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: | Seyed Moeen Tayebi |
Institution: | Amirkabir University of Technology |
Department: | Electrical Engineering |
Country: | |
Proposed Analysis: | Originally as my B.Sc. dissertation project, and under the supervision of Dr. Mohammad Bagher Menhaj, I aim to use the ADNI dataset to train a generative model for obtaining Positron Emission Tomography (PET) images with adequate accuracy from Magnetic Resonance Imaging (MRI) images (mainly T1w images, to be specific) to be used downstream in Alzheimer's diagnosis (classification). Firstly, I am planning on reproducing the work described in the paper "Three-dimensional self-attention conditional GAN with spectral normalization for multimodal neuroimaging synthesis" by Lan et al. (2021) to get familiar with the data and the preprocessing steps necessary for utilizing it in a deep learning model, as well as implementing their proposed Generative Adversarial Network as a baseline for this task. Following that, I have two paths in mind for extending that work; 1. Integrating the network proposed in the aforementioned paper with the idea of including the downstream task of AD diagnosis in the training of the generative model (e.g. by introducing a new punitive term in the loss function based on the result of the AD classifier network). 2. Implementing a Generative Teaching Network for the same task and to let the model choose what secondary modality it deems suitable to be able to achieve a good result in the AD diagnosis task. The importance of this work lies in the disadvantages of PET imaging, in contrast to the quicker, safer, and more widely available MR imaging. |
Additional Investigators |