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: | Dylan Turnley |
| Institution: | University of Leeds |
| Department: | Artificial Intellegence |
| Country: | |
| Proposed Analysis: | Masters of AI Project for the University of Leeds (UK) I propose to use the ADNI Alzheimer’s related: MRI, PET and clinical text data for analysis under a new AI model. This AI analysis will use a novel method called multimodal networks. These are specialised networks that can accept multimodal (multi input types, imaging, text, video/audio) data. As such, the hope is that providing the model a higher variance of data shall provide a more accurate representation of patients reality, thus increasing the predictive capability as well as trustworthiness of that prediction. We are also able to incorporate written/video data, which could mean that cognitive testing of individuals can be considered, especially at younger ages. This is good because cognitive test are often better early indicators of disease versus the symptomatic evidence associated with later stages of the disease. If metadata from bio samples is available, I would love to use that too. If that’s something held please let me know, as I only saw physical samples being available. The papers to describe the multimodal networks can be found here: Switch-BERT: https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960325.pdf MBERT: https://www.arxiv-vanity.com/papers/1908.05787/ MMFT-BERT: https://aclanthology.org/2020.findings-emnlp.417.pdf ____________________________________________________________________________ If this analysis proves to be beneficial, it is possible that, if we are able to collect enough data globally, we can create disease models services (think OpenAI's ChatGPT, but you can interact with it like a chat bot, and upload your own standardized MRI/PET/clinical data or cognitive tests and you data will contribute to the next learning cycle, as well as provide you as accurate as possible the prediction for your own disease state). This could work as a paid or free model, but ideally free service provided people contribute their anonymized standard data. Another benefit of this is that this model is that can incorporate related papers and works in Alzheimer’s’ research and build a collective model of knowledge for Alzheimer’s. Which poses the best chance of accelerating our understanding of it. Past this, a similar method could build a new network on other disease, like cancer, etc. Eventually, it could be that a single network could accept ANY disease data, being able to predict any possible disease. This work is most similar to another paper using your MRI and PET data, using unimodal networks though: https://adni.loni.usc.edu/adni-publications/Shanmugam_2021_Alzheimer_s%20disease%20classificat.pdf Please let me know if there are any questions or if there are any possible metadata for the Biospecimens data. D Turnley |
| Additional Investigators |
