Marjorie Villien1,
2, Alexandre Krainik3, Julien Bouvier1, Matthias
J.P. van Osch4, Laurent Lamalle5, Irne Troprs5,
Jan M. Warnking1
1Grenoble
Institut of Neurosciences, INSERM, Grenoble, France; 2Athinoula A.
Martinos center for biomedical imaging, Massachusetts General Hospital,
Charlestown, MA, United States; 3Clinique universitaire de
neuroradiologie et d'IRM, CHU Grenoble, Grenoble, France; 4Leiden
University Medical Center, Leiden, Netherlands; 5SFR1, Universit
Joseph Fourier, Grenoble, France
Our aim is to compare the performance of a variety of data analysis methods in order to maximize the robustness of ASL CVR mapping in the context of clinical exams and basic and clinical research. Here, we analyzed 56 sessions of ASL vasoreactivity data obtained in patients and healthy subjects using the classical block regressor and regressors based on the physiological state of the individual subjects. We also analyzed the effect of excluding data obtained during the transition periods between capnia levels, and of regressors modeling physiological noise. Regressors based on individual capnia timecourses consistently outperformed standard block regressors.