August 2021
Since diagnosing ECG-data manually by medical professionals is costly in terms of time, my bachelor thesis proposed an attempt to automate the process. A Convolutional Neural Network for classifying six different cardiac arrhythmias, including sinus arrhythmia, sinus bradycardia, supraventricular tachycardia, sinus tachycardia, atrial fibrillation and atrial flutter, and the normal sinus rhythm was developed. The already published Convolutional Neural Network by Acharya et al. provided the basis for the neural network, used in my thesis with the proposed dataset. This annotated dataset with raw and denoised data from a 12-lead ECG of more than 10 000 patients was used for training and validating the neural network. With a final accuracy of the entire network for all heart rhythms of 92,18 % using raw data and 91,60 % using denoised data of one lead, it was one of many different approaches to ECG-data classification with neural networks.
During my thesis, I took a close look at the medical and technical basics of ECG-data collection and processing and also deepened my knowledge of the structure and applications of convolutional neural networks.
If you are interested in any further details please feel free to contact me.