INVESTIGADOR(ES) PRINCIPAL(ES):

Hernán Darío Vargas Cardona

0 horas

CODIGO CIE

E6-16-5

Alvaro Angel Orozco Gutiérrez

Tutor

0 Horas

TIPO DE CONVOCATORIA

2015. Estudiantes De Posgrado

TIPO DE PROYECTO

Investigación Aplicada

OBJETIVO(S)

General objetive: To develop a methodology based on stochastic modeling of high dimensional data to enhance spatial resolution in diffusion tensor imaging (DTI). Specific objectives: -To process dMRI data for denoising, artifacts remotion and DTI fields estimation. - To model DTI fields with multi-output Gaussian processes (MOGP) applied to features extracted from tensorial data and to model DTI fields with Generalized Wishart processes (GWP) to perform interpolation over positive definite tensors. - To optimize relevant parameters in MOGP and GWP employing process convolution and approximate Bayesian inference. - To validate proposed methods in toy and real data evaluating accuracy in spatial resolution enhancement and clinical procedures: tractography and brain conductivity.

RESUMEN

Diffusion Magnetic Resonance Imaging (dMRI) is a non-invasive tool for probing the microstructure of fibrous nerve and muscle tissue. From dMRI it is possible to estimate 2-rank diffusion tensors imaging (DTI) fields. - DTI fields are widely used in clinical applications: tissue segmentation, fiber tractography, brain atlas construction, brain conductivity models, among others. - Due to hardware limitations of MRI scanners, DTI has the difficult compromise between spatial resolution and signal noise ratio (SNR) during acquisition. For this reason, the data are often acquired with very low resolution. To enhance DTI data resolution, interpolation provides an interesting software solution. - The aim of this work is to develop a probabilistic framework for DTI interpolation that enhance the spatial resolution of DTI fields. - We developed a probabilistic methodology to interpolate Diffusion Tensor Imaging (DTI) data. We model a DTI field as a Generalized Wishart process (GWP). We employ approximate Bayesian inference for optimizing the relevant variables in GWP. Results obtained with GWP in synthetic and real DTI data outperform to commonly used geometric methods. Also, our proposed method guarantees positive definite tensors, excellent accuracy and it avoids an issue in tensorial interpolation known as swelling effect. - We presented a feature-based methodology for interpolation of diffusion tensor fields. We decompose the tensors in eigenvalues and Euler angles for performing multioutput regression with Gaussian processes (MOGP). Results obtained with MOGP on synthetic and real dMRI data outperforms to the classical log-euclidean method and the feature-based scheme with linear regression. We evaluated accuracy of interpolation (Frobenius norm and Riemann distance) and preservation of fractional anisotropy (FA). MOGP guarantees positive definite tensors, preserves FA of DTs and it avoids the swelling effect. Also, we observed that probabilistic approaches have a better performance than classical geometric methods.

ESTADO

Concluido

FECHA DE INICIO

18/01/2016

FECHA DE FINALIZACION

18/01/2018

PRODUCTOS

Analysis of the geometry and electric properties of brain tissue in simulation models for deep brain stimulation

Artículos en revista A1 ó A2

Multi-output Gaussian processes for enhancing resolution of diffusion tensor fields

Artículos publicados en Revistas B, C ó D

POSTER: Multi-output Gaussian processes for enhancing resolution of diffusion tensor fields

Ponencia en evento especializado