A PET ANALYSIS PIPELINE FOR MEASURING MISFOLDED PROTEIN CONCENTRATIONS IN MULTICENTER IMAGING TRIALS USING MIAKATTM

Abstract

The pathophysiological process of Alzheimer's disease (AD) begins years before clinical symptoms with the preclinical phase providing the best opportunity for effective therapeutic intervention. New molecular imaging techniques provide the possibility to detect and follow the early disease process using β-amyloid (Aβ) and tau biomarkers to monitor the concentration of these misfolded proteins and the effects of drug therapy. Here we propose a PET image analysis pipeline that can derive quantitative measurements of misfolded protein concentrations for multicentre trials in AD. A particular analytical pipeline is implemented in MIAKAT™ to process static PET (Aβ or Tau) images, along with an associated structural T1-MR image. It uses images in NIFTI file format and derives quantitative imaging outcome measures from PET data, namely regional Standardized Uptake Value Ratios (SUVR) with respect to cerebellum grey as reference region and SUVR parametric images as follows: 1) The subject's T1-MR image is segmented to derive a grey matter (GM) probability map using SPM12. A neuroanatomical atlas containing 125 regions (CIC atlas) is nonlinearly warped to the subject's MR image using non rigid registration. 2) Individual frames within the PET acquisition are corrected for subject motion via a frame-to-frame (rigid) registration algorithm and averaged. The resulting motion corrected PET image is rigid-body registered to the subject's structural MR data. 3) The atlas is applied to the PET data in order to calculate regional SUVR values and SUVR parametric images. The analysis pipeline executes automatically, and allows for appropriate QC of these steps on completion. It provides full audit trailing of the analysis and QC status, maintaining provenance of SUVR outcome measures and enabling reproduction of results in the future from the primary data. Example analysis results of an Aβ ([18F]AV45) and a tau ([18F]AV1451) PET scan from an early AD subject are presented in Figure1 and Table1. The analysis process is fast enabling processing of a single scan in under 10 mins on a standard workstation. The proposed analysis pipeline provides a consistent, efficient and reproducible way to derive quantitative endpoints from misfolded protein PET images with their associated QC status. Four different images produced by the pipeline for an example early AD subject are shown in three orthogonal cross-sections including: subject's MR (1st row), subject's brain grey mask (2nd row), SUVR parametric image of subject's brain with respect to cerebellum grey from Aβ PET scan (3rd row) and from tau PET scan (4th row).