Data Assimilation Using Mri Data

Abstract

Contrast enhanced dynamic MRI data is important for classifying anomalies in tissue and blood circulation, e.g. tumors, stokes, kidney failure. The resulting sequence of images providing spatial-temporal representation of contrast agent concentration must be given meaningful clinical interpretation. Current processing typically relies on localized strategies, where the images are analyzed voxel by voxel. The consensus seems to be that taking direct advantage of the global structure of the images potentially will allow for more accurate and robust interpretations.Based on a workflow that has been extensively studied within both petroleum and ground water research, the current paper combines explicit porous media flow modelling and state of the art data assimilation. Starting from a multi-compartment global flow model, we explore MRI interpretation via synthetic cases where model generated images are used as input to a parameter identification algorithm based on ensemble based data assimilation like the ensemble Kalman filter (EnKF).From reasonable assumptions, the procedure demonstrates ability to identify relevant model parameters for two synthetic cases.This study indicates that flow based assimilation of MRI data is viable and that it might provide valuable supplementing insight compared to state of the art. Regularizations are made on a firm statistical basis, eliminating the need for ad hoc parameters, and leading to estimates that are rigorous in terms of error thresholds. A priori knowledge can easily be incorporated in the procedure. The flow model allows explicit representation of basic material properties, and adds regularization to the procedure by enforcing globally consistent flow fields.

A new framework for assessing subject-specific whole brain circulation and perfusion using MRI-based measurements and a multi-scale continuous flow model

“…ensemble Kalman filters [57]. An initial investigation on estimating parameters of a forward model utilizing a multi-compartment model to estimate perfusion is presented in [58]. The approach presented there is based on the ensemble Kalman filter which is a promising candidate for parameter estimation also in our setting.…”

“…The ensemble Kalman filter is used for parameter estimation of many large scale problems within geoscience (see e.g. [58]).…”

Perfusion estimation using MRI-based measurements and a porous media flow model

“…Let the subscript i ∈ {a, v} indicate the arterial or venous compartments respectively. As detailed in [19], we consider two porous flow models each governed by the Darcy constitutive relation…”

Ensemble Kalman Filters for Reliability Estimation in Perfusion Inference

“…Nevertheless, there has been previous work on the use of Ensemble Kalman Filters in the application scenario. In [25], the authors concentrate on the introduction of a tissue model that includes space-dependent information. To achieve this, a blood flow model is combined with an EnKF.…”

“…Hence, our methodology is considered as an extension to the existing standard methodology introducing the opportunity to derive statistical information on the computed solution. In addition to the different objective compared to [25], we also perform a large number of parameter studies and convergence tests, which are crucial to understand the properties of the method.…”