Introduction

To prevent colorectal cancer, it is crucial to accurately identify and classify precancerous colonic lesions. Previous studies have shown that the classification of precancerous lesions still depends on the individual examiner. In this study, we attempted to create realistic AI-generated images for training physicians in the assessment of colorectal lesions. Artificially generated realistic images make it independent of patient data, allowing to specifically create various precancerous lesions and practice their classification.

Aims & Methods

We trained an AI using 25 million endoscopic images, that can generate images of different precancerous colorectal lesions. To show that the AI is capable of producing realistic-looking images, we designed a study in which we presented artificial polyps and real images of polyps to eleven physicians and asked them to discriminate between them. They were also asked about their level of confidence for the assignment. We used 20 real and 20 AI generated images. We included five experts and six beginners in this study. Beginners were defined as endoscopists who have performed less than 100 examinations or have been practicing for less than 2 years.

Results

Real images were identified as such with an accuracy of 79%, while only 66% of AI-generated images were recognized as artificially generated. The p-value is 0.0034. The physicians had a higher confidence when choosing an answer for real images (63%) and a lower for fake images (56%). Especially the AI-generated images that clearly showed mucosal vessels were classified as real images by many physicians.

Conclusion

In this study we were able to generate synthetic images with a high degree of authenticity. We have identified features that make images appear particularly realistic and we will continue to train the AI with these specific characteristics to achieve even higher quality. The images that were identified as most realistic will be used in further projects to classify precancerous lesions.

Introduction

Adequacy of bowel preparation (ABP) directly affects colonoscopy performance and adenoma detection rate. Studies attributed a role to the dosing regimen of the bowel cleansing agent, social and economic factors and comorbidities. However, data on the impact of kidney failure and renal replacement therapy (RRT) are lacking.

Aims & Methods

This study investigated how ABP compares between patients on RRT and age-matched subjects with normal kidney function.
All patients on RRT undergoing an elective colonoscopy between January 2016 and December 2021 were identified from the medical records of a university hospital, and matched to 2 closest age controls undergoing colonoscopy at the same date ± 7 days. Study subjects and controls were compared for demographics, comorbidities including history of abdominal surgery, use of bowel transit-influencing drugs, residential care stay and colonoscopy-related data (bowel preparation regimen, sedation type, colonoscopy findings, ABP defined as a cumulative Bristol Bowel Preparation Score >= 6 and a score of >= 2 for each colon segment). Statistical significant contributors to ABP were assessed by logistic regression.

Results

Ninety-one patients on RRT (76 on haemodialysis and 15 on peritoneal dialysis) and 182 controls were identified. Patients on RRT were older (69±12 vs. 64±14y, p=.008) and less frequently women (34% vs. 50%, p=.013). They had more diabetes (40% vs. 24%, p=.008), more history of abdominal surgery (32% vs. 21%, p=.047) and took more bowel transit-influencing drugs (50% vs. 28%, p<.0001). Bowel preparation regimens were equally distributed between both groups. There was no difference in the rate of ABP between groups (87% vs. 81%, p=NS). Logistic regression analysis did not reveal an association between RRT and ABP. ABP was negatively influenced by residency in residential care, overnight stay in the hospital before colonoscopy, history of central nervous system disease, and positively by split-dose regimen (Nagelkerke R square 0.329, p=<.001).

Conclusion

Despite less favourable associated conditions, ABP in RRT patients remains comparable to controls. Split-dose regimen, neurological conditions and in-hospital stay are significant predictors of adequate bowel preparation.

References

Clark BT, Rustagi T, Laine L. What level of bowel prep quality requires early repeat colonoscopy: systematic review and meta-analysis of the impact of preparation quality on adenoma detection rate. Am J Gastroenterol. 2014;109(11):1714-23; quiz 24.
Hassan C, East J, Radaelli F, Spada C, Benamouzig R, Bisschops R, et al. Bowel preparation for colonoscopy: European Society of Gastrointestinal Endoscopy (ESGE) Guideline - Update 2019. Endoscopy. 2019;51(8):775-94.
Saumoy M, Jesudian AB, Aden B, Serur D, Sundararajan S, Sivananthan G, et al. High prevalence of colon adenomas in end-stage kidney disease patients on hemodialysis undergoing renal transplant evaluation. Clin Transplant. 2016;30(3):256-62.
24. Calderwood AH, Schroy PC, 3rd, Lieberman DA, Logan JR, Zurfluh M, Jacobson BC. Boston Bowel Preparation Scale scores provide a standardized definition of adequate for describing bowel cleanliness. Gastrointest Endosc. 2014;80(2):269-76.
25. Clark BT, Protiva P, Nagar A, Imaeda A, Ciarleglio MM, Deng Y, et al. Quantification of Adequate Bowel Preparation for Screening or Surveillance Colonoscopy in Men. Gastroenterology. 2016;150(2):396-405; quiz e14-5.
artel M, Barkun AN, Menard C, Restellini S, Kherad O, Vanasse A. Split-Dose Preparations Are Superior to Day-Before Bowel Cleansing Regimens: A Meta-analysis. Gastroenterology. 2015;149(1):79-88

Introduction

Ulcerative colitis (UC) is characterized by chronic immune mediated inflammation of the colon and rectum. Corticosteroid (CS) is a common treatment for patients with moderate to severe active UC, but its effect varies among patients, including resistance, response, and dependence. In addition, the longer CS therapy continues, the greater the concern regarding drug-related adverse effects. New immunomodulatory therapies are being developed for the treatment of UC drug showing promise in the treatment of patients with moderate UC refractory to aminosalicylic acid (5-ASA). Thus, the clinical benefit of being able to predict whether certain patients would benefit from CS therapy prior to administration cannot be overestimated.

Aims & Methods

We compared the clinical characteristics of patients with steroid-responder (SR) and steroid-non-responder (SNR) UC and retrospectively evaluated for biomarkers that could predict the efficacy of CS prior to administration in patients with moderate to severe active UC who failed to respond to 5-ASA.
We enrolled 100 UC patients presenting at Nippon Medical School Hospital with a history of CS administration prior to December 2023. The patients were divided into two groups, SR patients and SNR patients, and their clinical characteristics were determined according to medical records. Eosinophil counts were measured in the high-power field using Direct Fast Scarlet staining.

Results

Clinical characteristics of UC patients were as follows: the median age, 48±17 years; height 167±9.2 cm, and weight, 58±14 kg; including 61 males to 39 females. The most common disease extent type was the total colitis (63%), followed by left-sided colitis (32%), and proctitis (5%). Complete SNR was observed in 15 patients (15%), with remaining 85 patients (85%) classified as SR. There was no significant difference found in Mayo Endoscopic Score, first dose of CS, number of hospitalizations, number of relapses, propotion of patients with total-colitis type, or duration of CS treatment between the SR group and the SNR group. The non-remission rate, defined as failure to achieve mucosal healing, was significantly higher in the SNR group (66%) than in the SR group (36%) (P=0.03). The proportion of patients treated with azathioprine was significantly higher in the SR group (40%) than in the SNR group (39%). The proportion of patients treated with biologics or small molecular compounds was significantly higher in the SNR group (40%) than in the SR group (22%) (P=0.03).
We then compared hematological biomarkers in blood samples obtained prior to CS administration between the two groups. There was no significant difference in WBC counts, hemoglobin values, serum albumin concentrations, CRP values, and ESR between the two groups. However, eosinophil profiles (both total count numbers and ratio to white blood cell counts) were significantly higher in the SR group than in the SNR group (P=0.02). Neutrophil-lymphocyte ratio tended to be higher in the SR group than in the SNR group (P=0.06). The number of infiltrated eosinophils in the inflamed rectal mucosal tissue was not significantly different between the two groups. The number of infiltrated eosinophils in the rectal mucosa tended to decrease in the steroid-response groups (p = 0.07). However, there was no difference in the number of infiltrated eosinophils in the steroid-non-response groups.

Conclusion

Both eosinophil in the blood and in the rectal mucosa, and neutrophil-lymphocyte ratio in the blood could be potential biomarkers for steroid responsiveness prior to administration.

References

・British Society of Gastroenterology consensus guidelines on the management of inflammatory bowel disease in adults. Gut. 2019; 68: s1–s106
・Severe ulcerative colitis: at what point should we define resistance to steroids? WJG 2008 Sep 28;14(36):5504-7.
・A reproducible grading scale for histological assessment of inflammation in ulcerative colitis: Gut 47：404－409，2000
・Mucosal Eosinophil Abundance in Non-Inflamed Colonic Tissue Is Associated with Response to Vedolizumab Induction Therapy in Inflammatory Bowel Disease：J Clin Med Res,2022

Introduction

Standardized classification of the gastroesophageal valve, achievable using the Hill classification, is a common step during gastroscopy. Given the visual nature of endoscopic examinations, artificial intelligence (AI) can support this task. We conducted a single-center prospective study to compare the performance of a self-developed AI model in determining the Hill classification with that of examiners.

Aims & Methods

From October 2023 to January 2024 consecutive patients with unaltered anatomy of the gastroesophageal junction were enrolled in the study. The AI model ran in the background during routine gastroscopy, providing real-time predictions, while physicians assessed the Hill grade without knowledge of the model's output. The Gold standard was established through a majority vote of a three-member expert committee in the cases where a clear image of the hiatus was available. The primary outcome was accuracy in Hill grade assessment, with secondary outcomes including precision and sensitivity for each grade.

Results

Out of 150 patients that participated, 131 were eligible for analysis. Our AI model exhibited an accuracy of 84.7%, surpassing the physicians' accuracy with 61.8% (p<0.01). Regarding rating the different Hill grades the AI's lowest accuracy, precision and sensitivity were 88%, 53% and 60% respectively, while physicians achieved lowest scores of 66%, 20%, and 30%. Post hoc analysis of the videos using the systems capability for real-time explainability via heatmaps highlighting image regions influencing the classification, revealed that the model used the correct anatomical region for classification.

Conclusion

Our AI system demonstrates superior accuracy compared to physicians in Hill grade assessment. Future research should explore the impact of physician-AI interactions during examinations through a randomized controlled trial. (ClinicalTrials.gov, Trial register number: NCT06040723)

Clinical Case Summary

A 39-year-old woman was admitted on June 20, 2024, with a right middle cerebral artery stroke, nine weeks after starting Upadacitinib for ulcerative colitis. She presented with dysarthria, dysphagia, and severe left-sided weakness. CT imaging revealed a wedge-shaped infarction without intracranial hemorrhage (CT head/ Brain MRI).
She was diagnosed with ulcerative colitis in November 2023, she initially responded to Mezavent and steroids but was hospitalized on March 31, 2024 for moderate UC flare, requiring infliximab rescue therapy. The patient presented with four bloody bowel movements and elevated inflammatory markers (CRP>100), confirmed by a Mayo score of 2 on flexible sigmoidoscopy and cytomegalovirus (CMV) infection was excluded. Clostridium difficile tests were negative twice and a CT scan revealed pancolitis. Despite minimal response to rescue therapy, she opted for medical treatment over colectomy.
On April 23, 2024, she began an eight-week induction of Upadacitinib 45 mg, which she tolerated well. However, on June 20, she developed sudden stroke symptoms while eating dinner, including difficulty swallowing, choking, left-sided weakness, and numbness. She had no vascular risk factors or family history of premature cardiovascular conditions. Lupus, and vasculitis work up were unremarkable and extensive thrombophilia screenings were all negative.
Discussion: Thrombosis in IBD patients can result from various potential causes. First of all, Inflammation and hypercoagulability, driven by elevated cytokines like TNF, interleukins, and CRP, activate the coagulation cascade, increasing thrombosis risk.
Conclusion: IBD is a known risk of thromboembolic disease, this risk increased during flare up will raise future research questions regarding extended anticoagulation prophylaxis during flare and uses of Jak stats inhibitor.

References

1. Ferro JM,Oliveira Santos M. Neurology of inflammatory bowel disease. J Neurol Sci. 2021 May 15;424:117426. doi: 10.1016/j.jns.2021.117426. Epub 2021 Mar 27.

2. Solitano V.Thrombosis in IBD in the Era of JAK Inhibition. Curr Drug Targets. 2021;22(1):126-136. doi: 10.2174/1389450121666200902164240.

3. Guiu A,a new opportunistic pathogen associated with cystic fibrosis?]Enferm Infecc Microbiol Clin. 2014 Oct;32(8):497-501. Spanish.DOI: 10.1016/j.eimc.2013.08.003

4. Tan JT, Ng DP.Genetic variation in KCNQ1 associates with fasting glucose and beta-cell function: a study of 3,734 subjects comprising three ethnicities living in Singapore.Diabetes. 2009 Jun;58(6):1445-9.DOI: 10.2337/db08-1138

5. Reddy VB.The burden of diverticular disease on patients and healthcare systems.Gastroenterol Hepatol (N Y). 2013 Jan;9(1):21-7.

6. Kim WJ.Major constituents and antimicrobial activity of Korean herb Acorus calamus.Nat Prod Res. 2011 Aug;25(13):1278-81.DOI: 10.1080/14786419.2010.513333

7. Lou JS, Effect of expectancy and personality on cortical excitability in Parkinson's disease.Mov Disord. 2013 Aug;28(9):1257-62.DOI: 10.1002/mds.25471PMID: 23630185

8. Parra-Izquierdo I,Janus kinase inhibitors ruxolitinib and baricitinib impair glycoprotein-VI mediated platelet function.Platelets. 2022 Apr 3;33(3):404-415.DOI: 10.1080/09537104.2021.1934665Epub 2021 Jun 7.

9. Kwak SH, Incidentally found resectable lung cancer with the usage of artificial intelligence on chest radiographs.PLoS One. 2023 Mar 10;18(3):e0281690.DOI: 10.1371/journal.pone.0281690PMID: 36897865;

10. Nguyen GC, et al. Consensus Statements on the Risk, Prevention, and Treatment of Venous Thromboembolism in Inflammatory Bowel Disease: Canadian Association of Gastroenterology. Gastroenterology. 2014;146(3):835–848. doi:10.1053/j.gastro.2014.01.042

Introduction

The current gold standard for diagnosis of Inflammatory bowel disease (IBD) is based on clinical and endoscopic evaluation as well as histologic and radiologic evaluation. However, endoscopic evaluation can be uncomfortable and carry some risks, especially in the pediatric population that requires general anesthesia in order to perform the procedure.

Aims & Methods

Our aim was to investigate whether serum biomarkers can differentiate between pediatric patients with and without IBD. Secondary aims were to determine whether there are biomarkers that can differentiate between Crohn's disease (CD) and Ulcerative colitis (UC), and whether we can predict progression to biologic treatment within one year from diagnosis using these biomarkers.
Pediatric patients undergoing a first diagnostic colonoscopy at British Columbia Children's Hospital (BCCH) between December 2019 and June 2022 were invited to participate in the study. A 2 ml blood sample was taken from each participant at the time of colonoscopy. Demographic and clinical data were collected from each patient. Blood samples were analyzed using the Legendplex^TM flow cytometry kits investigating which included 50 different inflammatory cytokines and chemokines associated with IBD. A prediction model was built via a supervised learning method to determine how well we can predict the different groups.

Results

One hundred pediatric patients participated in the initial study. Five were excluded from the analysis due to uncertain diagnosis. Median age was 13.36 years, 39 were females (41%) and 54 (56.84%) were diagnosed with IBD, with 35 (64.81%) diagnosed with Crohn's disease. On a univariate analysis and Volcano plot 6 biomarkers were found to be strongly associated with the diagnosis of IBD: IL22, IL8, TARC, GROA, MIG, and IL18. The cross-validated AUROC was 0.70. Using SPLS-DA the classification accuracy was improved to 0.84, and MIG was found to be the key biomarker to discriminate between pediatric IBD patients and non-IBD patients.
On a univariate analysis only TSLP was found statistically significant as a predictor of CD, and RANTES and TARC were found as negative predictors for initiating biologic treatment within one year of diagnosis. However, the prediction model did not predict well the subclasses nor progression to biologic treatment.

Conclusion

Using serum biomarkers can predict the diagnosis of Pediatric IBD. MIG was found to be the key biomarker in the prediction model.