Quantification of tissue damage in AD using diffusion tensor and magnetization transfer MRI

Patients and methods.

We studied 18 patients (10 women, 8 men; mean age [range], 71.3 [64 to 81] years; median disease duration [range], 24 [12 to 40] months) who met the National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) criteria for clinically probable AD,8 and 16 healthy control subjects (9 women, 7 men; mean age [range], 69.0 [65 to 70] years). Major systemic, psychiatric, and other neurologic illnesses were carefully excluded. Subjects were excluded from the study if they had either one T2-hyperintense area with a diameter ≥5 mm or more than four T2-hyperintense areas smaller than 5 mm. Mean Mini-Mental State Examination (MMSE) score9 (corrected for age and level of education) was 17.7 (range, 5.0 to 25.4) for patients with AD and 27.7 (range, 26.3 to 30.7) for control subjects.

The following MRI sequences were obtained: 1) dual-echo turbo spin echo (SE) (TR = 3300, TE = 16/98, echo train length = 5); 2) T1-weighted SE (TR = 650, TE = 12); 3) 2D gradient-echo (GE) (TR = 640, TE = 12, flip angle = 20°), with and without an off-resonance radio-frequency (RF) saturation pulse (offset frequency = 1.5 kHz, Gaussian envelope duration = 7.68 msec, flip angle = 500°); and 4) pulsed-gradient spin-echo (PGSE) echo-planar (inter-echo spacing = 0.8, TE = 123), with diffusion gradients applied in eight noncollinear directions, chosen to cover three-dimensional space uniformly. Additional information about this sequence is reported elsewhere.7 For the dual-echo, T1-weighted and GE sequences, 24 contiguous, 5 mm-thick axial slices were acquired with an in-plane resolution of approximately 1 × 1 mm. The slices were positioned to run parallel to a line that joins the most inferoanterior and inferoposterior parts of the corpus callosum. For the PGSE scans, 10 5 mm-thick slices were acquired with the same orientation as the other scans, positioning the second-last caudal slice to match exactly the central slices of the other image sets.

Two observers, unaware to whom the scans belonged, identified any white matter hyperintensity on the dual-echo scans. Brain volumes were measured by a single observer, again unaware to whom the scans belonged, using the entire T1-weighted image set of slices and a segmentation technique based on signal intensity thresholding.10 Image postprocessing to obtain MTR, D̄, and FA maps was run as described elsewhere.7 Then, GM and WM were segmented using a technique based on FA thresholding.7 MTR and D̄ characteristics of the brain tissue (BT), cGM, and WM were studied using histogram analysis, as previously described.7 Using standard anatomic landmarks, the imaged portions of the temporal and occipital lobes were segmented on the coregistered T2-weighted images. The selected regions were then superimposed on the GM maps to produce the corresponding MTR and D̄ histograms. For each histogram, we derived the peak height, the peak location, and the average MTR and D̄.

Student’s t-test was used to compare histogram metrics from patients with AD and control subjects. Multivariable linear regression models were used to generate composite MR scores. Each score was computed using a linear combination of MR parameters, chosen on a priori biologic considerations. The weight of each MR parameter resulted from the coefficients estimated by the regression model. The correlations between MMSE and the composite MR scores were evaluated by Spearman correlation analysis. To reflect the large numbers of comparisons, a p value ≤0.01 was considered significant and a p value between >0.01 and ≤0.05 was considered a trend.

Results.

Aspecific WM hyperintense lesions were detected on scans from four controls and five patients with AD. Brain volume was lower (p < 0.001) in patients with AD (mean [SD] = 918 [82] ml) than in control subjects (mean [SD] = 1086 [99] ml). The peak heights of the BT D̄ and MTR histograms were lower and average brain D̄ higher in patients with AD than in control subjects (table 1). The same differences were found when considering cGM (table 2). A significance trend for the cGM D̄ histogram peak location was also found. The peak heights of D̄ and MTR histograms of temporal lobe GM were lower and the peak location of the D̄ histogram higher in patients with AD than in control subjects (table 3). A significance trend was found for temporal lobe average D̄. No significant difference was found between patients with AD and control subjects for any of the histogram-derived measures of the occipital lobe GM. WM average D̄ (0.89 ± 0.05 versus 0.85 ± 0.04 mm²s⁻¹; p = 0.007) and peak height of the D̄ histogram (126.0 ± 22.6 versus 143.9 ± 15.7; p = 0.01) were different between patients and controls.

Two composite MR models were created using quantities reflecting different aspects of AD pathology. The first item was brain volume, which reflects the extent of tissue loss, and the second item was either cGM D̄ histogram peak height or cGM MTR histogram peak height (both these quantities reflect intrinsic pathology of the residual cGM7). The r values of the correlations with the MMSE score were 0.21 for brain volume, 0.31 for cGM D̄ histogram peak height, and 0.58 (p = 0.01) for cGM MTR histogram peak height. The composite MR score derived from the second model (i.e., brain volume and cGM MTR histogram peak height) was strongly correlated with MMSE score (r = 0.65, p = 0.003). The r value of the correlation between the other composite MR score and MMSE score was 0.40.

Discussion.

In this preliminary study based on MTR and D̄ histogram analysis, we showed reduced MTR and increased D̄ in a large brain portion from patients with AD. This is likely to reflect two of the major pathologic events of AD, i.e., nerve cell loss in cGM and Wallerian degeneration of WM fibers. Although information related to the status of specific brain structures is lost, this approach might be desirable in the context of clinical trials of AD, in which it may be unfeasible to measure MR changes from several different regions of the brain.

The definition of MTR and D̄ characteristics of specific brain regions and tissues was achieved by segmenting GM and WM pixels using an automated technique based on a quantitative index of tissue microstructural organization.7 We showed abnormal water distribution and diffusivity in the cGM and in that of the temporal lobe taken in isolation. This is consistent with nerve cell loss in associative cortices of patients with AD. Admittedly, we cannot exclude that, because of brain atrophy, more pixels with partial volume effect from the cerebrospinal fluid might have been introduced in the patients’ GM pool, thus contributing to reduced MTR and increased D̄ values. Although we cannot exclude the presence of selective partial volume averaging, the demonstration that MTR and D̄ of the occipital lobe GM did not differ between patients with AD and control subjects suggests that partial volume averaging is likely not to have affected our results significantly.

Because GM histogram peak heights and brain volume reflect the intrinsic damage and the extent of residual tissue,7 we anticipated that aggregates of these MR measures might result in more accurate pictures of tissue damage in AD than those derived from the application of these MR techniques in isolation. Our results confirmed that, in patients with AD, a composite MR score is strongly related with a clinical measure of overall cognitive function.