Parametric Description of EEG Profiles for Assessment of Sleep Architecture in Disorders of Consciousness.

We propose a fully parametric approach to the assessment of sleep architecture, based upon the clas- sical electroencephalographic criteria, applicable also to the recordings of patients with disorders of consciousness (DOC). Sleep spindles and slow waves are automatically detected from the matching pur- suit decomposition of overnight EEG recordings. Their evolution can be presented in the form of EEG profiles, yielding a continuous description of sleep architecture, compatible with the classical criteria used in sleep staging. We propose assessment of these EEG profiles by five parameters, which can be com- bined by a linear classifier, assessing the quality of sleep architecture. Proposed methodology is evaluated on 59 overnight EEG recordings from 19 patients from a hospital for children with severe brain dam- age, in relation to their behavioral diagnosis according to the Coma Recovery Scale-Revised. Presented results indicate robustness of the proposed approach, which may serve as a valuable aid in diagnosis of DOC patients. Complete software environment for computing and presentation of EEG profiles is freely available from http://svarog.pl.