Feasibility of Implementing Standardized Fluorodeoxyglucose and Florbetapir Brain PET Processing Pipelines
M Naseri1,2*, S Ramakrishnapillai2, L Bazzano3, O Carmichael2, (1) Louisiana State University, Baton Rouge, LA, (2) Pennington Biomedical Research Center, Baton Rouge, LA, (3) Tulane School of Public Health and Tropical Medicine, New Orleans, LA,
Presentations
PO-GePV-I-77 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: To assess the feasibility of implementing the brain PET data processing pipelines described by a large, flagship research consortium, the Alzheimer's Disease Neuroimaging Initiative (ADNI), in single-lab research setting for application to an independent cohort collecting brain florbetapir (AV45) and fluorodeoxyglucose (FDG) PET data, the Bogalusa Heart Study (BHS).
Methods: All 4 preprocessing steps described in ADNI PET pipeline documentation were implemented in-house, validated on PET scans from 11 ADNI participants, and applied to 48 BHS participants. First, raw FDG and AV45 frames were coregistered to the first acquisition frame. Second, coregistered frames were averaged. Third, averaged-FDG images were resampled into a standardized voxel image grid. To create the AV45 common image grid, averaged-AV45 was coregistered to either FDG or MRI (when FDG was unavailable). Fourth, step-3 images were spatially smoothed. The outputs from step-4 were spatially normalized to MNI space and FDG uptake and amyloid deposition were calculated using standardized uptake value ratios (SUVRs) in the regions of interest. To validate the pipeline, FDG and AV45 scans of 11 ADNI subjects were run through the pipeline and SUVRs were compared against those calculated by ADNI. SUVRs were also calculated from the BHS scans.
Results: There were numerous implementation challenges in this study. Several intermediate steps, design choices, and parameter settings were not directly specified. Furthermore, AV45 analysis using MRI was not mentioned in the documentation. Nonetheless, in-house calculated SUVRs for BHS scans followed an expected pattern for both tracers based on participant age and brain health. There was excellent correlation and strong linearity between in-house and ADNI SUVRs (R-squared of 0.99, 0.99, and 0.98 for FDG, FDG-based AV45, and MRI-based AV45 analysis, respectively).
Conclusion: Although single-lab implementation of a standard PET processing pipeline is possible, there are numerous time-consuming obstacles. These obstacles motivate future research on scientific workflows.
Keywords
PET, FDG PET, Image Processing
Taxonomy
IM/TH- Image Analysis (Single Modality or Multi-Modality): Image processing
Contact Email