Rates of global and regional cerebral atrophy in AD and frontotemporal dementia

Subjects.

All patients were recruited through the Specialist Cognitive Disorders Clinic at the National Hospital for Neurology and Neurosurgery (London, UK). Fifty-four patients were diagnosed as having probable AD using the National Institute for Neurological and Communication Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association criteria11; 31 patients were diagnosed as having FTD according to consensus criteria.12 The diagnosis of AD or FTD was made on clinical grounds, and imaging was used to exclude space-occupying lesions, vascular disease, and other pathologies. FTD patients were subdivided into frontal-variant and temporal-variant FTD cases on the basis of neuropsychological findings without further reference to neuroimaging data. Similarly, the AD subjects were given a clinical diagnosis, which was supplemented only by the demonstration on neuropsychological testing of deficits in multiple cognitive domains.

Twenty-seven control subjects were drawn from volunteers as well as from relatives and caregivers of patients. None reported any subjective cognitive problems, and all had Mini-Mental State Examination (MMSE) scores of at least 28/30.

This study received local ethics committee approval.

MRI.

All subjects were scanned on a 1.5 T Signa MR scanner (General Electric, Milwaukee, WI). Scans included a sagittal T1-weighted scout sequence and an axial dual-echo sequence (T2-weighted and proton density weighted). T1-weighted volumetric images were obtained using a spoiled gradient echo technique with a 24-cm field of view and 256 × 128 matrix to provide 124 contiguous 1.5-mm-thick slices in the coronal plane. Scan acquisition parameters were as follows: repetition time = 3,500 ms; echo time = 5 ms; no. of excitations = 1; flip angle = 35°. All repeat scans were performed on the same scanner, using the same acquisition parameters. The mean time interval between scans was 12.8 months (range 3.6 to 45.1 months).

Image analysis.

Image data were transferred to a Sun workstation (Sun Microsystems, Mountain View, CA). All analyses were performed blind to patient details and diagnosis, with random presentation of control and patient scans to the operator. Rates of brain atrophy were calculated following registration of each subject’s pair of scans. Thirty-two subjects (14 control, 11 AD, 7 FTD) had three or more scans; in each of these cases, the rate of atrophy was calculated for each temporally adjacent registered scan pair, and the average of these values was taken as the overall rate of atrophy.

Regions defining whole brain were first obtained using semiautomated iterative morphologic techniques13 with the image intensity threshold for the boundary between brain and CSF set at 60% of mean brain image intensity. The inferior cutoff for the brain volume was taken through the brainstem at the level of the most inferior point of the cerebellum. Ventricular volumes comprised the lateral ventricles and their temporal horns.

The registration algorithm determines the rotations and translations required to obtain an optimum match over the whole brain. The optimization procedure minimizes the SD of the ratio of signal intensity from each voxel. In this fashion, all brain voxels contribute toward accurate registration.14 Changes in whole-brain volume are then calculated directly from the registered image by measuring the BBSI. All rates of atrophy are expressed in terms of percentage change in brain volume per annum.

Regional brain analysis.

A standard anatomic space defined by the MNI 305 average brain15 was split into four subvolumes (figure 1). An initial left–right split was made by dividing the whole volume along a line coincident with the interhemispheric fissure. These two subvolumes were each then divided midway along the anteroposterior axis to define left anterior (LA), left posterior (LP), right anterior (RA), and right posterior (RP) quadrants.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. (A and B) Quadrantic segmentation of a whole-brain volume. The baseline (initial) MRI scan is registered into standard space, and quadrant regions are transformed onto the baseline scan. Quadrantic brain boundary shift integral (BBSI) computations result in four regional BBSI measures. Note that the anterior quadrants encompass a large proportion of the frontal lobes as well as the anterior portion of the temporal lobes. (The images represent two axial sections from the same scan of a patient with AD.)

The quadrant regions were transformed onto the baseline scan of each subject and used to restrict BBSI computation to the volume of interest, resulting in four regional BBSI measures of change, namely, BBSI_LA, BBSI_LP, BBSI_RA, and BBSI_RP. These were normalized by the volume of brain enclosed by the relevant quadrant on the baseline scan and converted to annual rates of atrophy. Similarly, left and right annual rates of loss were calculated by normalizing to total left and total right hemisphere volume and anterior and posterior rates were normalized to total anterior or total posterior volume.

For validation, the total change reported by the quadrant-masked BBSI was compared with the standard whole-brain BBSI for the cohort and found to be in close agreement. Accuracy of quadrant BBSI segmentation was also tested by checking the concordance between the volume of the summed baseline quadrant volumes and that of the whole brain. In all patients, 100% concordance was achieved.

Reproducibility of quadrant volume measurements was tested by reanalysis of the brains of 10 study subjects. Measurement reproducibility was 100% in all cases.

Reproducibility.

Inter-rater reproducibility was assessed by measuring the brain and ventricular volumes in 10 and 15 subjects, with scans presented to two investigators blind to patient details and diagnosis. Intrarater reproducibility was assessed with five randomly chosen subjects measured twice. Reproducibility was expressed as the coefficient of variation (SD divided by mean).

BBSI reproducibility was determined by comparing the BBSI between a baseline scan and two “same-day” repeat scans performed on the same subject to obtain two independent measures of volume loss for the same subject over the same time interval. This reproducibility testing has been documented in full.16

Statistical analysis.

Data were analyzed using Stata version 6.0 (Stata Corp., College Station, TX) and SPSS version 8.0 (SPSS, Chicago, IL). The Mann–Whitney test was used to compare the rates of brain atrophy between dementia patients and control subjects and also to compare the rates of atrophy between men and women in each group. If this test revealed a significant difference between patients and controls, analysis of variance was used to assess the different patient groups.

Reproducibility.

For brain volumes, the mean coefficients of variation for inter- and intrarater reproducibility were 0.55% (range 0.07 to 1.1%) and 0.54% (range 0.03 to 1.5%). For ventricular volumes, the mean coefficients of variation for inter- and intrarater reproducibility were 0.13% (range 0 to 0.49%) and 0.04% (range 0 to 0.12%).

Clinical details of study subjects.

There was no difference between the ages of control subjects and demented patients (z = −1.4, p = 0.16) (table).

The mean MMSE scores in the AD and FTD groups were 19.6 (SD 5.2) and 21.9 (SD 6.0). There was a wide range of disease severity in both patient groups, as indicated by the range of scores on the MMSE (AD: range 11 to 29; FTD: range 6 to 28). One AD patient had a high MMSE of 29; this patient had autosomal dominant familial AD caused by a presenilin-1 mutation and at the time of the initial scan fulfilled the criteria for diagnosis of AD. (This patient has been described in a previous publication.17) There was no difference in the MMSE scores between AD and FTD patients (Mann–Whitney test: z = −1.6, p = 0.11).

Rates of brain atrophy.

There was a difference in the rates of brain atrophy between the control group and the dementia (AD + FTD) group (z = −7.20, p < 0.0005) (figures 2 and 3⇓). When the AD and FTD subgroups were compared individually with the control group, both patient groups were different from controls in terms of the rates of whole-brain atrophy (p < 0.0005), but AD and FTD groups did not differ from each other (z = −1.4, p = 0.16).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Distribution of cerebral atrophy in a patient with AD (A) and frontotemporal dementia (FTD) (B). Both patients had similar interscan intervals (11 months) and annualized rates of global cerebral atrophy (approximately 4%). Areas of volume loss are highlighted in red. Note the generalized distribution of atrophy in AD and the preferential anterior volume loss in FTD.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Rates of global cerebral atrophy. Squares = control subjects; triangles = AD; diamonds = frontotemporal dementia.

There was no effect of age (p = 0.51) or gender (p = 0.49) on the rates of atrophy.

The value of the measurement of the rate of whole-brain atrophy as a variable to discriminate between control subjects and patients was determined by calculation of its sensitivity and specificity. For a specificity of 85%, measurement of the rate of atrophy was found to discriminate between control subjects and all patients with dementia with a sensitivity of 96%.

The AD and FTD groups were then analyzed separately. The receiver operator characteristic curves for each group are shown in figure 4. For the AD group, discrimination from control subjects was achieved with a specificity of 89% and sensitivity of 94%. For the FTD group, discrimination from control subjects was obtained with a specificity of 89% and a sensitivity of 83%.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Receiver operator characteristic (ROC) curves for AD (A) and frontotemporal dementia (FTD) (B).

Rates of atrophy in different FTD subtypes.

There was considerable variation in the rates of atrophy in the FTD patients. In view of the fact that FTD is recognized to have clinically distinct frontal and temporal subtypes,12,18⇓ these two subtypes were analyzed separately (figure 5). To this end, the FTD patients were segregated into frontal (n = 17) and temporal (n = 13) groups. Frontal FTD patients (14 men, 3 women) presented with change in personality, impairment of social interpersonal skills, loss of insight, and perseverative behavior. Their clinical presentations were consistent with a diagnosis of FTD based on consensus criteria12 and with the syndromes described previously as frontal lobe degeneration,19 frontal lobe degeneration of non-Alzheimer type,20 and dementia of frontal type.21 All but one of the temporal-variant FTD cases (five men, eight women) had a diagnosis of semantic dementia,22 with anomia, loss of word meaning, and impaired word comprehension as the predominant presenting features. The other patient had a diagnosis of primary progressive nonfluent aphasia with a disorder of expressive language as the primary clinical feature.23 Although primary progressive nonfluent aphasia is typically associated with left perisylvian atrophy,24 it is considered here as a temporal variant of FTD in accordance with a previous study.18

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. Rates of global atrophy in frontotemporal dementia (FTD) subgroups. Squares = frontal FTD; triangles = temporal FTD.

Separate analyses revealed that there was a difference in the rates of atrophy between the two FTD subgroups. Whereas the patients with frontal-variant FTD (mean rate 3.7%, SD 2.5%) exhibited a wide range of atrophy rates (0.3 to 8.0%), the patients with temporal FTD were more homogeneous; and in terms of the mean rate of atrophy (2.5%, SD 1.0%) as well as the range of atrophy rates (0.9 to 4.0%), these latter were more similar to the AD patients (mean 2.4%, SD 1.0%, range 0.5 to 4.7%) than the frontal FTD patients.

Rates of atrophy in different brain quadrants.

The results of regional analysis of atrophy rates are provided in figure 6. In the patients with AD, there were no differences in the rates of atrophy affecting different quadrants (see figure 6A), whereas differences were noted for both frontal and temporal FTD groups (see figure 6B). In frontal FTD, the atrophy rate was greater in the two anterior quadrants than in the two posterior quadrants (LA versus LP: p = 0.003; RA versus RP: p = 0.004), but there was no interhemispheric difference anteriorly or posteriorly. In temporal FTD, the greatest loss of brain volume occurred in the LA quadrant, with a marked difference between LA and LP quadrants (p = 0.001); in addition, a difference was noted between the two anterior quadrants (LA versus RA: p = 0.008). There was also a difference between the LP and RP quadrants, but this did not reach significance (p = 0.05).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6. Annualized rates of atrophy in different brain quadrants, expressed in terms of mean group values (SD in parentheses). (A) AD and frontotemporal dementia (FTD); (B) frontal FTD and temporal FTD; (C) controls. The figure on top depicts the parcellation of the quadrants. LA = left anterior; RA = right anterior; LP = left posterior; RP = right posterior. Unpaired t-tests were used to compare pairs of quadrants. A post hoc Bonferroni-type correction was made to compensate for the multiple comparisons made for each patient group, and the level of significance was therefore set at p = 0.01. *Significant difference across hemispheres; †significant difference between anterior and posterior quadrants in the same hemisphere.

When compared with the quadrantic rates of atrophy in the control population (see figure 6C), the regional atrophy rates were significantly increased in all patient groups. Of particular interest is the observation that the patients with frontal FTD, in whom the volume loss is most marked for the anterior quadrants, were also found to have increased atrophy in the posterior cortex.

A comparison of AD and FTD revealed that there were differences between the two patient groups for the atrophy rates of both anterior quadrants (LA: p = 0.03; RA: p = 0.04) but not for the posterior quadrants (LP: p = 0.17; RP: p = 0.19).

Comparison of rates of atrophy in AD and FTD.

With regard to the rates of brain atrophy in the different dementia groups, two main observations were made. First, there is a large amount of overlap across these groups, to the extent that the rate of whole-brain atrophy cannot be employed as a variable to discriminate between AD and FTD. Second, the FTD group was associated with a far greater spread in terms of the rates of atrophy, with values ranging from 0.32% annual loss of brain volume (less than the mean control value) to 8.01%, which is more than twice the mean for the FTD group and approximately 70% greater than the greatest rate of atrophy in the AD group. It remains possible that the AD and FTD groups differed in terms of disease severity, although in this respect, it is of note that the two patient groups were similar in terms of their MMSE scores.

Atrophy rates in different FTD subtypes.

One interesting finding was the observation of a clear difference in the rates of atrophy in patients with frontal and temporal variants of FTD. Despite the fact that the patients with frontal FTD were relatively similar in terms of their presenting features and clinical progression, there was a large variation in their rates of atrophy; in fact, this subgroup accounted entirely for the wide range of values noted in the FTD group as a whole. This observation was particularly notable because none of the FTD patients in this study had been diagnosed as having FTD associated with motor neuron disease, since this clinical subset of FTD is associated with a more aggressive disease course.

By contrast, the patients with temporal-variant FTD were more homogeneous in terms of their atrophy rates than the patients with frontal FTD. The temporal-variant FTD patients were similar to the AD patients in terms of whole-brain atrophy rates, with mean annual volume losses of 2.5 and 2.4% of total brain volume. Perhaps importantly in this respect, the temporal FTD group was relatively homogeneous in clinical presentation, with semantic memory impairment in all cases and a diagnosis of semantic dementia in 12 of the 13 patients.

It is unclear why the patients with frontal FTD should have a wider variation in atrophy rates. One possibility is that the patients with temporal FTD in this study share the same underlying pathologic process, whereas the frontal FTD cases represent a variety of different pathologies. FTD is associated with two main histologic variants: The commoner variant is characterized by neuronal loss and microvacuolation, whereas the less common Pick-type histology is typified by tau-positive argyrophilic intraneuronal inclusion bodies and inflated neurons. More recently, a third histopathologic variant of FTD has been described that is associated with ubiquitin-positive tau-negative inclusion bodies.28,29⇓ At present, it is not known whether the different histologic subtypes of FTD are associated with different rates of disease progression, and the possibility exists that the large variation in the atrophy rates of the patients with frontal FTD may reflect different underlying pathologic processes. However, alternative explanations include the possibility that the frontal FTD cases exhibited a wider range of disease severity than was reflected in their MMSE test scores. With regard to the relative homogeneity of the atrophy rates in the temporal FTD group, either this might reflect the homogeneous clinical presentation of the temporal FTD cases (all but one of which had semantic dementia), or it might imply that the temporal FTD group shared similar underlying histologic features.

Rates of atrophy in different brain quadrants.

Analysis of the regional distribution of volume loss revealed striking differences between the patient groups. In the AD patients, the rate of atrophy was similar across all four brain quadrants, whereas in the FTD patients, there was a greater loss of volume in the two anterior quadrants. A comparison of the two patient groups revealed that atrophy rates in both LA and RA quadrants were significantly higher in the patients with FTD. Further analysis of the two FTD subgroups demonstrated that there was no difference between the LA and RA quadrants in the case of the frontal FTD patients, which contrasts with the distribution of volume loss in the patients with temporal FTD, which affected primarily the LA quadrant.

These data lend support to the notion that disease progression in AD is associated with widespread cortical atrophy that is evenly distributed between the two cerebral hemispheres. Furthermore, although FTD is known to be associated with focal cerebral atrophy, these results show that the burden of disease progression in FTD is more focal in nature and that the two clinical FTD variants are associated with different patterns of cerebral volume loss.

Analysis of quadrantic rates of atrophy revealed the interesting observation that in FTD there was significant volume loss in the posterior quadrants, with rates of atrophy comparable with that observed in the AD patient group. Although this may represent real tissue loss in these brain regions in FTD, it is possible that these atrophy rates reflect instead the structural readjustment of the brain to the marked degree of volume loss in the anterior brain regions. This would manifest as remodeling of the lateral ventricles, with redistribution of CSF from the anterior to the posterior portion of the ventricles, resulting in an apparent increased rate of atrophy posteriorly.

Study limitations.

Imaging studies of this type rely predominantly on antemortem diagnosis, based on clinical observations. The lack of histologic confirmation of diagnosis introduces an element of uncertainty into the study results. Of the 84 patients with dementia in this study, confirmatory pathologic data are available only in 3 cases so far.

In this study, the MMSE was used as a measure of disease severity. However, the MMSE has several drawbacks when used to compare AD and FTD patients, with the semantic dementia patients in particular being disproportionately disadvantaged on account of their impaired word comprehension. However, other measures of dementia severity such as the Clinical Dementia Rating Scale30 or the documentation of disease duration have their own flaws when used to compare the two diseases. As a result, the MMSE was used in this study in a more restricted role. The wide range of values within the AD and FTD groups was taken to represent the differing levels of disease severity within each group, and the MMSE was employed only as a covariate in the analysis of the relationship between study group and atrophy rate and was not analyzed as an independent variable in its own right.

As a final point, the possible limitations of the image registration technique need to be considered. The intraindividual registration technique is relatively robust to asymmetric atrophic change due to the enormous complexity of the brain. The degree of atrophy documented across the interscan intervals used in this study is such that there remains a large body of complex brain structure for the registration technique to utilize. However, the registration used to transform quadrant regions from the MNI template onto the subject scans is less well constrained as one-to-one correspondence of structure at the finest scale is lacking. Thus, there is the potential of introducing bias into the registration in cases where scans differ significantly from the standard. To minimize this bias, a Gaussian smoothing function was applied to the template registration step to minimize differences between scans: quadrant placement was subject to visual inspection, and regional atrophy has been quoted as a percentage of the enclosed quadrant brain volume. Different approaches may be required in the future to automate the detection of atrophy on smaller or more structure-specific levels.