P0726 - Artificial Intelligence and Machine Learning-Based System for Diagnosis of Chronic Hepatitis C

Annie Shergill, MD¹, V. Lakshmi Narasimhan, PhD², Emmanuella Budu, PhD student^2
1Clovis Community Medical Center, Fresno, CA; ²University of Botswana, Gaborone, South-East, Botswana

Introduction: Chronic Hepatitis C virus (HCV) infection is one of the most common causes of chronic liver disease worldwide. Diagnosis involves detection of antibodies to HCV and viral RNA. This study introduces an artificial intelligence (AI) and machine learning (ML)-based system for laboratory result interpretation and reporting of HCV, thereby minimizing operator-based errors. Our aim is to introduce the concept of computer-aided diagnosis (CAD) in the detection of Chronic Hepatitis C.

Methods: ML algorithms employed in this study are: Multilayer Perceptron (MLP), Random Forest Classifier (RF) and a Decision Tree (DT). Patient dataset utilized has been obtained from a public domain, includes patients diagnosed with chronic HCV. In the first phase, missing values in the dataset are replaced with zeroes. Values in the target variable in the dataset are encoded to transform the categorical values into integer values as algorithms cannot process non-numerical data. A '0' indicates absence of HCV and '1' indicates a possible case of HCV. The dataset is then divided into a training set, validation set and test set. To ensure uniformity, the values in the dataset are normalized on a scale of 0 to 1. Algorithms are implemented and executed on the dataset to obtain predictions on HCV.

Results: Results are evaluated based on the accuracy, specificity, and sensitivity of each ML algorithm (Table 1). All algorithms obtain a high accuracy of 80% and above, implying that they can correctly make predictions on HCV from the dataset. The RF and DT algorithms are highly sensitive with a score of 100% each, meaning that they are highly capable of detecting patients with HCV. MLP algorithm scores 0% on sensitivity indicating that this algorithm is incapable of diagnosing patients with HCV. All algorithms show a high specificity indicating that these algorithms will correctly diagnose patients without HCV.

Discussion: AI is revolutionizing the field of medicine with the potential of introducing CAD of various diseases. Currently, diagnosis of chronic HCV is reliant on physician interpretation of test results. We propose that using AI and ML-based technique in the diagnosis of chronic HCV can improve diagnostic efficacy by minimizing interpretation errors and delays in reporting results, thereby facilitating prompt initiation of treatment. Application of AI in diagnostic evaluation aims at increasing the efficacy of handling and reporting large volume of patient specific data.

METRIC	RANDOM FOREST CLASSIFIER	MULTILAYER PERCEPTRON	DECISION TREE
ACCURACY (%)	82.60	86.95	81.52
SPECIFICITY (%)	80	100	78.75
SENSITIVITY (%)	100	0	100

Table: Summary of the Results


Disclosures:
Annie Shergill indicated no relevant financial relationships.
V. Lakshmi Narasimhan indicated no relevant financial relationships.
Emmanuella Budu indicated no relevant financial relationships.


Annie Shergill, MD¹, V. Lakshmi Narasimhan, PhD², Emmanuella Budu, PhD student². P0726 - Artificial Intelligence and Machine Learning-Based System for Diagnosis of Chronic Hepatitis C, ACG 2021 Annual Scientific Meeting Abstracts. Las Vegas, Nevada: American College of Gastroenterology.