Complex dynamics for a reduced model of human EEG: implications for the physiological basis of brain activity

Tracking Electroencephalographic Changes Using Distributions of Linear Models: Application to Propofol-Based Depth of Anesthesia Monitoring

“…Although low-order ARMA (2, 1) models gave better testing performance than the model orders (6,3) and (8,5), which correspond to the model orders of physiologically based models of EEG [25], [29], these higher orders still gave reasonable testing sensitivity. This suggests that approaches based on physiologically based models may also work, however, one needs to remain aware of the effects of parameter overfitting on out-ofsample testing performance since these models can have large numbers of parameters [1], [2].…”

“…Based on prior physiological modeling, model orders of (p = 6, q = 3) and (8, 5) were considered. The (6,3) and (8,5) model orders correspond to the model orders of transfer functions of linearized versions of the six-dimensional state and eight-dimensional state Liley models of "resting" and "anaesthetized" EEG, respectively [25], [29]. Note that this is true for all linearizations about any physiologically plausible stable fixed point of the six-dimensional state and eightdimensional state Liley models.…”