Open AccessIterative analysis of cerebrovascular reactivity dynamic response by temporal decomposition
Author(s) -
Niftrik Christiaan Hendrik Bas,
Piccirelli Marco,
Bozinov Oliver,
Pangalu Athina,
Fisher Joseph A.,
Valavanis Antonios,
Luft Andreas R.,
Weller Michael,
Regli Luca,
Fierstra Jorn
Publication year - 2017
Publication title -
brain and behavior
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.915
H-Index - 41
ISSN - 2162-3279
DOI - 10.1002/brb3.705
Subject(s) - voxel , principal component analysis , dynamic functional connectivity , dynamic mode decomposition , dynamic data , computer science , mathematics , nuclear medicine , artificial intelligence , medicine , resting state fmri , radiology , machine learning , programming language
Objective To improve quantitative cerebrovascular reactivity ( CVR ) measurements and CO 2 arrival times, we present an iterative analysis capable of decomposing different temporal components of the dynamic carbon dioxide‐ Blood Oxygen‐Level Dependent ( CO 2 ‐ BOLD ) relationship. Experimental Design Decomposition of the dynamic parameters included a redefinition of the voxel‐wise CO 2 arrival time, and a separation from the vascular response to a stepwise increase in CO 2 (Delay to signal Plateau – DTP ) and a decrease in CO 2 (Delay to signal Baseline – DTB ). Twenty‐five (normal) datasets, obtained from BOLD MRI combined with a standardized pseudo‐square wave CO 2 change, were co‐registered to generate reference atlases for the aforementioned dynamic processes to score the voxel‐by‐voxel deviation probability from normal range. This analysis is further illustrated in two subjects with unilateral carotid artery occlusion using these reference atlases. Principal Observations We have found that our redefined CO 2 arrival time resulted in the best data fit. Additionally, excluding both dynamic BOLD phases ( DTP and DTB ) resulted in a static CVR , that is maximal response, defined as CVR calculated only over a normocapnic and hypercapnic calibrated plateau. Conclusion Decomposition and novel iterative modeling of different temporal components of the dynamic CO 2 ‐ BOLD relationship improves quantitative CVR measurements.