Table 3 Detection performance variability across subjects.

From: A robust deep learning detector for sleep spindles and K-complexes: towards population norms

Dataset

Detector

Coefficient of variation (%)

Mean ± SD

p-value

MASS2-SS-E1

SEED

4.6 ± 0.4

 

DOSED

6.9 ± 0.4

 < 0.001

A7

6.9 ± 0.1

 < 0.001

MASS2-SS-E2

SEED

4.0 ± 0.1

 

DOSED

5.2 ± 0.3

 < 0.001

A7

5.8 ± 0.1*

 < 0.001

MODA

SEED

9.2 ± 0.4

 

DOSED

11.6 ± 0.5

 < 0.001

A7

16.9 ± 0.4*

 < 0.001

MASS2-KC

SEED

3.9 ± 0.2

 

DOSED

5.2 ± 0.3

 < 0.001

Spinky

11.5 ± 0.1

 < 0.001

  1. The variability (less is better) is measured as the coefficient of variation of the F1-score. The coefficient of variation is the ratio between the standard deviation and the mean, expressed as a percentage. Due to repeated cross-validation, each subject has 3 available test performance values, allowing many possible coefficients of variations to be obtained by randomly selecting one value per subject. Leveraging that, its distribution was estimated by repeating such random selection 100 times. P-values are defined against SEED’s performance.
  2. * Data with non-normal distribution. See Methods/Statistics for details on the statistical tests.
Back to article page