Published November 15, 2020 | Version v1
Journal article Open

Intensity normalization methods in brain FDG-PET quantification

Authors/Creators

  • 1. Molecular Imaging Group, Radiology Department, Faculty of Medicine, Universidade de Santiago de Compostela, Galicia, Spain
  • 2. Molecular Imaging Unit, Centro de Investigaciones Médico-Sanitarias, General Foundation of the University of Málaga, Málaga, Spain
  • 3. R&D Department, Qubiotech Health Intelligence, SL., Rúa Real n° 24, Planta 1, A Coruña, Galicia, Spain
  • 4. Nuclear Medicine Department & Molecular Imaging Group, University Hospital (SERGAS) & Health Research Institute of Santiago de Compostela (IDIS), Travesía da Choupana S/N 15706, Santiago de Compostela, Galicia, Spain
  • 5. Nuclear Medicine Department, Hospital Clínic, Barcelona, Spain
  • 6. Biomedical Research Networking Center in Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN), Barcelona, Spain
  • 7. Positron Emission Tomography Research Centre, University of Hull, Hull HU6 7RX, United Kingdom

Description

This study evaluates the impact of different intensity normalization methods on brain FDG-PET quantification using realistic Monte Carlo–simulated images with known hypometabolic patterns derived from healthy subjects. Several reference-region–based and data-driven normalization approaches were assessed using single-subject SPM analysis, comparing their ability to recover introduced hypometabolism and to limit unspecific metabolic findings. All methods underestimated hypometabolic volumes, especially for mild reductions, but data-driven methods—particularly iterative histogram-based approaches—outperformed reference-region techniques. Proportional scaling showed notable limitations due to unspecific hypermetabolism, while other methods produced comparable levels of unspecific findings. Overall, the results demonstrate that improper intensity normalization can significantly bias FDG-PET quantification and increase false positives, leading to a recommendation for histogram-based methods, with reference-region approaches being reliable only when large and stable regions are used.

This is the published version of the article originally published in NeuroImage.
DOI: https://doi.org/10.1016/j.neuroimage.2020.117229
Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International.

Files

Intensity normalization methods in brain FDG-PET quantification.pdf

Files (4.6 MB)