Development and Validation of Embedded Device for Electrocardiogram Arrhythmia Empowered with Transfer Learning

Table 9

Proposed DVEEA-TL model compared with the state-of-the-art literature.

Studies	Hardware implementation	Data augmentation	Data fusion	Datasets	Method	Findings
Yeh et al. [1]	Yes	No	No	PTB DB	ResNet, AlexNet, and SqueezeNet	Predicted accuracies: 97%, 95%, and 75%
Wasimuddin et al. [2]	No	No	No	ECG-ID	CAD and machine learning	Predicted accuracy: 98.5%
Vijayakumar et al. [6]	No	No	No	No	Feature extraction to remove noise	Predicted accuracy: 96.5%
Hsu et al. [7]	No	No	No	MIT-DB	AlexNet and ResNet	Predicted accuracy: 94.4%
Gaddam and Sreehari [12]	No	No	No	MIT-DB	AlexNet	Predicted accuracy: 95.6%
Simjanoska et al. [14]	No	No	No	PTB DB	ML-train-validation-test evaluation	Predicted accuracy: 98%
Acharya et al. [15]	No	No	No	PTB DB	CNN layers	Predicted accuracy: 93.5% (for noise data); Predicted accuracy: 95.22% (for non-noise data)
Hammad et al. [17]	No	No	No	PTB	ResNet model	Predicted accuracy: 98.85%
Golany et al. [21]	No	No	No	MIT-DB	GAN-based model	Predicted accuracy: 97.5%
Sehirli et al. [28]	No	No	No	PTB-XL	RNN (LSTM and GRU)	Predicted accuracy: 97.7%
Strodthoff et al. [33]	No	No	No	PTB-XL	ResNet and inception	Predicted accuracy: 89.8%
Rahman et al. [45]	No	Yes	No	MIT-BIH	CAA-TL model (deep learning)	Predicted accuracy: 98.38%
Proposed DVEEA-TL model	Yes	Yes	Yes	BIH-RT (real-time dataset)	Transfer learning (AlexNet)	Training (99.9%)
						Validation (99.8%)