Table 3.

Confusion matrices With four past and two future epochs included in the model

		Radar nightstand		Radar ceiling		Actigraphy
Healthy volunteers test set*		Wake	Sleep	Wake	Sleep	Wake	Sleep
PSG†	Wake	7489	803	6551	1027	6961	1098
	Sleep	708	19 644	667	17 877	845	19 505
TPR/FNR‡		0.97/0.03		0.96/0.04		0.96/0.04
FPR/TNR§		0.10/0.90		0.14/0.86		0.14/0.86
		Radar nightstand				Actigraphy
Patients with sleep disorders||		Wake	Sleep			Wake	Sleep
PSG	Wake	3515	3070			4381	2204
	Sleep	2354	19 123			2352	19 125
TPR/FNR		0.89/0.11				0.89/0.11
FPR/TNR		0.47/0.53				0.33/0.67
With five past and zero future epochs included in the model
		Radar nightstand		Radar ceiling		Actigraphy
Healthy volunteers test set		Wake	Sleep	Wake	Sleep	Wake	Sleep
PSG	Wake	7465	827	6479	1099	6841	1220
	Sleep	756	19 596	717	17 827	829	19 521
TPR/FNR		0.96/0.04		0.96/0.04		0.96/0.04
FPR/TNR		0.10/0.90		0.15/0.85		0.15/0.85
		Radar nightstand				Actigraphy
		Wake	Sleep			Wake	Sleep
PSG	Wake	3483	3102			4350	2235
	Sleep	2309	19 168			2253	19 224
TPR/FNR		0.89/0.11				0.90/0.10
FPR/TNR		0.47/0.53				0.34/0.66

Confusion matrices of epoch-by-epoch classification performance for two selected models, both with the heuristic Cole–Kripke rescoring rules applied, over two datasets. Numbers on the main diagonals of the 2 × 2 matrices represent correctly classified epochs.

*n = 12, mean age ± SD: 23.0 ± 3.1 years, 5 male, 4 nights of PSG + actigraphy + two radars per participant. The participants were randomly assigned into a training set for model development (n = 24/22 for nightstand/ceiling), and a testing set for validation (n = 23/21 for nightstand/ceiling).

^†PSG, Polysomnography.

^‡TPR, true positive rate. FNR, false negative rate.

^§FPR, false positive rate. TNR, true negative rate.

^||Ambulatory sleep disorder patients. n = 28, mean age ± SD: 46.25 ± 13.98 years, 19 males.