Analyzing Online Patient-EKG Data Sets: A Novel Approach for the Early Detection of Heart Disease

Abstract:

Currently, cardiovascular diseases kill 17 million people worldwide each year. Treatment is often difficult for patients because they only discover their risk for the condition when it is too late. In this research project, a novel computational platform was designed to efficiently diagnose a patient's risk of cardiovascular disease (CVD) determined by age, gender, hypertension, and an ACE gene mutation. Through the application of machine learning, neural networks, and statistical modeling, this platform was equipped with the tools to accurately predict patients’ heart disease diagnoses.

In Phase 1, NCBI datasets were analyzed, including more than 9000 CVD-diagnosed and healthy patients, specifically examining age, gender, and hypertension. Consequently, in Phase 2, 221 patients from ACE gene data sets were analyzed by correlating age, gender, and hypertension to the genotype. A multilayer perceptron (MLP) neural network was developed using Octave that outputs a risk score and diagnosis for the patient.

It was discovered that females over 45 with hypertension and the homozygous mutated genotype of the ACE gene exhibited the highest risk of developing CVD. After cross-validation, the Phase 1 code achieved an F1 accuracy score of 0.8531. Phase 2 cross-validation allowed to distinguish between normal and diseased patients with a 0.965 F1 score. After the characteristics were plotted on a three-dimensional graph, it was found that patients with the ACE gene generally presented a higher risk of CVD regardless of age, gender, and hypertension. Using this CVD predictive tool, patients will be able to take precautionary measures before the disease’s severity increases.