Introduction

Colorectal anastomotic dehiscence (CAD) is one of the most severe postoperative complications in colorectal surgery, significantly impacting patient morbidity and mortality. Current risk assessment methods rely on clinical and intraoperative factors, but no real-time predictive tool exists. This study aimed to develop an artificial intelligence model based on intraoperative recording of the anastomosis for CAD prediction.

Aims & Methods

A convolutional neural network (CNN) was trained using a dataset of annotated frames from colorectal surgery videos across three international high-volume centers: IPO Lisboa, Hospital das Clínicas de Ribeirão Preto, São Paulo, Brazil and Royal Liverpool University Hospital. The dataset included a total of 5,356 annotated frames from 25 patients, 2,007 frames classified as CAD-positive and 3,349 as CAD-negative. Four CNN architectures (EfficientNetB0, EfficientNetB7, ResNet50, and MobileNetV2) were tested. Performance was evaluated using sensitivity, specificity, accuracy, and area under the receiver operating characteristic (AUROC) curve. Heatmaps were generated to identify key regions influencing predictions.

Results

The best-performing model achieved an accuracy of 99.55%, AUROC of 99.60%, sensitivity of 99.21%, specificity of 100%, PPV of 100%, and NPV of 98.94%. The model reliably identified CAD-positive frames and provided visual explanations through heatmaps.

Conclusion

To our knowledge, this is the first AI model developed to predict CAD using intraoperative video analysis. Its accuracy suggests the potential to redefine surgical decision-making by providing real-time risk assessment. Further refinement with a larger dataset and diverse surgical techniques could enable intraoperative interventions to prevent CAD before it occurs, marking a paradigm shift in colorectal surgery.

Introduction

Metabolic dysfunction-associated Steatotic Liver disease (MASLD) is complicated by various cardiometabolic (CMR) manifestations, and the prognosis is severely burdened by acute cardiovascular events (ACE).
Even though several drugs have been tested to target CMR comorbidities, lifestyle modifications remain the cornerstone intervention. Unfortunately, the dietary-behavioral prescription is constantly limited by poor compliance, with the sociodemographic-determinant factors impacting this remaining undetermined.
Moreover, despite emerging evidence supporting behavioral therapy's relevance, the real long-term benefits of offering motivational support to MASLD individuals represent an unexplored scenario in real life.

Aims & Methods

This study aimed to evaluate the efficacy of multidisciplinary management on clinical outcomes in MASLD, investigating compliance with specialist-tailored indications. We consecutively enrolled individuals presenting MASLD diagnostic criteria (n.286). After recruitment, patients were randomized into three cohorts: subjects (n.72) receiving generic hepatologist-provided advice (“H”); subjects (n.71) also receiving the nutritionists-prescribed individualized intervention (“HN”); patients (n.143) treated with an approach additionally involving cognitive/behavioral-based psychological support (“HNP”). H and HN represented the “standard of care” (n.143), whereas “HNP” constituted the “experimental group” (n.143). At baseline, anthropometric, biochemical, clinical, liver stiffness (LSM), controlled attenuation parameter (CAP), lifestyle habits (including sociodemographic determinants), and body composition values were recorded. Along 24 months, on semestral hepatological, nutritional (H and HN), and psychological (HNP) follow-ups, all parameters were reassessed, the occurrence of ACEs was recorded, and compliance was opportunely evaluated. Three time-study points were considered: baseline (T0), intermediate (i.e., after 18 months – T1), and end of the study (T2).

Results

Intermediate evaluations revealed a significantly higher prevalence of individuals achieving a >10% decrease in body weight (HNP:62.09%; HN:44.9%; H:35.8%; chi-square, HNP vs. HN, p:0.01; HNP vs. H, p:0.0002) after 18 months (T1) in the experimental group. At the end of the observation period (T2), in this setting, a further increase in the proportion of individuals presenting and preserving >10% decrease in body weight was reported (HNP: 72.03%; HN: 46.7%; H: 39.2%, chi-square T1, HNP vs. HN, p:0.02; HNP vs. H, p:0.0001; McNemar T1-T2 HNP vs HN, p:0.001, HNP vs H, p:0.001), as wells as a significant improvement in Homeostatic-model-assessment-for-insulin-resistance (p:0.001), HDL (p < 0.0001), LSM (p:0.007), CAP (p:0.002), and fat-mass (p < 0.0001) was observed (HNP, T0 vs T2).
Relevantly, HNP patients presented a significantly lower risk of ACEs [HR: 0.497, IC (0.236-0.751) 95%, p:0.04)]. Logistic regression analysis (adjusted for sex, age, drug administration, and cardiovascular risk factors) revealed that HNP adherence was significantly associated with lower ACE occurrence (aOR: 0.81; C.I. 95%: 0.55-0.97; p:0.02). Consistently, the loss of compliance rate was significantly lower in HNP (HNP:12.08%; HN:34.7%; H:45.8%; HNP vs HN, HNP vs H, p:0.001). Finally, the analysis of sociodemographic factors revealed large family (i.e., > 3 individuals), dynamic work (i.e., < 2 hours/die dedicated to main meals), and city life as social determinants impacting compliance (OR: 2.11, 1.78, 1.12) (all p < 0.05).

Conclusion

In the era of Precision Medicine, the correct management of MASLD patients should contemplate the integration of standard hepatological-nutritional with psychological support.

References

1. Romeo M, Dallio M, Cipullo M, Coppola A, Mazzarella C, Mammone S, Iadanza G, Napolitano C, Vaia P, Ventriglia L, Federico A. Nutritional and Psychological Support as a Multidisciplinary Coordinated Approach in the Management of Chronic Liver Disease: A Scoping Review. Nutr Rev. 2025 Feb 24:nuaf001. doi: 10.1093/nutrit/nuaf001. Epub ahead of print. PMID: 39992295.
2. Righetti R, Cinque F, Volpe MT, Sebastiani G. Integrating behavioral interventions into a holistic approach to metabolic dysfunction-associated steatotic liver disease. Expert Rev Gastroenterol Hepatol. 2024 Jul;18(7):303-313. doi: 10.1080/17474124.2024.2385487. Epub 2024 Aug 5. PMID: 39075881.
3. European Association for the Study of the Liver (EASL) Clinical Practice Guidelines on the management of metabolic dysfunction-associated steatotic liver disease (MASLD). J Hepatol. 2024 Sep;81(3):492-542. doi: 10.1016/j.jhep.2024.04.031. Epub 2024 Jun 7. PMID: 38851997.

Introduction

Colonoscopy is a widely accepted screening test for detecting colorectal cancer. Computer-aided detection (CADe) based on artificial intelligence (AI) may be the way forward to improve its quality. As AI becomes a common upgrade in endoscopic equipment, an increasing number of young endoscopists are trained in an AI environment. However, how AI will impact trainees' future endoscopy outcomes remains unclear.

Aims & Methods

The study aimed to evaluate quality indicators in the first 1000 colonoscopies performed by newly trained endoscopists, comparing those trained in an AI environment to those trained in a non-AI environment. A study included 6,000 patients over 18 years old who underwent a colonoscopy for various reasons. The procedures were carried out by six young endoscopists, of which three were trained without AI and three with the assistance of the AI-powered Olympus Endo-Aid CADe system. The study retrospectively evaluated the first 1,000 procedures performed by each endoscopist after completing training. These procedures were performed without AI enhancement and relied entirely on the endoscopists' detection skills. Quality indicators were assessed in non-AI-training and AI-training groups. The morphology of detected polyps was assessed to determine the influence of AI during training on laterally spreading tumors (LST) detection.

Results

A total of 6,000 participants were included in the analysis. Withdrawal time (WT) was longer in the AI-training group (8 vs 9 min, p<0.05). Adenoma detection rate (ADR; 30.3 vs. 35.5%, p<0.05), advanced adenoma detection rate (AADR; 4.6 vs. 5.1%, p<0.05), and adenoma per colonoscopy score (APC; 0.73 vs. 0.75, p<0.05) were higher in the AI-training group. AI-assisted endoscopy training significantly improved LST-NG detection (p<0.05) for lesions smaller than 20mm.

Conclusion

Implementing CADe during endoscopy training can positively impact quality indicators in the first 1,000 cases. After AI-enhanced training, the detection of LST-NG may improve. The potential of artificial intelligence in endoscopy training requires further investigation through prospective randomized controlled trials.

References

1. Kaminski MF et al. N Engl J Med. 2010 May 13;362(19):1795-803. doi: 10.1056/NEJMoa0907667. PMID: 20463339.
2. Repici A et al. Efficacy of Real-Time Computer-Aided Detection of Colorectal Neoplasia in a Randomized Trial. Gastroenterology. 2020 Aug;159(2):512-520.e7. doi: 10.1053/j.gastro.2020.04.062. Epub 2020 May 1. PMID: 32371116.
3. Vilmann AS et al. Computerized feedback during colonoscopy training leads to improved performance: a randomized trial. Gastrointest Endosc. 2018 Nov;88(5):869-876. doi: 10.1016/j.gie.2018.07.008. Epub 2018 Jul 19. PMID: 30031803.
4. Kaminski MF et al. Leadership training to improve adenoma detection rate in screening colonoscopy: a randomised trial. Gut. 2016 Apr;65(4):616-24. doi: 10.1136/gutjnl-2014-307503. Epub 2015 Feb 10. PMID: 25670810; PMCID: PMC4819605.
5. Forbes N et al. Association Between Endoscopist Annual Procedure Volume and Colonoscopy Quality: Systematic Review and Meta-analysis. Clin Gastroenterol Hepatol. 2020 Sep;18(10):2192-2208.e12. doi: 10.1016/j.cgh.2020.03.046. Epub 2020 Mar 30. PMID: 32240836.