Table 4 Correlation between experts and detectors for subject-level parameters of SSs and KCs.

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

Parameter

Detector

R-squared

Mean difference

Mean SS duration (s) (MODA)

SEED

0.62

 − 0.017

DOSED

0.47

0.179

A7

0.35

 − 0.020

SS density (epm) (MODA)

SEED

0.94

0.136

DOSED

0.90

 − 0.154

A7

0.88

0.177

Mean SS PP amplitude (μV) (MODA)

SEED

0.99

0.815

DOSED

0.98

1.542

A7

0.97

0.496

Mean SS frequency (Hz) (MODA)

SEED

0.95

0.063

DOSED

0.93

0.071

A7

0.77

0.087

Mean KC duration (s) (MASS2-KC)

SEED

0.80

 − 0.003

DOSED

0.64

0.140

Spinky

0.00

0.671

KC density (epm) (MASS2-KC)

SEED

0.91

0.112

DOSED

0.91

0.106

Spinky

0.82

 − 0.179

Mean KC PP amplitude (μV) (MASS2-KC)

SEED

0.93

1.157

DOSED

0.91

3.179

Spinky

0.88

9.642

  1. SS: sleep spindle; KC: K-complex; epm: events per minute; PP: peak-to-peak. Subject-level parameters correspond to whole-recording aggregates of event-level instances. The correlation is measured between values determined by expert annotations and detections. The difference is defined as the estimated value (by detections) minus the ground truth value (by expert annotations).
Back to article page