TABLE II.

Automatic regression methods for the potassium concentration. CNN: convolutional neural network, $T_{S/A}$: T downslope divided by T amplitude $T_{S/\sqrt{A}}$: T downslope divided by the square root of T amplitude, result: mean ± standard deviation of signed errors in mmol/l, result (abs): mean ± standard deviation of absolute values of errors in mmol/l, N_Pat: number of patients, N_Sess: number of HD sessions, n/r: not relevant if data were not from HD sessions, and n/a: not available/given.

Work	Lead (s)	Features	Model	Result (mmol/l)	Result (abs) (mmol/l)	Dataset (mmol/l)	NPat	NSess
Corsi et al.13	PCA	$T_{S/A}$	Polynomial second order	−0.09 ± 0.59	0.46 ± 0.39	n/a	45	128
Attia et al.81 personalized	V3−V5	$T_{S/\sqrt{A}}$	Polynomial first order	n/a	0.36 ± 0.34	4.2 ± 0.95	26	113
Attia et al.81 global	V3–V5	$T_{S/\sqrt{A}}$	Polynomial first order	n/a	0.50 ± 0.42	3.9 ± 0.8	26	113
Yasin et al.82	I	$T_{S/\sqrt{A}}$	Polynomial first order	n/a	0.38 ± 0.32	4.3 ± 0.8	18	n/a
Lin et al.15	12-Lead	CNN	CNN	n/a	0.53 ± n/a	n/a	40 180	n/r