[Development of a criterion for the automatic detection of sleep spindles in the infant].

The authors present a model for automatic detection of spindles during the different sleep stages. Baseline spindle-free EEG was modelized on a 10 sec recording using an autoregressive model. Spindles were thereafter detected using the quadratic error from this baseline model on successive 1 sec periods. This automatic detection is in a 95.7% agreement with visual analysis.