Agenda

Presenter: Asst. prof. Dr. Mohammed Ahmed Shakir

Host:

Title of the Seminar: (about Published Paper)

Improving ECG signals classification by using deep learning

techniques: A review

Objectives or Summary:

Heart diseases are serious global health concerns that could result in many deaths.

Detecting and classifying the heart diseases early is crucial for initiating treatment and

improving patient outcomes. ECG signals contain valuable information to analyze cardiac

functions. It can be argued that techniques of Deep learning (DL) are effective aid to

classify ECG signals accurately through learning from large amount of ECG data, ability

to extract hidden information, and achieving superior performance in detection heart

abnormalities. ECG signals processing involves three phases, preprocessing, extraction

features and classification. This paper intends to review several studies published from

2019 to 2024 in this field. It follows a method of comparative analysis, considering

specific performance metrics, preprocessing techniques, and the DL model used. The aim

is to determine the most accurate DL technique for classifying ECG signals. Eventually,

the paper indicated that the debate on the most accurate technique for classification

remains ongoing. However , the reviewed studies demonstrated that models based on CNN

and RNN can achieve significant level of accuracy in classifying ECG signals. On other

hand, according to the conducted comparative analysis, it is recommended to use VGG16

as a classifier for ECG signals. As a suggestion, the complexity of VGG16 can be reduced,

allowing for the implementation of a real-time application.

The presentation is made as a fellow up for a paper that was published as part of the

proceedings of ICACS24