Crow search optimization with deep transfer learning enabled ventricular fibrillation prediction model

Ventricular Fibrillation (VF) is a severely hazardous arrhythmia and it occurs just before cardiac arrest. Accurate and early prediction of VF may help save many lives. To classify an electrocardiogram (ECG) signal as normal or VF, this study offers a model that uses the U-Net deep transfer learning (DLT) method in conjunction with Crow Search Optimization (CSO) applied to the Support Vector Machine (SVM). The model starts with delineating the filtered and segmented QRS segment on the U-Net algorithm by transfer learning approach. Further, extracting advanced heart rate variability (HRV) and QRS complex area features of segmented ECG signal to predict VF. Finally, CSO is used for parameter tuning of the SVM model. The findings reveal the supremacy of the proposed model, showing significant improvements in accuracy (ACC) and sensitivity (SV) contrasted with other contemporary approaches. This model may greatly help the researcher decide the timing for the start of implantable cardiac defibrillators.