GIM.

Aims & Methods

This is a single-center, retrospective study from January 1, 2022 to December 31, 2023. Including patients aged ≥18 years who underwent esophagogastroduodenoscopy (EGD) with gastric biopsies confirming GIM, All identified cases of GIM were reviewed by gastrointestinal pathologists to confirm the diagnosis. Th primary outcome was to assess the prevalence, epidemiological characteristic, and clinical features of GIM. The secondary outcome was to evaluate adherence to recommended surveillance protocols for GIM.

Results

Among 3,663 patients who underwent EGD, GIM was identified in (4.17%, n=153). The mean age of patients with GIM was 60 years, Males represented more than half of patients (54.2%), the majority were classified as overweight (Body mass index (BMI) 25–29.9; 64%, n=97) or obese (BMI ≥30; 24.3%, n=37).Smoking was reported in (17.6%, n=27), while (5.2%, n=8) reported alcohol consumption. Moreover, (7.2%, n=11) had a history of prior gastric surgeries. Dyspepsia (33.3%) was the most common indication for EGD. Gastritis was the predominant endoscopic finding and the primary reason for performing gastric biopsies (73%, n=112 ).The most common GIM subtype was complete (52%), followed by incomplete (37%) and mixed (11%). In terms of GIM location, extensive GIM was observed in (28.7%), localized GIM in (17.6%), while the location remained unknown in (53.6%) due to the lack of a follow-up EGD for mapping. Coexisting autoimmune atrophic gastritis was observed in (12.4%, n=19), Helicobacter pylori infection was detected in (16.3%, n=25). Furthermore, Dysplasia was identified in (2.6%, n=4), with low-grade dysplasia and high-grade dysplasia each accounting for (1.3%, n=2) (Table 1). Three patients developed.gastric cancer (2%) , while (4.6%, n=7) were found to have neuroendocrine tumors (NET). First degree family history of gastric cancer was reported in (12.4%, n=19).

Conclusion

This study highlights a 4.17% prevalence of GIM among patients who underwent EGD. High BMI and male sex were observed in more than half of the patients. Dysplasia was detected in 2.6%, gastric cancer in 2%, and NET in 4.6% of cases. Many patients lacked follow-up EGD surveillance, limiting effective monitoring. Larger studies with extended follow-up are required to further investigate these associations and their clinical implications.

Introduction

The increasing accessibility of upper gastrointestinal endoscopy has led to a rise in its prescription. However, its overuse and inappropriate indications result in environmental and economic costs, potential iatrogenesis, and a paradoxical reduction in access. This study aims to assess the indications for elective diagnostic upper gastrointestinal endoscopy (UGE) requested in the hospital setting.

Aims & Methods

A retrospective analysis was conducted on elective diagnostic UGE requests from both outpatient and inpatient settings. The appropriateness of the requests was classified according to the recommendations of the American Society for Gastrointestinal Endoscopy and the European Society of Gastrointestinal Endoscopy.

Results

A total of 314 requests were analyzed over a three-month period: 90.8% (n=285) from outpatient consultations and 7.6% (n=24) from inpatient settings. A previous UGE had been performed in 53.8% (n=169) of patients, with an average interval of 27.13 ± 2.12 months between requests. Overall, 59.6% (n=187) of the requests were deemed appropriate, with the most frequent indications being the need for biopsies (15%, n=28), gastrointestinal bleeding or anemia (14.4%, n=27), and post-endoscopic intervention surveillance (10.7%, n=20).
Clinical information indicated the presence of alarm symptoms in 27.7% (n=87) cases, the most frequent being gastrointestinal bleeding or anemia (33.3%, n=29), age >50 years with new-onset symptoms or suspected structural disease (21.8%, n=19), and progressive dysphagia (17.2%, n=15). Conversely, 40.4% (n=127) of requests were considered inappropriate, mainly due to non-compliance with guidelines and surveillance intervals. The most common inappropriate indications included symptoms attributed to disorders of the gut-brain axis (20.5%, n=26), surveillance of atrophic gastritis, fundic gland polyps, hyperplastic polyps, and gastric intestinal metaplasia (11.8%, n=15), as well as post-gastric surgery surveillance for benign conditions in asymptomatic patients (8.7%, n=11). Additionally, 39 (28.3%) UGEs were requested for gastric cancer screening.

Conclusion

This study identified a significant proportion of UGE requests lacking appropriate and timely indications. These findings emphasize the need for well-defined criteria to ensure the rational and optimized use of this resource.

Introduction

Acute kidney injury (AKI) represents a complication associated with high morbidity and mortality, as well as progression to chronic kidney disease. Early recognition is critical for targeted treatment to reverse AKI. A high percentage of patients have been shown to experience volume overload and pulmonary edema, as volume re-expansion is systematically performed in all patients without prior evaluation of their volume status. A relationship between volume overload and mortality has been demonstrated. The VExUS protocol enables the identification and stratification of systemic venous congestion, complementing inferior vena cava diameter assessment with pulsed Doppler of the suprahepatic veins, portal vein, and interlobar renal veins, evaluating altered flow patterns observed in venous congestion.

Aims & Methods

General Objective: To evaluate the impact of volume overload measured by VExUS on mortality and morbidity in patients with hepatic cirrhosis and acute kidney injury.
Materials and Methods: A prospective, observational, and analytical cohort study was conducted in patients diagnosed with hepatic cirrhosis who developed AKI as a complication and required hospitalization at the Centro Médico Nacional de Occidente. These patients received intravenous albumin at a dose of 1 g/kg/day for 48 hours, followed by an evaluation of their response to treatment. Prior to albumin administration, ultrasound was performed with the VExUS protocol to determine the patients' venous congestion status. The study was carried out between July 2022 and August 2024. No sample size calculation was performed; instead, all patients meeting the inclusion criteria during the study period were included to encompass the entire population affected by this condition.

Results

A total of 38 patients treated at the Centro Médico Nacional de Occidente were included. Patients were classified according to the VExUS index into three levels: 0 (60.5%), 1 (23.7%), and 2 (15.8%). Clinical, biochemical, and outcome variables such as response to albumin, hospital mortality, and survival time were recorded.
The results showed that hospital mortality was 39.5%, with higher rates in patients with higher VExUS levels. Survival analysis revealed that patients with VExUS 0 had an average survival time of 497 days, compared to 207 days in VExUS 1 and 237 days in VExUS 2. Although the Chi-square analysis did not show a statistically significant association between VExUS and mortality (p = 0.361), a trend was observed suggesting a relationship between higher VExUS levels and worse outcomes. The CLIF-SOFA score emerged as a significant predictor of mortality, with a 10.5% increase in risk for each additional unit.

Conclusion

The VExUS index is positioned as a useful tool to evaluate volume overload and its impact on clinical outcomes, highlighting its potential to guide therapeutic decisions in patients with hepatic cirrhosis and AKI. Prospective and multicenter studies with larger sample sizes are recommended to confirm these findings and evaluate the impact of VExUS-guided interventions on personalized clinical management.

References

1. Velez JC et al. Point-of-Care Echocardiography Unveils Misclassification of Acute Kidney Injury as Hepatorenal Syndrome. Am J Nephrol. 2019;50(3):204-211. doi: 10.1159/000501299.
2. Koratala A et al. Point of Care Ultrasound in Cirrhosis-Associated Acute Kidney Injury: Beyond Inferior Vena Cava. Kidney360. 2022 Oct 5;3(11):1965-1968. doi: 10.34067/KID.0005522022.
3. Abhilash Koratala, Claudio Ronco, Amir Kazory. Albumin Infusion in Patients with Cirrhosis: Time for POCUS-Enhanced Physical Examination. Cardiorenal Med 2021;11:161–165.
4. David Jury and Andrew D. Shaw. Utility of bedside ultrasound derived hepatic and renal parenchymal flow patterns to guide management of acute kidney injury. Curr Opin Crit Care 2021, 27:587 – 592.
5. Bhardwaj V, et al. Combination of Inferior Vena Cava Diameter, Hepatic Venous Flow, and Portal Vein Pulsatility Index: Venous Excess Ultrasound Score (VEXUS Score) in Predicting Acute Kidney Injury in Patients with Cardiorenal Syndrome: A Prospective Cohort Study. Indian J Crit Care Med 2020;24(9): 783–789.
6. Koratala A, et al. Nephrologist-performed point-of-care venous excess Doppler ultrasound (VExUS) in the management of acute kidney injury. J Ultrasound. 2023 Mar;26(1):301-306. doi: 10.1007/s40477-022-00760-6
7. Banegas-Deras EJ, et al. Acute kidney injury and point-of-care ultrasound in liver cirrhosis: redefining hepatorenal syndrome. Clin Kidney J. 2024 Apr 15;17(5):sfae112. doi: 10.1093/ckj/sfae

Introduction

Device-assisted enteroscopy (DAE) offers comprehensive examination of the gastrointestinal tract, but its efficacy is dependent on adequate bowel preparation. While prior research explored AI models for colonoscopy cleansing, there is a lack of studies evaluating both small bowel and colon cleansing during DAE.

Aims & Methods

The aim of this study was to develop and validate a convolutional neural network (CNN) capable of automatically classifying bowel cleansing levels during DAE, encompassing both small bowel and colon regions. We retrospectively analyzed 28 procedures (single balloon, double-balloon and motorized spiral enteroscopy) from a high-volume center. Bowel preparation was graded as excellent (≥90% visible mucosa), satisfactory (50-90%), or unsatisfactory (<50%). A dataset of 88,623 images (training: 90%, testing: 10%) was used, covering both small bowel and colon areas. The CNN's performance was evaluated against expert consensus using sensitivity, specificity, accuracy, and AUC-ROC.

Results

We included 51948 frames that were considered excellent, 28820 frames that were considered satisfactory and 7855 frames that were considered unsatisfactory. The CNN demonstrated the following performance metrics: excellent cleansing (sensitivity: 97.8%, specificity: 80.3%, accuracy: 90.6%, AUC-ROC: 0.95), satisfactory cleansing (sensitivity: 81.8%, specificity: 97.9%, accuracy: 92.7%, AUC-ROC: 0.95), and unsatisfactory cleansing (sensitivity: 68.7%, specificity: 99.5%, accuracy: 96.8%, AUC-ROC: 0.96).

Conclusion

This study presents the first globally developed CNN for accurate classification of bowel cleansing throughout DAE, encompassing both small bowel and colon. This advancement represents a necessary step for optimizing diagnostic accuracy and therapeutic efficacy in gastrointestinal procedures, namely DAE, underscoring the pivotal role of innovative technologies in advancing medical practice