Heart disease diagnosis and categorization from ECG signals using hybrid Fuzzy-CNN machine optimized by meta-heuristic algorithms

Cardiovascular diseases are among the most important causes of global mortality, and their diagnosis is mainly based on ECG signals. The complexity and nonlinearity of these signals have limited the effectiveness of classical machine learning methods. In this study, an optimized hybrid model based on Fuzzy-CNN is presented using meta-heuristic algorithms for classifying 2D images obtained from ECG signals. First, raw signals are extracted from the MIT-BIH Arrhythmia Database (LTAF) and are ready to use after preprocessing including noise removal and normalization. Then, with the help of the Wild Horse Optimization (WHO) algorithm, optimized 2D images are generated from these signals. In the classification stage, a hybrid Fuzzy-CNN system is used that extracts image features through convolution layers and combines temporal and spectral information to perform multi-class classification using Takagi-Sugeno fuzzy logic in the fully connected layer. All the coefficients of the convolution filters and the parameters of the fuzzy system are simultaneously tuned and optimized using the Puma Optimization Algorithm (POA). The proposed model is capable of classifying seven clinically important arrhythmia classes according to the AAMI EC57 standard (including N, S, V, F, Q, LBBB and RBBB) and shows a very remarkable performance in accurately detecting VEB and S classes, which are associated with sudden cardiac death. The proposed model was evaluated on the MIT-BIH database with MATLAB, and the results include 99.71% accuracy, 97.18% precision, 97.87% recall, and an F1 score of 95.32%. This performance shows significant superiority over the baseline methods (such as standard CNN, SVM, and LSTM) and recent advanced models (MPA-CNN, hybrid metaheuristic-CNN, CNN-LSTM, and other optimized methods). In particular, the sensitivity of the critical classes V and S is significantly improved compared to previous studies.