AFibNet: an implementation of atrial fibrillation detection with convolutional neural network

Table 6 All performance of the AFibNet with several datasets

Dataset	Class	Number of subjects	Performance (%)
Dataset	Class	Number of subjects	Accuracy	Sensitivity	Specificity
Training and validation data
The 2017 PhysioNet/CinC challenge	N
China physiological signal challenge 2018	AF	8232	99.8	99.8	99.8
MIT-BIH atrial fibrillation
Unseen data testing
ECG long term AF	AF	38	100	100	–
Paroxysmal AF	AF	48	100	100	–
MIT-BIH Arrhythmia	AF	6	100	100	–
AF termination challenge	AF	10	100	100	–
Fantasia	N	24	100	100	–
Indonesian Hospital (ECG 1)	N	42	100	100	100
Indonesian Hospital (ECG 1)	AF	3
Indonesian Hospital (ECG 2)	AF	13	100	100	–
ECG recording from Chapman University and Shaoxing People’s Hospital	N	1646	98.86	98.88	98.88
	F	1780
All unseen data testing	N	1712	98.94	98.97	98.97
All unseen data testing	AF	1898

Training and validation dataset: The sample of data used to fit the and provide an unbiased evaluation of a model fit on the training dataset while tuning model hyperparameters. Unseen data: The unseen data can include data having an attribute not seen by the data set

ISSN: 1472-6947