Ultra-slow frequency bands reflecting potential coherence between neocortical brain regions

• We use the resting MEG records for schizophrenia classification.
• We adopt frequency band including Ultra Low Frequency (<1 Hz).
• Comparison on the phase locking values achieves p = 0.00008.
• The ultraslow frequency bands provide a higher rate of classification.

Recent studies of electromagnetic ultra-slow waves (⩽0.1 Hz) have suggested that they play a role in the integration of otherwise disassociated brain regions supporting vital functions (Ackermann and Borbely, 1997; Picchioni et al., 2010; Knyazev, 2012; Le Bon et al., 2012). We emphasize this spectral domain in probing sensor coherence issues raised by these studies using Hilbert phase coherences in the human MEG. In addition, we ask: will temporal–spatial phase coherence in regional brain oscillations obtained from the ultraslow spectral bands of multi-channel magnetoencephalograms (MEG) differentiate resting, “task-free” MEG records of normal control and schizophrenic subjects? The goal of the study is a comparison of the relative persistence of intra-regional phase locking values (PLVs), among 10, region-defined, sensors in examined in the resting multichannel, MEG records as a function of spectral frequency bands and diagnostic category. The following comparison of Hilbert-transform-engendered relative phases of each designated spectral band was made using their pair-wise PLVs. This indicated the proportion of shared cycle time in which the phase relations between the index location and reference leads were maintained. Leave one out, bootstrapping of the PLVs via a support vector machine (SVM), classified clinical status with 97.3% accuracy. It was generally the case that spectral bands ⩽1.0 Hz generated the highest values of the PLVs and discriminated best between control and patient populations. We conclude that PLV analysis of the oscillatory patterns of MEG recordings in the ultraslow frequency bands suggest their functional significance in intra-regional signal coherence and provide a higher rate of classification of patients and normal subjects than the other spectral domains examined.