There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Tensho Yamao |
| Institution: | Fukushima Medical University |
| Department: | Health Science |
| Country: | |
| Proposed Analysis: | Positron emission tomography (PET) imaging plays important role for detection of subtle pathetical hallmark in living subjects. National Institute on Aging Alzheimer’s Association reported the ATN classification which based on pathological process of amyloid beta plaques, pathologic tau and neurodegeneration (Jack CR, et al. Alzheimers Dement. 2018). Assessment based on biomarkers of multi-modality is required in more accurate characterization of subject’s etiology. Several deep learning strategies has recently been reported excellent classification accuracy and prediction of onset Alzheimer’s disease (AD) (Choi H, et al. Behav Brain Res. 2018; Reith F, et al. AJNR Am J Neuroradiol. 2020). Using multi-modal model of deep learning, we investigate classification accuracy and predict onset dementia with multiple information (i.e. amyloid PET images, tau PET images, magnetic resonance image: MRI and subject’s imformations). We use data underwent amyloid PET, tau PET and MRI from elderly controls, mild cognitive impairment (MCI) and AD. PET images are spatially normalized to Montreal Neurological Institute (MNI) space by transformation parameter calculated from MRI transformation on statistical parametric mapping (SPM) version 12. Then, voxel values of PET images are converted to Centiloid scale according to the Global Alzheimer's Association Interactive Network protocol (Klunk WE, et al. Alzheimers Dement. 2015). Images are simply augmented by rotating, flip vertical or horizontal, add noise and masking. In this study, we construct a deep learning model based on the three-stream convolutional neural network (Duohan L, et al. IEEE Xplore, 2019). This architecture is consisted of three phases. The first phase independently extracts features. The second phase is feature fusion. The third phase is a multi-task network for classification. This model is evaluated by accuracy, precision, F-measure and area under curve. |
| Additional Investigators |
