Parametric mapping using spectral analysis for 11C-PBR28 PET reveals neuroinflammation in mild cognitive impairment subjects

Abstract

Purpose: Neuroinflammation and microglial activation play an important role in amnestic mild cognitive impairment (MCI) and Alzheimer's disease. In this study, we investigated the spatial distribution of neuroinflammation in MCI subjects, using spectral analysis (SA) to generate parametric maps and quantify ¹¹C-PBR28 PET, and compared these with compartmental and other kinetic models of quantification.

Methods: Thirteen MCI and nine healthy controls were enrolled in this study. Subjects underwent ¹¹C-PBR28 PET scans with arterial cannulation. Spectral analysis with an arterial plasma input function was used to generate ¹¹C-PBR28 parametric maps. These maps were then compared with regional ¹¹C-PBR28 V_T (volume of distribution) using a two-tissue compartment model and Logan graphic analysis. Amyloid load was also assessed with ¹⁸F-Flutemetamol PET.

Results: With SA, three component peaks were identified in addition to blood volume. The ¹¹C-PBR28 impulse response function (IRF) at 90 min produced the lowest coefficient of variation. Single-subject analysis using this IRF demonstrated microglial activation in five out of seven amyloid-positive MCI subjects. IRF parametric maps of ¹¹C-PBR28 uptake revealed a group-wise significant increase in neuroinflammation in amyloid-positive MCI subjects versus HC in multiple cortical association areas, and particularly in the temporal lobe. Interestingly, compartmental analysis detected group-wise increase in ¹¹C-PBR28 binding in the thalamus of amyloid-positive MCI subjects, while Logan parametric maps did not perform well.

Conclusions: This study demonstrates for the first time that spectral analysis can be used to generate parametric maps of ¹¹C-PBR28 uptake, and is able to detect microglial activation in amyloid-positive MCI subjects. IRF parametric maps of ¹¹C-PBR28 uptake allow voxel-wise single-subject analysis and could be used to evaluate microglial activation in individual subjects.

Keywords: 11C–PBR28; Compartmental modelling; Logan graphic analysis; MCI; PET; Spectral analysis.

Fig. 1

a Kinetic spectrum for a healthy control subject, which revealed three different components [at position βi of 7.07e-04 s^-1 (blue), 1.58e-03 s^-1 (green) and 5.01e-03 s^-1 (red) with amplitude αi of 8.35e-04 s^-1, 1.03e-03 s^-1, and 1.21e-03 s^-1] with fractional blood volume (bv) of 0.065 (cyan). b Predicted curves using spectral analysis IRF (impulse response function) for tracer activity (dashed line) which was measured by the sum of three individual components of the spectrum. The blue (K_I), green and red curves corresponded to the three component peaks in the spectrum

Fig. 2

Individual IRF-90 and Logan V_T parametric maps demonstrated with corresponding MR image. Upper panel displays an MCI patient, middle panel shows a healthy control (HC), and lower panel demonstrates the group-wise average image (Mean) and standard deviation image (SD). The colour bar on the left represents the colour scale used for IRF-90 images, and the colour bar on the right represents the colour scale used for Logan V_T images

Fig. 3

Single-subject VOI analysis of IRF. The colour map represents the significant clusters of increased ¹¹C–PBR28 binding for each MCI patient compared to healthy control cohort. Aβ + and Aβ- represent amyloid-positive MCI and amyloid-negative MCI subjects respectively. The colour bar indicates the significant Z-score which was used for the colour-coded map

Fig. 4

¹¹C–PBR28 blood data. a Parent fraction in arterial plasma. b Plasma over blood ratio

Fig. 5

a The correlation between ¹¹C–PBR28 IRF-90 and Logan V_T in frontal lobe and temporal lobe. b The correlation between ¹¹C–PBR28 IRF and 2TCM4k-1 K V_T in frontal lobe and temporal lobe. c The correlation between ¹¹C–PBR28 IRF and 2TCM4k V_T in frontal lobe and temporal lobe. d The correlation between 2TCM4k-1 K V_T and 2TCM4k V_T in frontal lobe and temporal lobe