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: | Philipp Kopper |
| Institution: | Ludwig Maximilian University Munich |
| Department: | Institute of Statistics |
| Country: | |
| Proposed Analysis: | We are currently developing a deep learning approach for survival analysis with multimodal data. Our work will help to better understand medical (image) data. The ADNI data is particularly suitable for testing the new methodology as it combines classical tabular data with image data. Our approach builds on Bender et al. (2020) and Rügamer et al. (2020) and combines interpretable additive models with deep neural networks in a so-called semi-structured distributional regression approach similar to wide-and-deep neural networks. The set-up of Rügamer et al. (2020) guarantees that the components of the additive model (i.e. the model part explaining the tabular data) are still interpretable, even in the presence of a deep neural network. We are currently collaborating with Sebastian Pölsterl and colleagues who already analyzed your data in Pölsterl et al. (2019) with another wide-and-deep neural network. We are aiming to analyze the progression of mild cognitive impairment (MCI) to Alzheimer’s disease (AD). We are especially interested in modeling the underlying risk process in a time-to-event analysis style. We are going to use a combination of a convolutional neural network (CNN) and a piecewise additive mixed model (PAMM) to predict associated hazards for individuals. Using our model our approach can estimate the impact of the patients’ MRI images (either directly images or processed to point clouds) using the deep neural network and estimate the baseline hazards and as well as tabular covariate effects (e.g. some biomarkers or demographics) using the PAMM. Our model is interpretable by nature and can be used to identify the individuals at risk to a progression from MCI to AD. Our methodology, therefore, combines the benefits of classical survival analysis and deep learning while allowing for straightforward interpretation. We intend to make use of your data for two separate publications. The first one will focus on the new methodology and the second one on the data analysis itself. Our team consists of Philipp Kopper (M.S.), David Rügamer (Ph.D.), Andreas Bender (Ph.D.) and Prof. Bernd Bischl (Ph.D.). Both publications will be in collaboration with Sebastian Pölsterl and colleagues. Furthermore, Andreas Bender intends to offer a master thesis where the data is supposed to be used. In general, we are interested in a long-term partnership with ADNI and would want to analyze your data in multiple projects. Bender et al. (2020): https://arxiv.org/abs/2006.15442 Pölsterl et al. (2019): https://arxiv.org/abs/1909.03890 Rügamer et al. (2020): https://arxiv.org/abs/2002.05777 |
| Additional Investigators | |
| Investigator's Name: | David Rügamer |
| Proposed Analysis: | Already specified in the proposed analysis of Philipp Kopper |
| Investigator's Name: | Andreas Bender |
| Proposed Analysis: | Already specified in the proposed analysis of Philipp Kopper Additionally possibly a master thesis for a M.S. student. |
