| Summary: | 碩士 === 中華大學 === 資訊工程學系碩士班 === 93 === Electrocardiogram (ECG) is one of the most used tools for diagnosis of heart-related diseases due to its fast operational and noninvasive features. P wave, QRS complex, and T wave reflect the change of electrophysiological conditions of heart cells at various positions of heart. The clinician can diagnose different heart diseases basing on the characteristic of these wave forms, for example, Long QT syndrome, Acute Myocardial Infarction (AMI), and Atrial Fibrillation (Af), and so on. Generally, the R wave is the most prominent waveform within the ECG signal. After locating the R waves, the positions of P and T waves can found either. In addition, if the R waves can be located accurately, we can also calculate the R-R intervals, which then can be used for analysis of Heart Rate Variability (HRV). In recent years, many of researches have already put forward various methods to detect R wave, for example, filter banks, artificial intelligence algorithms, Hidden Markov Models (HMM), genetic algorithm and wavelet transform. Wavelet transform is the most promising method.
In this study we cooperate with Wei-Gong Memorial Hospital in Miao-Li County. About 10000 clinical SCP-ECG records have been collected from the emergency department of the Hospital. These records were decoded into text files through the SCP decoding program developed in our laboratory, and the ECG information and wave form data were stored within a database. Linked with the HIS and confirmed by clinical physicians, we have been able to construct disease-specific ECG Databases for AMI, Hyperkalemia, and Af, and so on. In order to analyze the features of ECG’s of these heart diseases, an R wave delineator was developed in MATLAB, which used a novel hybrid wavelet transform method (includes wavelet packet analysis and discrete wavelet transform (DWT)) to detect R waves. The results showed that the sensitivities of R wave detection were 100%, 99.51%, 99.72%, 99.65% for normal, AMI, Hyperkalemia, and Af ECG’s respectively; and the positive predictive value of R wave detection were 100%, 99.46%, 99.66%, 99.88% for previous four categories of ECG records.
As shown in the above results, the algorithms developed in this study can be applied directly to clinical 12-lead ECG records for waveform analyses with high accuracy of R wave detection in various leads and diseases, regardless of interferences embedded in an ECG record such as baseline wandering, muscle contraction noise, and patient movement. The R wave detecting tool can also be applied to Holter ECG systems for wave form analysis because of its robust ability for processing single-lead ECG signals.
|
|---|
