An Effective LSTM Recurrent Network to Detect Arrhythmia on Imbalanced ECG Dataset

Table 8

Comparison between the related work and the method proposed in this work.

Works	Year	Classes	Methods	ACC (%)	RE (%)	SP (%)
Martis et al. [35]	2013	5 beat types	DCT + PCA, PNN	99.52	98.69	99.91
Raj et al. [7]	2016	16 beat types	DOST, SVM-PSO	99.18	—	—
Sharma and Ray [36]	2016	6 beat types	EMD, HHT, SVM	99.51	98.64	99.77
Gutiérrez-Gnecchi et al. [37]	2017	8 beat types	PNN	98.89	—	—
Jung and Lee [38]	2017	4 beat types	WKNN	96.12	96.12	99.97
Li et al. [6]	2017	6 beat types	GA-BPNN	97.78	97.86	99.54
Rajesh and Dhuli [25]	2018	5 beat groups	DBB, AdaBoost	99.10	97.90	99.40
W. Li and J. Li [14]	2018	16 beat types	LDP, DNN	98.37	—	—
Yildirim. [20]	2018	5 beat types	DULSTM-WS2	99.25	—	—
Oh et al. [22]	2018	5 beat types	CNN-LSTM	98.10	97.50	98.70
Pławiak and Acharya [39]	2019	17 classes	DGEC	99.37	94.62	99.66
Yildirim et al. [21]	2019	5 beat types	LSTM	99.23	99.00	99.00
Our work	2019	8 beat types	LSTM, FL	99.26	99.26	99.14

DCT: discrete cosine transform; GMM + EM: Gaussian mixture modeling with enhanced expectation maximization; DOST: discrete orthogonal stockwell transform; SVM-PSO: PSO-tuned support vector machine; EMD: empirical mode decomposition; HHT: Hilbert–Huang transform; PNN: probabilistic neural network; WKNN: weighted k-nearest neighbor; NRSC: neighborhood rough set; DWT: discrete wavelet transform; GA-BPNN: genetic algorithm-backpropagation neural network; DNN: deep neural network; DULSTM-WS: deep unidirectional LSTM network-based wavelet sequences; DBLSTM-WS: deep bidirectional LSTM network-based wavelet sequences; LDP: local deep field; DBB: distribution-based balancing; FL: focal loss; DGEC: deep genetic ensemble of classifiers.