Deep Learning Based Models for Coronary Artery Disease Detection: A Review
Authors: Wani, J.Y., Wani, M.A.
Journal: Communications in Computer and Information Science
Publication Date: 01/01/2027
Volume: 2994 CCIS
Pages: 133-150
eISSN: 1865-0937
ISSN: 1865-0929
DOI: 10.1007/978-3-032-28478-5_10
Abstract:Coronary artery disease (CAD) has remained among the leading causes of death in the world hence urgent and accurate diagnosis is of great essence. The deep learning-based models are also a valuable approach to clinical decision assistance over the years, which offer high predictive power and automated feature extraction in comparison with conventional ones. This paper presents a systematic review of CAD detection models that are based on deep learning with the primary focus on the type of data, namely structured clinical records. To make a comparative analysis of these deep learning-based models, popular benchmark datasets like Cleveland, Framingham, NHANES and Z-Alizadeh Sani are used. We also classify and analyze deep learning-based models as artificial neural networks (ANNs), convolutional neural networks (CNNs) and hybrid models highlighting their main characteristics and the comparison of their performance. The benefits and drawbacks of these methods are outlined, and the performance of different methods are compared on the basis of similar datasets to find out the trends and difficulties. Overall, the imbalance between classes, limited sample sizes, overfitting and low generalizability remain as the problem of CNN and Hybrid models, although they show excellent results on clinical data. This paper also provides a general overview of the current state of DL on the field of coronary artery disease detection and the future developments of models that are robust, interpretable, and able to be clinically implemented.
Source: Scopus