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: | Junbeom Kwon |
| Institution: | Seoul National University |
| Department: | Psychology |
| Country: | |
| Proposed Analysis: | Though ADNI datasets was controlled to be equal among each group, clinical datasets usually confront the issue of imbalance among the control group and treatment group, which makes it difficult for the traditional models to have desired performances. To tackle this problem, I propose to predict Alzheimer's disease with unsupervised anomaly detection using 3-dimensional MRI. Anomaly detection is defined as the identification of test data deviating from the normal data distribution learned during training. Deep generative models such as GANs are usually used to map the distribution of the normal training datasets. After training our model with normal brain images, we can find the brain images with Alzheimer's diseases using anomaly scores, which represents the degree of deviance from the normal brain images. Recent studies have tried to apply anomaly detection to enhance the classification of the Alzheimer's disease using PET images. However, the effect of anomaly detection for the structural MRI of Alzheimer’s disease was not yet verified. It is expected that MRI containing anatomical information would provide great insight that was not included in PET images. Furthermore, whereas previous experiments with structural MRI took 2D slice-based approaches, I plan to fully utilize volumetric information in complex brain anatomy by inputting the 3D volumes to the model. |
| Additional Investigators |
