Amyloid PET quantification using low-dose CT-guided anatomic standardization

Abstract

Background: Centiloid (CL) scaling has become a standardized quantitative measure in amyloid PET because it facilitates the direct comparison of results across institutions, even when different analytical methods or tracers are used. Standard volumes of interest must be used to calculate the CL scale after the anatomic standardization of amyloid PET images using coregistered MRI; if the MRI is unavailable, the CL scale cannot be accurately calculated. This study sought to determine the substitutability of low-dose CT, which is used to correct PET attenuation in PET/CT equipment, by evaluating the measurement accuracy when low-dose CT is used as an alternative to MRI in the calculation of the CL scale. Amyloid PET images obtained using ¹⁸F-flutemetamol from 24 patients with possible or probable Alzheimer's disease were processed to calculate the CL scale using 3D T1-weighted MRI and low-dose CT of PET/CT. CL_MRI and CL_CT were, respectively, defined as the use of MRI and CT for anatomic standardization and compared. Regional differences in the CT-based and MRI-based standardized anatomic images were also investigated.

Trial registration: Japan Registry of Clinical Trials, jRCTs031180321 (registered 18 March 2019, https://jrct.niph.go.jp/latest-detail/jRCTs031180321 ).

Results: A Bland-Altman plot showed that CL_CT was slightly but significantly underestimated (mean ± standard deviation, - 1.7 ± 2.4; p < 0.002) compared with CL_MRI. The 95% limits of agreement ranged from - 2.8 to - 0.7. Pearson correlation analysis showed a highly significant correlation of r = 0.998 between CL_CT and CL_MRI (p < 0.001). The linear regression equation was CL_MRI = 1.027 × CL_CT + 0.762. In a Bland-Altman plot, Spearman correlation analysis did not identify a significant association between the difference in CL_MRI versus CL_CT and CL load (ρ = - 0.389, p = 0.060). This slight underestimation of CL_CT may derive from slightly higher uptake when the cerebellum is used as a reference area in CT-based anatomically standardized PET images versus MRI-based images.

Conclusions: Low-dose CT of PET/CT can substitute for MRI in the anatomic standardization used to calculate the CL scale from amyloid PET, although a slight underestimation occurs.

Keywords: Alzheimer’s disease; Amyloid imaging; Centiloid scale; Computed tomography; Magnetic resonance imaging; Positron emission tomography.

Fig. 1

Processing pipeline for quantitative measurements of ¹⁸F-flutemetamol accumulation in the cerebral cortex. The subject MRI or CT was oriented and coregistered to the MNI template. The subject PET was oriented and coregistered to the coregistered subject MRI or subject CT. Then, the coregistered subject MRI or subject CT was warped into MNI space using unified segmentation in SPM12. The parameters of the deformation field in this warping were applied to the coregistered subject PET for anatomical standardization into MNI space. SUVR was calculated from the ¹⁸F-flutemetamol PET counts in the cerebral cortical areas (Cortex VOI, CTX VOI) and in the whole cerebellum as a reference area (Reference VOI, WhlCbl VOI) using the GAAIN standard VOI template. Then, SUVR was converted to CL using a direct conversion equation. Processing with a black background was performed using Statistical Parametric Mapping (SPM) 12

Fig. 2

Coregistered (upper row, native space) and anatomically standardized (lower row, MNI space) MRI, PET, and CT images. Almost identical PET images were obtained after anatomic standardization between the MRI-based and CT-based approaches. The PET-only approach involved anatomic standardization using a mean atlas for a single template or an adaptive atlas for a positive or negative template

Fig. 3

Comparison of the SUVR and CL values obtained from the MRI-based, CT-based, and PET-only approaches. Bland–Altman plots a, d showed slight but significant underestimation of SUVR_CT and CL_CT compared with SUVR_MRI and CL_MRI, respectively (p < 0.002). Spearman correlation analysis did not show a significant association between the difference in SUVR_MRI versus SUVR_CT and SUVR load (ρ =  − 0.379, p = 0.051) and between the difference in CL_MRI versus CL_CT and CL load (ρ =  − 0.389, p = 0.060). Pearson correlation analysis g, j showed highly significant correlations of r = 0.998 between SUVR_CT and SUVR_MRI and between CL_CT and CL_MRI (p < 0.001). Bland–Altman plots b, e showed significant underestimation of SUVR_mPET and CL_mPET compared with SUVR_MRI and CL_MRI, respectively (p < 0.05). Spearman correlation analysis showed a significant association between the difference in SUVR_MRI versus SUVR_mPET and SUVR load (ρ =  − 0.713, p < 0.001) and between the difference in CL_MRI versus CL_mPET and CL load (ρ =  − 0.702, p < 0.001). Pearson correlation analysis h, k showed highly significant correlations of r = 0.971 between SUVR_mPET and SUVR_MRI and between CL_mPET and CL_MRI (p < 0.001). Bland–Altman plots c, f showed a tendency for overestimation of SUVR_aPET and CL_aPET compared with SUVR_MRI and CL_MRI, respectively (p > 0.2). Spearman correlation analysis showed a significant association between the difference in SUVR_MRI versus SUVR_aPET and SUVR load (ρ = 0.515, p < 0.001) and between the difference in CL_MRI versus CL_mPET difference and CL load (ρ = 0.515, p < 0.001). Pearson correlation analysis i, l showed significant correlations of r = 0.975 between SUVR_aPET and SUVR_MRI and between CL_aPET and CL_MRI (p < 0.001)

Fig. 4

Direct comparison of anatomically standardized amyloid PET images using CT and MRI. SPM analysis showed significantly (p < 0.001) higher and lower uptake of CT-based standardized PET images than MRI-based standardized PET images presented in a warm color scale (t value, from 0 to 10) and a cool color scale (t value, from 0 to − 10), respectively. The largest differences in the accumulation were visible in the brain stem. Higher uptake of CT-based standardized PET images was observed within the whole cerebellar VOI as a reference area (solid black area). In the supratentorial area, most of the significant differences in uptake were found outside the cortical target VOI (solid white area)