Deep Recurrent Learning for Heart Sounds Segmentation based on Instantaneous Frequency Features

ARINI, PEDRO DAVID; ÁLVARO JOAQUIN GAONA

REVISTA ELEKTRON

ELEKTRON AMBAR

Año: 2020

2525-0159

In this work, a novel stack of well-known technologies is presented to determine an automatic method to segment the heart sounds in a phonocardiogram (PCG). We will show a deep recurrent neural network (DRNN) capable of segmenting a PCG into their main components and a very specific way of extracting instantaneous frequencythat will play an important role in the training and testing of the proposed model. More specifically, it involves an Long Short-Term Memory (LSTM) neural network accompaniedby the Fourier Synchrosqueezed Transform (FSST) used to extract instantaneous time-frequency features from a PCG. The present approach was tested on heart sound signalslonger than 5 seconds and shorter than 35 seconds from freely-available databases. This approach proved that, with a relatively small architecture, a small set of data and theright features, this method achieved an almost state-of-the-artperformance, showing an average sensitivity of 89.5%, anaverage positive predictive value of 89.3% and an averageaccuracy of 91.3%.