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: | Yongjun Lee |
| Institution: | University of California, Irvine |
| Department: | Statistics |
| Country: | |
| Proposed Analysis: | For the prediction of Alzheimer's disease, deep learning (DL) models that fuse electronic health records (EHR) with imaging data such as magnetic resonance images (MRI) or positron emission tomography (PET) have proven to be more effective and yield higher performance than single modality prediction models due to the complementary characteristics of the datasets. Early fusion has been the predominant method of choice to accommodate the interdependent nature of the datasets. However, early fusion requires each EHR to be matched with its corresponding MRIs pertaining to the same patient. Thus, it limits the amount of data that can be utilized for the training due to the requirements that both data need to be present. which is not ideal for DL models. To overcome this issue, we propose a late fusion model that combines a graph neural network (GNN) model for EHR and a convolutional neural network (CNN) for extracting features from the MRI dataset. We expect GNN to capture the key features of EHR while identifying possible subgroups of the patients. Further, CNN models have been shown to provide a more effective representation of the visual features of MRI compared to manual feature extraction processes. Overall, we expect our proposed model to outperform models trained with the early fusion method while allowing for the model to utilize all the data that are available. |
| Additional Investigators |
