Statistics from Altmetric.com
In the paper by Du et al (this issue, pp441–447)1 an evaluation of the entorhinal cortex (ERC), hippocampus (HP), and total brain tissue volumes was conducted in controls, patients with Alzheimer's disease, and non-demented subjects with a clinical dementia rating (CDR) score of 0.5 implicating mild memory impairment. Although the major finding of the study was that the HP and ERC were significantly reduced in volume for the CDR 0.5 group compared with controls, there was a large degree of overlap between the two groups and the sensitivity and specificity of correctly classifying the two groups were not improved when ERC volumes were added to HP measures alone. Postmortem studies of those with mild memory impairment and early Alzheimer's disease have implicated the ERC as the first site to carry the burden of Alzheimer's disease pathology, which then moves into the HP proper. Brain imaging studies in aging and dementia are focusing on those brain regions that might demonstrate the earliest abnormalities underlying incipient AD. The atrophic hippocampus has yielded the best results in persons with mild impairment. Yet, with so much pathological evidence implicating the ERC as the germinal site of incipient Alzheimer's disease why has it escaped our grasp in antemortem studies?
Several factors could be responsible for the large variability of ERC imaging measures. Beyond the obvious variability of the outliner's hand and the patient's anatomy not being easily discernible due to imaging artifacts that are quite common in the medial temporal region, there is the intersubject variability of brain anatomy, normal or disease related, that confounds our outlining rules; add to this variable head positioning during scan acquisition and slice thickness that exceeds pixel size. All of these methodological issues pale if the Alzheimer's disease process itself is heterogeneous in its early phase with only some patients having a medial temporal disease focus. What strategies then are available to maximise the earliest disease detection?
The imaging community is now undergoing a radical advance in methodology. Studies that collapse the complexity of a brain region into a single value, such as volume, are incapable of discerning a subtle abnormality that may be present within a portion of the volume of interest. Thus, voxel based approaches are now being conducted to avoid such type II errors. Most voxel based approaches combine individual subjects within a common brain space, typically the Talairach space, for statistical analysis. Unfortunately, the aged or demented brain is not easily compatible with the Talairach space and thus anatomical mismatch occurs that hinders statistical comparisons across groups of brains. To consider this problem disease specific atlases have been created,2 with non-linear high order warping algorithms, in which surface based morphometry can be applied to detect subtle changes that escape regional volumetry.3Extracting the unique shape changes, presumably produced by disease, encoded in the high order warping field that brings one subject into the atlas template is termed tensor mapping. Tensor mapping currently holds the greatest promise in detecting the earliest morphological disease specific change for incipient AD.4-6
Armed with a better method for image analysis is only part of the solution. The poor sensitivity/specificity of correctly classifying the mild memory group from controls when ERC volumes were added to HP measures alone, as reported in the paper by Du et al,1 may result from the heterogeneity of the mildly impaired group. Not all persons with mild memory impairment have incipient Alzheimer's disease. Future studies should evaluate the ability of sensitive imaging methods to correctly identify the incipient patients with AD within the total group of mildly impaired persons. Now that many centres are collecting longitudinal clinical and imaging data on these “at risk” persons collaborative studies are urgently needed to increase the power of shared imaging methods across centres. Only through the sharing of common methods and data can a predictive imaging tool be validated for the identification of incipient Alzheimer's disease in the mildly impaired person who presents to a memory disorder clinic.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.