Big picture of ECG signal processing and machine learning based cardiac disease risk prediction and research directions

Cardiac Health Problems are one of the top main causes of death all over the world. Though Electrocardiograpy (ECG) analysis, Imaging Technics and Blood Tests are the methods to diagnose the problem and detect probable risks, due to spontaneous and sudden progression nature of cardiac diseases, monitoring the status of the heart before and after hospitalisation of the patient is of great importance. Recent Signal Processing and Machine Learning studies on ECG, regarding arrhythmia classification and sudden cardiac death prediction are footsteps of a revolution on cardiology. With the attained risk prediction and disease classification accuracy rates, such as 96.7% for Sudden Cardiac Arrest Risk (Murukesan et al., 2014), 99, 13% for Acute Chronary Syndrome diagnosis(Berikol et al., 2016), Cardiac Health problems and relevant sudden deaths seem to significantly diminish by means of preventive ECG monitoring in the future. Another medical domain which makes use of ECG Signal Processing and Machine Learning is the prediction of Sepsis like diseases, especially used in preterm neonatal disease prediction. Due to randomized control trails such as the one carried out on around 3000 patients in preterm neonatal infants (Moorman et al.,.2011), some of the academic studies started to be used in pediatric clinical routies. On the other hand, cardiac based studies are not prevalent in clinical routines yet due to lack of consensus regarding the most adequate methods. Features of ECG signal are significant clues regarding the prognosis of cardiac diseases. More effective research on ECG signal processing for feature extraction, with machine learning to extract the patterns relevant to cardiac diseases and randomized control trials seem to speed up this consensus.