Open AccessSoftware development for quantitative analysis of brain amyloid PET
Author(s) -
Matsuda Hiroshi,
Yamao Tensho
Publication year - 2022
Publication title -
brain and behavior
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.915
H-Index - 41
ISSN - 2162-3279
DOI - 10.1002/brb3.2499
Subject(s) - standardized uptake value , positron emission tomography , nuclear medicine , pipeline (software) , computer science , software , standardization , pattern recognition (psychology) , artificial intelligence , medicine , programming language , operating system
Centiloid (CL) scaling has become a standard quantitative measure in amyloid PET because it allows the direct comparison of results across sites, even when different analytical methods or PET tracers are used. Methods In the present study, we developed new standalone software to easily handle a pipeline for accurate calculation of the CL scale for the five currently available amyloid PET tracers— 11 C‐PiB, 18 F‐florbetapir, 18 F‐flutemetamol, 18 F‐florbetaben, and 18 F‐NAV4694. This pipeline requires reorientation and coregistration of PET and MRI, anatomic standardization of coregistered PET to a standardized space using a warping parameter for coregistered MRI, application of standard volumes of interest (VOIs) to the warped PET, calculation of the standardized uptake value ratio (SUVR) for the target VOIs, and finally conversion of the SUVR to the CL scale. The PET data for these tracers were collected from the publicly available Global Alzheimer's Association Interactive Network (GAAIN) repository. We also developed software to map Z‐scores for the statistical comparison of a patient's PET data with a negative control database obtained from young healthy controls in the GAAIN repository. Results When whole cerebellum or whole cerebellum plus brainstem was chosen as the reference area, an excellent correlation was found between the CL scale calculated by this software and the CL scale published by GAAIN. There were no significant differences in the detection performance of significant amyloid accumulation using Z‐score mapping between each 18 F‐labeled tracer and 11 C‐PiB. The cutoff CL values providing the most accurate detection of regional amyloid positivity in Z‐score mapping were 11.8, 14.4, 14.7, 15.6, and 17.7 in the posterior cingulate gyrus and precuneus, frontal cortex, temporal cortex, parietal cortex, and striatum, respectively. Conclusion This software is able to not only provide reliable calculation of the global CL scale but also detect significant local amyloid accumulation in an individual patient.