Abstract Introduction The glymphatic system regulates cerebrospinal fluid (CSF) flow to clear metabolic waste. But continuous monitoring of CSF dynamics during natural sleep remains challenging. We developed a wireless, flexible brain-water near-infrared spectroscopy (BW-NIRS) device for long-term, noninvasive measurement of CSF-linked brain water dynamics and associated physiological rhythms during sleep. Methods The BW-NIRS device integrates multi-wavelength LEDs (640, 680 and 950 nm) and a photodetector on a flexible circuit to track water fraction using a Beer–Lambert model. Monte Carlo photon transport and bioheat simulations guided source–detector geometry and confirmed thermal safety ( 41 °C). Device performance was validated in experiments, controlled exercise, and breath-holding paradigms, and at-home sleep recordings. Light-based data and commercial EEG data were acquired simultaneously for sleep-stage classification using a hybrid convolutional neural network–bidirectional LSTM and sigma-band thresholding. Results Bioheat simulations and infrared thermography showed surface temperature stabilized at ~40 °C after 60 min. Monte Carlo modeling demonstrated measurable photon sensitivity to CSF depth. ΔH₂O signals showed state-dependent oscillations during exercise and breath holding. Also, power spectral analysis identified respiratory signal (~0.3 Hz), slow oscillation-linked NIRS oscillation (0.6–0.7 Hz), and cardiac cycle (0.8–1.2 Hz) rhythms with stage-specific variations. Hybrid sleep-stage classification achieved 0.87 test accuracy. During natural sleep, accumulated brain water rose during transitions into NREM and declined during transitions into REM, paralleling EEG slow-wave activity. Conclusion The BW-NIRS device enables continuous, at-home monitoring of CSF-linked brain water dynamics and multiple physiological rhythms across sleep stages, providing a noninvasive observation of glymphatic system and potential early screening of sleep or neurodegenerative disorders. Support (if any) This work was supported by the WISH Center grant from the Georgia Tech Institute for Matter and Systems, the Global Industrial Technology Cooperation Center (GITCC) through a grant agreement with the Korea Institute for Advancement of Technology (KIAT), the National Science Foundation Research Traineeship (Grant No. NRT-FW-HTF 2345860), and the Institute of Information & Communications Technology Planning & Evaluation (IITP) grant funded by the Korea government (MSIT) (No. RS-2024-00443780).
Ban et al. (Fri,) studied this question.