×
  • Select the area you would like to search.
  • ACTIVE INVESTIGATIONS Search for current projects using the investigator's name, institution, or keywords.
  • EXPERTS KNOWLEDGE BASE Enter keywords to search a list of questions and answers received and processed by the ADNI team.
  • ADNI PDFS Search any ADNI publication pdf by author, keyword, or PMID. Use an asterisk only to view all pdfs.
Principal Investigator  
Principal Investigator's Name: Jaume Banus Cobo
Institution: INRIA
Department: EPIONE
Country:
Proposed Analysis: The modelling of cardiovascular factors in brain disorders is challenging since neuroimaging datasets often do not provide a wide representation of both heart and brain features. As a result, these studies often are limited to a reduced set of features, or to very small sample size with complete information. To overcome this issue, we propose a probabilistic framework for inference of a personalized cardiovascular model from brain information that can leverage on datasets with missing cardiac information. The inference scheme is formed by two nested models: 1) an imputation model of cardiac information from the available cardiac and brain features and 2) an emulator of cardiovascular dynamics derived from the imputed cardiac information. We develop a variational framework for the joint optimization of models 1) and 2), where the emulator is learned via Gaussian process (GP) regression, and the imputation scheme is based on approximated inference of cardiac features to maximize the marginal GP likelihood. After training, our approach allows us to impute missing cardiac features distributions from available cardiac and brain imaging information, and perform data assimilation in a lumped cardiovascular model to predict the most likely associated cardiovascular parameters, without the use of the integration of the dynamical cardiovascular model. Experimental results in data from the UK Biobank show that our model can be helpful to impute missing cardiac features in datasets containing minimal cardiac information, e.g. limited to systolic and diastolic blood pressures, while jointly estimating the emulated parameters of the lumped model with a mean squared error of less than 5%. This allows to further explore the heart-brain relationship through simulation of realistic cardiac dynamics corresponding to different scenarios of brain anatomy. For example, we have studies how the evolution of the cardiac dynamics effect the total volume of white matter hyperintensities. Our goal is to extend that analysis and leverage in the knowledge learnt in UK Biobank to employ our framework in the available data in ADNI. Hence, allowing us to study possible relationships between the parameters of the cardiovascular model with B-amyloid deposition and the clinical diagnosis of AD.
Additional Investigators