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: | Yujuan Feng |
| Institution: | Beijing University of Technology |
| Department: | School of Software Engineering |
| Country: | |
| Proposed Analysis: | Alzheimer's disease (AD) is an incurable neurodegenerative disease. Early diagnosis of AD is crucial to defer the progression of the disease and improve the living qualities of potential patients. The current state-of-the-art methods for early diagnosis of AD based on unimodal data are unsatisfactory. Confronted with the need for effective early diagnosis of AD based on multimodal data, this project aims to address some basic issues in aspects of strong heterogeneity, complex relation, incomplete modalities, inadequate numbers of samples and high-dimensional feature space of multimodal data by making use of advanced deep multimodal learning, transfer learning and self-supervised learning theories and methods. In this project, we will propose a deep memory network-based framework to capture complex higher-order relations and learn completeness-enhanced multimodal representations. Furthermore, we will propose a pre-trained incomplete multimodal fusion model to adaptively impute missing modalities by utilizing available data as much as possible. Finally, we will apply self-supervised learning to our multimodal fusion models. By using contrastive learning and data transformation methods, we aim to learn more generalized and effective multimodal representations with limited samples, thus improving the performance of early diagnosis of AD. We believe that this project is potential to be applied in many scenarios, e.g., clinical diagnoses of AD and other complex diseases. |
| Additional Investigators | |
| Investigator's Name: | Yiming Xu |
| Proposed Analysis: | Alzheimer's disease (AD) is an incurable neurodegenerative disease. Early diagnosis of AD is crucial to defer the progression of the disease and improve the living qualities of potential patients. The current state-of-the-art methods for early diagnosis of AD based on unimodal data are unsatisfactory. Confronted with the need for effective early diagnosis of AD based on multimodal data, this project aims to address some basic issues in aspects of strong heterogeneity, complex relation, incomplete modalities, inadequate numbers of samples and high-dimensional feature space of multimodal data by making use of advanced deep multimodal learning, transfer learning and self-supervised learning theories and methods. In this project, we will propose a deep memory network-based framework to capture complex higher-order relations and learn completeness-enhanced multimodal representations. Furthermore, we will propose a pre-trained incomplete multimodal fusion model to adaptively impute missing modalities by utilizing available data as much as possible. Finally, we will apply self-supervised learning to our multimodal fusion models. By using contrastive learning and data transformation methods, we aim to learn more generalized and effective multimodal representations with limited samples, thus improving the performance of early diagnosis of AD. We believe that this project is potential to be applied in many scenarios, e.g., clinical diagnoses of AD and other complex diseases. |
