Calculator for detection of colorectal adenomas by using artificial intelligence models in patients with chronic hepatitis C.

Abstract

70 Background: Hepatitis C virus (HCV) is known for its oncogenic potential, especially in hepatocellular carcinoma and non-Hodgkin lymphoma. On review, several studies have indicated that patients with chronic hepatitis C (CHC) have an increased risk of developing colorectal cancer (CRC). We developed an artificial intelligence (AI) based tool using machine learning (ML) algorithms to help stratify these patients into a higher risk of CRC/adenomas. Methods: The study was approved by the institutional review board. We developed an AI automated calculator uploaded to a graphical user interface (GUI), and we applied ML to train models to predict the probability and the number of adenomas detected on colonoscopy. Data collected were age, smoking history, significant alcohol consumption, aspirin intake, ethnicity, HCV status, gender, body mass index (BMI), and colonoscopy findings. The models can operate either in the presence or absence of the above parameters. Data sets were split into 70:30 ratios for training and internal validation. Scikit-learn StandardScaler was used to scale values of continuous variables. We used the colonoscopy findings as the gold standard and applied a deep learning architecture to train six ML models for prediction. The ML models used were Support Vector Classifier, Random Forest, Bernoulli Naïve Bayes (BNB), Gradient Boosting Classifier (GBC), Logistic Regression, and Deep Neural Networks. Additional regression models were trained and tested to predict the number of polyps. A Flask (customizable python framework) application programming interface (API) was used to deploy the trained ML model with the highest accuracy as a web application. Finally, Heroku was used for the deployment of the web-based API to https://adenomadetection.herokuapp.com. Results: Data was collected for 415 patients, of which only 206 had colonoscopy results. On internal validation with the remaining patients, BNB predicted the probability of adenoma detection with the highest accuracy of 56%, precision of 55%, recall of 55%, and F1 measure of 54%. Support Vector Regressor (SVR) predicted the number of adenomas with the least mean absolute error (MAE) of 0.905. Conclusions: Our AI-based tool shows an association between CHC and colorectal adenomas. This tool can help providers stratify patients with CHC for early referral for screening colonoscopy. Along with giving a numerical percentage, the calculator can also comment on the number of adenomatous polyps a gastroenterologist can expect while doing a colonoscopy, thus prompting a higher adenoma detection rate.