Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Alzheimers Dis. Author manuscript; available in PMC 2012 January 1.
Published in final edited form as:
PMCID: PMC3030673

Automated detection of β-amyloid-related cortical and subcortical signal changes in a transgenic model of Alzheimer’s disease using high-field MRI


In vivo imaging of β-amyloid load as a biomarker of Alzheimer’s disease (AD) would be of considerable clinical relevance for the early diagnosis and monitoring of treatment effects. Here, we investigated automated quantification of in vivo T2 relaxation time as a surrogate measure of plaque load in the brains of ten APP/PS1 transgenic mice (age 20 weeks) using in vivo MRI acquisitions on a 7T Bruker ClinScan magnet. APP/PS1 mice present with rapid-onset cerebral β-amyloidosis, and were compared with eight age-matched, wild-type control mice (C57Bl/6J) that do not develop Aβ-deposition in brain. Data were analyzed with a novel automated voxel-based analysis that allowed mapping the entire brain for significant signal changes. In APP/PS1 mice, we found a significant decrease in T2 relaxation times in the deeper neocortical layers, caudate-putamen, thalamus, hippocampus and cerebellum compared to wildtype controls. These changes were in line with the histological distribution of cerebral Aβ plaques and activated microglia. Grey matter density did not differ between wild-type mice and APP/PS1 mice, consistent with a lack of neuronal loss in histological investigations. High-field MRI with automated mapping of T2 time changes may be a useful tool for the detection of plaque load in living transgenic animals, which may become relevant for the evaluation of amyloid lowering intervention effects in future studies.

Keywords: Aβ, amyloid, transgenic mouse models, iron deposition, T2 relaxation time, high-field MRI, Alzheimer’s disease


In vivo imaging of the characteristic neuropathological hallmarks of Alzheimer’s disease (AD) would greatly enhance our ability to improve the early diagnosis of the disease, offering pathways for the evaluation and initiation of preventive treatments. A prominent target of in vivo imaging efforts in AD has been senile plaques, brain lesions that contain large amounts of aggregated β-amyloid (Aβ) peptides. Positron emission tomography (PET) technology using specific ligands such as 11C-PIB (Pittsburgh compound B, [N-Methyl-11C]2-(4′-methylaminophenyl)-6-hydroxybenzothiazole) can give a semiquantitative estimate of Aβ load in the human brain [1]. However, magnetic resonance imaging (MRI) would offer several advantages over PET if MRI could be employed to detect Aβ load in vivo: MRI provides better spatial resolution, is more widely available, requires lower costs and lacks radiation exposure as compared to PET. In addition, there is some uncertainty about the consistency with which PIB imaging can detect high Aβ-load in humans [2, 3], and the ligand does not bind avidly to Aβ deposits in experimental animals [4, 5].

There are three 18F markers presently being evaluated in phase III studies (see, namely as Florbetaben ([18F]-BAY94-9172) [6], Florbetapir ([18F]-AV-45) [7], and Flutemetamol (GE-067) [8]. These tracers have shown similar properties to 11C-PIB, however, with higher unspecific binding in the cerebral white matter and a lower rate of amyloid positive cognitively healthy control subjects [9]. More recent approaches aim at the measurement of fibrillar amyloid load in vivo using PET markers such as [18F]AZD4694. This and similar markers are presently undergoing Phase II studies. 18F-labeled PET tracers may help to make amyloid PET imaging wider available compared to 11C tracers, but uncertainties remain in respect to the specificity of the amyloid binding in cognitively healthy subjects.

In contrast to PET, MRI has not yet achieved the detection of senile plaques in the living human brain, although several strategies to enhance MR contrast have successfully been applied to β-amyloid precursor protein (APP) transgenic mouse brains both in vitro and in vivo [10, 11]. One important source of intrinsic contrast/paramagnetic susceptibility for MRI is iron, which is also associated with β-amyloid plaques [1215]. The widely held view that iron found in and around senile plaques is the dominant cause of the hypointensities seen in MR images has recently been challenged, but still a close association between T2 time-reductions and plaque load has been confirmed [16]. This intrinsic contrast mechanism allows the detection of single, large senile plaques of more than 100μm diameter at very high field strengths of more than 9T in living APP-transgenic mice without the aid of exogenous contrast agents [17, 18]. Measuring the decline of T2 relaxation times is another approach to exploiting the intrinsic paramagnetic properties of plaque-bound iron [19]. Although this method does not allow visualization of individual plaques, it is applicable at lower field strengths and can be easily quantified. As it is not yet clear to what extent the individual imaging of large senile plaques at very high field strengths serves as a reliable estimate of overall cerebral lesion load in transgenic mice, the use of accessible T2 relaxation time mapping is an easier and faster alternative for characterizing plaque load in vivo. This approach is of further interest, because it may be more easily transferable to in vivo studies in humans than the imaging of individual senile plaques. This will become even more relevant with the broader availability of human MR imaging at very high field strengths, such as 7T. The size of dense-core plaques in the human brain is typically smaller than the detection threshold for single plaques of more than 100μm diameter [20, 21] in living transgenic mice [17]. Indeed, the large majority of senile plaques were not visible in T2*-weighted MRI scans of post-mortem human brain specimens, even at 11 Tesla [22].

To date, measurement of T2 times in living transgenic mice has been mainly restricted to selected regions of interest in areas prominently affected by Aβ pathology, such as neocortex and hippocampus [17, 23, 24]. One study used automated image registration to facilitate ROI drawings, but did not apply voxel-based statistics to the co-registered T2 relaxation time maps [25]. The transgenic models investigated thus far develop considerable Aβ plaque load only after more than 1 year of age, extending the time required for larger scale longitudinal studies.

In the present study, we employed T2 time mapping, as originally proposed by Helpern et al. [19], in an APP/PS1 transgenic mouse model [26] that starts to develop cerebral β-amyloidosis by six weeks of age, and that shows considerable accumulation of Aβ at 10 weeks, even before the onset of neuronal loss at age 8 months and beyond. This rapid onset of cerebral β-amyloidosis enables the assessment of lesion load with in vivo imaging at a much earlier age than in previous studies. To extend previous studies of T2 time changes in transgenic models of AD, we complemented the ROI-based analysis of selected subregions of the brain with an automated voxel-based analysis in which the entire brain is mapped for changes in T2 relaxation times. This approach allows a rapid assessment of brain changes without the need for a priori selection of brain areas. We propose that the use of automated, voxel-based analysis of T2 time changes in a rapidly progressing transgenic model of cerebral β-amyloidosis is an easily accessible approach to investigating the potential utility of high-field MRI for the detection of senile plaque load in the living human brain.



We used 10 APP/PS1 heterozygote transgene-positive mice harboring a mutant human presenilin 1 (L166P variant) and a mutant human β-amyloid precursor protein (APPswe) in a C57BL/6J background driven by two separate Thy1 promoters [26], mean age 19.47 (SD 0.56) weeks, and 8 non-transgenic control mice (C57Bl/6J), mean age 20.18 (SD 1.44) weeks. Both groups were not different in age (T = 1.44, 16 df, p = 0.17). Mice were housed under 12h/12h light/dark cycles and had access to food and water ad libitum. For imaging, the mice were anaesthetized using 0.5–2% isoflurane in pure oxygen (flow 0.5–1L/min). The anaesthesia was controlled by continued mechanical measurement of diaphragm movement (30 to 50 movements per minute) and isoflurane concentration was adjusted accordingly. The experiments were conducted as recommended by the NIH and compliant with minimal standards as defined by the European Communities Council Directive of 24 November 1986 (86/609/EEC). The experiments were approved by the local institutional animal care and use committee (LALLF M-V/TSD/7221.3-2.4-020/08).

MRI acquisition

MRI measurements were performed on a 7T Bruker ClinScan magnet with a 20cm inner bore, capable of 290mT/m in 250μs (Bruker BioSpin MRI, Ettlingen, Germany). Images were received by a 2×2 phased-array RF coil, designed specifically for mouse brain studies, which was placed directly on the skull. For T2 relaxation time measurements, a multislice spin echo sequence was applied, with a field-of-view (FOV) of 2.0 × 2.0cm2, matrix size 256 × 256 pixels, two averages, a repetition time (TR) of 4000ms and an echo time (TE) ranging from 15.1ms to 211.4ms in steps of 15.1ms. Six horizontal slices of 0.5mm slice thickness were acquired for each of the 14 echoes.

For morphological analysis, we acquired a T2-weighted turbo spin echo (TSE) sequence with the same FOV and orientation as the SE sequence. Ten slices of 0.6 mm thickness each, with no gap between them, were acquired with TR = 5500 ms, TE = 33 ms, matrix size 640 × 640, 6 averages and echo train length = 7.


One day after the MRI scans, mice were killed by cervical dislocation and perfused with ice-cold, freshly prepared 4% paraformaldehyde, and then post-fixed for 48 hours. The brains were subsequently dehydrated, defatted and embedded in paraffin wax using a Histokinette (Leica Microsystems, Wetzlar, Germany). The paraffin blocks were cut into 4μm-thick sections that were baked for 60 minutes onto glass slides. Selected slides were stained with H&E, and others were immunostained as described previously [27, 28] using a BondMaxx Immunostainer (Menarini, Berlin, Germany) with antibodies against β-amyloid (clone 4G8, host: mouse, Millipore, Schwalbach, Germany), glial fibrillary acidic protein (GFAP) (host: mouse, Dako, Hamburg, Germany), and Iba-1 (host: rabbit, Wako Chemicals GmbH, Neuss, Germany). Initially, slides were dewaxed, pretreated with formic acid for 1 hour (only for the Aβ stain) and incubated with the primary antibody (dilutions: 4G8-Aβ 1:400, GFAP 1:1000, Iba-1 1:1000). Nonspecific binding sites on the tissue sections were blocked and the immunoreaction product then developed using a mixed-DAB-Refine Kit (Menarini). The sections then were counterstained with hematoxylin.

Sections were double-stained with Aβ and Iba-1 using a BondMaxx (Menarini) automated immunostaining system. Sections were first pretreated with 98% formic acid for 5min and immunostained for Aβ using the anti-human Aβ clone 6F3D (1:200, Dako) and the Bond Polymer Refine Detection kit (Menarini). Microglia were then immunostained on the same sections using anti-Iba-1 (1:1000, Wako) and the Bond Polymer AP-Red Detection kit (Menarini). For iron-labeling, slides were incubated with sodium hexacyanoferrate and H2O2 (Prussian Blue method) and enhanced using DAB according to [29]. The Campbell-Switzer (CS) and LuxolFastBlue-Eosin (LFBe) stains were performed as previously described [30, 31]. Finally, the slides were scanned at 230nm resolution using a MIRAX Midi slide scanner (Zeiss MicroImaging GmbH, Jena, Germany) and analyzed using the AxioVision software package (Zeiss).

MR data processing

Prior to further processing of the MR images, the brain was manually edited from the skull using MRIcro 1.40. We determined T2 time maps for each animal from the multiecho sequence by fitting a single exponential curve across the different TEs at each single voxel using a customized script in Matlab 7.0.1 (MathWorks, Natwick, USA) (Figure 2A).

Figure 2
Processing stream for voxel-based analysis of T2 time maps

T2 time measurements

Measurement of T2 relaxation was performed in two ways: (i) using manually placed regions of interest (ROI) on the T2 relaxation time maps in native space; and (ii) using automated voxel-based analysis across the entire grey matter in a group-specific common space. In a post-hoc analysis, after we had found strong effects in the cerebellum (see results section below), we used an approach to correct the T2 relaxation time maps for the amount of grey matter in each voxel based on the TSE scans in an attempt to control partial volume effects from CSF spaces.

ROI-based analysis

Manual measurement was implemented in Medical Image Processing, Analysis and Visualization, Version 4.1.3 (2008-10-15), (MIPAV, Center for Information Technology (CIT), National Institutes of Health (NIH); available at [32]. ROIs were placed in the ME sequence at TE 15.1 ms, because this sequence provided good contrast between cortical and subcortical grey matter and white matter. The locations of the ROIs were then transferred to the T2 relaxation time maps that had been derived from the ME sequence (see above). We placed circular ROIs across the neocortical ribbon including the cingulate cortex, as well as the hippocampus, and the caudate-putamen (CPu). ROIs were placed according to anatomical criteria with reference to a published mouse brain atlas (Figure 1) [33].

Figure 1
Regions of interest (ROIs) in native space

Each anatomical region was sampled in two consecutive slices, with the size of the ROIs adapted to the size of the anatomical area to be sampled. The following anatomical areas were defined:

  • Neocortical ribbon: Twelve ROIs, with an area of 0.2502mm2 each, were evenly distributed across the entire lateral neocortical band of each hemisphere in two consecutive slices. These slices were selected from the six available slices according to the clear visibility of the cortex. They corresponded to inter-aural distances (IAD) of 2.52 mm and 3.24 mm, respectively [33]. The three most anterior ROIs were grouped in the ventral insula across the two consecutive slices at IAD 2.52mm and 3.24mm. With the slice at IAD 2.52mm, we grouped the three ROIs of the middle anterior cortical portion, representing posterior aspects of the ventral insular cortex, the three ROIs of the middle posterior cortical portion, representing perirhinal and posterior insular cortex, and the most posterior three ROIs, representing entorhinal cortex. The latter three ROIs were averaged with the most posterior three ROIs in the more ventral slice at IAD 3.24mm, to represent entorhinal cortex. The middle-anterior three ROIs of slice IAD 3.24mm were averaged to represent primary and secondary somato-sensory cortex, and the following three ROIs (middle posterior) were grouped to represent auditory and temporal association cortex.
  • Hippocampal formation: The hippocampal formation was sampled with a single ROI with an area of 1.1169mm2 in each hemisphere in two consecutive slices. Of the six available slices, we selected the two in which the hippocampal formation could best be delineated, with IADs corresponding to 1.68 mm and 2.52 mm, respectively [33].
  • Caudate-Putamen: The CPu was sampled in two consecutive slices corresponding to IADs of 2.52mm and 3.24mm [33]. One ROI was placed in the anterior CPu with an area of 0.9399mm2, and one ROI was placed in the posterior CPu with an area of 0.5737mm2.
  • Cingulate gyrus: The cingulate gyrus (CG) was measured with a single ROI with an area of 0.2075mm2 in one slice, which was selected from six available slices in which the CG could best be defined, with an IAD corresponding to 3.82mm [33].

T2 relaxation times were averaged within each ROI. The selection of the consecutive slices out of the six available slices was unambiguous in all animals. To control for background intensity, we used a control ROI placed in masseter muscle. For this measurement, we derived T2 relaxation time maps from the ME sequence without prior editing of the brain from the skull. Additionally, based on the findings from the voxel-based analysis, we sampled ventricular CSF space using a single ROI with an area of 0.3174mm2 placed in the lateral ventricle in one slice. This slice was selected out of six slices in which the ventricular CSF space could best be outlined, with an IAD corresponding to 3.82mm [33].

Voxel-based analysis

Voxel-based analysis was implemented with Matlab 7.0.1 (MathWorks, USA) through Statistical Parametric Mapping [34, 35] (SPM 5, Wellcome Department of Imaging Neuroscience, London; available at and SPMMouse (Wolfson Brain Imaging Centre, University of Cambridge, UK), a Matlab-based modification of SPM5, available at ( [36, 37]. SPMMouse implements grey matter, white matter and CSF space probability maps from a high-resolution sequence that is in affine co-registration with a publicly-available digital C57BL/6J mouse atlas [38]. The processing of our data followed four subsequent steps that are illustrated (in a simplified fashion) in Figure 2.

In detail, in the first step, the ME sequence with a TE of 15.1ms was co-registered to the TSE sequence using affine transformation with a mutual information cost-function [39], implemented in SPM5 and adapted to murine data in SPMMouse. The affine transformation parameters were then applied to the T2 relaxation time maps that had originated from the ME sequence, resulting in T2 relaxation maps co-registered to the TSE sequence (Figure 2A).

In the second step, the TSE maps were segmented into grey matter and white matter maps. The SPM segmentation employs a mixture model cluster analysis (after correcting for non-uniformity in image intensity) to identify voxel intensities that match particular tissue types combined with a priori probabilistic knowledge of the spatial distribution of tissues derived from grey and white matter and CSF prior probability images (priors) [40]. Prior probability images were derived from the tissue maps implemented in SPMMouse that are in co-registration with a digital mouse atlas [36, 37].

In the third step, the segmented grey matter maps derived from the TSE sequence were normalized to the standard grey matter map provided in SPMMouse using a set of nonlinear basis functions [41, 42]. Normalized grey matter maps then were smoothed (0.5mm full width at half maximum isotropic Gaussian kernel) and averaged to obtain a group-specific grey matter template that was in co-registration with the digital C57 atlas [38] (Figure 2B and C).

In the fourth step, the segmented grey matter maps in native space were normalized to the group-specific grey matter template, the normalization parameters were applied to the grey matter maps themselves as well as to the T2 relaxation maps that were already co-registered with the TSE sequence (Figure 2C and D). This resulted in grey matter maps as well as T2 relaxation time maps that were spatially normalized into the group-specific standard space. The normalized maps were smoothed with a 0.5mm–full-width at half maximum isotropic Gaussian kernel before subsequent statistical analysis.

Post-hoc analysis of CSF partial volume effects

To our knowledge, there is no established method for partial volume effect (PVE) correction of T2 relaxation time data. However, to assess a potential effect of CSF partial volume effects on the cerebellum signal, we adapted an algorithm for PET atrophy correction based on the Müller-Gärtner algorithm [43]. The basic idea of the Müller-Gärtner algorithm in a more recent adaptation [4446] includes the following processing steps:

  • The PET scan is spatially aligned to the high resolution anatomical MRI scan.
  • The anatomical MRI scan is automatically segmented into grey and white matter and CSF compartments.
  • The segmented images are convoluted with the point spread function representing the spatial resolution of the PET camera.
  • The white matter compartment is subtracted from the PET scan to obtain the grey matter PET scan, assuming that CSF radioactivity is negligible.
  • The grey matter PET scan is divided by the convoluted grey matter MRI map, resulting in a PVE corrected PET scan.

Several factors are different in the case of PVE correction of the T2 relaxation time maps compared to the PET PVE correction:

  • There is no equivalent for the point spread function of the PET camera in MRI. Therefore, we performed a sensitivity analysis where we used the T2-weighted turbo spin echo (TSE) sequence convoluted with the voxel resolution of the T2 relaxation time maps in a first analysis and the original voxel resolution of the TSE sequence in a second analysis. Since results were essentially unchanged between both analyses, in the following we only report the results of the unconvoluted TSE scans.
  • In PET, grey matter voxel would have a higher, CSF voxel a lower signal. Here, signal was expected to decrease both in grey matter and CSF. Therefore, it would not be useful to divide voxel intensity of the T2 relaxation time maps by the grey matter probability maps. Therefore, binarized the segemented grey matter maps, where voxel with a grey matter probability of 0.5 and higher were assigned a value of 1, al other voxel were assigned a value of zero. We then multiplied the T2 relaxation time maps by the binarized grey matter maps. We used sensitivity analysis to determine the effect of the threshold used for binarization. Using a threshold of 0.33 probability for grey matter as well as a threshold of 0.66 probability did not essential change the results so that here we report the results from the 0.5 threshold.
  • The white matter compartment in the mouse brain is difficult to separate from grey matter and is of small spatial extent. Therefore, we considered no separate white matter compartment in our analysis.

In summary, we co-registered the T2 maps (via the multiecho sequence with a TE of 15.1ms) with the TSE sequence, as described above. The TSE sequence was segmented into grey and white matter maps as described above. The segmented grey matter maps then were binarized at a grey matter probability threshold of 0.5. Then the coregistered T2 maps were multiplied with the binarized grey matter maps to obtain PVE corrected maps which then were normalized into standard space, as described above.


ROI data

According to the Shapiro-Wilk Test and visual inspection of scatterplots, the data from the transgenic group were not normally distributed. This was mainly due to one transgenic animal whose values lay more than 1.5 interquartile ranges above the third quartile for muscle, hippocampus and CPu measurements, corresponding to a mild outlier, and more than three interquartile ranges above the third quartile for all cortical measurements, corresponding to an extreme outlier. Upon checking the data quality and measurement accuracy, there was no indication of a problem with data acquisition or measurement in this animal. Therefore, we decided to leave the animal in the analysis and to use non-parametric tests for statistical inference to mitigate the effect of the outlier. This is a conservative decision, as the values of the outlier were in an a priori unexpected direction. However, effect-size calculations were repeated with the animal removed from the data set to assess the influence of the outlier on our findings.

ROI-based measurements were compared between groups using the Mann-Whitney U-test. The level of significance was set at p < 0.05.

Automated voxel-based analysis

For statistical analysis, we employed the general linear model on a voxel basis, where we assessed the effect of genotype on T2 relaxation times and grey matter density, respectively. Smoothed T2 relaxation times and grey matter maps were masked for tissue outside of the grey matter using the segmented grey matter map from the group-specific brain template. Results were thresholded at a p level of < 0.05, corrected for multiple comparisons using false discovery rate (FDR), and an extent threshold of 50 contiguous voxels was applied. FDR correction ensures that, on average, not more than 5% of the significant voxels are false positives [47]. We assessed effects in both directions, i.e. an increase as well as a decrease in T2 relaxation times and grey matter density in APP/PS1 transgenic mice relative to wildtype controls. The animal showing extreme values in the ROI-based analysis was kept in the voxel-based analysis as well for two reasons. First, the values were higher than expected, so that the risk of false positives would be decreased. Second, the average intensity of this animal after preprocessing was highest within the transgenic group, but not an outlier.

Post-hoc analysis of CSF partial volume effects

We employed a voxel-based linear model to determine group differences in T2 time measurements based on the smoothed normalized PVE corrected T2 maps. The p-value was set to a liberal level of 0.05, uncorrected, in order to determine whether T2 relaxation times would be different between groups in the cerebellum signal to the same extent than in cortical areas. Thus we controlled the type II error of erroneously not detecting an effect in the cerebellum after PVE correction.


ROI analyses

In the APP/PS1 transgenic mice, Mann-Whitney U-tests revealed significant reductions in T2 relaxation times in all sampled brain regions except the dorsalmost cortical area representing entorhinal cortex (p = 0.146) (Table 1). The ventricular CSF also showed significantly reduced T2 relaxation time in the transgenic mice compared to wildtype controls. In contrast, the control region in the masseter muscle showed no significant difference between groups (p = 0.146, M-W U = 23). Effect size estimates are shown in Table 1, both with and without the values from the one animal presenting as an outlier.

Table 1
Mean values and effect sizes for regional T2 time measurements

Voxel based analyses

Data analysis using a linear model with automated VBM revealed extended reductions of T2 times in deeper cortical layers, hippocampus and CPu of transgenic mice at p < 0.05, FDR-corrected for multiple comparisons. Additionally, the APP/PS1 mice had significant clusters of diminished T2 times in the thalamus, septal nuclei and cerebellum (Figure 3). Effects also were found in ventricular CSF spaces. The opposite contrast, i.e., reduced T2 times in wildtype mice, showed no significant clusters anywhere in the brain at p < 0.05, FDR corrected.

Figure 3
Voxel-based analysis of T2 time reductions in APP/PS1 mice as compared to wildtype mice

The voxel-based analysis of the grey matter maps derived from the TSE sequences showed no significant increase or decline of grey matter volume in any brain region of the transgenic mice compared to wildtype controls at p < 0.05, FDR corrected.

Histological analyses

Mouse brains were analyzed using different conventional and immunohistochemical stains (Figures 4 and and5).5). The H&E stain revealed no obvious pathological or gross morphological changes (Figure 4A). Senile plaques were visualized using the Campbell-Switzer silver-staining method and Aβ-immunohistochemistry. The areal density of plaques was greater in deeper cortical layers compared to superficial cortical layers (Figures 4B and C). GFAP-positive astrocytes were located in close proximity to senile plaques and in central cerebellar regions. The density of astrocytes was higher in deeper cortical layers than in the superficial layers (Figure 4D). Microglial staining labelled activated microglia only in proximity to Aβ deposits (Figure 4E). The Luxol-Fast Blue myelin stain showed no indication of myelin pathology or subcortical fibre tract reductions (Figure 4F). DAB-enhanced Prussian Blue staining showed that the senile plaques and adjacent microglia accumulate iron (Figure 5F).

Figure 4
Histological analysis of an APP/PS1 mouse at the age of 20 weeks
Figure 5
Detailed histological presentation of senile plaques in deeper cortical layers

Post-hoc analysis of CSF partial volume effects

Similar to the uncorrected data, in the voxel-based analysis using the PVE corrected data we found reductions in T2 relaxation times in cortical areas, however, the cerebellum was nearly devoid of any T2 relaxation time changes after PVE correction (Figure 6).

Figure 6
Voxel-based analysis of T2 time reductions in APP/PS1 mice compared to wildtype mice after correction for partial volume effects


We investigated 20-week old APP/PS1 transgenic mice developing early cerebral plaque-like β-amyloidosis using high-field MRI at 7T prior to the onset of neurodegeneration (after age 34–36 weeks). We found a significant reduction of T2 relaxation times in the transgenic animals compared to wildtype controls in deeper neocortical layers, hippocampus and caudate-putamen using both manual ROI measurements in native space as well as automated voxel-based analysis in a group-specific common space. Moreover, this effect was independent of local atrophy, as grey matter maps did not differ in wildtype and APP/PS1-transgenic animals at this timepoint. We focussed with our investigation on the detection of early changes at the beginning of the cerebral amyloidosis prior to the onset of marked neurodegeneration. The rapid onset of amyloidosis as compared to APP-only models (e.g. Tg2576) does not replicate the development of AD, but it helps to investigate in vivo changes due to cerebral amyloidosis within a shorter time frame. The localization of the T2 relaxation time effects mapped the spatial distribution of Aβ-plaques and activated microglia in histological sections of the brain (Figure 4B, C, E).

To our knowledge, this is the first description of the high-field MRI features of the Thy1-promoter-driven APP/PS1 double mutant mouse model in vivo. In this model, cerebral β-amyloidosis has been shown to start by 6 to 8 weeks of age, and to lead to significant plaque load in neocortex, hippocampus, thalamus and striatum by the age of 5 months [26]. At age 8 months, the entire forebrain is laden with Aβ-positive plaques and loss of neurons becomes detectable.

Consistent with the early onset and spatial distribution of plaques in the transgenic mice, our animals showed significantly reduced T2 relaxation times at age 4.4 to 5 months in neocortex, hippocampus, CPu, and thalamus. Several studies have demonstrated an association of iron with plaques in humans and in mouse models of AD pathology [1215]. The iron has been localized to senile plaques and adjacent microglia, where it leads to paramagnetic susceptibility and consequent reductions of T2 relaxation times [17]. Consistently, we identified iron accumulation in the center of the amyloid plaques in our transgenic mice (Figure 5F).

In agreement with the report that significant neuron loss does not begin until 8 months of age in this double transgenic model [26], we did not find significant reductions of grey matter density in the transgenic mice when compared to the wildtype mice, and histological analysis with cell stains further indicates that neuronal loss is minimal or absent at age 5 months (Figure 4A). Moreover, there was no evidence of myelin changes or gross changes in the corpus callosum (Figure 4F) that would also indicate significant neuronal loss in the cortex. It should be noted, however, that other mechanisms could compensate volumetrically for minor neuronal loss, for example, the volume effect of β-amyloid plaques and accompanying astrogliosis (Figure 4D) or microgliosis (Figure 4E). Neuronal loss in this model has been reported at age 8 months and later [26]. For comparison, APP/PS1 models resulting from intercrosses of two independent APP- and PS1-transgene strains exhibit atrophy at age 24 months and later that was detected by in vivo MRI and histological evaluation [48]. A PDAPP mouse model by [49] showed significant reductions of commissural fibers already at 3.6 months of age that became more pronounced at 17 months of age, as detected by MRI. It has been suggested that the very early changes of commissural connections in this PDAPP model represent a neurodevelopmental abnormality induced by the PDAPP transgene construct rather than an effect of neurodegeneration induced by amyloid pathology.

Unexpectedly, we found reduced T2 relaxation times in the ventricular CSF spaces of the transgenic animals. Though the explanation for this outcome remains uncertain, one study in transgenic mice overexpressing the London mutant form of APP at age 24 months found high levels of Aβ in CSF [50]. Therefore, high levels of Aβ, or possibly other substances related to cerebral Aβ deposition, in CSF may lead to reduced T2 relaxation times in CSF spaces, in agreement with the observation that higher viscosity of fluids decreases T2 relaxation times [51]. This issue clearly merits further study.

In addition, we found significantly reduced T2 relaxation times in the cerebellum in the voxel-based analysis. The spatial distribution of the effects matched the extension of the cerebellar folia. In the histological analysis, these areas were devoid of amyloid plaques and amyloid-related microgliosis (Figure 4H). Astrogliosis occurred in cerebellar structures, but was localized to the cerebellar peduncles, sparing the cerebellar folia (Figure 4G). The close association between CSF spaces and grey matter folia in the cerebellum suggests that the observed signal reductions in the cerebellum are most likely due to partial volume effects from CSF spaces that have decreased T2 relaxation times, as seen in the forebrain regions. This interpretation is supported by the finding that after correcting the T2 relaxation time maps for the content of grey matter within each voxel as assessed from the TSE data, cortical areas showed reduced T2 relaxation times, but the cerebellum was devoid of almost any signal reductions. Segmentation of TSE scans into different tissue compartments is less accurate than for human structural MRI data and therefore induces additional variability into the voxel-based analysis, which can also bee seen by the reduced T-values compared to the uncorrected analysis. However, the lack of effects in the cerebellum is not only due to the reduced sensitivity of the analysis, since the effects in the cortex, although with reduced effect sizes, remained preserved.

In our histological analysis, we found iron accumulation in the center of the plaques (Figure 5F), supporting the association between T2 relaxation time reductions and plaque-related iron accumulation in forebrain regions. One study challenged the specificity of T2 relaxation time changes [24]. In this longitudinal study, T2 relaxation times declined significantly over a one-year follow-up in the hippocampus and cortex of an APP/PS1 model that shows slightly later onset of cerebral β-amyloidosis than does the model used in our study [26, 52]. However, PS1 single-transgenic mice, which do not develop Aβ plaques, showed a significant, age-associated decline of T2 relaxation times as well. At the same time, however, these mice were reported not to differ from wildtype controls at baseline, as well as at both follow-up time points in cross-sectional analyses. As the wildtype mice showed no significant longitudinal decline in T2 relaxation times, it remained unclear why the PS single-transgenic mice had progressive decline of T2 relaxation times but never differed from wildtype mice at any time point in the cross-sectional analyses. A likely reason is the severe attrition of animal numbers over time, since the study began with 23 PS1 single-transgenic mice at baseline and ended with only four animals at the second follow-up. This attrition renders the estimates of rates of change unreliable. In contrast, cross-sectional analyses were performed with reasonable group sizes after data from single time-point acquisitions had been added.

The transgenic mice employed in our study represent a valuable model to investigate early onset cerebral β-amyloidosis. However, the morphology and composition of amyloid plaques differs across different mouse models and between transgenic models and AD patients [53, 54]. These differences may limit the translation of imaging findings in transgenic mouse models to human studies. The first attempts to undertake ultra-high-field MRI in human studies in AD at 7T suggest that plaque-related susceptibility alterations can be detected in vivo in cortical areas [55]. T2 relaxation time reductions were found to be associated with senile plaques, both in human brain tissue as well as in post mortem APP/PS1 transgenic mouse brains [16]. T2* contrast reductions were well-matched with iron deposits within and close to senile plaques in post mortem samples of AD patients’ brains. In contrast, there were reductions of the T2* signal that corresponded to the location of plaques in APP/PS1 mice that develop cerebral β-amyloidosis after 1 year of age, but in the absence of significant iron deposits. The authors suggested that the detection of the murine plaques via the T2 effect was related to their higher fibrillar organization and density compared to human plaques. So different morphological changes may lead to similar signal changes in MRI, limiting the inference from MRI signal changes on the underlying plaque morphology. In our transgenic model, we consistently found a significant amount of iron within and in proximity to senile plaques, suggesting that iron is a significant contributor to T2 relaxation time reduction in this context.

The absolute T2 values we measured in our study were higher than those reported in other animal studies [17, 19], but are comparable to those observed in a human investigation [56]. These differences may partly be due to the pulse sequence parameters used. Additionally, the processing of the data likely had an influence. In previous experiments, multiecho data were fitted to an exponential function, but no details were given as to whether the data were processed to exclude overall noise levels [17, 19]. The human study did not employ noise level corrections [56]. We used the raw data without correcting for the effects of noise as well. However, we also estimated noise levels through the signal intensity in an ROI that was placed at the maximum distance from the measured object. Signal intensity in this ROI was nearly identical across the different echo times. Therefore, when considering noise levels, the signal was relatively more reduced with higher echo times, yielding a steeper decay function of the T2 signal. After correcting for noise levels, our T2 values were close to those in the previous animal studies [17, 19]. However, since the between-group effects were not affected by the noise level correction of the data, we decided to use the T2 time measurements derived from the raw data.

There is a wide range of MRI based diagnostic markers in AD, from visual rating of hippocampus atrophy [57] through manual [58] and automated [59] hippocampus volumetry, and cortical thickness measurement [60] to automated detection of high resolution differences in brain morphology using deformation based morphology and multivariate analysis such as principal component analysis [61] or machine learning algorithms such as support vector machines [62]. These approaches yield an accuracy of about 80% in the prediction of AD in at risk subjects such as patients with the clinical syndrome of amnestic mild cognitive impairment [63]. In vivo amyloid imaging techniques will help in future to shift the diagnosis of AD into presymptomatic stages. A recent model on the dynamic use of biomarkers in AD [64] suggests that amyloid pathology precedes the loss of cortical neurons, as detected by structural MRI. However, the specificity of amyloid load findings in the brain for predicting AD in cognitively normal subjects remains to be shown. Additionally, amyloid imaging in combination with functional and structural markers of brain changes based on MRI will help us to better understand the dynamic sequence of molecular changes, such as amyloid accumulation, synaptic dysfunction and neuronal loss. In future, the combined use of these markers will help to improve our models on the development of AD pathology in the brain.

In summary, we found significant reductions of T2 relaxation times in cortical and subcortical brain areas that corresponded to the pattern of Aβ accumulation and reactive gliosis shown in subsequent histological analyses (Figure 4). The APP/PS1-related effects on T2 relaxation times occurred in the absence of obvious morphological brain changes (besides the occurrence of senile plaques), as shown both in vivo in the grey matter density maps and post mortem in histological sections. Our findings suggest that the APP/PS1 transgenic model can serve as a useful model for the high-field MRI analysis of brain changes related to cerebral β-amyloidosis. Due to the earlyonset and rapid progression of Aβ pathology, the APP/PS1 mouse also could be employed for the efficient testing of therapeutic interventions that would be expected to influence Aβ load and associated MR signal. Moreover, our data suggest that voxel-based analysis can readily be employed as an observer-independent and automated approach to detecting in vivo surrogate markers of β-amyloid load and related tissue changes throughout the brain.


Funding sources

Part of this work was supported by grants from the Interdisciplinary Faculty, Department “Ageing Science and Humanities”, University of Rostock, to S.J.T. and J.P., from the Hirnliga Foundation, Wiehl, Germany, to S.J.T., from the NIH to L.C.W. (RR-00165), and from the Alzheimer Forschung Initative e.V., Düsseldorf, Germany, to J.P.

The authors are grateful to Dr. Stephen Sawiak, Behavioural and Clinical Neuroscience Institute and Wolfson Brain Imaging Centre, University of Cambridge, UK, for helpful discussions and providing support for his public domain Matlab toolbox. The authors also want to thank Dr. Markus Becker, Bruker, Ettlingen, Germany, for helpful discussions.


There are no conflicts of interest associated with the work presented in this article.


1. Klunk WE, Engler H, Nordberg A, Wang Y, Blomqvist G, Holt DP, Bergstrom M, Savitcheva I, Huang GF, Estrada S, Ausen B, Debnath M, Barletta J, Price JC, Sandell J, Lopresti BJ, Wall A, Koivisto P, Antoni G, Mathis CA, Langstrom B. Imaging brain amyloid in Alzheimer’s disease with Pittsburgh Compound-B. Ann Neurol. 2004;55:306–319. [PubMed]
2. Rosen RF, Ciliax BJ, Wingo TS, Gearing M, Dooyema J, Lah JJ, Ghiso JA, LeVine H, 3rd, Walker LC. Deficient high-affinity binding of Pittsburgh compound B in a case of Alzheimer’s disease. Acta Neuropathol. 2010;119:221–233. [PMC free article] [PubMed]
3. Cairns NJ, Ikonomovic MD, Benzinger T, Storandt M, Fagan AM, Shah AR, Reinwald LT, Carter D, Felton A, Holtzman DM, Mintun MA, Klunk WE, Morris JC. Absence of Pittsburgh compound B detection of cerebral amyloid beta in a patient with clinical, cognitive, and cerebrospinal fluid markers of Alzheimer disease: a case report. Arch Neurol. 2009;66:1557–1562. [PMC free article] [PubMed]
4. Klunk WE, Lopresti BJ, Ikonomovic MD, Lefterov IM, Koldamova RP, Abrahamson EE, Debnath ML, Holt DP, Huang GF, Shao L, DeKosky ST, Price JC, Mathis CA. Binding of the positron emission tomography tracer Pittsburgh compound-B reflects the amount of amyloid-beta in Alzheimer’s disease brain but not in transgenic mouse brain. J Neurosci. 2005;25:10598–10606. [PubMed]
5. Rosen RF, Walker LC, Levine H., 3rd PIB binding in aged primate brain: Enrichment of high-affinity sites in humans with Alzheimer’s disease. Neurobiol Aging. 2009 [Epub ahead of print] [PMC free article] [PubMed]
6. Rowe CC, Ackerman U, Browne W, Mulligan R, Pike KL, O’Keefe G, Tochon-Danguy H, Chan G, Berlangieri SU, Jones G, Dickinson-Rowe KL, Kung HP, Zhang W, Kung MP, Skovronsky D, Dyrks T, Holl G, Krause S, Friebe M, Lehman L, Lindemann S, Dinkelborg LM, Masters CL, Villemagne VL. Imaging of amyloid beta in Alzheimer’s disease with 18F-BAY94-9172, a novel PET tracer: proof of mechanism. Lancet Neurol. 2008;7:129–135. [PubMed]
7. Yao CH, Lin KJ, Weng CC, Hsiao IT, Ting YS, Yen TC, Jan TR, Skovronsky D, Kung MP, Wey SP. GMP-compliant automated synthesis of [(18)F]AV-45 (Florbetapir F 18) for imaging beta-amyloid plaques in human brain. Appl Radiat Isot. 2010 epub ahead of print. [PubMed]
8. Nelissen N, Van Laere K, Thurfjell L, Owenius R, Vandenbulcke M, Koole M, Bormans G, Brooks DJ, Vandenberghe R. Phase 1 study of the Pittsburgh compound B derivative 18F-flutemetamol in healthy volunteers and patients with probable Alzheimer disease. J Nucl Med. 2009;50:1251–1259. [PubMed]
9. Rowe CC. Standardization of amyloid imaging. Alzheimer’s & Dementia. 2010;6:S87.
10. Poduslo JF, Ramakrishnan M, Holasek SS, Ramirez-Alvarado M, Kandimalla KK, Gilles EJ, Curran GL, Wengenack TM. In vivo targeting of antibody fragments to the nervous system for Alzheimer’s disease immunotherapy and molecular imaging of amyloid plaques. J Neurochem. 2007;102:420–433. [PubMed]
11. Ramakrishnan M, Wengenack TM, Kandimalla KK, Curran GL, Gilles EJ, Ramirez-Alvarado M, Lin J, Garwood M, Jack CR, Jr, Poduslo JF. Selective contrast enhancement of individual Alzheimer’s disease amyloid plaques using a polyamine and Gd-DOTA conjugated antibody fragment against fibrillar Abeta42 for magnetic resonance molecular imaging. Pharm Res. 2008;25:1861–1872. [PMC free article] [PubMed]
12. LeVine SM. Iron deposits in multiple sclerosis and Alzheimer’s disease brains. Brain Res. 1997;760:298–303. [PubMed]
13. Jack CR, Jr, Wengenack TM, Reyes DA, Garwood M, Curran GL, Borowski BJ, Lin J, Preboske GM, Holasek SS, Adriany G, Poduslo JF. In vivo magnetic resonance microimaging of individual amyloid plaques in Alzheimer’s transgenic mice. J Neurosci. 2005;25:10041–10048. [PMC free article] [PubMed]
14. Falangola MF, Lee SP, Nixon RA, Duff K, Helpern JA. Histological co-localization of iron in Abeta plaques of PS/APP transgenic mice. Neurochem Res. 2005;30:201–205. [PubMed]
15. Vanhoutte G, Dewachter I, Borghgraef P, Van Leuven F, Van der Linden A. Noninvasive in vivo MRI detection of neuritic plaques associated with iron in APP[V717I] transgenic mice, a model for Alzheimer’s disease. Magn Reson Med. 2005;53:607–613. [PubMed]
16. Meadowcroft MD, Connor JR, Smith MB, Yang QX. MRI and histological analysis of beta-amyloid plaques in both human Alzheimer’s disease and APP/PS1 transgenic mice. J Magn Reson Imaging. 2009;29:997–1007. [PMC free article] [PubMed]
17. Braakman N, Matysik J, van Duinen SG, Verbeek F, Schliebs R, de Groot HJ, Alia A. Longitudinal assessment of Alzheimer’s beta-amyloid plaque development in transgenic mice monitored by in vivo magnetic resonance microimaging. J Magn Reson Imaging. 2006;24:530–536. [PubMed]
18. Jack CR, Jr, Garwood M, Wengenack TM, Borowski B, Curran GL, Lin J, Adriany G, Grohn OH, Grimm R, Poduslo JF. In vivo visualization of Alzheimer’s amyloid plaques by magnetic resonance imaging in transgenic mice without a contrast agent. Magn Reson Med. 2004;52:1263–1271. [PMC free article] [PubMed]
19. Helpern JA, Lee SP, Falangola MF, Dyakin VV, Bogart A, Ardekani B, Duff K, Branch C, Wisniewski T, de Leon MJ, Wolf O, O’Shea J, Nixon RA. MRI assessment of neuropathology in a transgenic mouse model of Alzheimer’s disease. Magn Reson Med. 2004;51:794–798. [PubMed]
20. Hyman BT, Marzloff K, Arriagada PV. The lack of accumulation of senile plaques or amyloid burden in Alzheimer’s disease suggests a dynamic balance between amyloid deposition and resolution. J Neuropathol Exp Neurol. 1993;52:594–600. [PubMed]
21. Benveniste H, Einstein G, Kim KR, Hulette C, Johnson GA. Detection of neuritic plaques in Alzheimer’s disease by magnetic resonance microscopy. Proc Natl Acad Sci USA. 1999;96:14079–14084. [PubMed]
22. Dhenain M, Privat N, Duyckaerts C, Jacobs RE. Senile plaques do not induce susceptibility effects in T2*-weighted MR microscopic images. NMR Biomed. 2002;15:197–203. [PubMed]
23. Helpern JA, Jensen J, Lee SP, Falangola MF. Quantitative MRI assessment of Alzheimer’s disease. J Mol Neurosci. 2004;24:45–48. [PubMed]
24. Falangola MF, Dyakin VV, Lee SP, Bogart A, Babb JS, Duff K, Nixon R, Helpern JA. Quantitative MRI reveals aging-associated T2 changes in mouse models of Alzheimer’s disease. NMR Biomed. 2007;20:343–351. [PubMed]
25. Falangola MF, Ardekani BA, Lee SP, Babb JS, Bogart A, Dyakin VV, Nixon R, Duff K, Helpern JA. Application of a non-linear image registration algorithm to quantitative analysis of T2 relaxation time in transgenic mouse models of AD pathology. J Neurosci Methods. 2005;144:91–97. [PubMed]
26. Radde R, Bolmont T, Kaeser SA, Coomaraswamy J, Lindau D, Stoltze L, Calhoun ME, Jaggi F, Wolburg H, Gengler S, Haass C, Ghetti B, Czech C, Holscher C, Mathews PM, Jucker M. Abeta42-driven cerebral amyloidosis in transgenic mice reveals early and robust pathology. EMBO Rep. 2006;7:940–946. [PubMed]
27. Pahnke J, Wolkenhauer O, Krohn M, Walker LC. Clinico-pathologic function of cerebral ABC transporters - implications for the pathogenesis of Alzheimer’s disease. Curr Alzheimer Res. 2008;5:396–405. [PMC free article] [PubMed]
28. Vogelgesang S, Warzok RW, Cascorbi I, Kunert-Keil C, Schroeder E, Kroemer HK, Siegmund W, Walker LC, Pahnke J. The role of P-glycoprotein in cerebral amyloid angiopathy; implications for the early pathogenesis of Alzheimer’s disease. Curr Alzheimer Res. 2004;1:121–125. [PMC free article] [PubMed]
29. Nguyen-Legros J, Bizot J, Bolesse M, Pulicani JP. “Diaminobenzidine black” as a new histochemical demonstration of exogenous iron. Histochemistry. 1980;66:239–244. [PubMed]
30. Pahnke J, Walker LC, Scheffler K, Krohn M. Alzheimer’s disease and blood-brain barrier function-Why have anti-beta-amyloid therapies failed to prevent dementia progression? Neurosci Biobehav Rev. 2009;33:1099–1108. [PMC free article] [PubMed]
31. Baumann F, Pahnke J, Radovanovic I, Rulicke T, Bremer J, Tolnay M, Aguzzi A. Functionally relevant domains of the prion protein identified in vivo. PLoS One. 2009;4:e6707. [PMC free article] [PubMed]
32. McAuliffe MJ, Lalonde FM, McGarry D, Gandler W, Csaky K, Trus BL. Medical image processing, analysis and visualization in clinical research. IEEE COMPUTER-BASED MEDICAL SYSTEMS (CBMS) 2001:381–386.
33. Franklin KBJ, Paxinos G. The Mouse Brain in stereotaxic coordinates. Elsevier Inc; Amsterdam: 2007.
34. Friston K, Holmes AP, Worsley K, Poline J-B, CDF, Frackowiak RSJ. Statistical parametric maps in functional imaging: a general linear approach. Hum Brain Mapp. 1995;2:189–210.
35. Friston K, Ashburner J, Frith CD, Poline J-B, Heather JD, Frackowiak RSJ. Spatial registration and normalization of images. Hum Brain Mapp. 1995;2:165–189.
36. Sawiak SJ, Wood NI, Williams GB, Morton AJ, Carpenter TA. Voxel-based morphometry in the R6/2 transgenic mouse reveals differences between genotypes not seen with manual 2D morphometry. Neurobiol Dis. 2009;33:20–27. [PubMed]
37. Sawiak SJ, Wood NI, Williams GB, Morton AJ, Carpenter TA. Use of magnetic resonance imaging for anatomical phenotyping of the R6/2 mouse model of Huntington’s disease. Neurobiol Dis. 2009;33:12–19. [PubMed]
38. Ma Y, Hof PR, Grant SC, Blackband SJ, Bennett R, Slatest L, McGuigan MD, Benveniste H. A three-dimensional digital atlas database of the adult C57BL/6J mouse brain by magnetic resonance microscopy. Neuroscience. 2005;135:1203–1215. [PubMed]
39. Wells WM, 3rd, Viola P, Atsumi H, Nakajima S, Kikinis R. Multi-modal volume registration by maximization of mutual information. Med Image Anal. 1996;1:35–51. [PubMed]
40. Ashburner J, Friston K. Multimodal image coregistration and partitioning--a unified framework. Neuroimage. 1997;6:209–217. [PubMed]
41. Ashburner J, Friston KJ. Voxel-based morphometry-the methods. Neuroimage. 2000;11:805–821. [PubMed]
42. Ashburner J, Neelin P, Collins DL, Evans A, Friston K. Incorporating prior knowledge into image registration. Neuroimage. 1997;6:344–352. [PubMed]
43. Müller-Gärtner HW, Links JM, Prince JL, Bryan RN, McVeigh E, Leal JP, Davatzikos C, Frost JJ. Measurement of radiotracer concentration in brain gray matter using positron emission tomography: MRI-based correction for partial volume effects. Journal of Cerbral Blood Flow and Metabolism. 1992;12:571–583. [PubMed]
44. Matsuda H, Ohnishi T, Asada T, Li ZJ, Kanetaka H, Imabayashi E, Tanaka F, Nakano S. Correction for partial-volume effects on brain perfusion SPECT in healthy men. J Nucl Med. 2003;44:1243–1252. [PubMed]
45. Ibanez V, Pietrini P, Alexander GE, Furey ML, Teichberg D, Rajapakse JC, Rapoport SI, Schapiro MB, Horwitz B. Regional glucose metabolic abnormalities are not the result of atrophy in Alzheimer’s disease. Neurology. 1998;50:1585–1593. [PubMed]
46. Bokde AL, Teipel SJ, Drzezga A, Thissen J, Bartenstein P, Dong W, Leinsinger G, Born C, Schwaiger M, Moeller HJ, Hampel H. Association between Cognitive Performance and Cortical Glucose Metabolism in Patients with Mild Alzheimer’s Disease. Dement Geriatr Cogn Disord. 2005;20:352–357. [PubMed]
47. Genovese CR, Lazar NA, Nichols T. Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage. 2002;15:870–878. [PubMed]
48. Delatour B, Guegan M, Volk A, Dhenain M. In vivo MRI and histological evaluation of brain atrophy in APP/PS1 transgenic mice. Neurobiol Aging. 2006;27:835–847. [PubMed]
49. Gonzalez-Lima F, Berndt JD, Valla JE, Games D, Reiman EM. Reduced corpus callosum, fornix and hippocampus in PDAPP transgenic mouse model of Alzheimer’s disease. Neuroreport. 2001;12:2375–2379. [PubMed]
50. Van Dorpe J, Smeijers L, Dewachter I, Nuyens D, Spittaels K, Van Den Haute C, Mercken M, Moechars D, Laenen I, Kuiperi C, Bruynseels K, Tesseur I, Loos R, Vanderstichele H, Checler F, Sciot R, Van Leuven F. Prominent cerebral amyloid angiopathy in transgenic mice overexpressing the london mutant of human APP in neurons. Am J Pathol. 2000;157:1283–1298. [PubMed]
51. Kamman RL, Go KG, Brouwer W, Berendsen HJ. Nuclear magnetic resonance relaxation in experimental brain edema: effects of water concentration, protein concentration, and temperature. Magn Reson Med. 1988;6:265–274. [PubMed]
52. Holcomb L, Gordon MN, McGowan E, Yu X, Benkovic S, Jantzen P, Wright K, Saad I, Mueller R, Morgan D, Sanders S, Zehr C, O’Campo K, Hardy J, Prada CM, Eckman C, Younkin S, Hsiao K, Duff K. Accelerated Alzheimer-type phenotype in transgenic mice carrying both mutant amyloid precursor protein and presenilin 1 transgenes. Nat Med. 1998;4:97–100. [PubMed]
53. Duyckaerts C, Potier MC, Delatour B. Alzheimer disease models and human neuropathology: similarities and differences. Acta Neuropathol. 2008;115:5–38. [PMC free article] [PubMed]
54. Güntert A, Dobeli H, Bohrmann B. High sensitivity analysis of amyloid-beta peptide composition in amyloid deposits from human and PS2APP mouse brain. Neuroscience. 2006;143:461–475. [PubMed]
55. Nakada T, Matsuzawa H, Igarashi H, Fujii Y, Kwee IL. In vivo visualization of senile-plaque-like pathology in Alzheimer’s disease patients by MR microscopy on a 7T system. J Neuroimaging. 2008;18:125–129. [PubMed]
56. Campeau NG, Petersen RC, Felmlee JP, O’Brien PC, Jack CR., Jr Hippocampal transverse relaxation times in patients with Alzheimer disease. Radiology. 1997;205:197–201. [PubMed]
57. Wahlund LO, Julin P, Johansson SE, Scheltens P. Visual rating and volumetry of the medial temporal lobe on magnetic resonance imaging in dementia: a comparative study. J Neurol Neurosurg Psychiatry. 2000;69:630–635. [PMC free article] [PubMed]
58. Teipel SJ, Ewers M, Wolf S, Jessen F, Kölsch H, Arlt S, Luckhaus C, Schönknecht P, Schmidtke K, Heuser I, Frölich L, Ende G, Pantel J, Wiltfang J, Rakebrandt F, Peters O, Born C, Kornhuber J, Hampel H. Multicentre variability of MRI-based medial temporal lobe volumetry in Alzheimer’s disease. Psychiatry Res Neuroimaging. 2010 accepted. [PubMed]
59. Csernansky JG, Wang L, Swank J, Miller JP, Gado M, McKeel D, Miller MI, Morris JC. Preclinical detection of Alzheimer’s disease: hippocampal shape and volume predict dementia onset in the elderly. Neuroimage. 2005;25:783–792. [PubMed]
60. Lerch JP, Pruessner J, Zijdenbos AP, Collins DL, Teipel SJ, Hampel H, Evans AC. Automated cortical thickness measurements from MRI can accurately separate Alzheimer’s patients from normal elderly controls. Neurobiol Aging. 2008;29:23–30. [PubMed]
61. Teipel SJ, Born C, Ewers M, Bokde AL, Reiser MF, Moller HJ, Hampel H. Multivariate deformation-based analysis of brain atrophy to predict Alzheimer’s disease in mild cognitive impairment. Neuroimage. 2007;38:13–24. [PubMed]
62. Plant C, Teipel SJ, Oswald A, Bohm C, Meindl T, Mourao-Miranda J, Bokde AW, Hampel H, Ewers M. Automated detection of brain atrophy patterns based on MRI for the prediction of Alzheimer’s disease. Neuroimage 2009 [PMC free article] [PubMed]
63. Hampel H, Frank R, Broich K, Teipel SJ, Katz RG, Hardy J, Herholz K, Bokde AL, Jessen F, Hoessler YC, Sanhai WR, Zetterberg H, Woodcock J, Blennow K. Biomarkers for Alzheimer’s disease: academic, industry and regulatory perspectives. Nat Rev Drug Discov. 2010;9:560–574. [PubMed]
64. Jack CR, Jr, Knopman DS, Jagust WJ, Shaw LM, Aisen PS, Weiner MW, Petersen RC, Trojanowski JQ. Hypothetical model of dynamic biomarkers of the Alzheimer’s pathological cascade. Lancet Neurol. 9:119–128. [PMC free article] [PubMed]