Comparison of 18F-florbetaben quantification results using MR-based and MR-less capaibl: Validation against histopathology

Background: 18F-florbetaben is currently approved for the visual rule out Aβ-amyloid (Aβ) pathology, but Aβ imaging is also used for recruitment and as outcome measure in therapeutic trials, requiring accurate and reproducible quantification of Aβ burden in the brain. We compared the accuracy of MR-based and MR-less CapAIBL® quantification of 18F-florbetaben against histopathology as the standard of truth. Methods: Data from seventy-nine subjects (58%M, aged 80.4±10.3 years) who underwent a 18F-florbetaben PET scan and MRI during life and neuropathological assessment at autopsy (PET scans were acquired less than a year before death) were assessed. Image analysis was performed blind to post-mortem results. Fifteen subjects were randomly chosen for test-retest validation. Three subjects had an inadequateMRI and were excluded from analysis. Aβ scans were normalised with CapAIBL® and scaled using the Centiloid cortical mask and the Centiloid whole cerebellum (WC) and cerebellar grey matter (CGM) reference regions. A 1.43 SUVRCGM threshold was used to differentiate high (Ab+) and low(Ab-) Aβ burden.While this threshold was previously reported, it has not been optimized for the current population. At autopsy, neuritic plaques were scored on Bielschowsky silver stained brain sections as “none”, “sparse”, “moderate” and “frequent”. “None” or “sparse” were categorized to be low, while “moderate” and “frequent” as high neuritic plaques. Results: Test-retest analysis yielded identical SUVR values for both MR and MR-less CapAIBL approaches. The differences between MR-based and MR-less CapAIBL SUVRWC and SUVRCGM were 3% and 5%, respectively. Correlations between SPM and MR-less CapAIBL for SUVRWC and SUVRCGM were 0.96 and 0.89, respectively. Against the histopathological standard of truth, the MR-based approach had a 84% sensitivity, 78% specificity, 80% accuracy, and an effect size of 1.5, while MR-less CapAIBL SUVRWC had a 88% sensitivity, 81% specificity, 84% accuracy, and the effect size was 1.7. Conclusions: Quantifications of 18Fflorbetaben PET scans using MR-based and MR-less CapAIBL approaches showed high agreement. Strong concordance was found between SPM, MR-based and MR-less CapAIBL approaches, suggesting that reproducible quantification of Aβ levels in the brain can be obtained without an MRI.