A pioneering approach for early prediction of sudden cardiac death via morphological ECG features measurement and ensemble growing techniques

Abstract

Sudden cardiac death (SCD) is a devastating cardiovascular condition that occurs suddenly within 1 hour of onset, usually without warning. The primary cause is a disruption in the heart's electrical system, leading to the cessation of blood flow and oxygen delivery to vital organs. Despite medical advancements, SCD prognosis remains poor, necessitating risk identification for lifesaving interventions. Hence, in this study, we analyse the morphological changes in electrocardiogram (ECG) signals associated with various cardiac conditions, including SCD and other conditions that can lead to SCD development. The ECG signals were pre-processed using a two-stage filter technique involving wavelet transform (WT) and progressive switching mean filter (PSMF) to eliminate noise and outliers. The denoised signals were then segmented and utilized for extracting temporal and amplitude features related to the P-wave, QRS complex, and T-wave components. These extracted features are further refined and given to the novel Ensemble Growing (EG) technique, which enhances the classification accuracy of different cardiac conditions. Examination of experimental findings revealed that the temporal features play an important role in the development of SCD. In particular, the prolonged durations of t_P-wave, t_QRS complex, t_T-wave, t_PpRp, t_RpSp, t_RpTp, t_PpQp, t_PpSp,t_PpTp, t_QpSp, and t_QpTp are closely associated with SCD. Furthermore, by incorporating significant temporal and amplitude features along with EG technique, produced an impressive SCD prediction accuracy of 99.82 % for 1 hour before its onset. This method offers advantages, including efficient handling of multiple cardiac conditions and real-time predictions, representing a major advancement towards proactive cardiac care and early SCD prediction.