Session PB9.1

Arrhythmia Detection Using Image Representation of Electrocardiogram

F Wang, T Syeda-Mahmood*, D Beymer

IBM Almaden Research Center
San Jose, CA, USA

There has been an increased need for automatic processing of the electrocardiogram (ECG), the rapid development of powerful microcomputers promoted the widespread application of such automatic algorithms, such as 12-lead ECG analysis in cardiological devices. These applications require an accurate detection of the heart rate by the ECG. Furthermore, heart rate estimation is a prerequisite step to identify arrhythmia diseases such as bradycardia and tachycardia, the ability to automatically identify these diseases from ECG recordings is important for clinical diagnosis and treatment. Much of the prior work has focused on determining the heart rate from digital ECG time series signals. There are a large number of ECGs still in paper form. Their digital records are created as scanned images. Not much prior art exists in the determination of period from ECG images obtained from scanner paper ECGs. Although many hospitals now have digital ECG records, much of the ECG data is in paper form. Unlocking these ECG records printed on paper and exposing them to digital analysis would be useful. Another source of ECGs is the synchronization ECG used in echocardiograms. While it is difficult to estimate the heart cycle directly from the depicted heart region in a video, it is relatively easy to estimate the heart rate from a synchronizing ECG. Hence, there is a need for a method and apparatus for determining a heart period using image representation obtained from an ECG.
In this paper, we present a computational algorithm as well as medical decision support tool for finding the heart period from ECG images. Specifically, we use the knowledge that when ECG trace is scanned or rendered in videos, the peaks of the waveform (R-wave) is often traced thicker due to pixel dithering. We exploit the pixel thickness information, for the first time, as a reliable feature for determining periodicity. Existing period estimation algorithm, such as estimation of R peaks as largest positive peaks often leads to detection error when R peaks are lower than the P or T peaks. While our method is relatively robust to these cases since our features latched on pixel dithering is more consistent among corresponding R peaks.
In our experiment, we used our algorithm to recognize the arrhythmia patients from the ECG database, which contains 16, 613 12-channel ECG scans. Our method found 2, 538 cases of bradycardia, which corresponds to an accuracy of 95.88%. It also found 1, 541 cases of tachycardia with an accuracy of 92.33%. The accuracy of our approach reached 94.5% for arrhythmia patients. In all cases accuracy was computed by comparing with disease labels from an expert following the American Heart Association guidelines.

(Abstract Control Number: 49)