Interpretable hybrid model for an automated patient-wise categorization of hypertensive and normotensive electrocardiogram signals

Hypertension is critical risk factor of fatal cardiovascular diseases and multiple organ damage. Early detection hypertension even at pre-hypertension stage helpful in preventing the forthcoming complications. Electrocardiogram (ECG) has been attempted to observe changes electrical activities hearts hypertensive patients. To automate ECG assessment hypertension, an interpretable hybrid model proposed this paper. The framework consists one dimensional - Convolutional Neural Network architecture with four blocks convolutional layers, maxpooling followed by dropout layers fused Support Vector Machine classifier final layer. implemented made explainable using Local Interpretable Model-agnostic Explanations (LIME) method. developed trained tested for patient-wise classification ECGs online Physionet datasets hospital data. method achieved highest accuracy 81.81% datasets, 93.33% as normotensive hypertensive. visualization results showed only patient's misclassified (predicted) hypertensive, identification patient number, among 15 patients (8 7 hypertensive) tested. In addition, LIME demonstrated explanation predictions highlighting features location waveform responsible it, thus making decision more interpretable. Furthermore, our system assisting automated software tool called, HANDI (Hypertensive And Normotensive Detection Interpretability) real-time validation clinics early capture proper monitoring