Sleep transforms the brain. Neurons that fire in synchrony during wakefulness break into new patterns as an animal drifts into non-REM sleep, then reorganize again during REM. But capturing exactly how those transitions unfold across large populations of cells has been a persistent technical challenge. Most studies record from a single cortical region or lack the temporal resolution to track individual neurons across sleep stages. A preprint posted July 24 on bioRxiv offers a powerful remedy: a publicly available dataset of neuronal activity from the mouse cortex spanning wakefulness, NREM sleep, REM sleep, and isoflurane anesthesia, all at single-cell resolution.
The dataset, created by Ikumi Oomoto, Daiki Kiyooka, Masafumi Oizumi, and Masanori Murayama at RIKEN Center for Brain Science in Japan, is a resource paper, not a discovery paper. Its purpose is to give the research community raw material rather than finished findings. And by the numbers, that material is substantial.
What the dataset covers
Using wide-field two-photon calcium imaging, the team recorded from layers 2/3 of the mouse cortex across multiple sessions, capturing between 4,000 and 10,000 neurons per session at a frame rate of 7.65 Hz. Each neuron comes with spatial coordinates, meaning researchers can examine not just who fires when, but where those cells sit in the cortical sheet.
The recordings cover four distinct brain states. During wakefulness, mice moved freely through their environment. Natural sleep sessions captured both NREM sleep, marked by slow-wave activity, and REM sleep, with its desynchronized cortical dynamics. A fourth condition, isoflurane anesthesia, provides a pharmacologically induced state for comparison. This four-state design is one of the dataset’s most valuable features, because it allows direct within-subject comparisons across physiological and pharmacological conditions.
The repository distributes data in several formats. Raw imaging movies are available as TIFF stacks, giving labs the option to run their own preprocessing pipelines. Processed data in MATLAB format includes deltaF/F fluorescence signals, deconvolved spike estimates, and Gaussian-smoothed spike trains. Behavioral state annotations and comprehensive metadata accompany each session. Electrophysiological recordings are also provided in MATLAB format, offering a complementary signal modality.
Why it matters
The dataset arrives at a moment when the field of sleep neuroscience is increasingly focused on population-level dynamics. Classical sleep staging relies on EEG signatures, but those signals reflect summed activity across millions of neurons. To understand how individual cells and local circuits orchestrate sleep state transitions, researchers need recording techniques that resolve single neurons while covering enough of the cortex to see spatial patterns. Wide-field two-photon microscopy does exactly that, and the RIKEN team has done the difficult, expensive work of collecting these data across four brain states.
For labs that lack two-photon setups, the dataset is a gateway. A computational neuroscientist studying how cortical activity patterns change between NREM and REM can now download ready-to-analyze recordings spanning thousands of neurons. A methods developer testing a new spike-sorting algorithm can benchmark it against ground-truth calcium imaging data across multiple behavioral states. A team exploring the spatial organization of slow-wave activity can examine the raw coordinates of individual neurons rather than relying on proxy measures.
The dataset also fills a gap in open neuroscience. While large-scale recording projects exist for the mouse visual cortex during wakefulness, few publicly available resources combine single-cell resolution with natural sleep. This preprint helps correct that imbalance by placing a high-quality, multi-area sleep dataset under a CC-BY 4.0 license, meaning it can be freely reused, modified, and redistributed.
Limits to note
As a preprint, this work has not yet undergone peer review. The dataset is described and documented, but independent verification of the preprocessing steps and state annotations will come only as other groups begin working with the data.
The recordings come from layers 2/3 of the cortex, which are the superficial layers. Deep-layer dynamics, which may play distinct roles in sleep regulation, are not captured. The frame rate of 7.65 Hz is sufficient to track population activity but cannot resolve individual action potentials directly; the spike estimates are deconvolved from calcium signals, a process that introduces uncertainty, particularly during high-frequency firing.
The sample is also limited to mice. Whether the spatial and dynamical patterns observed in the rodent cortex generalize to humans remains an open question, though the fundamental architecture of cortical layers is conserved across mammals.
The bottom line
Oomoto, Kiyooka, Oizumi, and Murayama have released a reusable, multi-area calcium imaging dataset that captures the mouse cortex across wakefulness, NREM sleep, REM sleep, and isoflurane anesthesia at single-cell resolution. With 4,000 to 10,000 neurons per session, raw and processed data, and open licensing, it is the kind of resource that enables discoveries rather than reporting them directly. For researchers studying how brain states reshape cortical activity at the population level, it is worth downloading.
Source
Oomoto, I., Kiyooka, D., Oizumi, M., & Murayama, M. Multi-area single-cell calcium imaging dataset of the mouse cortex across wakefulness, sleep, and anesthesia. bioRxiv. https://doi.org/10.64898/2026.07.20.739676 (Preprint posted July 24, 2026, not yet peer-reviewed.)