An application of semiparametric model estimation under shape invariance to fmri data

Tasks become easier with practice and we all can subjectively assess this. We propose a method for measuring this difference by comparing fMRI activations during saccadic task trials before and after practice sessions. We treat each subject's pre-practice and post-practice activations as functions. We estimate the functions using local polynomial regression with kernel smoothing, establish a shape invariant model for the relevant function estimates and use evolutionary algorithms to estimate shape invariant model parameters. We apply hypothesis testing on the estimates of the shape invariant parameters to decide if attenuation is significant for different practice groups. We see significant attenuation where we expect to see it and have similar results to those of a different method.