Marjorie Villien1,2, Julien Bouvier3, Irne Tropres3, Matthias J. P. van Osch4, Christoph Segebarth1,2, Jean-Franois Le Bas5, Alexandre Krainik1,5, Jan Martin Warnking1,2
1Centre de Recherche Inserm, U836, Grenoble, France; 2Grenoble Institut des Neurosciences, Universit Joseph Fourier, Grenoble, France; 3IFR 1, Universit Joseph Fourier, Grenoble, France; 4Department of Radiology, Leiden University Medical Center, Leiden, Netherlands; 5Service de Neuroradiologie, CHU Grenoble, Grenoble, France
Robust MRI methods to measure cerebral vasoreactivity in patients are increasingly sought. A challenge in such studies is to correctly model the signal during a capnic stimulus, potentially varying with subject compliance and response. This is especially critical in ASL with limited SNR, as any mismatch will further decrease sensitivity. Here we compare the performance of regressors derived from hypercapnia data collected during each scanning session for the analysis of ASL data to a standard block regressor. Capnia-derived regressors slightly but consistently outperformed block regressors, more than compensating any variability in the capnia measurement and increasing robustness to experimental variability.