Agreement of Cerebral Blood Flow Estimates for Large-Scale Cortical Networks Across Multiple Resolution Parcellations

Arterial spin-labeling (ASL) is a non-invasive MRI approach for in-vivo cerebral blood flow (CBF) quantification, a crucial measure for gaining insights into brain function and neurovascular coupling. Cortical parcellation schemes are widely used to quantify mean CBF in cortical parcels and large-scale networks. However, the impact of the parcellation resolution on CBF estimate in cortical networks has not been investigated yet. This study aims to assess the agreement between CBF estimates in 7 cortical networks through three different parcellation resolutions (100, 200, and 400 parcels). Fifty-one healthy subjects underwent MRI including 3D- Tl-weighted and 3D-multi-delay pseudo-continuous ASL sequences. CBF maps were calculated and used to obtain mean CBF in networks from network-labeled parcels for all resolution schemes. The agreement of mean CBF per network across resolutions was assessed with intraclass correlation coefficient (ICC) and Bland-Altman analysis, while the variability was assessed using inter-subject and inter-parcellation coefficient of variation (CV). ICC revealed excellent reliability in all cases when changing the resolution, whereas Bland-Altman analysis revealed a perfect agreement and no proportional bias between measurements in the following cases: visual network between 100 and 200 parcels; ventral attention both between 100 and 200, and 200 and 400; somatomotor, frontoparietal and default mode networks between 200 and 400. Inter-subject CV (range: 20-26%) revealed a negligible parcellation resolution impact on estimates, whereas an inter-parcellation CV < 4% was found for all cortical networks. This agreement analysis provides a first insight into the impact that the choice of parcellation resolution may have in the quantification of CBF in cortical networks.