Skip to main content
. 2022 Jul 19;9:421. doi: 10.1038/s41597-022-01545-6

Table 6.

Sleep stage classification results of our baseline algorithm applied to different age groups.

Automated score sleep stage
W N1 N2 N3 R
Manual score sleep stage, N W (661,645) 63.1 0. 34.0 1.5 1.4
N1 (127,602) 23.9 0.9 68.1 2.1 5.0
N2 (1,375,678) 4.4 0. 88.6 5.8 1.1
N3 (871,200) 1.7 0. 27.2 70.7 0.
R (608,180) 6.7 0. 76.6 1.5 15.1
Manual score sleep stage, N W (52,979) 89.5 0.1 8.2 0.5 1.7
N1 (8,263) 37.5 2.5 47.4 0.6 12.1
N2 (80,275) 5.6 0.1 89.1 2.9 2.3
N3 (30,612) 2.6 0. 18.3 79.1 0.
R (24,006) 9.2 0. 24.7 0.6 65.5
Manual score sleep stage, N W (63,041) 83.3 0. 2.4 2.8 11.4
N1 (4,579) 28.7 1.1 24.7 6.2 39.2
N2 (38,525) 9.4 0. 62.9 10.2 17.4
N3 (64,512) 4.5 0. 3.7 83.3 8.5
R (60,167) 11.1 0. 5.0 7.1 76.8

(a) All age groups. 3,928 sleep studies and 3,644,305 samples. Overall accuracy is 64.4%.

(b) 18 years and older. 222 sleep studies and 196,135 samples. Overall accuracy is 81.1%.

(c) 0–1 year olds. 242 sleep studies and 230,824 samples. Overall accuracy is 76.6%.

One sample is a 30-second epoch of sleep. Cell (row i, column j) of the normalized confusion matrix indicates the percentage (%) of samples in stage i (manually scored by NCH technician) that were predicted to be in stage j (by our automated algorithm). Each row adds to 100%. Bolded diagonal entries are the percentages of samples in each stage that were correctly classified. Overall accuracy is the total number of correctly classified samples divided by the total number of samples in %. All numbers reported are averaged over 3-fold stratified cross validation trials and rounded to one decimal point. Standard deviation was <1% for all entries except one and not shown here.