Bayesian Classification Models for Premature Ventricular Contraction Detection on ECG Traces

Table 4

Comparisons between related works.

Comparison with other works
Work	Year	Features	Classifier	Classes	Acc	Se	PPV
Nazarahari et al. [8]	2015	Wavelet + distances measures	Multilayer perception	Normal, PVC, APC, paced, LBBB, RBBB	97.51	—	—
Martis et al. [9]	2013	QRS, bispectrum, PCA	SVM NN	N, LBBB, RBBB, APC, VPC	93.48	—	—
Afkhami et al. [10]	2016	RR interval, HOS, GMM	Decision trees, ensemble learnes	AAMI, all classification in MIT-BIH	99.7	100	100
Javadi et al. [11]	2013	Wavelet + morpho-logical and temporal features	Mixture of experts, negative correlation learning	N, PVC, other	96.02	92.27	79.4
Kamath [12]	2011	Teager energy functions in time and frequency domains	Neural network	N, LBBB, RBBB, PVC, paced beats	100	100	100
Martis et al. [13]	2013	DWT + PCA + ICA + LDA	SVM, NN, PNN	AAMI	99.28	—	—
Sharma and Ray [14]	2016	Hilbert–Huang transform, statistical features	SVM	N, LBBB, RBBB, PVC, paced, APC	99.51	99.36	100
Banerjee and Mitra [15]	2014	Cross wavelet transform	Heuristic classification	Abnormal versus normal	97.6	97.3	98.8
Oliveira et al. [16]	2016	Dynamic Bayesian networks	Dynamic threshold	PVC versus others	99.88	99	99
Work		FSC, SFE	QDA	NB, PVC, OB	98.3	100	98