Session S84.4

An Improved Spectral Method of Detecting and Quantifying T-Wave Alternans

TW Shen, YT Tsao*

Tzu Chi University
Hualien, Taiwan

Objectives: T-wave alternans (TWA) was first observed by H.E. Hering one hundred years ago. Basically, TWA is a pattern in the ECG characterized by two distinct forms of T-waves appearing in alternation. Currently, TWA is widely understood as an important indicator and a predictor of risk of sudden cardiac death (SCD) which is responsible for an estimated millions of deaths worldwide. This research provides a method of detecting and quantifying T-Wave alternans in order to predict SCD events. The algorithm were developed by using fifty records from MIT/BIH databases, including Long-Term ST Database (18 records), Sudden Cardiac Death Holter Database (17 records), and BIDMC Congestive Heart Failure Database (15 records). Finally the T-Wave Alternans Challenge Database (TACD) was used for testing the performance of the algorithm.
Methods: An improved spectrum method (SM) is developed to estimate TWA. First, the preprocessing stage was involved to remove arterial interference and to limit the ECG bandwidth between 1Hz and 50Hz. Second, Pan and Tompkins method is used to find R waves as indexes to obtain T waves. Once the R points are found, most of T points would appear within a 450ms period backward from each R point. Those T points can be detected by searching extreme values of all the zero-crossing points of the first derivative ECG (dECG). Third, totally 128 T waves were selected from each record to generate beat-to-beat fluctuations in its amplitude. Unlike traditional alternans ratio method, we enhanced the power spectra by summarizing of weighted an aggregate spectrum, a spectrum at 0.1 second before T points, and a spectrum at 0.1 second after T points, and by applying AR model. Finally, the so-called enhanced alternans ratio was generated to evaluate the risk of SCD.
Results & Discussions: According to developing databases, the traditional alternans ratio only picks up 17 out of 50 subjects with ratio higher 2.5. By comparison, our enhanced alternans ratio detected 23 out of 50 subjects with the same threshold. For the TACD database, our method detected 42 out of 100 subjects with ratio higher 2.5. Moreover, the TWA estimation for subjects in the SCD database has the overall higher rank than other database. Hence, we expect that our enhanced alternans ratio has better effect to evaluate the risk of SCD than the traditional SM method.

(Abstract Control Number: 370)