| dc.description.abstract | Sudden cardiac death (SCD) is responsible for 200,000-450,000 adult deaths each year in the United States. Since sudden cardiac arrest (SCA) can happen unexpectedly, implantable-cardioverter defibrillators (ICDs) are inserted into patients who are at high risk of SCA so that they can provide immediate defibrillation when SCAs occur. Even though ICDs can be life-saving, there are still many problems that need to be solved. Firstly, how does one determine whether a person should receive an ICD? An ICD installed but never used is a waste of resources. On the other hand, if the patients need ICDs but do not get them, very likely they will lose their lives through SCDs. Secondly, ICDs do not prevent life-threatening cardiac arrhythmias (LTCAs), but simply terminate such arrhythmias after they have occurred. As a result, the patients suffering from these arrhythmia can be in danger if, for example, they are driving. It would be ideal if ICDs can issue warnings for impending LTCAs. Last but not least, shocks are very painful and decrease the quality of life of patients. If one can predict the onset of these arrhythmias, it may be possible to treat the patients with pacing or modulation of autonomic nervous system thus can decrease the number of shocks received by patients. To solve these problems, we hypothesized that the patients’ R-R interval statistics can be used for risk stratification for SCAs and prediction of SCAs. In addition, algorithms from machine learning were used to predict the occurrences of SCA with R-R interval statistics and demographic information of patients as features. Our study sample consists of patients who enrolled in Sudden Cardiac Death – Heart Failure Trial (SCD-HeFT). Our work shows that R-R interval statistics, particularly the short-term and long-term fractal scaling exponents from detrended fluctuation analysis (DFA), are indeed correlated to the occurrences of SCAs. Such findings certainly will aid the patient selection for receiving ICDs and will help create a new generation of ICDs which can issue warnings for the occurrences of SCAs. | |
