High field magnetic resonance microscopy of the human hippocampus in Alzheimer's disease: quantitative imaging and correlation with iron

Abstract

We report R(2) and R(2) in human hippocampus from five unfixed post-mortem Alzheimer's disease (AD) and three age-matched control cases. Formalin-fixed tissues from opposing hemispheres in a matched AD and control were included for comparison. Imaging was performed in a 600MHz (14T) vertical bore magnet at MR microscopy resolution to obtain R(2) and R(2) (62 μm×62 μm in-plane, 80 μm slice thickness), and R(1) at 250 μm isotropic resolution. R(1), R(2) and R(2) maps were computed for individual slices in each case, and used to compare subfields between AD and controls. The magnitudes of R(2) and R(2) changed very little between AD and control, but their variances in the Cornu Ammonis and dentate gyrus were significantly higher in AD compared for controls (p<0.001). To investigate the relationship between tissue iron and MRI parameters, each tissue block was cryosectioned at 30 μm in the imaging plane, and iron distribution was mapped using synchrotron microfocus X-ray fluorescence spectroscopy. A positive correlation of R(2) and R(2)* with iron was demonstrated. While studies with fixed tissues are more straightforward to conduct, fixation can alter iron status in tissues, making measurement of unfixed tissue relevant. To our knowledge, these data represent an advance in quantitative imaging of hippocampal subfields in unfixed tissue, and the methods facilitate direct analysis of the relationship between MRI parameters and iron. The significantly increased variance in AD compared for controls warrants investigation at lower fields and in-vivo, to determine if this parameter is clinically relevant.

Figure 1

Schematic of measurement configuration for the MRI measurement of unfixed tissue blocks. The tissue is suspended in approximately 2 ml Fluorinert™ (FC-43) and positioned from above with gauze to prevent sample movement during data acquisition. Care was taken to avoid introduction of air bubbles, especially at the gauze-tissue interface, and any small pockets of trapped air were released with a fine plastic rod.

Figure 2

Guide to the segmentation process adopted for the cross sections through the blocks of hippocampal tissue. a) Where image resolution permitted, regions of interest were selected from within the boundaries shown for the following subfields: Cornu Ammonis regions (CA1-CA3); Cornu Ammonis region 4 and dentate gyrus (CA4-DG); alveus (ALV); polymorphic layer (PL); stratum granulosum (SG); stratum moleculare (SM); stratum pyramidale (SP); stratum radiatum (SR); subiculum (sub, not evident in this particular slice). Some layers could not be segmented reliably at this resolution (e.g. stratum lacunosum), the position of the divisions between CA1, CA2, and CA3 cannot be determined in these MRI sections, and the stratum moleculare for the CA and DG was also treated as one region. The numbering and labelling system is from Fig.7 in Duvernoy and Bourgouin, illustrating segmentation of the human hippocampus matched to a high resolution MRI image (Duvernoy and Bourgouin, 1998). For histogram and R₁ analysis, the primary subfields CA1-CA3 and CA4-DG were selected as shown in b).

Figure 3

Decay curves from regions of interest in fixed and unfixed tissues, including examples from the sequences used to obtain R₂* and R₂. A mono-exponential decay is fitted in Paravisionv 4.0 as outlined in Section 2.3.

Figure 4

Box plots of R₂* values for AD (n=5) and control (n=3) hippocampus subfields from unfixed tissue. The middle line represents the median, the bottom and top rectangles represent the 25th and 75th percentile of the R₂* distribution respectively. Bars represent the maximum and minimum R₂* values seen from the data. Diamonds represent average values calculated from adjacent slices in each tissue block in regions where the principal hippocampus structures could be clearly identified and segmented according to the schematic in Figure 2.

Figure 5

Analysis of histograms for the R₂ and R₂* distributions, using Levene’s test for equality of variances. Pooled data for the AD and control cases are shown for R₂* and R₂ in the two primary subfields, CA1-CA3, and CA4-DG. The variance is significantly different between AD and control for each condition illustrated in a) d), with more scatter in the AD cases than in the controls. In each condition, the Levene test output shows the 95% Bonferroni confidence intervals for standard deviations, with the outlier scatter (individual data points as black asterisks) either side. The pattern is also evident at the level of individual cases, as shown with two age- and gender-matched pairs in e), where the complete histograms (with frequency on a logarithmic scale) are accompanied by the Levene’s test results showing the detail of the tails above and below the 95% confidence intervals. The variances are not equal in all instances, at significance level p < 0.001.

Figure 6

R₂* (s⁻¹) and R₂* maps (s⁻¹) of representative MRI datasets from an AD case and a control.

Figure 7

Qualitative comparison of the a) R₂* map, b) R₂ map, c) unstained 30 μm thick tissue section cut from the corresponding region within the block of unfixed control case C3, and d) the iron fluorescence intensity map of the same section as measured with μXRF (60 μm in-plane resolution, linear intensity scale with black to white indicating low to high fluorescence intensity). Sixteen sites used for quantitative analysis (where the central 9 pixels were selected at each site) are outlined on the paired images.

Figure 8

Plot of R₂ and R₂* versus iron fluorescence intensity, using data sampled from regions throughout the gray and white matter of the hippocampus cross section as illustrated in Figure 7. There is evidence of a positive correlation in both examples, although it is clear (and consistent with expectations) that not all R₂ and R₂* contrast in the MR microscopy maps can be directly attributed to iron concentration.