Page 68 - Magyar Gasztroenterológiai Társaság 2020. november 6–7. – ONLINE KONGRESSZUS

Előadáskivonatok / Abstracts

Introduction: Type 2 diabetes mellitus (T2DM) has an Aims: Our aim was to prospectively evaluate the precision
increased prevalence in chronic hepatitis C (CHC) patients. of our recently developed Polypbrain® software and to
Presence of T2DM is known as a risk factor for severe compare final histology and real-time optical diagnosis of
complications such as high-grade fibrosis, liver cirrhosis and subcentimetric polyps in our everyday colonoscopy practice.
failure, hepatocellular carcinoma (HCC) and liver related Methods: Polypbrain® is an AI-DSS using deep learning
mortality in CHC patients, but slight Hungarian data are neural network trained on our anonymous electronic
available. database from a total of 1800 histologically identified sub-
Aims: To compare the frequency of severe fibrosis, cirrhosis, centimetric colorectal polyps and 26000 HD, electronic
liver failure, HCC, and liver-related mortality in our non- chromo-endoscopic images(malignant,juvenile,inflammatory
diabetic and diabetic CHC genotype 1 patients using a and sessile serrated polyps were excluded).For this on-going
rigorous match-pared analysis method. prospective study, in every consecutive patients with 5-10
Methods: In the first round we verified 408 CHC patients mm polyp, HD images were captured and histological
(178 male, M/ 230 female, F; all HCV genotype 1) from 2006- diagnosis prediction(probability of hyperplasia or
2019. We followed these subjects from the first detection of adenoma,LGD/HGD)was done real-time with Polypbrain®
HCV infection. Each of them was tested by fasting and during colonoscopy.All polyps were removed and
postprandial serum glucose values and oral glucose finally,results were compared to histological diagnosis.
tolerance test for T2DM; under 40 years we did not found Results: 16 HD-BLI polyp
diabetic patients. From the rest of 302 CHC patients we images(hyperplastic/adenoma:4/12 (dysplasia LG/HG:8/4)
found 124 male and 124 female subjects without significant of 16 patients (male/female:11/5, mean age: 58.98 years)
differences in demographic parameters. In our matched- were included into this preliminary analysis. Mean count of
paired study these males and females were randomized in subimages from each polyps were 160.31(5-452). Feasibility
two groups, the only difference was the presence or lack of of the predicted diagnosis was 96.48% for all
T2DM. Fibrosis stages were evaluated with liver biopsies or polyps(adenoma: 96.42%, hyperplasia: 95.65%).
Fibroscan method. Liver failure was diagnosed by laboratory Confidence in categorizing adenomas according to dysplasia
data and clinical symptoms. HCC was diagnosed by grade was 69.94% and 69.69% in LGD and HGD group.
ultrasonography and/or other imaging diagnostic tools and Concordance between predicted and final histological
confirmed by histology. Mortality data also were registered. diagnosis for hyperplasia or adenoma was 93.75%.
Results: In diabetic vs. non-diabetic comparison, severe Conclusion :Polypbrain® could provide a highly accurate
fibrosis stage (≥F3, Ʃ=142) was detected in 65% vs. 35% tool for real-time optical diagnosis of polyps and to predict
(p<0.001); the cirrhosis rate (Ʃ =142) was 69% vs. 31% HGD. AI-DSS could not only support resect&discard strategy
(p<0.001). HCC rate (Ʃ=18) was 78% vs. 22% (p=0.014). but might improve efficacy of endoscopic therapy to reduce
The frequency of severe liver insufficiency (Ʃ= 44) was 89% polyp recurrence.
vs. 11% (p<0.001), while the mortality rate (Ʃ =28) 90% vs.
10% (p<0.0001).The relative risk (RR) of severe fibrosis, 89. NEOPLASZTIKUS ÉS NEM-NEOPLASZTIKUS
cirrhosis, HCC, liver insufficiency and death was 1.85, 2.26, POLYPOK VALÓS IDEJŰ OPTIKAI DIAGNÓZISA
3.5, 7.8 and 8.33 in diabetic vs. non-diabetic patients, VASTAGBÉLTÜKRÖZÉS SORÁN MESTERSÉGES
respectively. The strongest liver-related mortality factor was INTELLIGENCIA ALAPÚ DÖNTÉSTÁMOGATÓ
the presence of HCC. RENDSZER (POLYPBRAIN®) SEGÍTSÉGÉVEL
4
1,2
4
4
4
Conclusion: Diabetic patients had more serious liver Lovász B. , Szalai M. , Zsobrák K. , Finta Á. , Oczella L. ,
3
4
fibrosis, higher rate of cirrhosis, HCC, liver insufficiency and Dubravcsik Z. , Madácsy L.
mortality than non-diabetic subjects in this matched-paired 1. Semmelweis Egyetem, Belgyógyászati és Onkológiai
analysis. Our data suggest that occurrence of diabetes is a Klinika, Budapest; 2. Semmelweis Egyetem, ETK,
strong negative predictive factor in the HCV-related liver Egészségtudományi Klinikai Tanszék, Budapest; 3. Bács-
diseases and requires a close surveillance even after Kiskun Megyei Kórház, Kecskemét; 4. Endo-Kapszula
successful antiviral treatment. Magánorvosi Centrum, Székesfehérvár

88. PROSPECTIVE REAL-LIFE PREDICTION OF FINAL Bevezetés: A neoplasztikus és nem-neoplasztikus polypok
HISTOLOGY AND REAL-TIME OPTICAL DIAGNOSIS OF elkülönítése különösen fontos a colorectalis rák
NEOPLASTIC AND NON-NEOPLASTIC POLYPS DURING mortalitásának csökkentése céljából. Az új mesterséges
COLONOSCOPY USING A NEW ARTIFICIAL intelligencia alapú döntéstámogató számítógépes
INTELLIGENCE DECISION SUPPORT SYSTEM rendszerek segíthetnek a „vedd le és dobd el” stratégia
(POLYPBRAIN®) kivitelezésében.
4
4
4
Lovász B. , Szalai M. , Zsobrák K. , Finta Á. , Oczella L. , Célkitűzés: A célunk az általunk a közelmúltban fejlesztett
4
1,2
Dubravcsik Z. , Madácsy L. Polybrain® program kipróbálása volt a polypok valós-idejű
3
4
1. Department of Internal Medicine and Oncology, diagnosztikájában vastagbéltükrözés során polypectomián
Semmelweis University, Budapest; 2. Institute of Applied átesett egymást követő ambuláns betegekben.
Health Sciences, Semmelweis University, Budapest, Módszerek: A Polypbrain® egy mesterséges intelligencia
Hungary; 3. Bács Kiskun County Teaching Hospital alapú, mélytanuló neurális hálót használó program, melyet a
University of Szeged, Kecskemét; 4. Endo-Kapszula saját anoním, 1800 szövettanilag diagnosztizált colorectalis
Endoscopy Unit, Székesfehérvár polypról készült 26000 HD chromoendoszkópiával készült
képet tartalmazó, elektronikus adatbázisunk
Introduction: Precise differentiation between non- felhasználásával tanítottunk (malignus, juvenilis,
neoplastic and neoplastic polyps with high/low-grade gyulladásos és serrated polypok nem szerepeltek benne). A
dysplasia is important to assist optimal endoscopic jelenleg is zajló vizsgálatban vastagbéltükrözés során talált
therapy.New artificial intelligence-based decision support 5-10 mm-es polypokról készült polyp képek valós idejű
systems(AI-DSS) can help to support resect&discard szövettani diagnózisát végeztük (a program hyperplasia,
strategy. adenoma elkülönítését, low/high grade dysplasia
66

66 Central European Journal of Gastroenterology and Hepatology
Volume 6, Supplementum 2 / November 2020