Hello. I am just looking for opinions on the way I have calculated percent signal change (PSC) in my BOLD fMRI data. My data has five "on" periods which each evoke a short spike in the BOLD signal. These are separated by about 36 seconds of "off" time. I fit my model to the data as: BOLD=b1*X + b0, where X is my "on-off" design and b0 captures the mean in the data. I can calculate PSC = (b1_est-b0_est)/b0_est*100. This bases the PSC on the model fit. Upon looking at the data there are some spikes that are the same height as the data and some higher or lower. So is it resonable to calculate the PSC based on the mean values of the BOLD signal at the locations of the spikes in the model fit? PSC = [mean(BOLD(find(fit==max(fit)))- b0_est]/b0_est*100 Is one method better than another? And if so what would be the reasons. All the best and thank you. Jason. PS I posted this to SPM and FSL lists. __________________________________ Yahoo! Mail - PC Magazine Editors' Choice 2005 http://mail.yahoo.com