From:  Multi-class valvular heart disease classification and severity detection with enhanced accuracy using convolutional neural network

 Method and performance comparison.

StudyYearMethodSounds (n)Classes (n)Results (%)
[16]2020Auto-encoder Neural Networks449NormalAcc = 100.00
Noisy
NormalSen = 100.00
Extrasystole
MurmurSpe = 100.00
Noisy murmur
409AbnormalAcc = 99.8
NormalSen = 99.65
Spe = 99.13
[17]2020MFCC features, CNN, recurrent neural networks3,240AbnormalAcc = 98.34
Spe = 98.01
NormalSen = 98.66
[18]2020Custom CNN architecture1,081AbnormalAcc = 85.65
NormalSpe = 86.73
Sen = 84.75
[19]2019Custom CNN architecture8,272AbnormalAcc = 89.22
NormalSpe = 89.94
Sen = 86.35
[20]2022Data augmentation, spectrogram, MFCC, chromagram, CNN656Normal
Noisy
Normal
ExtrasystoleAcc = 97
Murmur
Noisy murmur
Unlabeled
[21]2019MFCC statistical and frequency features, decision tree XGBoost3,240AbnormalAcc = 92.9
NormalSpe = 94.5
[22]2021Time domain, MFCC, artificial neural network, linear discriminant analysis3,126AbnormalAcc = 93.33
Normal
[23]2020Deep WaveNet model1,000AS
MRAcc = 97.0
MS Sen = 92.5
MVPSpe = 98.1
Normal
[24]2019Spectrogram and custom 1D CNN13,015Abnormal Acc = 99.01
NormalFscr = 99.10
[25]2021Bi-directional LSTM, CNN1,000AS
MRAcc = 99.32
MSSen = 98.30
MVPSpe = 99.5
Normal
[26]2022DSTP, DWT, INCA, SVM10,366Severe aortic stenosis
Mild aortic stenosisAcc = 99.58
Mild mitral stenosisPre = 99.55
Moderate mitral stenosisRec = 99.59
Severe mitral regurgitationFscr = 99.57
Mild mitral regurgitation
Moderate mitral regurgitation
Severe tricuspid
Moderate tricuspid
Healthy
Proposed methodCNN10,366Severe aortic stenosis
Mild aortic stenosis
Mild mitral stenosisAcc = 99.81
Moderate mitral stenosisPre = 99.84
Severe mitral regurgitationRec = 99.82
Mild mitral regurgitationFscr = 99.85
Moderate mitral regurgitation
Severe tricuspid
Moderate tricuspid
Healthy
Return to article view

Acc: accuracy; Sen: sensitivity; Spe: specificity; Pre: precision; Rec: recall; Fscr: F1 score.