A DIFFERENT WAY TO DIAGNOSIS ACUTE APPENDICITIS: MACHINE LEARNING

Abstract

BackgroundMachine learning is a branch of artificial intelligence based on the idea that systems can learn from data, identify patterns and make decisions with minimal human intervention.. Our aim is to predict acute appendicitis, which is the most common emergency surgery indication, using machine learning algorithms with an easy and inexpensive method.Materials and Methods:Patients who were treated surgically with a prediagnosis of acute appendicitis in a single-center between 2011 and 2021 were analyzed. Patients with right lower quadrant pain were selected. 189 positive and 156 negative appendectomies were found. Gender and hemogram were used as features. Machine learning algorithms and data analysis were made in Python (3.7) programming language.ResultsNegative appendectomies were 62%(n=97) female and 38%(n=59) male. Positive appendectomies were 38% (n=72) female and 62% (n=117) male. The accuracy in the test data was 82.7% in logistic regression, 68.9% in support vector machines, 78.1% in k-nearest neighbors, 83.9% in neural networks, The accuracy in the voiting classier created with logistic regression, k-nearest neighbor, support vector machines and artificial neural networks was 86.2%. In Voting classifier, sensitivity was 83.7% and specificity was 88.6%.ConclusionThe results of our study showed that ML is an effective method in diagnosing acute appendicitis. This study presents a practical, easy, fast and inexpensive method to predict the diagnosis of acute appendicitis.