Screening of Abdominal Aortic Aneurysms based on Machine Learning Approach

Abstract

An abdominal aortic aneurysm (AAA) is identified as a localized expansion of the abdominal aorta with a 50% increase in the aortic diameter. This cardiovascular condition is usually asymptomatic and rupture can occur at any time without warning, making it difficult for patients to seek help and treatment. As a consequence, a robust early AAA predictor is highly relevant for minimizing the mortality rate due to the aortic wall rupture. This work aims to identify potential attributes for AAA prediction and to find a machine learning model that can be used in a clinical setting for prediction of AAA. In this study, N=424 subject’s data (age and maximum aortic diameter) were collected from the works of literature (includes healthy and AAA population).70% and 30% of N=424 subject’s data were used for training and testing of the model respectively. Five models named K-Nearest Neighbor (KNN), Decision Tree (DT), Random Forest (RF), and Logistic Regression (LR), and Naïve Bayes (NB) were employed with 10-fold cross-validation for better evaluation. For the prediction of AAA, among five models, RF achieved the best classification accuracy of 99.22 %, precision of 0.98, recall of 0.98 and f1-score of 0.99 with features (age and maximum aortic diameter) incorporation. KNN achieved the second best classification accuracy of 98.22 %, precision of 0.99, recall of 0.98 and f1-score of 0.98 with same features incorporation. Future work will incorporate additional attributes that will improve the sensitivity of the models and help clinicians in their decision-making.