A new vision of a simple 1D Convolutional Neural Networks (1D-CNN) with Leaky-ReLU function for ECG abnormalities classification

Lakhdari, Kheira and Saeed, Nagham ORCID: https://orcid.org/0000-0002-5124-7973 (2022) A new vision of a simple 1D Convolutional Neural Networks (1D-CNN) with Leaky-ReLU function for ECG abnormalities classification. Intelligence-Based Medicine, 6. p. 100080. ISSN 2666-5212

[thumbnail of Lakhdari_and_Saeed_2022_j.ibmed._A_new_vision_of_a_simple_1D_CNN_with_Leaky-ReLU_function_for_ECG_abnormalities_classification.pdf]
Lakhdari_and_Saeed_2022_j.ibmed._A_new_vision_of_a_simple_1D_CNN_with_Leaky-ReLU_function_for_ECG_abnormalities_classification.pdf - Published Version
Available under License Creative Commons Attribution.

Download (6MB) | Preview


Artificial Intelligence (AI) is increasingly impacting the healthcare field, due to its computational power that reduces time, cost and efforts for both healthcare professionals and patients. Diagnosing cardiac abnormalities using AI represents a very attractive subject for both medical and technical professionals. Cardiac abnormalities are characterized by the ECG signal, which is known by its variable morphology and intense affection by noises and artifacts. In this context, the presented study aims to propose a simple yet efficient version of Convolutional Neural Networks (CNN) to classify those abnormalities. This version increases the ability to detect several heart rate arrhythmias and severe cardiac abnormalities based only on the original 1D format of the ECG signal, which reserve the main feature of this signal and can be very suitable for ready-to-use and real-time applications. The main used training datasets are the MIT-BIH arrhythmias and the PTB databases. The proposed architectures are mainly inspired by the most recent CNN models and introduce several modifications on functions and layers, such as the use of the Leaky-ReLU instead of the ReLU activation function. The results of the proposed model are varying from an accuracy of 97%–99% in classifying Normal (n), Supraventricular (s), Ventricular (v), Fusion of ventricular and normal (f), and noisy (q) beats, in addition to the Myocardial Infarction (MI) case. A continuous performance was achieved while testing the model on real data, and after its migration to real mobile devices.

Item Type: Article
Identifier: 10.1016/j.ibmed.2022.100080
Keywords: AI, Cardiac abnormalities, Classification, CNN, Deep learning, ECG, Healthcare, Leaky-ReLU, MI, ReLU
Subjects: Computing > Intelligent systems
Related URLs:
Depositing User: Nagham Saeed
Date Deposited: 07 Nov 2022 06:53
Last Modified: 06 Feb 2024 16:12
URI: https://repository.uwl.ac.uk/id/eprint/9601


Downloads per month over past year

Actions (login required)

View Item View Item