Novel Approaches for Predicting Risk Factors of Atherosclerosis

TL;DR

This paper introduces a novel machine learning approach using particle swarm optimization to predict atherosclerosis risk factors from clinical data, achieving high accuracy and identifying physical inactivity as a key risk factor.

Contribution

The paper presents a new predictive methodology with an in-built imputation algorithm and PSO, outperforming existing machine learning techniques in identifying atherosclerosis risk factors.

Findings

Achieved 99.73% accuracy in risk prediction.

Identified physical inactivity as a significant risk factor.

Outperformed other machine learning methods on STULONG dataset.

Abstract

Coronary heart disease (CHD) caused by hardening of artery walls due to cholesterol known as atherosclerosis is responsible for large number of deaths world-wide. The disease progression is slow, asymptomatic and may lead to sudden cardiac arrest, stroke or myocardial infraction. Presently, imaging techniques are being employed to understand the molecular and metabolic activity of atherosclerotic plaques to estimate the risk. Though imaging methods are able to provide some information on plaque metabolism they lack the required resolution and sensitivity for detection. In this paper we consider the clinical observations and habits of individuals for predicting the risk factors of CHD. The identification of risk factors helps in stratifying patients for further intensive tests such as nuclear imaging or coronary angiography. We present a novel approach for predicting the risk factors of atherosclerosis with an in-built imputation algorithm and particle swarm optimization (PSO). We compare the performance of our methodology with other machine learning techniques on STULONG dataset which is based on longitudinal study of middle aged individuals lasting for twenty years. Our methodology powered by PSO search has identified physical inactivity as one of the risk factor for the onset of atherosclerosis in addition to other already known factors. The decision rules extracted by our methodology are able to predict the risk factors with an accuracy of $99.73%$ which is higher than the accuracies obtained by application of the state-of-the-art machine learning techniques presently being employed in the identification of atherosclerosis risk studies.