A DEDICATED DIAGNOSTIC DISTORTION MEASURE FOR THE EVALUATION OF ECG SIGNAL COMPRESSION TECHNIQUES ADOPTED FOR REMOTE DIAGNOSIS OF CARDIOVASCULAR DISEASES

Abstract

In the current century diseases of the heart, brain and blood vessels, termed Cardiovascular Diseases (CVDs), represent the main death reasons [1-2]. The CVDs can be diagnosed through the usage of the features extracted from the P-waves, PR-intervals, QRS-complexes, ST-segments, T- and U-waves. The irregularity of beat phases is generally called arrhythmia and some arrhythmias are very dangerous for patients. The types of arrhythmias that can be diagnosed using the Electrocardiogram (ECG) features include: sinus bradycardia, ventricular tachycardia, sinus arrhythmia, atrial premature contraction. In addition, right and left bundle branch block, atrial fibrillation and flutter, heart block as well as coronary artery diseases such as ST- and non ST-elevation myocardial infarction can be diagnosed using ECG signal features. This paper introduces automatic ECG signal interpretations adopted for detecting CVDs. It investigates the effect of compressing the ECG signals on evaluating the performance of diagnosing CVDs. This has been performed by introducing a Dedicated Diagnostic Distortion Measure (3DM) that is based on comparing the complex features of the original and the reconstructed ECG signals while keeping a pre-determined diagnostic error and good reconstructed signal quality. In fact, the allowable pre-determined diagnostic error is a disease dependent. The diagnostic features considered in this paper include duration features, amplitude features and shape features. Consultations with cardiologists through Mean Opinion Score (MOS) test show that the proposed 3DM is most informative compared to the well-known percentage root-mean square difference in the evaluation of compression techniques oriented for CVDs diagnosing. The price paid for the usage of 3DM as a good performance metric is the computational complexity in calculating the diagnostic features. However, the calculated features can be adopted for diagnosing purposes.