Computer Based Detection of ECG Characteristic Wave Points

Elektrokardiogram (EKG) je zapis električne aktivnosti srca. Karakteristični dijelovi signala EKG-a su: P-val, QRS-kompleks i T-val. Detekcija tih karakterističnih dijelova predstavlja nužan faktor za proučavanje rada srca. U ovom radu predstavljene su osnove anatomije srca i opisana je električna aktivnost srca. Nakon toga, ukratko su opisani osnovni koraci implementacije većine algoritama detekcije karakterističnih veličina EKG-a. Opisana je fazorska transformacija kao metoda analize signala te je razvijen algoritam detekcije karakterističnih veličina signala EKG-a. Za obradu signala korištena je fazorska transformacija (Phasor transform). Predstavljene su najznačajnije baze signala EKG-a za testiranje algoritama detekcije karakterističnih veličina EKG-a te je provedena usporedba anotacija P valova za 12 signala preuzetih iz MIT-BIH baze signala za aritmiju u odnosu na Elgendijev zapis anotacija P valova. Nakon toga, provedeno je testiranje algoritma na standardnim bazama signala EKG-a za aritmiju, MIT-BIH i QT bazi signala. Uspješnost algoritma procijenjena je izračunom osjetljivosti (Se) i pozitivne prediktivnosti (P+) uz dopušteni interval pogreške između anotacija i detektiranih pozicija vrhova karakterističnih valova iznosa 150 ms. Nakon toga, provedena je usporedba iznosa osjetljivosti za različite iznose intervala pogreške nad signalima iz MIT-BIH i QT baze signala. Sam algoritam implementiran je u programskom okruženju Matlab.The electrocardiogram (ECG) is a record of the heart’s electrical activity. Typical parts of an ECG signal are: P wave, QRS complex and T wave. Detection of these characteristic parts of the ECG signal is an important factor for the study of heart functions. This paper presents the basics of heart anatomy and explains the electric activity of the heart. The basic steps of implementing most algorithms for detecting characteristic ECG waves are briefly described. Phasor transformation as a method of signal analysis is described and an algorithm for detecting characteristic parts of ECG is developed. Phasor transform was used for signal processing. The most important bases of ECG signals for testing algorithms of detection characteristic ECG waves were presented and a comparison of P waves annotations for 12 signals taken from MIT-BIH arrhythmia signal base in relation to Elgendi's P wave annotation record was performed. After that, the algorithm was tested on standard ECG signal bases for arrhythmia, MIT-BIH and QT signal base. The success of the algorithm was estimated by calculating the sensitivity (Se) and positive predictivity (P+) with an allowable error interval of 150 ms between the annotations and the detected positions of the peak of characteristic waves. After that, a comparison of the sensitivity for different error intervals over the signals from the MIT-BIH and QT signal base was performed. An algorithm is implemented within Matlab programming environment