Investigation of Applying Machine Learning and Hyperparameter Tuned Deep Learning Approaches for Arrhythmia Detection in ECG Images

Table 1

Literature study.

Author	Model	Approach	Dataset	Evaluation
Acharya et al. [12]	Convolutional Neural Network	Deep CNN with 11 input layers and 4 output neurons	MIT-BIH	Accuracy—92.5%
Isin and Ozdalili [13]	Artificial Neural Network and transferred deep learning	Transferred deep CNN is used to extract features and then applied Artificial Neural Network (ANN)	MIT-BIH	Accuracy—92%
Zubair et al. [14]	Convolutional Neural Network and LSTM	Dropout regularization	MIT-BIH	Accuracy—91.8%; Appl. Sci. 9, 14 (2019), 2921. DOI: 10.3390/app9142921
Ribeiro et al. [15]	Deep Neural Network	Stacked transformations	CODE	F1-Score—above 80%
Porumb et al. [16]	CNN	Multilayer perceptron for raw signal classification	MIT-BIH(250 samples)	Accuracy—97%
Khan et al. [17]	Deep Neural Network	MobileNet	Manual dataset	Accuracy—98%
Atal and Singh [18]	Deep CNN	Bat-Rider Optimization algorithm	MIT-BIH	Accuracy—93.19%
Zheng et al. [19]	Ensemble models	Hyper-tuned classification model	Manually generate dataset	F1-Score—97%
Mathunjwa et al. [20]	CNN	Hyperparameter tuning	MIT-BIH	Accuracy—95. 3%
Jun et al. [21]	CNN	AlexNet, VGGNet	MIT-BIH	Accuracy—99.05%
Hu et al. [7]	CNN-Transformer-based model	Classification and positioning	MIT-BIH	Accuracy—99.49%
Yıldırım et al. [22]	1D-CNN	ECG signal fragments based on one lead	MIT-BIH	Accuracy—91.33%
Li et al. [23]	SE-ResNet deep learning model	19-layer deep squeeze-and-excitation residual network	MIT-BIH	Accuracy—99.61%
Sharma et al. [10]	LSTM	Rhythm-based method	MIT-BIH	Accuracy—90.07%
Simonyan and Zisserman [24]	Deep learning	ConvNet	ILSVRC-2012	Accuracy—93.2%
Damaševičius et al. [25]	Machine learning	KNN	Time series dataset	Accuracy—86%
Naz et al. [26]	Deep learning	AlexNet, VGG-16, Inception-V3	MIT-BIH	Accuracy—97.6%