Table 2.

Summary of detection statistics^† and performance^* for human reviewers (A, B, C) and the automated detector (X): true positives (TP), false positives (FP), false negatives, mean precision (P̄), mean recall (R̄), and mean F-measure (F̄). Ground truth set included all events declared as HFOs by two or more detectors during identification or review. Mean values were calculated from 20 partitions of the ground truth set.

	Patient 1				Patient 2				Combined
	A	B	C	X	A	B	C	X	A	B	C	X
TP	187	242	269	347	406	338	240	243	593	580	509	590
FP	0	5	9	69	40	19	35	25	40	24	44	94
FN	203	148	121	43	92	160	258	255	295	308	379	298
P̄	1.00	0.98	0.96	0.83	0.91	0.95	0.89	0.92	0.96	0.97	0.93	0.88
R̄	0.48	0.64	0.71	0.99	0.89	0.71	0.52	0.52	0.68	0.67	0.61	0.75
F̄	0.64	0.75	0.80	0.90	0.89	0.81	0.64	0.64	0.77	0.78	0.72	0.77

^†(TP + FN) is constant for a patient, and is equal to the number of ground truth events

^*Best values for each row and patient are emphasized in bold