Reliability of modeling evoked flow responses

Visually evoked flow characterises the relative changes of blood flow velocity in the posterior cerebral artery in response to visual stimulation. The present study is the first to address the reliability of model fitting to evoked flow responses, rigorously defined by Cronbach's alpha.

We fitted two models to the evoked flow responses recorded from 19 subjects on two different days. Model 1 characterises a harmonic oscillator with frequency omega and damping coefficient xi which (after a delay DeltaT) is driven from zero towards a new equilibrium K by an impulse with magnitude T. Model 2 is the sum of a first order system subjected to a step and a transient smoothed pulse, both again delayed by DeltaT.

Model 1 exhibited slightly smaller fit errors and convergence was less dependent on starting values for the parameters. As judged from the residual noise in the evoked flow response, there was no clear indication of sustained oscillations characterising model 1 exclusively. Both models showed considerable retest errors. Nonetheless, Cronbach's alpha was significant for K and omega, and highest for K.

Retest errors were considerable, particularly so for the damping coefficient xi and impulse magnitude T. A physiological interpretation of these parameters is limited by our findings.