What EEG doesn’t see: a whole-brain MEG map of sleep spindles and slow oscillations

Sleep spindles and slow oscillations (SOs) are hallmarks of non-rapid eye movement (NREM) sleep, linked to memory consolidation and restorative functions. They are usually detected in one or a few scalp electroencephalography (EEG) channels with parameters that vary across studies, and serve as the reference even in magnetoencephalography (MEG), fMRI and intracranial studies. We argue that these choices filter which neural activity is captured. Nineteen young adults took a daytime nap during simultaneous MEG and EEG. Using MEG source imaging across 416 cortical, subcortical and cerebellar regions, we mapped the amplitude and density of SOs (0.5–4 Hz, and 0.5–2 and 2–4 Hz separately) and of spindles (9–18 Hz, and slow, 9–12.9 Hz, and fast, 13–18 Hz, separately), and related each region’s activity to a single scalp derivation (Cz–M1). Both rhythms were detected throughout the brain, including the thalamus, hippocampus, amygdala, basal ganglia and cerebellum. The balance between bands varied across regions: fast spindles predominated over slow spindles along the dorsal midline, around the central sulcus and in the thalamus, slow spindles predominated in occipital cortex and the cerebellar hemispheres, and 2–4 Hz SOs were relatively most prominent in occipital cortex. Amplitude and density formed reliable but distinct maps: regions with larger fast spindles also had more of them, whereas SOs, and slow spindles once source depth was controlled, were larger where they were less frequent. More SOs were detected in every region than at Cz–M1, whereas more fast spindles were detected at Cz–M1 than in any single region. Around the SO, slow-spindle activity had the same timing at the scalp and in the sources, but the up-state increase in fast-spindle activity seen at the scalp was absent from local sources. Slow spindles were the most strongly shared events, and sensorimotor sources the most scalp-like. Thus, a single scalp derivation offers a selective and partly composite view of sleep oscillations, with implications for interpreting their functional roles and for targeting them with closed-loop auditory stimulation.