Automated detection of atrial fibrillation using Bayesian paradigm
Knowledge-Based Systems
| Hi-index | 0.00 |
Usage of compressed ECG for fast and efficient telecardiology application is crucial, as ECG signals are enormously large in size. However, conventional ECG diagnosis algorithms require the compressed ECG packets to be decompressed before diagnosis can be performed. This added step of decompression before performing diagnosis for every ECG packet introduces unnecessary delay, which is undesirable for cardiovascular diseased (CVD) patients. In this paper, we are demonstrating an innovative technique that performs real-time classification of CVD. With the help of this real-time classification of CVD, the emergency personnel or the hospital can automatically be notified via SMS/MMS/e-mail when a life-threatening cardiac abnormality of the CVD affected patient is detected. Our proposed system initially uses data mining techniques, such as attribute selection (i.e., selects only a few features from the compressed ECG) and expectation maximization (EM)-based clustering. These data mining techniques running on a hospital server generate a set of constraints for representing each of the abnormalities. Then, the patient's mobile phone receives these set of constraints and employs a rule-based system that can identify each of abnormal beats in real time. Our experimentation results on 50 MIT-BIH ECG entries reveal that the proposed approach can successfully detect cardiac abnormalities (e.g., ventricular flutter/fibrillation, premature ventricular contraction, atrial fibrillation, etc.) with 97% accuracy on average. This innovative data mining technique on compressed ECG packets enables faster identification of cardiac abnormality directly from the compressed ECG, helping to build an efficient telecardiology diagnosis system.