Cognitive Aging, Executive Function, and Fractional Anisotropy: A Diffusion Tensor MR Imaging Study

Subjects

Eight-seven healthy subjects spanning 6 decades were drawn from the Brain Resource International Database (BRID), a standardized data base combining demographic, psychometric, physiologic, and anatomic information.^39,40 Exclusion criteria were any known neurologic disorder, previous head injury, mental retardation, DSM-IV Axis 1 psychiatric diagnosis,⁴¹ and history of drug dependence. The DASS (Depression Anxiety Stress Scale) and SPHERE (Somatic and Psychological Health Report) questionnaires were used to screen for persons with a vulnerability to anxiety or depressive disorder.⁴² The age range of the sample was 20–73 years. Each subject provided written informed consent in accordance with the ethical guidelines of the National Health and Medical Research Council of Australia.

Image Acquisition

MR images were acquired using a 1.5T Sonata scanner (Siemens, Erlangen, Germany) at Perrett Imaging, Flinders University, Australia. 3D T1-weighted images were acquired in the sagittal plane using a 3D magnetization-prepared rapid acquisition of gradient echo (MPRAGE) sequence (TR, 9.7 ms; TE, 4 ms; echo train, 7; flip angle, 12°; TI, 200 ms; NEX, 1). A total of 180 contiguous 1-mm sections were acquired with a 256 × 256 matrix with an in-plane resolution of 1 × 1 mm resulting in isotropic voxels. Proton density and T2-weighted images were acquired using a dual-echo sequence (TR, 7530 ms; TE, 15/105 ms; echo train, 7; flip angle, 180°; NEX, 1). Forty-five contiguous 1-mm sections are acquired in an axial orientation, with an in-plane matrix of 256 × 256 at a resolution of 0.86 × 0.86 mm. DTIs were acquired using a DTI-EPI sequence (TR, 160 ms; TE, 88 ms; fat saturation; NEX, 4). A baseline image (b = 0) and 12 different diffusion orientations were acquired with a b-value of 1250. Thirty-two contiguous 6.5-mm sections were acquired with an in-plane matrix of 128 × 128 at a resolution of 1.72 × 1.72 mm.

Volumetric MR Imaging Measurements

The details of the global MR imaging analysis procedure are described in Grieve (2005).⁹ MPRAGE images were normalized to MNI space and segmented using Statistical Parametric Mapping (SPM2; Wellcome Department of Imaging Neuroscience, London; http://www.fil.ion.ucl.ac.uk/spm), running under MATLAB 6.5 (MathWorks, Natick, Mass). Voxels were corrected after normalization to allow for volume quantification.⁴³ Global measurements were obtained by summing each segmented voxel type after the application of a whole-brain mask.

Segmentation of T2-Weighted/Proton Density MR Imaging Data

Coregistration and normalization was performed using SPM. Customized GM and WM template images which were created from the averaged T2-weighted images of a first pass-segmentation were used for calculations.⁸ For each subject, T2-weighted data were normalized to the custom T2-weighted template. The segmentation protocol used a cluster analysis method to separate pixels based on intensity differences, together with a priori knowledge of spatial tissue distribution patterns in healthy subjects.^8,43,44 After segmentation, WM maps were saved for use as a mask to exclude GM voxels from the subsequent FA analysis. A threshold fractional value of 0.5 was used, implying greater than 50% probability of WM in the voxel. Over 98% of WM voxels in each segmented WM image used in this study have a fractional value between 0.8 and 1.0.

Diffusion Tensor Analysis

DTI data were processed using a custom written routine (by author S.M.G.) in MATLAB 6.5 (MathWorks). FA was calculated in native space from the b = 0 image and 12 diffusion-weighted imaged images (b = 1250 s/cm⁻²). FA was defined as: (3/2)^1/2 × [((λ₁ − λ_av)² + (λ₂ − λ_av)² + (λ₃ − λ_av)²) / (λ₁² + λ₂² + λ₃²)]^1/2, where λ_n = the eigenvalues describing the diffusion tensor, and λ_av is the mean diffusivity ((λ₁ + λ₂ + λ₃)/3).

T2-Weighted and ADC Image Normalization

The FA image was coregistered to the T2-weighted image using a minimization-of-residuals approach.⁴⁴ The DTI dataset and dual-echo images were acquired using the same section orientation and position. Use of the FA and T2-weighted image pair in the coregistration step gave the best results (data not shown). As described above, T2-weighted images were first normalized to MNI space. The affine transformation parameters used in this transform were saved and applied to the coregistered FA images to generate normalized FA images in MNI space.

Gross Region of Interest Parcellation

Frontal, parietal, temporal, and occipital lobe regions were defined so as to be consistent with the automated anatomic labeling (AAL) structures defined by Tzourio-Mozoyer et al (2001) but with borders extending into WM, deep into this cortical parcellation.⁴⁵ The AAL parcellation was chosen as the basis for the segmentation because it is defined in Montreal Neurologic Institute (MNI) space, a standardized coordinate system in common usage. The landmarks used for delineating these lobes are clarified below. All lobe assignments were made with reference to each other to be internally consistent. The frontal lobe was defined as all cerebrum anterior to the central sulcus and superior to the Sylvian fissure. Parcellation was performed in an axial view, and assignments were cross-checked with AAL parcellation to ensure consistency. At the deepest point of the central sulcus, a straight line was drawn to the interhemisphereic fissure. Where the anterior horns of the lateral ventricles were visible, a line was drawn from the central sulcus to the anterior limit of the ventricles, then back to the interhemispheric fissure. The corpus callosum was separately outlined to exclude it from the frontal lobe. In sections containing the insula, with reference to the AAL parcellation, this region was excluded, and a line was drawn from the Sylvian fissure to the interhemispheric fissure. Inferior regions of the brain were clearly separated from temporal lobe structures by CSF. The parietal lobe was defined as all cerebrum superior and anterior to the parieto-occipital sulcus, posterior to the central sulcus, and superior to the corpus callosum and was traced in a sagittal plane with reference to both AAL and other lobe assignments. The occipital lobe was defined as everything posterior to the boundaries described below. Parcellation of the occipital lobe was performed from the midline in a parasagittal view. The anterosuperior border was defined by the parieto-occipital and the temporo-occipital sulci. Lateral to the midline, from where the parieto-occipital sulcus was no longer prominent, a straight line was drawn between it and the horizontal ramus of the superior temporal sulcus. The posteroinferior border was defined by the anterior calcarine sulcus, the collateral sulcus, and the posterior transverse collateral sulcus and included the lingual gyrus. The temporal lobe included the superior, middle, and inferior gyri. The borders were defined by the Sylvian fissure, occipitotemporal sulcus, superior temporal sulcus, and the lateral border of the parahippocampal gyrus.

Statistical Analysis

Descriptive data are expressed as mean ± SD or as percentage values, as appropriate. Before correlation analysis, we grouped cognitive measures into quartiles according to cognitive performance. Statistical analysis used SPSS for Windows Version 12.0 (SPSS, Chicago, Ill). Correlations were performed using Pearson product-moment correlation, controlling for sex and age where indicated. Quantitative estimates of associations were performed using multiple linear regression and are quoted as β coefficients (95% confidence interval [95% CI]). In all analyses, a value of P < .05 was considered significant.

Diffusion Tensor Analysis

Voxel-based analysis of the FA images was performed using SPM. Using the WM segmentation of the T2-weighted/proton density dataset, WM masks were used to create FA images containing WM only. This was done because GM and WM have characteristically different FA values and are therefore not correctly analyzed using the parametric assumptions inherent in VBM analysis using SPM. A 12-mm Gaussian filter was applied to the FA datasets before analysis using VBM, but not before region-based analysis.

The relationship between aging and FA was investigated at a whole-brain level using regression model with age as a continuous variable and sex as a nuisance covariate. To investigate the relationship between FA and cognitive performance, a reduced field of interest including only the frontal lobe was defined (see above). Executive maze and the composite psychometric score described above were included in the model as continuous variables.

Psychometric Testing

Psychometric tests were accessed from the BRID, and the acquisition and validation of these tests has been have been fully described elsewhere.^39,40,46,47 The psychometric tests were part of a fully computerized and standardized battery, “IntegNeuro.”^39,40 The IntegNeuro test battery has been validated against conventional psychometric testing.⁴⁶ The cognitive tests used were chosen to index the brain’s core cognitive competencies: sensorimotor (tapping and choice reaction time), memory (digit span), executive function (switching of attention task, executive maze), and estimated intelligence as a global measure of cognition (spot the real word). Some details of these tests are given below.

Patient Characteristics

Differences in subject characteristics across age were tested for by dividing the subjects into 2 groups. The average age of the younger group was 25 ± 5 years and that of the older group was 51 ± 9 years. No differences in sex, handedness, education, alcohol consumption, or frequency of positive SPHERE scores were detected between the groups.

Psychometric Tests and Volumetric MR Imaging

Table 1 summarizes the age-related trends observed for the psychometric and volumetric MR imaging measures. Of the psychometric data, only the performance in the executive maze (P < .001) and in the attention-switching task (P = .004) were significantly different between groups. Analysis of the volumetric MR imaging data scores revealed significantly lower GM volume (P < .001), and a nonsignificantly higher CSF volume (P = .072) in the elder subjects, a result consistent with that previously demonstrated in the BRID cohort from which the subjects in the current study were drawn.

Region of Interest Measurements of FA

The global average FA was 0.320 ± 0.017. By lobe, the average FA values were: frontal, 0.311 ± 0.024; parietal, 0.324 ± 0.020; temporal, 0.325 ± 0.018; and occipital, 0.268 ± 0.017. Correlations between FA scores and age and between FA and a range of psychometric test scores are presented in Table 2. Except for the occipital lobe, FA was strongly related to age and was independent of sex for all brain regions (P < .001). FA in the frontal, parietal, and temporal lobes was strongly correlated to executive maze completion time and to the attention-switching task error rate (P < .01). The attention-switching–FA relationship was also significant in the occipital lobe and survived correction for age and sex in all lobes. The cognition-FA associations were strongest for executive maze; however, executive maze-FA associations were dependent on age. Improved tapping speed was positively correlated to FA in the temporal and frontal lobes, as well as reaction time to frontal lobe FA; however, these correlations were weaker than those described above. It is noteworthy that tapping speed appeared to be independently associated with FA in the temporal lobe. No significant correlations were detected between estimated IQ and digit span and any FA measurement.

Scatterplots of frontal FA data are presented in Fig 1 versus age (Fig 1A), and executive maze completion time (Fig 1B). Figure 1A shows the relationship between frontal FA and age can be approximated by a linear trend (β = 0.01/year; 95% CI, 0.007–0.013/year; r² = 0.371; P < .001). This corresponds to a decline from an average frontal FA of 0.34 in the third decade (20–30 years of age) to a value of 0.30 in the seventh (60–70 years of age), or approximately 3% per decade. Figure 1B shows the relationship between FA and executive maze, with executive maze completion time plotted as the dependent variable. This relationship can also be approximated as a linear trend (β = −24 seconds per 0.01 FA; 95% CI, 16–33 s/0.01 FA; r² = 0.274; P < .001). Post hoc analysis (Tukey test) of the attention switching data grouped the data into 2 homogenous subsets: poor performers (greater than one error, lower FA) and good performers (one error or less, higher FA). The difference in FA between these groups was 0.017 (P < .001) in the frontal and parietal lobes and 0.014 (P < .001) in the occipital and temporal lobes.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Scatterplots displaying the relationship between FA and age (A). FA decreases linearly with age (β = 0.010/year; 95% CI, 0.007–0.013 year; r² = 0.371; P < .001).

B, FA and time to completion for the executive maze test. Higher FA is associated with faster completion of the maze, a relationship described by a linear trend (β = −24 seconds per 0.01 FA unit; 95% CI, 16–33 seconds per 0.01 FA; r² = 0.274; P < .001).

Voxel-Based Morphometry of FA: Age Effects

The relationship between age and FA was analyzed at a voxel level by using VBM to characterize any focal age-related regional changes in FA. The results of a whole-brain, voxel-based analysis demonstrated significant regions of reduced FA associated with advanced age in the frontal WM tracts. No significant voxels were detected subcortically or in the temporal, parietal, or occipital lobes. Bilateral regions were seen for a negative relationship between FA and age (ie, FA is lower in older age groups) in the WM tracts adjacent to the middle frontal gyrus and the dorsolateral superior frontal gyrus. Further significant regions were detected over the WM adjacent to the left inferior frontal gyrus. These regions are presented in Fig 2 as color T-score maps superimposed on a representative anatomic MR image registered in MNI space. Figure 2A shows a series of axial views 10 mm apart (ordered left to right, top to bottom in an inferior to superior direction) that show the anterior to posterior extent of these bilateral clusters. Sections are labeled by their MNI section coordinate. Figure 2B shows a series of coronal section views (ordered left to right, top to bottom in a posterior-to-anterior direction). The bilateral clusters extend over the middle frontal, superior frontal, and inferior frontal gyri. This result was adjusted for sex by the inclusion of sex as a nuisance covariate.

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

A, A series of axial images at 10-mm intervals showing T-scores (in color) superimposed over a representative high-resolution anatomic magnetization-prepared rapid acquisition of gradient echo (MPRAGE) image in standardized MNI space. Bilateral cluster of voxels are present in the frontal and parietal lobes (sections +28 to +48 mm); in addition, significant voxels connect these regions to the anterior thalamus (sections −2 to +28 mm).

B, A series of coronal sections at 10-mm intervals showing the connection between the frontal and parietal significant voxels with the anterior thalamus (section −6 mm). Bilateral clusters of voxels represent brain regions where a significant negative association between FA and age exists. These clusters indicate regions in which FA loss is greatest in advancing age.

Voxel-Based Morphometry of FA: Psychometric Data

Of the psychometric tests, only executive maze and attention- switching error rate showed a significant difference between the younger and elder groups (Table 1); hence, our VBM analysis was limited to these measures. Figure 3 shows 2 sets of serial images displaying the significant voxel clusters detected in the VBM analysis. A number of significant FA voxel clusters associated with executive maze performance were detected and included, and they exceed those seen in the age-FA model, with extensive significant clusters extending to the parietal lobe and the thalamus. These clusters represent regions in which a positive FA/executive maze relationship (ie, higher FA is associated with faster executive maze completion). No negative maze performance-FA clusters were detected. Figure 3A shows a series of axial views 10 mm apart (ordered left to right, top to bottom in an inferior-to-superior direction) that show the anterior to posterior extent of these bilateral clusters. Sections are labeled by their MNI section coordinate. Figure 3B shows a series of coronal sections views (ordered left to right, top to bottom in a posterior-to-anterior direction) that demonstrate the corticothalamic and corticocortical connections between cluster maxima. Cluster maxima are seen in WM regions adjacent to the inferior parietal gyrus, the precentral and postcentral gyri, the prefrontal cortex, the middle frontal gyrus, and in the anterolateral portions of the thalamus, adjacent to the inferior left parietal gyrus.

• Download figure
• Open in new tab
• Download powerpoint

Fig 3.

A, A series of axial images at 10-mm intervals showing T-scores (in color) superimposed over a representative high-resolution anatomic MPRAGE image in standardized MNI space. Bilateral cluster of voxels are present in the frontal and parietal lobes (sections +28 to +48 mm), in addition significant voxels connect these regions to the anterior thalamus (sections −2 to +28 mm).

B, A series of coronal sections at 10-mm intervals showing the connection between the frontal and parietal significant voxels with the anterior thalamus (section −6 mm). Color voxels represent regions of association between voxel FA and overall time score in an executive function task is significant.

Common areas of significant clusters are shared with the age-FA model (bilaterally) adjacent to the middle frontal, superior frontal, and inferior frontal gyri. A cluster was also observed in the anterolateral aspect of the left thalamus extending both ventrally and caudally. Although they have been described for the purposes of location as clusters, Fig 3A,-B clearly show that the clusters described are contiguous. Figure 4 compares the overlap between the age-FA and maze performance-FA clusters by superimposing the age-FA clusters in color over the significant maze performance-FA clusters (gray-scale) in a maximum intensity projection. The overall volume of the age-FA clusters was 29.4 mL compared with 49.5 mL for the maze performance-FA clusters.

• Download figure
• Open in new tab
• Download powerpoint

Fig 4.

A maximum intensity projection image displaying the age-related FA changes superimposed over the significant regions of FA associated with altered performance in executive function task. The volume of the age-FA cluster was 29.4 mL versus 49.5 mL in the executive function-FA association cluster.

No significant regional clusters representing a structurally restricted association between attention-switching error rate and FA were detected using VBM. The highly skewed distribution of errors in the attention-switching task necessitated the use of categories (good versus poor performers) in this analysis.