Introduction

Artificial intelligence (AI) showed various performance among test sets with different diversity due to sample selection bias, which can be a stumbling block for AI application. We previously tested AI named ENDOANGEL diagnosing gastric neoplasms and early gastric cancer (EGC) on single-center videos in a man-machine competition. We aimed to retest ENDOANGEL on multi-center videos to explore challenges applying AI in multiple centers, then upgrade ENDOANGEL and explore solutions of solving the challenge.

Aims & Methods

ENDOANGEL was retested on multi-center videos retrospectively collected from 12 institutions and compared with performance in previously-reported single-center videos. We then upgraded ENDOANGEL to ENDOANGEL-2022 with more training samples and novel algorithms, and conducted competition between ENDOANGEL-2022 and endoscopists. ENDOANGEL-2022 was then tested on single-center videos and compared with performance in multi-center videos; the two AI systems were also compared with each other and endoscopists.

Results

Forty-six EGCs and 54 non-cancers were included in multi-center video cohort. On diagnosing EGCs, compared with single-center videos, ENDOANGEL showed stable sensitivity (97.83% vs. 100.00%) while sharply decreased specificity (61.11% vs. 82.54%); ENDOANGEL-2022 showed similar tendency while achieving significantly higher specificity (79.63%, p <0.01) making fewer mistakes on typical lesions than ENDOANGEL. On detecting gastric neoplasms, both AI showed stable sensitivity while sharply decreased specificity. Nevertheless, both AI outperformed endoscopists in the two competitions.

Conclusion

Great increase of false positives is prominent challenge for applying EGC diagnostic AI in multiple centers due to high heterogeneity of negative cases. Optimizing AI by adding samples and using novel algorithms is promising to overcome this challenge.

Introduction

In recent years, there has been a surge in the development of artificial intelligence (AI) systems for endoscopic applications, such as Barrett’s neoplasia detection and polyp characterization. While these systems have shown promising results, the vast majority is trained and validated on small, retrospective, single-center or expert-acquired data sets. Current AI systems are particularly prone to a so-called domain-shift: a system only performs well on data it has been trained on. As real-world scenarios in the endoscopy suite can be complex and diverse, the risk of poor generalization of AI should not be underestimated.

Aims & Methods

In this study we aim to provide a quantitative analysis of this domain shift and propose several strategies to improve generalization to clinical practice. We trained three AI systems for commonly used applications in gastrointestinal endoscopy, using publicly available data sets with high quality standards. In order to identify the limitations of these systems, all test sets were artificially degraded in a stepwise manner with clinically prevalent artifacts: blur, inadequate exposure and color variability. These degradations represent the heterogeneity of image quality in daily clinical practice. Severity of degradation was clinically calibrated by a panel of expert endoscopists. Two levels of degradation were identified: 1) mild degradation, still considered as high quality data; 2) moderate degradation, considered as image quality that could still be encountered in daily clinical practice. The trained models were subsequently evaluated for their appropriate performance metrics on the gradually degraded test sets. After implementation of several training strategies to mitigate potential performance loss, the models were re-evaluated on the same test sets.

Results

All three models display a considerable decrease in performance when exposed to various types of clinically relevant image degradation. Mean loss was 7.4% and 16.6% for respective mild and moderate degradation. Implementation of specific training strategies to improve robustness resulted in substantial reduction of performance loss on all degraded test sets.

Conclusion

This study highlights the potential limitations of current AI systems in endoscopic applications which may lead to poor generalization. Specific training strategies can mitigate this risk and improve robustness of these systems. Development of robust and reliable AI systems that deliver consistent and accurate performance is crucial to ensure safe implementation into daily clinical practice.

Introduction

The decision to stop biologic medication in Crohn’s Disease (CD) in apparent clinical remission is challenging. We investigated MR enterography (MRE) observations associated with future relapse.

Aims & Methods

49 CD patients discontinuing biologics due to clinical remission and undergoing MRE within 12-months prior, or 1-month after, were identified. Two blinded radiologists subjectively assessed small bowel disease presence/activity and disease length, wall-thickness, fat-wrapping [1] and sMARIA score [2]. 'Relapse' status was assigned based on full clinical records including patient symptoms, biochemistry, imaging and drug prescription. Per-patient analysis was performed using the most severely affected segment and groups were compared with Chi-Square and Wilcox test.

Results

Mean patient age was 30.1 (SD: 13.0, 19 female). Median follow-up was 55 months (6-110). Twenty-five (51%) had disease relapse (DR) and 24 (49%) no relapse (NDR). Sixteen of 24 (67%) and 15/25 (60%) of the NDR and DR respectively still had small bowel disease present on MRE prior to biologic cessation. This was subjectively deemed active in 10/15 (66%) DR vs just 5/16 (31%) of NDR. One patient was excluded from the quantitative analysis due to image artifacts. Maximum wall thickness in the most severely affected segment was the only MRE observation associated with subsequent relapse (Table 1).

Conclusion

Simple presence or absence of residual small bowel abnormality does not in itself predict future relapse. However, in those with residual disease, simple bowel wall thickness is greater in those who relapse and may be a simple tool to guide the decision to discontinue biologic medication.

References

[1] Makanyanga, J. C.. Evaluation of Crohn’s disease activity: Initial validation of a magnetic resonance enterography global score (MEGS) against faecal calprotectin. European Radiology 24, 277–287 (2014).

[2] Ordás, I.. Development and Validation of a Simplified Magnetic Resonance Index of Activity for Crohn’s Disease. Gastroenterology 157, 432–439.e1 (2019).

Disclosure

H E Fitzke has ongoing research collaborations with Motilent, the medical imaging analysis company who develop the image analysis tools and previously received funding from the BBSRC to undertake two internships with Motilent. S A Taylor has ongoing research collaborations and holds share options in Motilent.

Introduction

Abdominal surgery frequently results in postoperative ileus (POI), a neuroinflammation-based disorder of the gastrointestinal tract, characterized by motility disturbances leading to prolonged hospital stay and increased health care costs. The underlying neuroinflammation involves activation of resident macrophages and monocyte-derived macrophages (MDM) infiltration into the intestine's muscularis externa (ME). Immune checkpoint proteins, including PD-1 and its ligands PD-L1 and PD-L2, play a prominent role in cancer immunology, while less is known about their role in non-oncological inflammatory diseases.

Aims & Methods

The aim of this study was to investigate the role of PD-1 signaling in POI development. POI was induced by a surgical intestinal manipulation in wildtype (wt), PD-1^-/-, PD-L1^-/- and PD-L2^-/- mice. We performed immunohistochemistry and flow cytometry to identify the cellular sources of PD-1 and PD-L1 and used a blood cell transfer model to determine the role of a leukocyte-restricted PD-1 deficiency on immune cell infiltration. By qPCR, bulk 3´mRNA sequencing and use of the SMART-Seq2 technology in sorted CX3CR1^GFP/+ resident and monocyte-derived macrophages, we analysed transcriptional and functional differences between wt and PD-1^-/- macrophage populations. Furthermore, we quantified neuronal cell loss, in vitro and in vivo macrophage phagocytosis, GI transit measurements and metabolic activity by Seahorse XF Real-Time ATP rate assays to further characterize functional changes linked to PD-1 deficiency.

Results

PD-1, PD-L1 and PD-L2 mRNA levels were induced in the postoperative ME upon intestinal manipulation. PD-1^-/- and PD-L1^-/-, but not PD-L2^-/- mice, were protected from POI. PD-1^-/- mice showed a 50% reduction of MDM infiltration and reduced inflammation-mediated neuronal loss. A cell transfer experiment confirmed a 60% reduction of postoperative ME infiltration of PD-1 deficient leukocytes compared to wt leukocytes. Ly6C^-CX3CR1⁺ resident macrophages and Ly6C⁺CX3CR1⁺ MDM were identified as the primary cellular source of PD-1 in POI. PD-1 deficient Ly6C⁺CX3CR1⁺ MDM showed reduced in vivo phagocytosis, which was confirmed in bone-marrow derived macrophage cultures. Gene ontology analysis of postoperative ME samples of PD-1 and wt mice revealed lower enrichment levels of genes regulating metabolic respiratory chain processes, oxidative stress, cell death and immune functions. Moreover, GSEA analyses of sorted PD-1^-/-CX3CR1⁺Ly6C⁺ MDM showed reduced enrichment of genes associated with various immune functions, host defense and regulation of immune and neuronal cell death as well as an alteration of metabolic genes in POI. Seahorse XF Real-Time ATP rate measurement showed no difference between naïve monocytes from wt and PD-1^-/- mice. However, PD-1^-/- monocytes identified an increased oxidative chain reaction-dependent ATP generation during LPS stimulation.

Conclusion

Our data provide new evidence on the role of PD-1 signaling in MDM during postoperative intestinal neuroinflammation following abdominal surgery. PD-1 deficiency prevents POI, results in a metabolic switch and reduces inflammation-mediated neuronal loss as well as infiltration capacity and phagocytosis of MDM. Interaction in PD-1 signaling might be a potential target in prevention of POI.

Introduction

Motorized spiral enteroscopy (MSE) has shown to be safe and effective for deep small bowel enteroscopy in patients with and without prior abdominal surgery, including patients with Crohn’s disease.
This case series reports our first results regarding the feasibility and safety of the MSE with endoscopic balloon dilatation in patients with small bowel strictures due to Crohn’s disease.

Aims & Methods

Data were collected retrospectively on patients who underwent MSE for endoscopic balloon dilatation of small bowel strictures due to Crohn’s disease. Strictures were either postoperative (P.O.) or post-inflammatory (P.I.).

Results

A total of ten consecutive patients with Crohn’s disease underwent 13 procedures. Propofol sedation without endotracheal intubation was used in all procedures. Demographic and procedure related data are shown in the table. One patient underwent three procedures with 3–4 month intervals. During the third procedure triamcinolone acetonide was injected at four quadrants of the stricture. Fluorescence imaging was used during six procedures. The procedure was technically successful in 12/13 of procedures. In one case, dilatation was not performed because of a longer than anticipated and a sharply angulated stricture.
The number of treated strictures ranged from one to five per procedure. Dilatation of multiple strictures in two patients (4 and 5) was achieved within a relatively short procedure duration of 55 and 70 minutes. Two retained video capsules were retrieved.
Besides the anticipated mild abdominal discomfort and/or sore throat in three patients, no adverse events occurred.

Conclusion

These retrospective data show that MSE with CRE dilatation is feasible and safe in a cohort of patients with strictures due to Crohn’s disease, postoperatively as well as post-inflammatory. The fact that the distal 16 cm of the endoscope are not covered by the spiral overtube makes cautious introduction of the endoscope beyond a recently dilated stricture feasible and safe. Therefore, dilatation of multiple strictures in one session is possible, provided that they are at a short distance from each other. Fluorescence imaging guidance may be helpful in severe stenosis, multiple strictures, or sharp angulations.

Introduction

Low-volume bowel preparation solutions, including 1 L polyethylene glycol plus ascorbate (PEG-A), have been developed to improve tolerability because of the low tolerability of high-volume polyethylene glycol agents. However, 1 L PEG-A still has its typical odor that is not pleasant for most patients. The oral sodium sulfate tablet (OST) is a new agent with simethicone as a preloaded component. OST may have the better tolerability because of little odor due to its tablet formula.

Aims & Methods

We aimed to investigate the efficacy, safety, and tolerability of OST compared to 1 L PEG-A. We randomized patients aged 18 years or older who underwent colonoscopy into the OST (group A) and 1 L PEG-A (group B) groups. Bowel preparation efficacy was assessed on the Boston bowel preparation scale (BBPS) and bubble scale. Bowel preparation success was defined as BBPS score ≥ 6. The bubble scores ranged from 0 to 3 per segment (total from 0 to 9), with a lower score indicating lower bubbles in the colonic lumen. Safety and tolerability were evaluated using a systematized questionnaire.

Results

A total of 171 patients (group A: 87, group B: 84) were included in the final analyses. The average BBPS score of group A was 8.0±1.3, which was similar to that of group B (7.7±1.4) (p=0.215). The proportion of patients showing bowel preparation success was 95.4% in the group A whereas that in the group B was 96.4% (p=0.736). The adenoma detection rate was also similar between groups A (59.6%) and B (41.9%) (p=0.087). The bubble scale in the group A was 0.2±0.9, which was significantly better that that of the group B (1.9±1.7) (p<0.001). All adverse events were mild in both groups. The group A showed lower frequency of nausea compared to the group B (14.9% vs. 38.1%, p=0.001). Overall satisfaction was better in the group A (8.1 ± 2.1 vs. 6.4 ± 2.8, p<0.001).

Conclusion

Although both OST and 1 L PEG-A were efficacious, safe, and tolerable for bowel preparation of colonoscopy, the OST showed fewer bubbles and slightly better tolerability.

Introduction

Bowel preparation is recognized as the most burdensome aspect of the whole colonoscopy procedure, sometimes even worse that the examination itself, and represents a major barrier to screening colonoscopy. Mannitol could significantly improve patient acceptability as it is rapid, requires a single dose, a low volume and has a pleasant taste. The SATISFACTION study compared oral mannitol 100g/750ml with standard split-dose 2L PEG-ASC (MoviPrep®).

Aims & Methods

The SATISFACTION Phase III study was an international, multicentre, randomized (1:1), parallel-group, endoscopist-blinded non inferiority trial. Primary endpoint was the proportion of patients with adequate bowel cleansing. Secondary endpoints concerned efficacy (adenoma detection rate, caecal intubation rate, time of evacuation), safety (intestinal gases concentration, haematobiological parameters, adverse events) and patient satisfaction. The study included 703 patients (352 treated with mannitol and 351 with PEG-ASC) undergoing elective colonoscopy.

Results

Mannitol was not inferior to PEG-ASC for proportion of patients with adequate cleansing (91.1% and 95.5% respectively; p value for the non-inferiority =0.0095). There was no significant difference for secondary efficacy endpoints. The profile of acceptability was significantly better in mannitol group for easy to use, taste, willingness to use (p<0.0001 for all parameters). The concentration of intestinal gases (H₂, CH₄) was similar in the two groups and well below those potentially critical.

Conclusion

Mannitol was not inferior to PEG-ASC for proportion of patients with adequate cleansing (91.1% and 95.5% respectively; p value for the non-inferiority =0.0095). There was no significant difference for secondary efficacy endpoints. The profile of acceptability was significantly better in mannitol group for easy to use, taste, willingness to use (p<0.0001 for all parameters). The concentration of intestinal gases (H₂, CH₄) was similar in the two groups and well below those potentially critical.

Disclosure

Prof. Gian Eugenio Tontini and Prof. Maurizio Vecchi received a consultant fee from NTC Pharma. Anna Orsatti is a current NTC Pharma employee