CWT Scalogram+AlexNet

Sik-Ho Tsang

·

Follow

2 min read

·

Jan 13, 2024

--

Listen

Share

Classification of short unsegmented heart sound based on deep learning
CWT Scalogram+AlexNet, by NERIST; Tripura University
2019 I2MTC, Over 20 Citations (Sik-Ho Tsang @ Medium)

Heart Sound Classification
2013 … 2023 [2LSTM+3FC, 3CONV+2FC] [NRC-Net] [Log-MelSpectrum+Modified VGGNet] [CNN+BiGRU]
==== My Other Paper Readings Are Also Over Here ====

• The main aim of this paper is to eliminate the segmentation process and to measure the benefit for accurate and detailed classification of short unsegmented 5 second PCG recordings.
• CWT scalogram is extracted as features input to AlexNet for normal and abnormal heart sound classification.

Outline

1. CWT Scalogram+AlexNet
2. Results

1. Scalogram+AlexNet

1.1. Data

• PhysioNet 2016 Challenge dataset is used.
• The minimum 5-second of length from each of the PCG recordings is chosen.

1.1. Preprocessing

• Filtering: To eliminate high and low-frequency noise, a Butterworth bandpass filter is applied with a cut off frequency of 25 Hz and 400 Hz [3] as illustrated in Fig. 2(a).
• Spike removal: To remove the redundant part of the amplitude of PCG also known as the frictional spike, Naseri et al. [1] is used to recognize and eliminate the spikes as illustrated in Fig. 2(b).

1.2. CWT Scalogram

• One scalogram image is generated for every PCG recording.
• Hence a total of 3240 scalogram images were produced for training and validating the CNN model.

1.3. AlexNet Model

• AlexNet is used in which the final layer is replaced with binary classifier as the primary target is to classify between abnormal and normal heart sound.

2. Results

The proposed method achieved a validating accuracy of 90% with a sensitivity of 90% and specificity of 90% respectively.

The proposed method improves in overall performance compared with other methods.