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: | Nguyen Tuan |
| Institution: | VinUniversity |
| Department: | College of Engineering & Computer Science |
| Country: | |
| Proposed Analysis: | The primary objective of this research is to develop a federated learning framework that combines neuroimaging, biomarker, and lifestyle data from the AIBL, ADNI, and ADNIDOD datasets for accurate Alzheimer's disease prediction. Specifically, we aim to: 1. Design a federated learning architecture: Develop a secure and privacy-preserving federated learning framework that allows collaborative model training using distributed data sources. 2. Feature extraction and selection: Employ advanced artificial intelligence techniques, such as deep learning and feature engineering, to extract relevant features from neuroimaging scans, biomarker measurements, and lifestyle factors. 3. Model development and optimization: Train machine learning models, such as deep neural networks, on the distributed datasets using the federated learning framework. Optimize the models to achieve high predictive accuracy while respecting data privacy constraints. 4. Alzheimer's disease prediction: Utilize the trained models to predict the progression of Alzheimer's disease in individuals, considering demographic factors, genetic profiles, neuroimaging markers, and lifestyle indicators. 5. Evaluation and validation: Evaluate the performance of the proposed federated learning approach in predicting Alzheimer's disease progression using appropriate metrics and cross-validation techniques. Validate the findings against existing clinical measures and benchmarks. |
| Additional Investigators | |
| Investigator's Name: | Nguyen Son |
| Proposed Analysis: | The primary objective of this research is to investigate and develop personalized federated learning methods that cater to the specific needs of participants while maintaining high model performance and ensuring privacy preservation. Specifically, we aim to: 1. Personalized model initialization: Develop techniques to personalize the initial model parameters for each participant in the federated learning setup. This initialization process will take into account the participant's local data distribution and characteristics, enabling faster convergence and improved performance. 2. Adaptive model aggregation: Explore adaptive aggregation strategies that adaptively weigh the contribution of each participant during the model aggregation phase. This approach will give higher importance to participants with richer and more representative data, while still considering the privacy constraints of individual participants. 3. Privacy-preserving techniques: Investigate privacy-preserving mechanisms, such as secure multi-party computation and differential privacy, to safeguard sensitive participant data during the federated learning process. The aim is to strike a balance between model performance and privacy, ensuring that participants' data remains confidential and protected. 4. Personalized model updating: Develop mechanisms to personalize the model update process for each participant based on their specific needs. This can include prioritizing certain data samples or adjusting learning rates to better adapt to local data characteristics, thereby improving individual model performance. 5. Evaluation and validation: Evaluate the performance of the proposed personalized federated learning framework using benchmark datasets and real-world scenarios. Assess the impact of personalization on model convergence, prediction accuracy, and privacy preservation. Compare the personalized federated learning approach with traditional federated learning methods to showcase its advantages. |
| Investigator's Name: | Nguyen Dung |
| Proposed Analysis: | The primary objective of this research is to investigate and develop personalized federated learning methods that cater to the specific needs of participants while maintaining high model performance and ensuring privacy preservation. Specifically, we aim to: 1. Personalized model initialization: Develop techniques to personalize the initial model parameters for each participant in the federated learning setup. This initialization process will take into account the participant's local data distribution and characteristics, enabling faster convergence and improved performance. 2. Adaptive model aggregation: Explore adaptive aggregation strategies that adaptively weigh the contribution of each participant during the model aggregation phase. This approach will give higher importance to participants with richer and more representative data, while still considering the privacy constraints of individual participants. 3. Privacy-preserving techniques: Investigate privacy-preserving mechanisms, such as secure multi-party computation and differential privacy, to safeguard sensitive participant data during the federated learning process. The aim is to strike a balance between model performance and privacy, ensuring that participants' data remains confidential and protected. 4. Personalized model updating: Develop mechanisms to personalize the model update process for each participant based on their specific needs. This can include prioritizing certain data samples or adjusting learning rates to better adapt to local data characteristics, thereby improving individual model performance. 5. Evaluation and validation: Evaluate the performance of the proposed personalized federated learning framework using benchmark datasets and real-world scenarios. Assess the impact of personalization on model convergence, prediction accuracy, and privacy preservation. Compare the personalized federated learning approach with traditional federated learning methods to showcase its advantages. |
