Table 6.

Studies conducted to distinguish normal from CAD subjects using various signals.

Author	Data used	Method/features	Classifiers	Cross validation	Accuracy
using HSS signals
Karimi et al., 2005	5 CAD and 5 normal	DWT, WPD (some statistical features)	ANN	No	90%
Zhao and Ma, 2008	40 CAD and 40 normal	EMD, TEO (some statistical features)	BPNN	No	85%
using ECG signals
Lewenstein, 2001	479 CAD and 297 normal	(slope of an ST segment, blood pressure, load during the test)	Radial basis function neural networks	No	97%
Babaoǧlu et al., 2010a	480 CAD	principle component analysis	SVM	5-fold	79.1%
Babaoglu et al., 2010b	480 CAD	binary particle swarm optimization and genetic algorithm	SVM	5-fold	81.46%
Acharya et al., 2017c	7 CAD and 40 normal	Higher-Order Statistics and Spectra (HOS)	KNN,DT	10-fold	98.99%
Kumar et al., 2017	7 CAD and 40 normal	Flexible analytic wavelet transform (cross information potential)	LS-SVM	10-fold	99.6%
Normal and CAD using HRV signals
Lee et al., 2007	99 CAD, 94 Normal	Linear (time domain, frequency domain) and non-linear methods (Poincare plot, approximation entropy)	Support vector machine (SVM)	10-fold	90%
Lee et al., 2008	99 CAD, 94 Normal	Linear (time domain, frequency domain) and non-linear methods (Poincare plot, the hurst exponent, Detrended fluctuation analysis)	CPAR & SVM:	10-fold	85–90%
Dua et al., 2012	10 CAD and 10 normal subjects	Non-linear methods (recurrence plots, Shannon entropy) and principal component analysis (PCA)	multilayer perceptron (MLP)	5-fold	89.5%
Giri et al., 2013	10 CAD and 15 normal subjects	DWT and Independent Component Analysis (ICA)	Gaussian Mixture Model (GMM)	3-fold	96.8%
Patidar et al., 2015	10 CAD and 10 normal subjects	TQWT and PCA (correntropy)	LS-SVM	3-fold	99.72%
Kumar et al., 2016	10 CAD and 10 normal subjects	FAWT and entropy	LS-SVM	10-fold	100%
In this work	40 normal and 7 CAD subjects	RdisEn and WPD (statistical features)	KNN and SVM	10 times 10-fold	97.5%