Table 2.

Comparison of automated STEMI detection methods.

Study	Modality	Wearable	Method	Patients	Duration of	Sensitivity	Specificity
		Device		(STEMI/Control)	Recordings
Heden et al., (1997) [12]	12-lead ECG	No	ANN	1120/10,452	NA	46.2%	95.4%
Heden et al., (1997) [12]	12-lead ECG	No	Rule-based	1120/10,452	NA	30.7%	95.4%
Haraldsson et al., (2004) [13]	12-lead ECG	No	Bayesian ANN	1119/1119	NA	63.3%	85.0%
Haraldsson et al., (2004) [13]	12-lead ECG	No	Hermite functions	1119/1119	NA	61.5%	85.0%
Haraldsson et al., (2004) [13]	8-lead ECG	No	Bayesian ANN	1119/1119	NA	56.3%	85.0%
Haraldsson et al., (2004) [13]	8-lead ECG	No	Hermite functions	1119/1119	NA	59.3%	85.0%
Green et al., (2006) [14]	12-lead ECG	No	ANN ensemble	130/504	NA	95.0%	41.4%
Green et al., (2006) [14]	12-lead ECG	No	Logistic regression	130/504	NA	95.0%	33.7%
Olsson et al., (2006) [15] *	12-lead ECG	No	Feed-forward ANN	736/3264	NA	95.0%	88.0%
Tripathy et al., (2019) [16]	12-lead ECG	No	Fourier-Bessel DNN	100/52	24 h	99.9%	99.6%
Van Heuverswyn et al., (2019) [17]	3-lead ECG	Yes	Rule-based (ST-seg.)	59 (5011 rec.)	NA	87–100%	96.0%
Spaccarotella et al., (2020) [18]	1-lead ECG x 9	Yes	Human expert	54/19	5.8 min	93.0%	95%
Ours **	MCG (1-lead ECG)	Yes	Logistic regression	41/49	15 min	85.7%	73.9%

* transmural ischemia only, ** sinus rhythm recordings only.