Brief Review — Classifier identification using deep learning and machine learning algorithms for the detection of valvular heart diseases

Modified Xception

Sik-Ho Tsang
3 min readNov 23, 2023

Classifier identification using deep learning and machine learning algorithms for the detection of valvular heart diseases
Modified Xception
, by Haldia Institute of Technology, Eureka Scientech Research Foundation, Indian Institute of Technology (ISM)
2022 J. BEA (Sik-Ho Tsang @ Medium)

Heart Sound Classification
2013
2021 [CardioXNet] 2022 [CirCor] [CNN-LSTM] [DsaNet] 2023 [2LSTM+3FC, 3CONV+2FC] [NRC-Net]
==== My Other Paper Readings Are Also Over Here ====

  • The extracted features are Root Mean Square, Energy, Power, Zero Crossing Rate, Total Harmonic Distortion, Skewness, and Kurtosis in the time domain.
  • The modified CNN-based Xception is used for classification.

Outline

  1. Datasets & Feature Extraction
  2. Modified Xception
  3. Results

1. Datasets & Feature Extraction

1.1. Datasets

Left: Yaseen GitHub, Middle: PASCAL Dataset B, Right: PhysioNet
  • 3 datasets are used, which as shown above.
  • A bandpass filter having a bandwidth of 30 to 500 Hz is used to get rid of the noise.
  • A window frame of 5 s is selected and kept as fixed for every heart sound sample undergoing preprocessing.
  • Each dataset has been divided into training data (85%) and test data (15%).
  • Further, training data is decomposed into validation data (15%) and the rest for training the model.

1.2. Feature Extraction

  • Features of the heart sound considered for the entire study has been limited to:
  1. Root Mean Square (RMS)
  2. Signal Energy and Power
  3. Zero-Crossing Rate (ZCR)
  4. Total Harmonic distortion(THD)
  5. Skewness and Kurtosis

2. Modified Xception

Modified Xception
Modified Xception
  • Following Deep Learning Methods are used to classify the above features:
  1. LeNet-5
  2. AlexNet
  3. VGG16
  4. VGG19
  5. DenseNet121
  6. Inception Net (There are several Inception versions …)
  7. Residual Network
  8. Proposed Modified Xception Network (as shown above)

3. Results

3.1. Yaseen GitHub Dataset

Yaseen GitHub Dataset

Xception (The modified one I believe) obtains the highest training and validation accuracies.

Other Metrics
  • (I believe the above table shows the test set results.)

Again, Xception (The modified one I believe) obtains the highest accuracy.

3.2. ML Method Results

  • (There are results for traditional ML methods, please read the paper directly for more details.)

--

--

Sik-Ho Tsang

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