Scientists/MR physicists/Neurologist/Radiologists/Nuclear Medicine Physicians with interest in Alzheimer’s disease, dementia and PET-MR studiesStudy the metabolic changes using 18F-FDG-PET, perfusion using arterial spin labeling (ASL) and white matter integrity using diffusion tensor imaging (DTI) in the hippocampal subfields of patients suffering from mild cognitive impairment (MCI) and Alzheimer’s disease (AD).

Cohort & Image acquisition

12 subjects with memory complaints were recruited as part of an ongoing study, in accordance with IRB and HIPAA. Stratification was based on clinical dementia rating (CDR) scores. 4 subjects had probable AD (CDR=2, 68±4 yr), 4 subjects had MCI (CDR=1, 65±5 yr) and 4 subjects did not have any clinical evidence of MCI/AD and were considered controls (CDR=0, 70±6 yr). Patients underwent a 75-minute FDG PET-MRI scan on a simultaneous time-of-flight 3T PET-MR scanner (SIGNA, GE , Waukesha, WI, USA) following a 5 mCi intravenous injection of 18F-FDG. The MRI sequences are described in Table 1.

Image processing

Pixel-wise quantitative cerebral glucose uptake maps (CMRGlu) were computed based on pharmacokinetic modeling of the dynamic PET acquisition to minimize errors encountered in semi-quantitative standardized uptake value reconstruction. The carotid artery was delineated on a 10-second PET image during the initial tracer uptake and used to compute an arterial input function (AIF) (Figure 1). Kinetic modeling was performed using a two-compartment model in Pmod [1]. Cerebral blood flow (CBF) was computed from the pseudo-continuous ASL. High-resolution diffusion tensor imaging (DTI) was eddy and distortion corrected using FSL 5.0. Our whole-brain T1 IR-SPGR and FLAIR were input into a Freesurfer pipeline to parcellate the brain and compute intracranial volume (ICV).

Subfield segmentation

Hippocampal subfields were segmented using 3D T1-weighted IR-SPGR and high-resolution oblique coronal thin-section T2-weighted images with ASHS [2], which automatically segments the following subfields: cornu ammonis (CA) 1, CA2, CA3, dentate gyrus (DG)+CA4, subiculum (SUB) and entorhinal cortex (ERC) (Figure 2). CA2+CA3 were combined into a single subfield because of their small size. The whole hippocampus was all of these subfields combined.

Statistical analysis

An ANOVA was performed between the three groups (Controls, MCI and AD) separately for volumetry (adjusting for ICV), CMRGlu (adjusting for cerebellar metabolism), CBF (adjusting for cerebellar perfusion), fractional anisotropy (FA), and mean diffusivity (MD) within each subfield. Post hoc Tukey’s test was subsequently performed when the F-test was significant. Statistical analysis was performed in Matlab and Stata.

Whole hippocampus: The analysis did not yield significant results in the one-way ANOVA between the groups for volumes, CMRGlu, CBF, FA, or MD.

Hippocampal subfields: The subiculum was the only subfield to show a significant volume decrease (p=0.037, F=5.46). CMRGlu was reduced significantly in AD and MCI patients with the highest effect for metabolic change seen in DG+CA4 (p=0.0158, F=6.81) (Figure 3). Post hoc Tukey’s test showed that CMRGlu in the AD group was significantly lower than controls in the DG+CA4 subfield (no significance was found between controls and MCI). A significant decrease in CBF across the three groups was also found in the subiculum (p=0.043, F=4.93) (Figure 4). No significant differences were found in the analysis of FA and MD.

Histological data suggests that the hippocampal subfields are selectively affected in AD [3]. However, hippocampal subfield structures have not been explored with PET largely due to poor spatial resolution and the complexity of spatial alignment of small structures [4]. The simultaneous imaging nature of integrated PET-MR scanners decreases the registration complexity and facilitates a multimodal analysis. While the hippocampal subfields are at the limit of PET spatial resolution, we demonstrate the utility of subfield analysis: in this small cohort, a significant decline in CMRGlu was only detected with the subfield analysis. The association of volume and perfusion changes and dissociation of both with glucose metabolism suggests separable processes may accompany the neurodegeneration. The glucose metabolic changes may be secondary to reduced entorhinal perforant input into the hilus, whereas the subicular decreases may be a downstream effect.Our preliminary results demonstrate that metabolic and perfusion changes can be simultaneously assessed at the subfield level and that subfield analysis may be more sensitive to these pathological changes than global hippocampal assessment. This study highlights the utility of simultaneous MR-PET as a tool for discovering novel biomarkers for early diagnosis of AD and disentangling microstructural and metabolic derangements accompanying neurodegeneration. Future work will include a larger patient cohort and assessment of resting-state connectivity.