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: | Zhiyuan Song |
| Institution: | University of Cambridge |
| Department: | Department of Psychiatry |
| Country: | |
| Proposed Analysis: | The pathogenesis of AD presents and progresses from a healthy level of cognitive and social functioning to amnestic symptoms developing through a pre-clinical phase before a diagnosis is made. Early identification could be critical because recent clinical trials of disease-modifying agents suggest that therapeutics work best if started at earliest stages. In the past decade, the advances in machine learning (ML) have opened new avenues for the detection of the spectrum of AD before clinical diagnosis. The performance of ML models depends heavily on the representation quality of the original training samples. The data-centric approach of representation learning requires voluminous and diversified training data that can cover all possible appearances of a target semantic (e.g., a disease, an object, and a document). Unfortunately, the availability of such data is often limited, especially when it comes to medical data, which can be sensitive and difficult to obtain. This can make it challenging to build accurate and effective machine learning models for Alzheimer's disease. To solve this problem, this project will explore a variety of advanced training data generation and aggregation methods with the ADNI dataset, including data augmentation, transfer learning, and knowledge-guided generative adversarial networks (GANs) to enhance training data in terms of both volume and diversity. The lack of robustness and interpretability in AI (artificial intelligence) models has limited their performance in clinical trials. Therefore, more advanced ML architectures, such as graph neural networks (GNNs), could be utilized as a powerful class of model for representation learning on relational data and non-Euclidean structured signal. Meanwhile, since each biomarker modality could offer valuable and complementary information for AD, this project will exploit the combination of multi-modality data to identify AD in preclinical stage using the ADNI dataset. Several studies have pointed out that a lack of deep interdisciplinary collaboration between computer scientists and clinicians is one of the greatest challenges to overcome. Therefore, this project will investigate and design explainable scientific ML models which incorporate clinical knowledge, and eventually develop a precision medicine approach which integrates a variety of features from different biomarkers for early diagnosis and prognosis of AD. |
| Additional Investigators | |
| Investigator's Name: | Shahid Zaman |
| Proposed Analysis: | The pathogenesis of AD presents and progresses from a healthy level of cognitive and social functioning to amnestic symptoms developing through a pre-clinical phase before a diagnosis is made. Early identification could be critical because recent clinical trials of disease-modifying agents suggest that therapeutics work best if started at earliest stages. In the past decade, the advances in machine learning (ML) have opened new avenues for the detection of the spectrum of AD before clinical diagnosis. The performance of ML models depends heavily on the representation quality of the original training samples. The data-centric approach of representation learning requires voluminous and diversified training data that can cover all possible appearances of a target semantic (e.g., a disease, an object, and a document). Unfortunately, the availability of such data is often limited, especially when it comes to medical data, which can be sensitive and difficult to obtain. This can make it challenging to build accurate and effective machine learning models for Alzheimer's disease. To solve this problem, this project will explore a variety of advanced training data generation and aggregation methods with the ADNI dataset, including data augmentation, transfer learning, and knowledge-guided GANs to enhance training data in terms of both volume and diversity. The lack of robustness and interpretability in AI (artificial intelligence) models has limited their performance in clinical trials. Therefore, more advanced ML architectures, such as graph neural networks (GNNs), could be utilized as a powerful class of model for representation learning on relational data and non-Euclidean structured signal. Meanwhile, since each biomarker modality could offer valuable and complementary information for AD, this project will exploit the combination of multi-modality data to identify AD in preclinical stage using the ADNI dataset. Several studies have pointed out that a lack of deep interdisciplinary collaboration between computer scientists and clinicians is one of the greatest challenges to overcome. Therefore, this project will investigate and design explainable scientific ML models which incorporate clinical knowledge, and eventually develop a precision medicine approach which integrates a variety of features from different biomarkers for early diagnosis and prognosis of AD. |
