Deep learning-based algorithm accurately classifies sleep stages in preadolescent children with sleep-disordered breathing symptoms and age-matched controls

Introduction

Visual sleep scoring has several shortcomings, including inter-scorer inconsistency, which may adversely affect diagnostic decision-making. Although automatic sleep staging in adults has been extensively studied, it is uncertain whether such sophisticated algorithms generalize well to different pediatric age groups due to distinctive EEG characteristics. The preadolescent age group (10-13-year-olds) is relatively understudied, and thus, we aimed to develop an automatic deep learning-based sleep stage classifier specifically targeting this cohort.

Methods

A dataset (n = 115) containing polysomnographic recordings of Icelandic preadolescent children with sleep-disordered breathing (SDB) symptoms, and age and sex-matched controls was utilized. We developed a combined convolutional and long short-term memory neural network architecture relying on electroencephalography (F4-M1), electrooculography (E1-M2), and chin electromyography signals. Performance relative to human scoring was further evaluated by analyzing intra- and inter-rater agreements in a subset (n = 10) of data with repeat scoring from two manual scorers.

Results

The deep learning-based model achieved an overall cross-validated accuracy of 84.1% (Cohen's kappa κ = 0.78). There was no meaningful performance difference between SDB-symptomatic (n = 53) and control subgroups (n = 52) [83.9% (κ = 0.78) vs. 84.2% (κ = 0.78)]. The inter-rater reliability between manual scorers was 84.6% (κ = 0.78), and the automatic method reached similar agreements with scorers, 83.4% (κ = 0.76) and 82.7% (κ = 0.75).

Conclusion

The developed algorithm achieved high classification accuracy and substantial agreements with two manual scorers; the performance metrics compared favorably with typical inter-rater reliability between manual scorers and performance reported in previous studies. These suggest that our algorithm may facilitate less labor-intensive and reliable automatic sleep scoring in preadolescent children.

Should Young Children with Traumatic Brain Injury Be Compared with Community or Orthopedic Control Participants?

Pediatric traumatic brain injury (TBI) research depends on comparisons of profiles and outcomes between brain-injured individuals and groups consisting either of injured controls (e.g., orthopedic injuries, OI) or uninjured, typically developing children recruited from the community (community controls, CC). Children with OI are thought to provide optimal comparisons for individuals with TBI because they share injury-related experiences and pre-morbid characteristics; however, a study by Mathias and colleagues (2013) ¹ in adults has called into question the added value of injury control groups in TBI research. The comparability of these control groups has not been established in young children. Seventy-two children with OI and 84 CC aged between 18 and 60 months were compared on a range of demographic variables, developmental and medical history, pre-injury behavioral and adaptive profiles, as well as on measures of adaptive functioning, behavior, family functioning, post-concussive symptoms, and cognition (intellectual functioning, verbal abilities, executive functioning, social cognition) 6 months after the OI. There were no statistically significant differences between the OI and CC groups on any of the variables tested, whether they related to pre-injury or post-injury characteristics. The findings are applicable to studies seeking to identify appropriate control groups in the context of preschool TBI research, and suggest no clear advantage in recruiting OI controls based on the variables studied and the methodology used. However, further work is necessary to verify additional factors and outcomes relevant to pediatric TBI research, as well as to compare outcomes between these two groups at more acute stages (i.e., prior to 6 months post-injury).

Stress, cortisol, and B lymphocytes: a novel approach to understanding academic stress and immune function

Animal and human in vitro models suggest that stress-related B lymphocyte decrements are due to high levels of glucocorticoids which cause apoptosis of pre-B-cells as they emerge from the bone marrow. The present study sought to explore the relationships among distress, salivary cortisol, and human B lymphocytes in vivo. Distress (perceived stress, negative affect, depressive symptoms), lymphocyte phenotype, and salivary cortisol were assessed among first-year graduate students (n = 22) and a community control sample (n = 30) at the start of classes in the fall and the week immediately before spring preliminary exams. Compared to controls, students reported greater distress on all measures at each time point except baseline perceived stress. Hierarchical linear regression with necessary control variables was used to assess the effect of student status on the three measures of distress, the four measures of lymphocyte phenotype, and cortisol AUC and CAR over time (T1-T2). Student status was associated with a significant decrease in CD19 + B lymphocytes and flattened cortisol awakening response (CAR). Change in CAR was associated with the decrease in CD19 + B lymphocytes. Results indicated that there are significant associations among student status, flattening of CAR, and decrements in CD19 + lymphocytes.

Lipoarabinomannan-specific TNF-α and IFN-γ as markers of protective immunity against tuberculosis: a cohort study in an endemic setting

Lipoarabinomannan (LAM) is a virulent factor used for entry and survival of Mycobacterium tuberculosis (Mtb) in macrophages. Although the role of LAM for the diagnosis of tuberculosis (TB) has been extensively investigated, its cytokine response during natural Mtb infection in humans is largely unknown. In this study, LAM-specific IFN-γ, TNF-α, and IL-10 levels following whole blood assay were measured in untreated pulmonary TB patients, their contacts and community controls at baseline. In treated patients and contacts, cytokines were also measured at 6 and 12 months. At entry, 52.8% and 74.8% of controls and contacts were QFT-GIT positive, respectively. At baseline, untreated TB patients and contacts had significantly lower IFN-γ and TNF-α response compared to community controls (p < 0.0001). Besides, untreated patients had significantly higher TNF-α and IL-10 response compared to their contacts (p < 0.0001). At 6 months, contacts and treated TB patients had significantly increased INF-γ and TNF-α response (p < 0.0001). In TB patients, IFN-γ increased 10-fold following chemotherapy suggesting its potential role for treatment monitoring. The data suggests that LAM might have an anti-inflammatory effect during clinical TB and early Mtb infection. The data also suggests that LAM-induced IFN-γ and TNF-α could be used as biomarkers of protective immunity.