Effects of MR perfusion parameters selection on the prediction of tissue infarction in acute human stroke
Wu O, Koroshetz W, Gonzalez R, Benner T, Lopez C, Zhu M, Cacciola J, Melinosky C, Ay H, Singhal A, Sorensen A
Massachusetts General Hospital
Acute stroke patients not receiving thrombolysis (n=80) were retrospectively analyzed to determine whether multiparametric algorithms could be used to objectively select perfusion parameters for predicting tissue infarction. Models were generated with PWI parameters calculated by deconvolving with AIFs from ipsilateral or contralateral hemisphere, using standard or delay-insensitive methods. Separate models incorporating tracer-arrival time (DELAY) as a covariate were developed. The inclusion of DELAY as a parameter significantly (p<.01) increased performance accuracy (∆AUC=.01.02) for all models, suggesting that the transit time for tracer arrival is an important indicator of tissue at risk of infarction in patients not receiving aggressive therapeutic intervention.