Prediction of Coronary Artery Disease using Deep Learning – A Basic study

Abstract

Purpose: Cardiovascular disease continues to be a major contributor to global mortality and morbidity. Forecasting cardiovascular illness is regarded as a crucial concern studying clinical data. Due to the increasing volume of data, the task of analyzing and processing it has become more complex. This is particularly challenging when it comes to maintaining e-healthcare data. Furthermore, the utilization of artificial intelligence in this research field regards the prediction model to be a crucial aspect.  The field of computer-assisted diagnosis is characterized by its dynamic and rapid development. Given the possible risks linked to inaccurate medical diagnoses, considerable efforts have been directed towards enhancing computer programs that aid clinicians in achieving precise diagnoses. Artificial intelligence (AI) will assist in discovering new connections between data and disease. The implementation of AI is intended to mitigate human fallibility, decrease the expenses of medical care, and streamline the exchange of professional data in diverse formats. A branch of Machine Learning (ML) called Deep learning (DL) has a set of algorithms and software that enable a machine to autonomously acquire skills and execute tasks. DL machines utilize Neural Networks (NN) to process vast quantities of data. DL employs numerous layers to represent different levels of abstraction. ML is a highly encouraging technique for creating sophisticated and automated algorithms to investigate complex and diverse bio-medical data with multiple dimensions. This research investigates two datasets pertaining to heart disease. The study employs Neural Networks and evaluates their predictive accuracy and several other performance parameters.

Design/Methodology/Approach: The study employs datasets sourced from Kaggle containing heart-related information. The analysis employs Dense Neural Network (DNN) models using Python, Sklearn, Pandas, Tensorflow, and Jupyter Notebook, and their results have been analyzed.

Findings/Results: The ANN models built for both the datasets have performed well. The performance metrics of the model for Cleveland Dataset which is a balanced dataset is extremely good with an accuracy of 0.99. The performance for Framingham Dataset which is an imbalanced dataset is lesser with an accuracy of 0.88.

Originality/Value: The study involved an analysis of two cardiac databases from Kaggle. The distribution of data, interdependence of multiple features, and the impact of all features on the prediction of heart disease have been observed. DNN have been utilized to build and train models. Investigation has also been conducted to comprehend the significance of feature selection and its impact on the accuracy of predictions. This is a fundamental examination of Neural Networks and their application and appropriateness for forecasting cardiovascular illnesses.

Paper Type:  Research and Analysis with basic study