Validation of brain 18FDG-PET and pCASL joint analysis tool: a ROI-based approach

Introduction: Multimodal neuroimaging techniques allow to adequately address diagnosis questions issued by clinicians and researchers about human brain behavior. Multimodality means combining different type of information coming from functional and morpho-structural images. This is particularly important in case of neurodegenerative diseases, starting at Mild Cognitive Impairment (MCI) stage, ending to the effective progression to dementia. Brain hypometabolism patterns detected by 18FDG-PET can be combined with hypoperfusion information that nowadays can be non-invasively acquired with new MRI functional sequences, such as pseudo Continuous Arterial Spin Labeling (pCASL). This may allow pCASL to easily become an additional biomarker for diagnosis of dementia. A joint PET-pCASL analysis tool, optimized in preprocessing, was developed and then validated versus the conventional SPM12 toolbox. Material and Methods: An optimized protocol of preprocessing and ROI-based analysis was defined, starting from an SPM12 standard pipeline and using customized Matlab routines. Volumes were all normalized in MNI space. Normalization parameters for pCASL images were defined on the high resolution source image, co-registered to the T1-weigthed image. A mean Gray Matter (GM) tissue probability map, obtained from 662 healthy subjects in the patients range of age, was used for GM segmentation and for PVE correction in 18FDG-PET scans. The developed tool was validated on 20 MCI patients (mean age 76 yr.), assessing outcomes agreement vs a standard SPM12 analysis, using a single subject versus group analysis. This validation was performed on 18FDG-PET data only, where a sample of 111 healthy subjects 18FDG-PET scans was available as control group. In the SPM analysis a general linear model was applied, age was considered as covariate, and significance level was set at 0.05 (FWE-corrected) at a cluster level. Indirect validation for pCASL data, where normative data and a control group were not available, was derived by cross-correlation with 18FDG-PET results. Results: 18FDG-PET Standardized Uptake Values (SUVr) and pCASL Cerebral Blood Flow values (CBFr), both normalized on cerebellar mean values, were calculated in 22 macro-regions obtained by combining sub-areas of the Automated Anatomic Labeling template version 3 (AAL3). For 18FDG-PET, regions that fall outside normality range were the same found to belong to signifi cant clusters in the SPM referenceoutcome. Thereby, a comparison between 18FDG-PET and pCASL relying on SUVr and CBFr values incorresponding regions can be assessed (Fig. 1b). Both regional values and relative ratios between region scan be evaluated. Discussion: The validation of our tool on 18FDG-PET will allow it to be used for further investigation of correlation between SUVr and CBFr values in the 22 macro-regions. References: [1] Y. H. K. Cha, M. A. Jog, Y. C. Kim, S. Chakrapani, S. M. Kraman, and D. J. J. Wang, “Regional correlation between resting state FDG PET and pCASL perfusion MRI,” J. Cereb. Blood Flow Metab., vol. 33, no. 12, pp.1909–1914, Dec. 2013. [2] L. Yan et al., “Regional association of pCASL-MRI with FDG-PET and PiB-PET in people at risk for autosomal dominant Alzheimer’s disease,” NeuroImage. Clin., vol. 17, pp. 751–760, 2017.