A Combination of ECG Intervals and Logistic Regression in Automatic Detection of Coronary Heart Disease

Aim. ECG intervals precisely describe cardiac cycle. Its phases characterize the status of the heart in respect of deviation from the norm and detection of pathologies, risk of sudden death. The aim of the study was to find the best combination of heart rate, systolic, diastolic, and repolarization parameters of ECG wave for the detection of coronary heart disease.

Methods. 12-lead ECG at rest was recorded in three groups of males (n=129): I group – CHD-free (reference), II group – arterial hypertension patients, and III group – angina pectoris due to CHD. Coronary artery stenosis was confirmed by coronary angiography. RR, QT, JT, and TpTe intervals were analyzed by lead II with pronounced morphology of QRS complex and T-wave. Automatic measurements of 5 subsequent beat-to-beat intervals were performed to obtain an average. TQ was extracted by length of the RR and QT intervals from the same cardiac beat. Corrected JTc and QTc were calculated using Bazett method. A set of intervaltime series of patients with known diagnoses was used for logistic regression in order to perform binary classification for patients with unknown CHD.

Results. The accuracy of single measurements was significantly different (e.g. from 30% in QTc to 70% in RR and TQ interval in the II group). JTc+TQ combination was the most precise (above 74% in group II, and 100% in group III). Adding other parameters, such as TpTe reduced the accuracy of classification (e.g. JTc+TQ+ QTc – 54% in group II, 94% – in group III).

Conclusions. For early diagnosis of CHD, a combination of ECG signs should be used. Along with traditional parameters, it is important to analyze TQ interval, which characterizes the diastole. The results suggest that the combination of ECG ventricular repolarization and diastolic intervals is the informative marker of CHD.