Cardiovascular disease (CVD) is the single leading cause of death in both developed and developing countries. The most deadly CVD is heart attack, which 7,900,000 Americans suffer each year, and 16% of cases are fatal. The Electrocardiogram (ECG) is the most widely adopted clinical tool to diagnose and assess the risk of CVD. Early diagnosis of heart attacks, by detecting abnormal ST segments within one hour of the onset of symptoms, is necessary for successful treatment. In clinical settings, resting ECGs are used to monitor patients automatically. However, given the sporadic nature of heart attacks, it is unlikely that the patient will be in a clinical setting at the onset of a heart attack. While Holter-based portable monitoring solutions offer 24 to 48-hour ECG recording, they lack the capability of providing any real-time feedback for the thousands of heart beats they record, which must be tediously analyzed offline.Processing ECG signals on a smartphone-based platform would unite the portability of Holter monitors and the real-time processing capability of state-of-the-art resting ECG machines to provide an assistive diagnosis for early heart attack warning. Furthermore, smartphones serve as an ideal platform for telemedicine and alert systems and have a portable form factor. To detect heart attacks via ECG processing, a real-time, accurate, context aware ST segment monitoring algorithm, based on principal component analysis and a support vector machine classifier is proposed and evaluated. Real-time feedback is provided by implementing a state-of-the-art, multilevel warning system ranging from audible notifications to text messages to points of contacts with the GPS location of the user. The smartphone test bed makes use of a novel, real-time verification system using a streaming database to analyze the strain of heart attack detection system under normal phone operation. Furthermore, the entire system is prototyped and fully functional, running on a smartphone to demonstrate the real-time, portable functionality of the platform. Experimental results show that a classification accuracy of 96% for ST segment elevation of individual beats can be achieved and all ST episodes were correctly detected during testing with the European ST database.
展开▼