Internet of things-based health monitoring system for early detection of cardiovascular events during COVID-19 pandemic

Table 1 The most important features of electrocardiogram signals in the UCI dataset

Features	Values
Age	Yr
Sex	Male = 0, female = 1
Height	cm
Weight	Kg
QRS length	Average QRS length in milliseconds
Distance P-R	Average time interval between the start of waves P and Q in milliseconds
Distance Q-T	Average time interval between start of wave Q and end of wave T in milliseconds
Distance T	Average time interval of wave T in milliseconds
Distance P	Average P wave distance in milliseconds
QRS	Degree vector angles on the screen
T	Degree vector angles on the screen
P	Degree vector angles on the screen
QRST	Degree vector angles on the screen
J	Degree vector angles on the screen
Heart rate	Heart rate per minute

Table 2 Cardiac arrhythmia classes in the UCI dataset

Class No.	Class name	Number of classes
C1	Normal	245
C2	Ischemic changes (coronary artery diseases)	44
C3	Old anterior myocardial infarction	15
C4	Old inferior myocardial infarction	15
C5	Sinus tachycardy	13
C6	Sinus bradycardy	25
C7	Ventricular premature contraction (pvc)	3
C8	Supraventricular premature contraction	2
C9	Left bundle branch block	9
C10	Right bundle branch block	50
C11	1 Degree antrioventricular block	0
C12	2 Degree AV block	0
C13	3 Degree AV block	0
C14	Left ventricule hypertrophy	4
C15	Atrial fibrillation or flutter	5
C16	Others	22

Table 3 The confusion matrix

		True results
		Positive	Negative
Test results	Positive	TP	FP
	Negative	FN	TN

TP: True positive; TN: True negative; FP: False positive; FN: False negative.

Table 4 Long short-term memory model training and test times with/without rough set theory feature selection

Time	LSTM	RST-LSTM	Time reduction
Training	217154 ms	69247 ms	68.11%
Test	23854 ms	3856 ms	83.83%

LSTM: Long short-term memory; RST: Rough set theory.

Table 5 Positive prediction value of detection of the proposed system by cardiac arrhythmia classes

Class No.	LSTM	RST-LSTM	Class No.	LSTM	RST-LSTM
C1	97.65	98.44	C9	NaN	NaN
C2	89.54	90.23	C10	98.49	99.83
C3	98.76	99.08	C11	NaN	NaN
C4	99.14	99.12	C12	NaN	NaN
C5	98.26	98.74	C13	NaN	NaN
C6	94.29	96.63	C14	98.16	98.94
C7	NaN	NaN	C15	NaN	NaN
C8	NaN	NaN	C16	87.63	89.90
Average PPV	LSTM	95.76	Average PPV	RST-LSTM	96.77

NaN: Not a number; LSTM: Long short-term memory; RST: Rough set theory; PPV: Positive prediction value.

Table 6 Negative prediction value of detection of the proposed system by cardiac arrhythmia classes

Class No.	LSTM	RST-LSTM	Class No.	LSTM	RST-LSTM
C1	96.74	98.14	C9	NaN	NaN
C2	84.27	86.45	C10	97.45	98.32
C3	98.79	99.16	C11	NaN	NaN
C4	97.56	98.87	C12	NaN	NaN
C5	98.23	98.12	C13	NaN	NaN
C6	90.29	92.64	C14	98.34	99.05
C7	NaN	NaN	C15	NaN	NaN
C8	NaN	NaN	C16	83.11	85.36
Average NPV	LSTM	93.86	Average NPV	RST-LSTM	95.12

NaN: Not a number; LSTM: Long short-term memory; RST: Rough set theory; NPV: Negative prediction value.

Table 7 Sensitivity of detection of the proposed system by cardiac arrhythmia classes

Class No.	LSTM	RST-LSTM	Class No.	LSTM	RST-LSTM
C1	98.54	99.16	C9	NaN	NaN
C2	86.73	88.57	C10	98.24	98.65
C3	99.54	99.52	C11	NaN	NaN
C4	97.98	98.86	C12	NaN	NaN
C5	98.36	98.87	C13	NaN	NaN
C6	89.56	92.19	C14	97.92	99.03
C7	NaN	NaN	C15	NaN	NaN
C8	NaN	NaN	C16	81.57	82.87
Average sensitivity	LSTM	94.27	Average sensitivity	RST-LSTM	95.30

NaN: Not a number; LSTM: Long short-term memory; RST: Rough set theory.