Skip to main content
. 2023 Apr 6;46(6):zsad101. doi: 10.1093/sleep/zsad101

Figure 7.

Figure 7.

Examples of agreement, acceptable performance and true errors generated by SleepEns. (a) Example of a NREM to REM sleep transition showing the sensitivity of SleepEns to REM sleep transition. SleepEns detected the transition earlier than human scorers did, most likely due to the increased proportion of theta (θ) frequency in the EEG. The difference was estimated to be an acceptable performance (grey bar). Traces from top to bottom are Expert 0 scoring, SleepEns scoring, EMG and EEG signals. (b) Example of transition between NREM sleep and wake showing both a true error (red bar) and an acceptable performance (grey bar) of SleepEns. Note that SleepEns wrongly identified an active wake (AW) epoch that was scored as quiet wake (QW) by human scorers. This was likely due to the abrupt change in EEG frequencies and the context of the following epoch of AW. The epoch marked as acceptable performance was an epoch comprised of both NREM sleep and quiet wake characteristics in equal manner. (c) A 95s example showing a true error (red bar) and an acceptable difference (grey bar) in classification as the animal transition from REM sleep, wake and NREM sleep. Note that in this example SleepEns identified a REM sleep to wake transition earlier than Expert 0. Importantly in this instance, both Expert 1 & 2 made a similar identification as SleepEns (data not shown).