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: | Yang Yang |
| Institution: | Shanghai Jiao Tong University |
| Department: | Computer Science |
| Country: | |
| Proposed Analysis: | Genetic only explains a portion of the risk for developing AD, so the accuracy achievable for prediction on genetics alone is limited. The ADNI database provides chances to analyze genetic variations (SNPs) and neuroimaging data jointly. Benefitting from recent advances of deep neural networks, various kinds of multi-task and multi-modal deep learning methods have emerged in computer vision and natural language processing fields, while the existing studies combining multi-modal imaging data from ADNI are mostly traditional machine learning methods. To apply multi-task multi-modal deep learning to ADNI data, we have the following study plan: 1.We will develop a multi-modal deep neural network, which accepts multiple type of data, including SNPs, MRI, PET, etc. The network consists of specific layers to extract features for each type of data and map them into a unified feature embedding space, and also shared layers to make the AD-related prediction. The multi-modal deep neural network is expected to reveal both modal-specific features and modal-shared features. 2.Considering the limited data scale for deep learning, we will develop self-supervised learning model to perform data augmentation and improve the prediction performance. |
| Additional Investigators | |
| Investigator's Name: | Rui Hu |
| Proposed Analysis: | We will develop a multi-modal deep neural network, which accepts multiple type of data, including SNPs, MRI, PET, etc. The network consists of specific layers to extract features for each type of data and map them into a unified feature embedding space, and also shared layers to make the AD-related prediction. The multi-modal deep neural network is expected to reveal both modal-specific features and modal-shared features. |
| Investigator's Name: | Yuzhang Xie |
| Proposed Analysis: | Considering the limited data scale of ADNI for deep learning, we will develop self-supervised learning model to perform data augmentation and improve the prediction performance. Especially, we will design contrastive learning methods to train the deep neural network. |
| Investigator's Name: | Jinyi Xiang |
| Proposed Analysis: | Considering the limited data scale of ADNI for deep learning, we will develop self-supervised learning model to perform data augmentation and improve the prediction performance. Especially, we will design contrastive learning methods to train the deep neural network. |
