A Comparison and Validation of BIDS Apps for Preprocessing of UCB-J PET data

Abstract

Background Preprocessing of Positron Emission Tomography (PET) data can be performed in many different ways and often involves manual steps, resulting in reduced comparability across analyses and pipelines. Apps working on PET data organized according to the Brain Imaging Data Structure (BIDS) seek to automate preprocessing, enabling the application of standardized pipelines that increase comparability and reproducibility. In this study, two of these apps were validated against in-house software. Methods Data were acquired in-house from 14 subjects who received placebo in a randomized controlled clinical trial and underwent two PET measurements using the radioligand [11C]UCB-J. Preprocessing was performed with two BIDS Apps, PETPrep and PICNIC, as well as an in-house software. Arterial blood data were processed using Bloodstream, and [11C]UCB-J total distribution volume (VT) values were calculated using a 1-tissue compartment model using kinfitr. Results VT values for all 9 regions of interest were highly consistent across tools with estimates of Intraclass Correlation Coefficients for consistency ranging between 0.94 and >0.99, although bias between tools measured as signed Percentage Difference was present, ranging from 0.1% - 10.2% depending on the region. Tools performed similarly on test-retest metrics. We observed a negative test-retest bias in [11C]UCB-J VT of unknown origin. Conclusion BIDS Apps produce reliable and consistent results when compared both with each other and with an in-house software optimized for processing and analysis of our in-house data. Region-dependant bias between tools warrants caution for direct comparison of VT between tools. The usage of these apps could enable better comparability across publications