Synthetic Generation as Denoiser for P Mitrale Contour Extraction from a Noisy Electrocardiogram Dataset
Abstract
Akin to software as service, synthetic generation as denoiser addresses explain ability while detecting left atrial enlargement (LAE) in a noisy ECG. In our approach to denoise the ECG, we have taken an unconventional look at synthetic generation. Our objective caters to bring down cost per function by synthetic generator mimicking principles of cardiac atrial depolarization. To this effect, our results demonstrate the effective use of Gaussian function convolution for right and left atrial depolarization vector. Gaussian lobe inherently captures the underlying signal characteristics favoring explainable computation, which served our objective as well. The hump tracer and spread analyzer algorithms showcase simple, effective way to extract parameters from a noisy ECG dataset. This also opens up a portal for patient specific modeling solutions, which have gained much traction. We demonstrate our approach by delineating P contour and detecting the P mitrale (pathology) pattern from the PTB-XL (ECG) dataset. The low compute complexity of our approach O(K), K discrete samples, ensures real time process amenability. Our contribution also brings out trade-offs in generalized denoising solutions such as Empirical mode decomposition (EMD) and neural schemes as compared with tailored solutions such as synthetic generation as denoiser (SGAD).