MDensNet201-IDRSRNet: Efficient cardiovascular disease prediction system using hybrid deep learning

Cardiovascular diseases (CVDs) are common diseases that impact the heart or vascular system. Since early discovery significantly improves survival chances, precise prediction techniques are essential. There are new paths for more accurate CVD prediction due to emerging technologies like machine learning (ML). Heart disease may now be identified in its early stages using several machine learning algorithms, which can aid in future treatments. However, none of the existing algorithms achieve high accuracy and frequently fail because of bias and over-fitting. To improve the prediction accuracy of cardiovascular disease, a new innovative approach is proposed in this research by utilizing deep learning techniques to identify significant features. For efficient CVD prediction, we propose a hybrid deep-learning intelligent system. Tests and assessments have been conducted using the five benchmark datasets for cardiac disease from the UCI repository. Three data processing techniques are first utilized in the pre-processing stage to improve the dataset's quality by preventing undesired distortions: outlier removal, replacing missing values, and resolving data imbalance problems. Next, deep learning-based Modified DenseNet201 (MDenseNet201) extracts the disease-related features. Relief and Least Absolute Shrinkage and Selection Operator (LASSO) approaches are used to select the appropriate features. Finally, a deep learning-based improved deep residual shrinkage network (IDRSNet) is employed to predict cardiovascular disease. The accuracy of the proposed model on the University of California Irvine (UCI) machine learning repository dataset is 99.12%. Based on experimental results, the proposed hybrid deep learning system produced more excellent accuracy for CVD prediction than existing approaches. The combined intelligent system (MDensNet201-IDRSNet), which generates the best practical solution out of all input prediction models considering performance criteria, makes it possible for physicians and radiologists to diagnose cardiac patients more accurately.