Brief Review — A Novel Deep Learning CNN for Heart Valve Disease Classification Using Valve Sound Detection

CWT+MFCC+DWT+CNN+MLP

Sik-Ho Tsang
3 min readDec 17, 2023
Cardiac acoustics and sounds from several phono cardiographic recordings.

A Novel Deep Learning CNN for Heart Valve Disease Classification Using Valve Sound Detection
CWT+MFCC+DWT+CNN+MLP
, by Princess Nourah bint Abdulrahman University, King Saud University, University of Louisiana at Lafayette
2023 MDPI J. Electronics (Sik-Ho Tsang @ Medium)

Heart Sound Classification
20132023 [2LSTM+3FC, 3CONV+2FC] [NRC-Net] [Log-MelSpectrum+Modified VGGNet]
==== My Other Paper Readings Are Also Over Here ====

  • The current work presents the development and application of deep convolutional neural networks for the binary and multiclass categorization of multiple prevalent valve diseases and typical valve sounds.
  • 3 alternative methods were taken into consideration for feature extraction: mel-frequency cepstral coefficients and discrete wavelet transform.

Outline

  1. Proposed CWT+MFCC+DWT+CNN
  2. Results

1. Proposed CWT+MFCC+DWT+CNN

1.1. Yaseen GitHub Dataset

  • The signals were from an Yaseen GitHub dataset with 200 entries for each class. These were converted to digital form using an 8 kHz sampling rate, with each record lasting at least one second.
  • Each record was divided into 7000 data points (0.88 s) to ensure consistency in the data throughout the analysis. These windows must each contain at least one full cardiac cycle.
Prior Arts on Yaseen GitHub Dataset
  • Some SOTA results on the same dataset are shown above.

1.2. Proposed CWT+MFCC+DWT+CNN

Proposed CWT+MFCC+DWT+CNN
  • 3 separate models are employed: CWT, MFCC, and DWT.
CNN
  • For CWT and MFCC, they go through the CNN for deep feature extraction.
  • For DWT, MLP is used for deep feature extraction.
  • The outputs from the 3 separate networks were combined in the second stage as input to a multilayer classifier.
  • For multiclass classification, a probabilistic ReLu is used. The activation function for binary classification is a sigmoid function.

2. Results

Precision, recall, F1 score, specificity, and accuracy

Using all 3 features obtains the best performance.

(Not much explanation in the paper for the result)
(Not much explanation in the paper for the result)
  • The entire model’s F1 scores and binary accuracy attained values just over 95% and 99%, respectively.

--

--

Sik-Ho Tsang

PhD, Researcher. I share what I learn. :) Linktree: https://linktr.ee/shtsang for Twitter, LinkedIn, etc.