Abstract. This paper discusses ECG classification after parametrizing the ECG waveforms in the wavelet domain. The aim of the work is to develop an accurate classification algorithm that can be used to diagnose cardiac beat abnormalities detected using a mobile platform such as smart-phones. Continuous time recurrent neural network classifiers are considered for this task. Records from the European ST-T Database are decomposed in the wavelet domain using discrete wavelet transform (DWT) filter banks and the resulting DWT coefficients are filtered and used as inputs for training the neural network classifier. Advantages of the proposed methodology are the reduced memory requirement for the signals which is of relevance to mobile applications as well as an improvement in the ability of the neural network in its generalization ability due to the more parsimonious representation of the signal to its inputs.
IntroductionEarly diagnosis of heart diseases enables patients to improve their quality of life through more effective treatments [1]. Analysis and classification of ECG signals can be particularly helpful to identify the initiation of heart conditions such as atrial fibrillation or flutter, multifocal atrial tachycardia, palpitations, paroxysmal supraventricular tachycardia, reasons for frequent fainting, slow heart rate (bradycardia) or ventricular tachycardia. Normally patients will be given a Holter monitor and wear the monitoring electrodes over a period of 24-48 hours. The data-logged signals are postprocessed and examined for cardiac beat abnormalities by doctors over the following days. There are restrictions, however, to how often one should perform such measurements. It is not uncommon for patients to complain to their doctors that the arrhythmias they suffered prior to the examination period were not present during the monitoring process, making early diagnosis more difficult. If one is prepared to wear the appropriately placed electrodes more often, it would be possible to use mobile phones as data logging devices directly. Bluetooth emitters such as the RN-42 from Microchip can provide a direct input from a small footprint battery-operated mobile data acquisition card to which the electrodes are interfaced. With the ever-increasing capabilities of smart-phones, portable ECG telemonitoring is likely to become a common feature for these devices, performing data-logging functionality for the ageing population. Beyond their data-logging functionalities, smart -phones can also use Multimedia Messaging Services (MMS) to enable the recorded signals to be sent directly for diagnosis by experts through current mobile networks or perform directly classification tasks. With the number of patients increasing due to current population sedative lifestyles and unhealthy eating habits, an increased expectation by patients for personalized medical treatment, as well as the envisaged wider proliferation of ECG data-logging devices, it is widely anticipated that there will soon be an overwhelming requirement for expert...