Artificial Neural Network Application in Classifying the Left Ventricular Function of the Human Heart Using Echocardiography

Abstract

Abstract The human heart is one of the most important life-giving organs in the human body. According to the World Health Organization (WHO), heart diseases are considered world’s number one cause for deaths worldwide. As per the statistics of Ministry of Health of Sri Lanka, the number of heart patients admitted and the annual deaths caused by heart diseases has increased making heart diseases the leading cause of hospital deaths in Sri Lanka as well. Emergency medicine is the discipline focused on treating patients with urgent medical conditions who are admitted to the Emergency Treatment Unit (ETU) of a hospital. Due to the high number of deaths, the condition of the heart is considered as one of the most critical aspects in emergency medicine. Echocardiography is a widely accepted medical test performed to diagnose the heart condition in non-invasive manner. Generally, the echocardiographic examinations are conducted by acute care physicians who are trained specifically for emergency medicine. Yet there is a chance of them making incorrect decisions due to the lack of clinical experience and expertise in cardiology. Therefore,, an accurate evaluation of the heart condition is highly challenging within the emergency medicine settings. Left Ventricular (LV) function is a crucial factor when diagnosing the cardiac abnormalities. Several parameters such as LV diameter values and Ejection Fraction are considered to determine whether the patient’s LV function is normal or abnormal. Inspired by the recent studies, we carried out several experiments to investigate the possibility of classifying the LV function of the human heart using echocardiography readings. During our study, we considered several image processing and feature extraction methods to extract the important parameters from echocardiography images. The extracted parameters were subjected to train an Artificial Neural Network (ANN) to classify the LV function as normal or abnormal. From our research we obtained a high accuracy for the final result which proves the feasibility of using this methodology to determine the LV heart condition for clinical evaluations. Keywords: Emergency medicine, Artificial Neural Networks, Image Processing, Echocardiography and Left Ventricular Systolic function