Establishing sleep stages using Delta, Theta and Gamma oscillations from long term Local Field Potential (LFP) recordings in mice

Most current methods determine sleep and awake stages in rodents by using subjective thresholding methods. In contrast, we implemented objective thresholding algorithms based on intrinsic characteristics of local field potential (LFP) signals. We recorded brain LFP and infrared video of freely moving adult mice. Continuous long-lasting (several weeks) recordings were performed during 12/12 hours of dark and light cycles. The movement information from video recordings was recorded and processed using iSpy Open Source Video Surveillance Software, by subtracting each video frame from the previous, thus obtaining an activity measure (A). We used Igor Pro technical graphing and data analysis software to process the LFP recording by band-pass FIR filtering: 1-4 Hz for Delta oscillations, 5-12 Hz for Theta and 30-120 Hz for Gamma. Using an 8 s sliding window (in steps of 1 or 0.5 s) over 12 hour periods we calculated the Root Mean Square (RMS) values of each filtered wave, obtaining the DeltaRMS, ThetaRMS and GammaRMS. The histograms of these RMS values showed bimodal normal (Gaussian) distributions. Similarly, the distributions of GammaRMS/DeltaRMS (G/D) and ThetaRMS/DeltaRMS (T/D) were also bimodal. Using a Gaussian probability-weighting mechanism, we defined the NREM (non-REM) sleep segments. This first findings were adjusted by the GammaRMS intensity for each period. The remaining periods were classified to REM or A(wake) depending on the bimodal distribution of ThetaRMS/DeltaRMS function together with a post-hoc classification based on the motor activity record.