Search

International audience; Following the release of the SLEEP corpus during the Interspeech 2019 paralinguistic continuous sleepiness estimation challenge, a paper presented at Interspeech 2020 by Huckvale et al. examined the reasons for the poor performance of the models proposed for this task. Careful analyses of the corpus led to the conclusion that its bias makes it hazardous to use for training machine learning systems, but a perceptual experiment on a subset of this corpus seemed to indicate that human hearing is however able to estimate sleepiness on this corpus.In this study, we present the results of the Endymion replication study, in which the same samples were rated by thirty French-speaking naive listeners. We then discuss the causes of the differences between the two studies and examine the effect of listener and sample characteristics on annotation performances

The concept of sleep health encompasses variables such as sleep disturbances, sleep behaviors and symptoms of sleep disorders. It is an epidemiological and public health challenge, particularly because integrating multifactorial variables makes the modelisation of sleep health difficult. However, we can advance our understanding of sleep health by network analysis, which has already been used to conceptualize the interaction of variables in other health domains including sleep diagnostic criteria and the underlying structure of sleep disorder classification. A network analysis of sleep variables related to health shows the empirical associations between heterogeneous variables, and to identify central variables involved in sleep health. The aim of this study is to investigate sleep health through a multivariable network analysis approach. Studied sleep variables (n = 39), extracted from an online questionnaire accessible via the “Reseau Morphee”, are symptoms of sleep disorders, sleep disturbances, sleep behaviors, and comorbidities, collected in a large French population of subjects concerned by their sleep, who responded to an online questionnaire. We follow the network guidelines for the computational analysis of network estimation, inferences, and robustness. Complete data was available for 35,808 participants (mean age = 42.7 (SD=15.7); mean BMI = 24.7 (SD=5.1); average ESS score = 9.48 (SD=4.9; N = 14,600 ≥ 11 – 40.7%); average ISI score = 16.3 (SD=5.2; N = 21,926 ≥ 15 – 61.2%); average HAD-A = 9.6 (SD=3.9; N = 14,099 > 11 – 39.4%); average HAD-D = 6.6 (SD=3.9; N = 5,998 > 11 – 16.7%)). Among the 39 sleep variables, the four most central variables with a strong influence in the network in terms of strength are: “Non-Restorative Sleep” (belonging to the group “Sleep disturbances”), “Excessive Daytime Sleepiness” (belonging to the group “Symptoms of sleep disorders”), “Circadian irregularity” and “Chronic Sleep Deprivation” (both belonging to the group “Sleep behaviors”). The identification of these central variables has many implications, especially as they are not always included in health models and organization of sleep health prevention.

International audience; To evaluate the value of apnoea + hypopnoea index versus self-reported sleepiness at the wheel in anticipating the risk of sleepiness-related accidents in patients referred for obstructive sleep apnoea. A cross-sectional analysis of the French national obstructive sleep apnoea registry. 58,815 subjects referred for a suspicion of obstructive sleep apnoea were investigated by specific items addressing sleepiness at the wheel and sleepiness-related accidents. Apnoea + hypopnoea index was evaluated with a respiratory polygraphy or full polysomnography. Subjects had a median age of 55.6 years [45.3; 64.6], 65% were men, with a median apnoea + hypopnoea index of 22 [8; 39] events/h. Median Epworth sleepiness scale score was 9 [6; 13], 35% of the patients reported sleepiness at the wheel (n = 20,310), 8% (n = 4,588) reported a near-miss accident and 2% (n = 1,313) reported a sleepiness-related accident. Patients reporting sleepiness at the wheel whatever their obstructive sleep apnoea status and severity exhibited a tenfold higher risk of sleepiness-related accidents. In multivariate analysis, other predictors for sleepiness-related accidents were: male gender, ESS, history of previous near-miss accidents, restless leg syndrome/periodic leg movements, complaints of memory dysfunction and nocturnal sweating. Sleep apnoea per se was not an independent contributor. Self-reported sleepiness at the wheel is a better predictor of sleepiness-related traffic accidents than apnoea + hypopnoea index.

Objective/background It has been shown that actigraphy may have a discriminant function (DS) for the diagnosis of narcolepsy type 1 patients (NT1), based on a combination of nighttime and daytime parameters. Here, we aimed to test those findings using another actigraph model with a different clinical sample as control (ie, primary insomniacs, PI), carrying out a secondary analysis of previously collected data. Patients/methods The study sample consisted of 13 NT1 (nine females; mean age 39.38 ± 11.48), 13 PI (nine females; mean age 38.69 ± 10.72) and 13 Healthy Controls (HC) (nine females; mean age 38 ± 10.77). Participants wore the Actiwatch AW64 (Cambridge Neurotechnology Ltd, Cambridge, UK) around the non-dominant wrist for seven consecutive days. Results Significant differences between groups were observed with a higher number of episodes of wakefulness (wake bouts, WB) in PI than HC, a higher fragmentation index (FI) in NT1 than HC and PI, a higher duration of the longest nap (LNAP) in NT1 than HC and PI and higher DS in PI and NT1 than HC. A new DS (NDS), with LNAP and FI as independent variables, was proposed; which was higher in NT1 than HC and PI. Conclusions The present study confirms that actigraphy discriminates NT1 from HC. However, considering PI, a new discriminant function NDS which takes into account LNAP and FI is better for this actigraph model.

NA