Review on hybrid deep learning-based cardiovascular prediction systems

Cardiovascular diseases CVDs account for nearly one-third of the deaths of people around the world. For this reason, developing advanced diagnostic systems for the smart detection and prediction of the risk of onset of CVDs is paramount. The development of Artificial Intelligence (AI), Machine Learning (ML), and more recently, Deep Learning (DL), has enabled the automated analysis of ECG signals and other clinical data, which has the potential to radically change the diagnosis of CVDs. In particular, convolutional neural networks (CNN), recurrent neural networks (RNN), Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), Transformers and attention-based models, have all shown great promise in the classification of cardiac arrhythmias and the assessment of CVDs risk. Moreover, hybrid deep learning models, which consist of many of the above architectures, have been shown to have superior performance in the retrieval of spatial and temporal contextual features from large and complex biomedical datasets. This work focuses on review of hybrid deep learning-based systems for the prediction of CVDs. We also summarize the challenges that need to be addressed for the development of prediction systems of CVDs that are robust, explainable, and safe for use.