AI-Enhanced Cardiovascular Diagnostics: Arrhythmia Detection with CNN and SVM

Arrhythmia is characterized by abnormally beating heart a serious condition that the requires timely and accurate identification. In this study, a novel architecture consisting of both CNN and SVM has been designed and implemented for the classifier, for the diagnosis of arrhythmias based on the ECG inputs. The hybrid model combines applying CNN for feature extraction and employing SVM for classification to address subtle feature problems and class imbalance. The performance of the model was tested with the help of benchmark database called MIT-BIH Arrhythmia Database. The results of the experiments showed that the efficiency of the proposed hybrid CNN-SVM model was 94.7% more than the pure CNN (92.4%) and SVM (87.6%) models. The proposed hybrid model achieved high classification accuracy at 93.9%, recall of 92.8%, F1 measure of 93.3%, as well as an AUC-ROC of 0.96. Using the hybrid model, the underrepresented classes namely Bradycardia and Tachycardia have been successfully identified with recall scores of 91.2% and 89.8% respectively as against 85.4%, and 87.6% corresponding to the solo CNN. We found that the hybrid model shows much faster convergence in the training stage than the solo model; after 15 iterations, the hybrid model stability of the validation accuracy is achieved that takes 20 iterations for the solo CNN model. As the results estimated, the effectiveness of the proposed method is sufficiently high and proves its applicability for clinical practice, as well as their stability for the further usage. Future work will focus on further refining the model, testing it with more extensive data collections, and applying the techniques of explainable AI to enhance the paper's clinical applicability.