EP-653 Diagnosing Acute Appendicitis using Machine Learning: A Systematic Review

Abstract

Abstract Introduction Acute appendicitis is a surgical emergency that usually presents in the younger population. The mortality risk for uncomplicated acute appendicitis is less than 0.1%, but this rises to 0.6% should gangrene or a perforation develop during its disease course. The clinical diagnosis of acute appendicitis remains challenging, with systems such as the Alvarado score not specific enough to exclude a diagnosis. The aim of this study is to review the literature on the use and effectiveness of machine learning (ML) in the clinical diagnosis of acute appendicitis, with a particular focus on the parameters used to train the ML models. Methods A systematic review was conducted using PubMed and OvidSP using search terms ‘appendicitis’, ‘artificial intelligence’ and ‘machine learning’. Results There were 255 articles identified, which after excluding duplicates and screening, 14 articles were reviewed in detail and 6 articles included in the final review. Parameters used to train ML machines included patient demographics such as age and gender, clinical assessment such as temperature, presenting symptoms and examination findings, and biochemical and haematological results such as white cell count and C-Reactive Protein. Discussion The parameters used to train ML models are based on limited clinical and biochemical components from established scoring systems such as the Alvarardo score. Despite this limitation, ML models were generally better in predicting acute appendicitis in terms of sensitivity and specificity and positive- and negative-predictive values. Future ML models should use more contemporary markers to further improve diagnostic accuracy.