Advertisement

AUTOMATED REPORTING OF AMYLOID PET QUANTIFICATION ON BRAIN SURFACE THROUGH A WEB INTERFACE

      Background: Molecular brain imaging using Positron Emission Tomography (PET) is a robust diagnostic tool for which several tracers labelled with either 11C or 18F are available. For visual inspection of the images, cortical surface based visualisation presents the advantage of providing a compact and more convenient display than volumetric scans. We have developed an automated reporting tool to display the semiquantitative PET signal for several common tracers without the need of a MRI and using several standard normalisation methods (i.e. using the Pons, the whole Cerebellum WCB and the cerebellum cortex CB-CTX). This tool is available on-line to registered users and the processing is performed remotely (http://milxcloud.csiro.au/capaibl). The method was validated by comparing the surface signal computed with and without MRI.
      Methods: Three hundred and thirty nine participants from the AIBL cohort underwent MRI and PET scans with different tracers: 18F-Flutemetamol (n=101), 11C-PIB (n=91), 18F-Florbetapir (n=77), 18F-FDG (n=38), 18F-Florbetaben (n=14) and 18F-NAV4694 (n=18). Each individual PET image was spatially normalised to the MNI space and SUVR corrected with a common mask. Radiotracer retention was then estimated vertex-wise within several GM prior atlases. Atlas selection and Bayesian fusion were then used to estimate retention on the cortical surface. For comparison with MR-based approach, radiotracer retention was also estimated for each vertex with the individual GM segmentation and projected onto the individual cortical surface. The difference in radiotracer estimation between the MRI-dependent and CAPAIBL (Computational Analysis of PET from AIBL) approaches was measured by absolute difference of SUVR values at each vertex. The SUVR differences were averaged over vertices and over the total subjects.
      Results: Visual inspection revealed high concordance between PET only and MRI based surface projection; the surface projection was defined on 8 standard views for consistent reporting (Fig 1). Across the 6 tracers tested, the average absolute error over the brain surface with and without MRI was 0.12, whereas the average variance was 0.018.
      Conclusions: The proposed MRI-less surface projection method demonstrated better estimation of 11C-PIB retention than recently published methods displaying similar accuracy for various 18F labelled radiotracers. CAPAIBL provides an efficient reporting tool for PET imaging easily accessed remotely through a web interface.
      Figure thumbnail fx1
      Fig 1Example of CAPAIBL inspection on a positive 18F-Florbetapir. The report displays the radiotracer retention z-score when compare to a normal negative population followed by normalisation with cortical cerebellum, whole cerebellum and Pons.