Design and evaluation of an automated algorithm for cardiac risk stratification
MetadataShow full item record
This item's downloads: 1 (view details)
In the United States, sudden cardiac death (SCD) claims approximately 450,000 lives each year, more lives than stroke, lung and breast cancer and AIDS combined. In the United Kingdom and Ireland, the combined figure is approximately 105,000 lives each year. SCD is a condition in which people experience sudden uncontrolled lethal arrhythmias. In the majority of reported cases, death is due to the onset of ventricular tachycardia and ventricular fibrillation. The population at risk of SCD is broad, it may include younger adults and the elderly. However, the population that is considered at higher risk of SCD are the post-myocardial infarction and heart failure patients. The existing therapy for patients at risk of SCD is a combination of anti arrhythmic drugs and the use of the Implantable Cardioverter Defibrillator (ICD). A number of medical guidelines exist to identify patients at risk but despite their availability the accurate selection of patients that actually benefit from ICD therapy is difficult. In recent years, the use of ICD devices has been controversial because of the associated costs of the therapy and the adverse health risks to patients. An important clinical study1 reported that during a 5 year follow up period, 518 (62.5%) out of 829 heart failure patients with implanted ICD devices did not experience any defibrillator shocks. The findings from this study contradicted the observations made by the five major ICD survival trials conducted in the early 1990¿s, which concluded that heart failure patients are the main population that benefits from ICD therapy. The findings from this study suggested that ICD therapy was not needed in the majority of heart failure patients and that the current ICD selection guidelines are too broad. In this thesis, we describe the design, evaluation and implementation of a novel diagnostic algorithm for risk stratification of sudden cardiac death. Our algorithm belongs to a class of non-invasive heart rate variability methods which are used by researchers and cardiologists in clinical studies. The clinical motivation and benefit in introducing such an algorithm are two-fold: (1) suspected high risk patients can be more accurately selected for ICD therapy, (2) identifying those patients that would benefit from ICD therapy from those who do not would prevent unnecessary procedures which would in turn reduce therapy costs and health risks to patients. As part of our research, we conducted two evaluations which tested the classification accuracy and clinical utility of the algorithm using public ECG databases. The first evaluation consisted of 142 patients. The goal of this evaluation was to test the classification accuracy of the algorithm for discriminating low-risk and high-risk arrhythmia patients. The second evaluation consisted of 208 patients. The goal of this evaluation was to compare the classification accuracy of the algorithm against existing heart rate variability algorithms using public ECG recordings. In addition to these evaluations, we engineered two systems: (1) a smartphone application which implements our diagnostic algorithm, and (2) a software plat- form for running the algorithm remotely. As part of the smartphone application, we have implemented a novel QRS detector algorithm for extracting the R-R interval series from the ECG and validated the performance of the application using our diagnostic algorithm. In the second system, we implemented a distributed software platform for performing cardiac diagnostic testing using our algorithm and the smartphone application. We validated the scalability of the platform us- ing a configuration of several computer servers. In summary, the research contributions of this thesis are: (1) the design of a novel diagnostic algorithm for SCD stratification, (2) two evaluations which tested the clinical utility of the algorithm, (3) the development of a smartphone application which implements our diagnostic algorithm, (4) the design and implementation of a connected health platform for providing cardiac diagnostic services to patients.