Poster Presentations P1
T279
baseline to follow-up data using paired t-tests in SPM2. Complementary correlative analyses were performed between-modality as well as with cognitive decline to better understand the relationships between these on-going changes and their impact on cognition. Results: The profiles of progression mainly involved the whole hippocampus for GM atrophy, the cingulum and fornix for WM atrophy, and the precuneus for hypometabolism. There was a significant positive relationship between the rate of hippocampal atrophy, alteration of the caudal part of the cingulum bundle, and metabolic decrease in the precuneus, that were significantly related to memory deficits. Conclusions: As a whole, this study highlights the on-going brain changes that accompany progression from MCI to AD. It suggests that the well known progression of hippocampal atrophy is paralleled by the disruption of its projection tract to the PCC, which may itself be involved in the progression of hypometabolism in the precuneus, these related changes being responsible for memory deficits in early AD. P1-230
FRONTAL INFARCTS ARE ASSOCIATED WITH EXECUTIVE AND LANGUAGE IMPAIRMENTS IN PATIENTS WITH ALZHEIMER’S DISEASE
Soo-Jin Cho1, Nikolaos Scarmeas2, Karen Marder3, Lawrence S. Honig3, 1Hallym Medical College of Medicine, Seoul, Republic of Korea; 2Columbia University, New York, NY, USA; 3 Columbia Univerisity, New York, NY, USA. Contact e-mail:
[email protected]
P1-229
RATE OF HIPPOCAMPAL ATROPHY IS RELATED TO WHITE MATTER ATROPHY AND HYPOMETABOLISM PROGRESSION IN MILD COGNITIVE IMPAIRMENT: A MULTIMODAL LONGITUDINAL STUDY
Gae¨l Chetelat1, Florence Mezenge1, Brigitte Landeau1, Vincent de la Sayette1, Fausto Viader1, Jean-Claude Baron2, Francis Eustache1, Be´atrice Desgranges1, 1Inserm - EPHE - Universite´ de Caen Basse-Normandie, Unite´ U923, GIP Cyceron, CHU Coˆte de Nacre, caen, France; 2Dept of Neurology, University of Cambridge, Cambridge, United Kingdom. Contact e-mail:
[email protected] Background: There have been several longitudinal studies in patients with mild cognitive impairment (MCI) showing the progression of grey matter (GM) atrophy (mainly involving the hippocampus), and only one assessing the evolution of hypometabolism (highlighting the involvement of frontal and posterior cingulate (PCC) cortices). More recently, DTI studies have highlighted the alteration of white matter (WM) tracts, more specifically in the cingulum, corpus callosum, fornix and perforant path, but there is no longitudinal study upon the evolution of WM alterations to date. In the present study, our aim was to assess the progression over time of GM and WM atrophy as well as hypometabolism in the same patients with MCI and their inter-relationships so as to better understand the evolving pathophysiological processes underlying early Alzheimer’s disease (AD). Methods: MRI-T1 and FDGPET data were obtained in seventeen patients with MCI both at inclusion and eighteen months later. Grey matter (GM) and white matter (WM) data were obtained from the optimized VBM procedure while PET data were corrected for partial volume effect, spatially normalized, scaled (using the vermis as reference) and smoothed. Progression of GM and WM atrophy and hypometabolism was assessed by comparing
Background: Cerebral infarcts are common in patients with Alzheimer’s disease (AD). However, the nature of the influence of cerebral infarcts on AD symptoms is unclear. Methods: Patients with clinically probable or possible AD from the Columbia University Alzheimer Disease Research Center database, were selected if they met the following inclusion criteria: presence of information about stroke history and a brain MRI or CT neuroimaging study encoded in the database. For these 1001 patients, we used scores from 12 neuropsychological tests to calculate Z scores for memory, language, executive, abstract reasoning, and visuospartial cognition. We used logistic regression to evaluate the influence of location of cerebral infarcts on cognitive performance in AD patients. The models were adjusted for age, gender, education, and ethnicity. Results: Mean age was 76 ⫾ 9.4 years; 66.5% were women; average education was 11.1 ⫾ 5.1 years. Of the 308 patients with infarcts, based on neuroimaging study, 195 had single infarct and 249 had multiple infarcts. The locations of infarcts were as follows: 34 frontal, 14 temporal, 28 parietal, 17 occipital cortex, 119 focal white matter, 104 basal ganglia, 51 thalamus, and 75 the others. Poorer scores on executive testing (based on categorical fluency test and phonemic fluency) and language testing (based on 15-item naming, repetition and comprehension subtests of Boston Diagnostic Aphasia Examination) were related to frontal infarcts with odds ratios 2.54 (95% CI: 1.09-5.92) and 1.37 (95% CI 1.03-1.84), respectively. There were no other significant associations between infarct location and cognitive scores. Conclusions: The concomitant presence of frontal infarcts in AD patients was associated with worse executive and language performance. These findings suggest an influence of cerebral infarct location on the cognitive profiles of AD patients. P1-231
CORRELATION OF AMYLOID DEPOSITION WITH LOCAL AND DISTAL GLUCOSE METABOLISM IN COGNITIVELY NORMAL ELDERLY, MCI AND ALZHEIMER’S DISEASE
Ann D. Cohen, Scott Ziolko, Howard Aizenstein, Robert D. Nebes, Judith A. Saxton, Chester A. Mathis, Julie C. Price, Wenzhu Bi, Lisa A. Weissfeld, Beth E. Snitz, Edith Halligan, Steven T. DeKosky, William E. Klunk, University of Pittsburgh, Pittsburgh, PA, USA. Contact e-mail:
[email protected]
T280
Poster Presentations P1
Background: Amyloid deposition is hypothesized to be involved in the events leading to Alzheimer’s disease (AD), but the mechanism is unclear. Inverse correlations between amyloid deposition and glucose metabolism in parietal regions of AD patients have been shown, suggesting local amyloid toxicity. However, the complexity of brain connectivity suggests that amyloid effects could be far removed as well. Objective: Determine associations of amyloid deposition with metabolism in amyloid-positive normal control subjects, MCI and AD patients. Methods: PiB PET and FDG PET scans were performed on 60 normal control subjects (by standard neuropsychological battery). Sixteen controls (26%; 56-80yrs) with amyloid deposition were included. Of 24 MCI patients, 14 (58%) were amyloid-positive (65-82yrs) and were included, along with 18 AD patients (100% amyloid positive; 56-94yrs). Regional PiB retention (DVR: cerebellum reference, DVR⫽VT/VND) and glucose metabolism (SUVR: cerebellum reference) measures were determined and atrophy-corrected using an MR-based method. PiB retention and metabolism were compared across cortical regions using Pearson’s correlations. Results: Similar to previous reports, glucose metabolism in parietal and precuneus cortices of AD patients was negatively correlated with PiB retention locally, and with PiB retention in frontal and lateral temporal cortices. Positive correlations between PiB retention and metabolism were uncommon in AD (mainly anterior cingulate). In controls, a few negative correlations were found in frontal cortex, while numerous positive correlations were found between metabolism in anterior cingulate and precuneus and PiB retention in several brain areas. In MCI, this effect was more pronounced. There were no significant negative correlations observed in MCI subjects. However, glucose metabolism in anterior cingulate showed positive correlations with PiB retention in most brain areas. Metabolism in precuneus and parietal cortex also was positively correlated with PiB retention in most posterior regions. Conclusions: The association of amyloid deposition and glucose metabolism appears to vary with increasing cognitive severity. In controls with amyloid deposition, we observed positive correlations of glucose metabolism in anterior cingulate and precuneus with global amyloid deposition. In MCI, this phenomenon became exaggerated and predominant. In AD, these positive correlations were rare, and negative correlations become predominant, particularly in precuneus and parietal cortex. P1-232
LONGITUDINAL PROGRESSION OF ALZHEIMER’S DISEASE-LIKE PATTERNS OF BRAIN ATROPHY IN A NORMAL ELDERLY COHORT AND IN MILD COGNITIVE IMPAIRMENT : A HIGH-DIMENSIONAL PATTERN CLASSIFICATION STUDY
Christos Davatzikos1, Feng Xu1, Susan M. Resnick2, 1University of Pennsylvania, Philadelphia, PA, USA; 2Laboratory of Personality and Cognition, National Institute on Aging, Bethesda, MD, USA. Contact e-mail:
[email protected] Background: MRI is an established AD biomarker. Methods for computational neuroanatomy, including high-dimensional pattern analysis and classification, have been demonstrated by several studies to achieve excellent classification of individuals, thereby offering the potential for diagnosis and prognosis. This study was based on two large neuroimaging studies of normal aging and AD: the Baltimore Longitudinal Study of Aging (BLSA) and the Alzheimer’s Disease Neuroimaging Initiative (ADNI). We investigated longitudinal progression of AD-like patterns of atrophy, determined from ADNI, in the BLSA cohort of cognitively normal (CN) elderly and of MCI. Methods: A high-dimensional pattern classifier was trained on 66 CN and 56 AD ADNI patients, and was subsequently applied to 109 CN and 15 MCI individuals from the BLSA study over a period of 9 years. The longitudinal progression of AD-like patterns of atrophy was determined for different age brackets. Results: 98.7% of all BLSA participants that remained
CN were correctly classified as CN, thereby cross-validating the accuracy of ADNI-derived classification on datasets from a different study. CN subjects of ages above 80 progressively displayed AD-like patterns of brain atrophy. The rates of change of classification-derived abnormality scores of CN’s were fairly well clustered around 0, except a small subgroup of them (especially older subjects), generally indicating lack of progression of CN towards AD-like phenotypes. In contrast, rates of change of individuals that developed MCI were more variable and positive, indicating gradual progression of many, but not all, to AD-like structure. Moreover, cognitive scores of CN and MCI that were determined to have AD-like classification scores were significantly lower than their counterparts classified as normal-like. Conclusions: A biomarker of structural abnormality distinguishing CN from AD was derived using sophisticated high-dimensional pattern classification, and was tested on longitudinal MRI scans from cognitively normal elderly and of MCI individuals. Although most CN’s that remain cognitively stable display normal and stable patterns of atrophy, individuals that develop MCI show steady increases in AD-like atrophy patterns. Structural abnormality scores and their rates of change define subgroups of CN and MCI individuals whose cognitive scores differ significantly, further indicating the clinical relevance of this structural biomarker. P1-233
MRI-BASED HIGH-DIMENSIONAL PATTERN CLASSIFICATION OF THE ALZHEIMER’S DISEASE NEUROIMAGING INITIATIVE ALZHEIMER’S DISEASE, MILD COGNITIVE IMPAIRMENT, AND HEALTHY CONTROLS REVEALS PATTERNS OF ATROPHY USED FOR INDIVIDUAL CLASSIFICATION, AND PREDICT SUBSEQUENT COGNITIVE DECLINE
Christos Davatzikos, Yong Fan, Christopher M. Clark, University of Pennsylvania, Philadelphia, PA, USA. Contact e-mail:
[email protected] Background: Imaging biomarkers can potentially play an important role in early diagnosis and treatment monitoring of Alzheimer’s Disease. The ability to detect spatio-temporal patterns of brain atrophy that are characteristic of MCI, and even more importantly of MCI patients that convert to AD, is of high clinical value. Detecting these patterns on an individual patient basis is of highest importance for diagnostic purposes. Methods: This study applies advanced computer-based image analysis methodologies to MR images from ADNI MR images, and examines group differences between AD and MCI patients, and healthy controls. Moreover, high-dimensional pattern classification methods are applied to determine whether individual patient MRI scans can be correctly classified, with sufficient sensitivity and specificity. Results: Eighty-eight MCI patients, 66 healthy controls and 56 AD patients were selected from the ADNI database. The regional distribution of GM, WM and CSF was determined using the RAVENS approach for voxelbased analysis. A high-dimensional pattern classification approach was then used to determine the optimal differentiation between the three groups, and to establish a way to classify individual scans. Leave-oneout cross validation was used to estimate the generalization ability of the classifier, and showed 94% correct classification rate of individual AD patients and healthy controls; the MCI group was relatively more overlapping with the other two. Finally, an abnormality score was derived for each MCI participant, indicating the similarity of the participant’s spatial pattern of brain atrophy with that of AD patients. 2/3 of the MCI subjects had AD-like patterns of atrophy, whereas 1/3 had normal-like patterns. The former showed significant subsequent MMSE decline within a year (-2.31 points), whereas the latter did not. Conclusions: Significantly reduced grey matter volumes in AD and MCI patients of the ADNI cohort were measured in several brain regions known to be affected by AD. High-dimensional pattern classification