suggest that combined PET imaging with an amyloid and dopamine targeted tracer can help distinguish AD from DLB patients. The study is ongoing and additional results will be presented. P2-106 NO ASSOCIATION BETWEEN PRESENCE OF CAROTID DISEASE AND 11C-PIB-PET IN ACUTE STROKE PATIENTS Ramesh Sahathevan 1 , Brodtmann Amy 2 , Victor Villemagne 3 , Paul Yates 4 , Leonid Churilov 2 , John Ly 5 , Thomas Linden 6 , Christopher Rowe 7 , Geoffrey Donnan 2 , 1 Florey Neuroscience Institutes/Universiti Kebangssan Malaysia, Melbourne, Australia; 2 Florey Neuroscience Institutes/Melbourne University, Melbourne, Australia; 3 Austin Health/Mental Health Research Institute, Melbourne, Australia; 4 Austin Health/Melbourne University, Melbourne, Australia; 5 Florey Neuroscience Institutes/Monash University, Melbourne, Australia; 6 Gothernburg University, Gothenburg, Australia; 7 Austin Health, Heidelberg, Melbourne, Australia. Background: Vascular risk factors and stroke are associated with an in- creased risk of dementia. We hypothesised that an acute stroke might trigger amyloid deposition. Deposition of amyloid was measured using 11PiB- PET. In secondary analysis, we examined the association between the presence of carotid disease and retention of 11CPiB. Methods: Eligible patients admitted with stroke underwent 11CPiB-PET within 21 days of stroke. Image analysis was based on standard uptake value ratio (SUVR) using the cerebellar cortex as a reference region. A ratio exceeding 1.4 was considered high. Carotid ultrasonography (CUS) was performed to as- sess the presence of atheromatous plaque and stenosis. Significant stenosis was defined as occlusion > 70%. Results: 51 patients underwent PET scans. 27 patients had normal scans, 8 showed diffuse retention of 11CPiB while 16 showed localised retention in the stroke area. 3 patients in the localised group did not have CUS. The prevalence of carotid disease in the group is shown in Table 1. Comparison between 11CPiB retention and presence of carotid disease showed no significant statistical difference. Localised retention vs normal OR 0.71 (95%CI 0.14-3.71; p ¼ 0.442); dif- fuse retention vs normal OR 1.33 (0.18-16.19; p ¼ 0.568). Conclusions: Our results show no association between the presence of carotid disease and retention of 11CPiB in acute stroke patients. However the sample size is too small to suggest a definitive relationship. The association be- tween vascular disease and amyloid deposition is suitably weighted to war- rant further investigation. P2-107 LOCALIZATION OF HIPPOCAMPAL ATROPHY IN ALZHEIMER’S DISEASE Kaikai Shen 1 , Kaikai Shen 2 , Jurgen Fripp 2 , Fabrice Meriaudeau 3 , Ga€ el Chetelat 4 , Olivier Salvado 2 , Pierrick Bourgeat 2 , 1 Australian eHealth Research Centre, Herston, Australia; 2 Australian eHealth Research Centre, ICT Centre CSIRO, Herston, Queensland, Australia; 3 LE2I UMR 5158, Universit e de Bourgogne, Le Creusot, France; 4 Inserm-EPHE-Universit e de Caen/Basse-Normandie, Unit e U923, GIP Cyceron, Caen, France. Background: The hippocampus presents the highest rate of atrophy in the early stage of Alzheimer’s disease (AD), with more pronounced neuron loss reported in CA1 and subiculum. The aim of this study is to increase the dis- crimination power of hippocampal shape analysis between AD and normal controls (NC) by focusing on the subregions with atrophy associated with AD and describing the localized shape changes using statistical shape models (SSMs). Methods: The hippocampal segmentations used in this ab- stract are provided by Alzheimer’s Disease Neuroimaging Initiative (ADNI), which are divided into a training and a testing set. The training set consists of 60 AD subjects with average age of 75.2(6.7) and 60 NC sub- jects of 77.0(4.8). The testing set consists of 39 AD subjects of 77.8(7.3) and 78 NC of 76.3(5.2). The hippocampal SSM is built on the training set. Hotel- ling’s T2 test is performed on each landmark to evaluate the difference be- tween the Procrustes-aligned points of the NC and AD subpopulations. A subset of landmarks is selected at a given significance level. A Principal Component Analysis (PCA) is performed on the selected landmarks. The first 15 principal shape components explaining approximately 90% of the total variation are used as features. Hippocampal volume corrected for total intracranial volume is added as independent feature. The shape components are evaluated in the AD classification using bagged SVM with 25 resam- plings. Results: Using hippocampal volume gave on average 83.47% accu- racy (average of 20 repetitions). Using the features produced by SSMs alone provided the best performance (87.36%) when computed on the landmarks selected at p. Methods: Data were obtained from Alzheimer’s disease neu- roimaging initiative. Subjects (AD ¼ 26, MCI ¼ 53, HC ¼ 46; aged 55-91 years old) who had a baseline and a 24-month follow-up MRI scan at 3 Tesla were included. AD diagnosis/conversion was made by site physicians in Poster Presentations P2 S343