PI: Singh, Vikas (contact); Johnson, Sterling C
Email: VSINGH@BIOSTAT.WISC.EDU
Institution: University of Wisconsin-Madison
Title: Manifold-valued statistical models for longitudinal morphometric analysis in preclinical Alzheimers disease (AD)
Grant #: EB022883
Status: Completed
Deliverables:
Dual stream flow based generative models for manifold-valued data (ICCV 2019)
tl; dr: Modality transfer for manifold-valued images
ODE for Panel data (UAI 2021)
tl; dr: Analyzing panel/longitudinal imaging data using deep mixed effects ordinary differential equations
GANs for FreeSurfer cortical surface data (PAMI 2021)
tl; dr: When is it sensible to use synthetic brain imaging data for statistical analysis? Can the results be trusted?
Differential equations based modeling of longitudinal amyloid scans (ISBI 2020)
tl; dr: Neural differential equations can be used to model and make prognosis for longitudinal amyloid images
Sampling Free Uncertainty Estimation for Longitudinal Data (UAI 2019)
tl; dr: If GRUs are used to model longitudinal data, sampling is not required for uncertainty estimates (under mild assumptions)
Random Effects CNN (IPMI 2019)
tl; dr: Random effects to model dependent/longitudinal data is possible via moderate adjustments in CNN models
Dilated CNNs for Manifold valued data (ICCV 2019)
tl; dr: Sequential manifold measurements (such as morphometric change images or diffusion MR) can be modeled using dilated CNN
Conditional Recurrent Flow (ICCV 2019)
tl; dr: Longitudinal amyloid and morphometric change data can be modeled using invertible flow models
A Statistical Recurrent Unit (SRU) for SPD Matrices (NIPS 2018)
tl; dr: Provides a theoretical formulation to deal with sequential manifold-valued measurements
Riemannian Mixed Effects Model (CVPR 2017)
tl; dr: A formulation for efficient parameter estimation of mixed effects model on manifold-valued measurements
NIBIB Math Project - Singh.pptx [deprecated]