Year: 2023
Type: review
Abstract: The surge of Artificial Intelligence-related new technologies combined with the growing amount of available health data has led to many advances in modern medicine. In particular, the development of new classification tools derived from Machine Learning has facilitated the detection of pathologies, often in the early stages. These algorithms are all the more interesting because they can only require patients' physiological data, which can be easily collected through new devices such as smartwatches. As many diseases of cardiac or non-cardiac origin can have an impact on the electrical activity of the heart, the electrocardiogram is becoming more and more popular in screening and diagnosis. This paper aims to provide a broad view of the existing literature on these novel electrocardiogram-based classification models. The entire design process including signal processing methods, feature extraction and selection, training and evaluation protocols as well as the limitations of these models have been reviewed, through 411 articles. This pipeline is valid for both machine and deep learning algorithms. Unlike many reviews in the literature, this study addresses the topic of disease detection from ECGs from a broader and more varied scientific perspective, including technical and clinical specificities. Furthermore, this review falls within a rapidly growing field, and the challenge has been to provide a structured overview of the different methods and algorithms, as well as the increasing stakes involved. Overall, this study shows that deep learning solutions offer promising results, although they are still limited today by the lack of publicly available data and by concerns about interpretability of the models and their predictions.
Source: HAL
Language: English
Topic: ECG Monitoring and Analysis
Subfield: Cardiology and Cardiovascular Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7151172281