Neural network analysis of sleep stages enables efficient diagnosis of narcolepsy

Abstract

Analysis of sleep for the diagnosis of sleep disorders such as Type-1 Narcolepsy (T1N) currently requires visual inspection of polysomnography records by trained scoring technicians. Here, we used neural networks in approximately 3,000 normal and abnormal sleep recordings to automate sleep stage scoring, producing a hypnodensity graph—a probability distribution conveying more information than classical hypnograms. Accuracy of sleep stage scoring was validated in 70 subjects assessed by six scorers. The best model performed better than any individual scorer (87% versus consensus). It also reliably scores sleep down to 5 s instead of 30 s scoring epochs. A T1N marker based on unusual sleep stage overlaps achieved a specificity of 96% and a sensitivity of 91%, validated in independent datasets. Addition of HLA-DQB1*06:02 typing increased specificity to 99%. Our method can reduce time spent in sleep clinics and automates T1N diagnosis. It also opens the possibility of diagnosing T1N using home sleep studies.

Deep Transfer Learning for Single-Channel Automatic Sleep Staging with Channel Mismatch

“…However, these databases cannot be simply added into these studies due to channel mismatch. The problem of channel mismatch arises when different studies uses different channel layouts [2] or when novel electrode placements might be explored [6]. Moreover, it also happens when a study investigates a particular sleep abnormality, poor performance can be obtained when the automated diagnostic procedure is only trained on healthy volunteers [7].…”

DOSED: A deep learning approach to detect multiple sleep micro-events in EEG signal

“…Finally, events of interests such as spindles might exhibit specific temporal dynamic patterns, with the likelihood of a spindle occurring in a sample being related to the occurrence of spindles in a previous sample. Integrating a temporal context using a recurrent neural network as performed for EEG processing [38] or for sleep stage classification [39,40,41,5,42] might enhance detection performance for some events. In all cases, however, the proposed approach has the considerable advantage of simultaneous multi-event detection, a crucial feature that should allow to build more easily additional event detection methods on the same architecture.…”

“…In other cases however, when one needs to go beyond sleep scoring, they must be counted and specifically annotated. It is notably the case when the aim is to understand sleep physiology [3,4] or to study the pathophysiology of specific sleep or neuropsychiatric disorders [5,6,7]. The identification of micro-architectural events in the EEG is traditionally performed by trained sleep experts, also called scorers, who visually investigate the recorded signals over a night and annotate the relevant events with their respective start times and durations.…”

Automatic detection of cortical arousals in sleep and their contribution to daytime sleepiness

“…This also has the effect of simplifying the preprocessing and does not require manual annotations. The sleep/wake prediction performance of the [3,34]. The wake accuracies reported in these studies were measured using smaller test sets of 70 and 82 PSG.…”

“…The architecture of the proposed network is based on similar studies of sleep staging [3,33,34,36]. The general idea is to use CNNs to automatically design a set of features describing the preprocessed EEG, EOG, and EMG signals in 1 second bins.…”