TABLE I.

Median (IQR) performance metrics of previous solutions in the literature, evaluated on the study dataset.

	Performance by segment			Performance by patient
Algorithm	F1 (%)	Se (%)	PPV (%)	F1 (%)	Se (%)	PPV (%)
Proposed algorithm	89.1 (70.8 – 99.6)	93.3 (75.0 – 100.0)	90.0 (68.5 – 100.0)	84.1 (69.0 – 93.9)	86.5 (71.6 – 95.1)	85.4 (68.3 – 94.7)
Preliminary design [28]	85.2 (66.4 – 96.4)	87.5 (64.0 – 100.0)	87.9 (68.8 – 100.0)	80.3 (65.2 – 90.7)	80.0 (59.6 – 91.5)	83.9 (69.0 – 93.4)
Risdal et al. [23]	75.0 (53.3 – 90.0)	94.1 (75.0 – 100.0)	71.4 (46.7 – 91.0)	70.0 (56.3 – 82.8)	87.3 (71.7 – 95.8)	65.2 (48.1 – 82.1)
Edelson et al. [16]	66.7 (40.0 – 85.7)	83.5 (50.0 – 100.0)	71.4 (43.3 – 100.0)	62.1 (46.5 – 70.5)	77.3 (52.2 – 91.2)	64.3 (45.8 – 83.6)
Alonso et al. [26]	75.0 (55.2 – 92.3)	100.0 (85.7 – 100.0)	66.7 (42.9 – 90.0)	68.6 (56.1 – 85.5)	92.5 (84.6 – 97.1)	60.2 (43.4 – 79.5)