15 April 2026 in SLEEP Advances
Study Objectives: Sleep research in rodents often involves large-scale electroencephalography (EEG) recordings that are traditionally scored manually—a labor-intensive process prone to scorer bias and inconsistency. We developed the REST-Rodent EEG Sleep Transformer, a neural network trained for automated sleep stage classification based on EEG and electromyography (EMG). Materials and Methods: EEG and EMG signals were segmented into 4-second epochs, transformed via short-time Fourier transform to generate representations with five time frames and 130 frequency bins per epoch. We used a two-stage model: the first stage analyzes individual epochs to extract meaningful features, while the second stage captures sleep patterns over time by modeling the relationships between consecutive epochs. The model was trained based on 116 days of artifact-free, pre-labeled recordings from 94 male Fmr1 knockout mice (C57BL/6 background). It was then tested on 38 separate 24-hour recordings from the same mice and on three other mouse strains (C57, DBA, and CD-1). REST was also tested against a convolutional neural network (CNN)–based model (Accusleep). Results: REST achieved a Cohen’s kappa of 0.873 ± 0.005 and F1-scores for Wake (0.936 ± 0.003) and non-rapid eye movement sleep (0.926 ± 0.003), with strong performance for rapid eye movement sleep (0.913 ± 0.005). REST showed robust cross-strain compatibility and generalized effectively to external datasets. Conclusion: REST is a fast, accurate, and accessible tool for automated sleep stage classification in mice. Its Transformer architecture provides long-range temporal awareness, enabling consistent and reliable analysis of rodent sleep in large-scale datasets. Statement of Significance Manual scoring of rodent sleep is a major bottleneck in preclinical sleep research, especially for large-cohort studies with weeks-long recordings. REST, a Transformer-based neural network model, provides reliably accurate, automated sleep stage classification and is adaptable across different mouse strains and disease models. REST maintains high performance without retraining or manual fine-tuning, making it practical for translational research. Its application in mice subjected to traumatic brain injury revealed pharmacologically induced sleep changes, demonstrating its utility in both mechanistic and therapeutic studies. REST addresses a key gap in rodent sleep research by enabling fast, consistent sleep analysis. Future work will expand its capability to detect artifacts and extend to other species, enhancing its value for cross-strain investigations in neurological disorders.
6 February 2026 in SSRN Electronic Journal
Background: Obstructive sleep apnea (OSA) is a sleep disorder associated with neurocognitive morbidity. Sleep homeostasis (SH), is a physiological process critical for synaptic downscaling. We hypothesized that OSA disrupts SH in a severity-dependent manner. Methods: Polysomnography data from 940 adults in the Wisconsin Sleep Cohort with mild, moderate, severe OSA or no OSA were analyzed. Sleep parameters including apnea-hypopnea index (AHI) were derived from prescored files and slow wave activity (SWA 0.5-4 Hz) was extracted from electroencephalography (EEG) in non-rapid eye movement (NREM) sleep. SH was assessed with overnight decay of SWA, slopes of slow waves and change in SWA with wake after sleep onset (WASO). Associations between SH measures, OSA severity, and clinical variables were examined using linear regression and generalized linear models.Findings: SWA decayed across successive NREM clusters in participants with no, mild, and moderate OSA but was attenuated in severe OSA. Higher AHI (p<0·0001) and greater WASO (p=0·04) were associated with a flatter decay of SWA. For a given level of SWA decay, males exhibited higher AHI than females (p<0·0001). Severe OSA also impaired homeostatic decrease in slopes of slow waves from the first to last hour of NREM sleep. In generalized linear models, AHI, WASO, sedative use, N3 sleep fraction, caffeine, and antidepressant use independently predicted impaired SWA decay, whereas AHI was the sole predictor of reduced slow wave slope (p=0·03).Interpretation: SH was impaired mainly by severe OSA, evident in SWA dynamics. EEG-based SH metrics may provide mechanistic insight into OSA-related neural vulnerability and serve as physiological markers for therapeutic response.