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: | Luca Giancardo |
Institution: | University of Texas Health Science Centerat Houston |
Department: | School of Biomedical Informatics |
Country: | |
Proposed Analysis: | The first steps showing the feasibility of brain imaging-based biomarkers for early detection of neurodegeneration have been shown. However, validated imaging-based biomarkers for progression remain elusive, particularly for the prodromal phase. We believe that this is due to the lack of datasets and the limitation of the classical analytical MRI imaging approaches, where seeds are placed on regions of interest defined by a priori hypotheses and analyzed with univariate statistical methods. Such methods can be time consuming and costly, due to the research time spent in manually defining these regions with multiple iterations of trial and error, coupled with the complexity of modelling the time component. Recently, data-driven processes able to discover imaging biomarker candidates from brain data are starting to achieve exciting results, particularly due to machine learning (ML) based pipelines and brain connectomes. These approaches involve training a ML model which will attempt to learn the combination of brain connections able to characterize a particular condition based on a loss function. However, these methodologies would not be able to directly identify any pattern from the temporal information, which, we believe, is a key aspect for measuring brain neuroplasticity and creating progression metrics. This is due to the inability of the ML model to capture the temporal information. Recently, we have experimented with a new data representation able to encode temporal patterns in connectomes derived from Diffusion MRI, we call them “longitudinal connectomes”. Such representation allows ML models to learn multivariate temporal patterns, which can be too complex to define a-priori, thereby allowing the measurement of disease progression. This approach also has the potential of being less sensitive to inter-scanner variability, since the information shown to the model is a relative measurement among the same subject. Using the ADNI data we will refine our models and test it on subjects with Alzheimer's Disease and Dementia. |
Additional Investigators | |
Investigator's Name: | Danilo Peña |
Proposed Analysis: | Support the PI with the longitudinal connectome analysis to larger cohorts. Tests new machine learning models that learn from image changes |
Investigator's Name: | Xiaoqian Jiang |
Proposed Analysis: | The aim is to develop new integrated phenotypes that combine imaging with clinical data. We will study how to merge computational phenotypes acquired from imaging and clinical data to boost the performance of risk analysis. The availability of multi-modality data in the biomedical field opens new horizon to utilize complementary information to find meaningful association and improve the prediction performance of complex disease. We will combine state-of-the-art supervised tensor factorization framework with imaging pipelines to investigate this challenging problem. |
Investigator's Name: | Yejin Kim |
Proposed Analysis: | The aim is to develop new integrated phenotypes that combine imaging with clinical data. We will study how to merge computational phenotypes acquired from imaging and clinical data to boost the performance of risk analysis. The availability of multi-modality data in the biomedical field opens new horizon to utilize complementary information to find meaningful association and improve the prediction performance of complex disease. We will combine state-of-the-art supervised tensor factorization framework with imaging pipelines to investigate this challenging problem. |