Anomalies in the classification and detection of heart rate using deep learning
Authors
Kishor, Sakshi
Issue Date
2021
Degree
MSc in Data Analytics
Publisher
Dublin Business School
Rights holder
Rights
Arrhythmias, which in non-medical terms is commonly known as the irregularity in a heartbeat, contributes to approximately 300,000 sudden deaths per year. This incidence rate is higher than the other lethal and dangerous disease like lung cancer, stroke, or a breast cancer. Cardiologists can help patients cope with many types of arrhythmias, but researchers and doctors are kept awake at night by sudden and lethal arrhythmias. This study is an effort to bring down the incidence rate due to heart rate arrhythmias by identifying the irregularities in a heartbeat at early stages and help the cardiologists with their patients. The 5 different categories of heartbeats are classified in this study from publicly available MIT-BIH database which consists of ECG data from various patients. The data has been preprocessed with the help of wavelet transformation tool in MATLAB. To help achieve the classifications of the heartbeats, logistic regression and convolution neural networks algorithms are used. These two models’ evaluations are compared in the form of recall, precision, accuracy, and computational time. A comparative study between the two models is done with the help of tableau. Based on the evaluation, the deep learning model, CNN, gives out the accuracy of 99.09 percent.
Abstract
Arrhythmias, which in non-medical terms is commonly known as the irregularity in a heartbeat, contributes to approximately 300,000 sudden deaths per year. This incidence rate is higher than the other lethal and dangerous disease like lung cancer, stroke, or a breast cancer. Cardiologists can help patients cope with many types of arrhythmias, but researchers and doctors are kept awake at night by sudden and lethal arrhythmias. This study is an effort to bring down the incidence rate due to heart rate arrhythmias by identifying the irregularities in a heartbeat at early stages and help the cardiologists with their patients. The 5 different categories of heartbeats are classified in this study from publicly available MIT-BIH database which consists of ECG data from various patients. The data has been preprocessed with the help of wavelet transformation tool in MATLAB. To help achieve the classifications of the heartbeats, logistic regression and convolution neural networks algorithms are used. These two models’ evaluations are compared in the form of recall, precision, accuracy, and computational time. A comparative study between the two models is done with the help of tableau. Based on the evaluation, the deep learning model, CNN, gives out the accuracy of 99.09 percent.