Abstract:Perfusion magnetic resonance imaging (pMRI) is becoming increasingly important for evaluation of brain tumor grade and radiation therapy response. A frequently encountered problem in the applications of pMRI is that the tumor volume, even a single image voxel, contains more than one tissue type, i.e. a tissue mixture, where each tissue type may have different perfusion property each other. This is mainly caused by the fact that the pMRI data are usually acquired with a resolution much lower than that of routine structural T1/T2 images. The purpose of this study is to investigate an innovative approach of estimating the tissue mixture perfusions inside each image voxel from the measured perfusion signals at that voxel. A more accurate estimation is expected to be more sensitive to measure abnormal blood flows for disease diagnoses and radiation therapy assessment. A 4D expectation-maximization algorithm was developed for the estimation of tissue mixture perfusions from the human brain arterial spin labeled (ASL) perfusion scans. Both T1/T2-weighted structural and ASL functional MR images were available for each studied subject. Firstly, the T1/T2 images were used to initiate the tissue mixtures in each voxel via multi-spectral image segmentation. Then the initial mixtures were down-sampled and mapped onto the low-resolution ASL image array via co-registration. Finally, the tissue mixtures on ASL image array were fed into the proposed 4D algorithm for the estimation of the blood perfusion mean, variance, and mixture proportion for both gray matter (GM) and white matter (WM) in each brain voxel. Lilliefors test was used to verify the statistical normality assumption of ASL data on each brain voxel along the time dimension. The performance of the 4D algorithm can be evaluated by comparing the estimated perfusion image with the conventional ASL image processing toolbox. Statistical normality test on the ASL data showed the rejection rate of brain voxels is below 7.7%, which indicated the appropriateness of our statistical model. Furthermore, we conducted experiments to reveal the feasibility of the 4D algorithm for estimating the tissue mixtures’ contributions (e.g., from GM or WM) to the total perfusion in each voxel. The estimated perfusion-mean images are shown to be consistent (error rate < 1%) with the total perfusion from the ASL toolbox. However, different from the ASL toolbox, the 4D algorithm can further generate perfusion-variance and perfusion-percentage images, which provide more details on the variation and mixture of the perfusion signals. We have developed a novel image analysis method for tissue mixture perfusion estimation in ASL pMRI. The estimated mixture perfusion images can be regarded as activation maps to show abnormal blood flow activities for the diagnosis and therapeutic assessment of age-related brain diseases. The presented 4D algorithm should be readily applicable to improve the functional pMRI of other diseases.

Exploiting Magnetic Resonance Angiography Imaging Improves Model Estimation Of Bold Signal

Exploiting Mr Venography Segmentation for the Accurate Model Estimation of Bold Signal

Nonlinear estimation of BOLD signals with the aid of cerebral blood volume imaging

Quantitative Evaluation of Activation State in Functional Brain Imaging

Reliable Off-Resonance Correction in High-Field Cardiac MRI Using Autonomous Cardiac B0 Segmentation with Dual-Modality Deep Neural Networks.

BOLD Dynamic Model of Functional MRI

An Expectation-Maximization Approach For Partial Volume Estimation Of Arterial Spin Labeled Mri Data: A Feasibility Study

Influence of Resting Venous Blood Volume Fraction on Dynamic Causal Modeling and System Identifiability

Nonlinear Analysis of the BOLD Signal

Effects of Reconstructed Magnetic Field from Sparse Noisy Boundary Measurements on Localization of Active Neural Source

Evaluation of MRI Models in the Measurement of CMRO2 and Its Relationship with CBF.

Comparison of compartmental analytical Blood-Oxygen-Level-Dependent functional Magnetic Resonance Imaging models against Monte Carlo simulations performed over cortical micro-angiograms

Predicting the macrovascular contribution to resting-state fMRI functional connectivity at 3 Tesla: A model-informed approach

Axon initial segment plasticity caused by auditory deprivation degrades time difference sensitivity in a model of neural responses to cochlear implants

Expectation-Maximization Algorithm-Based Tissue Mixture Quantification For Perfusion Mri

An Extended Convolution Dynamic Model of Fmri BOLD Response

Blood oxygenation level-dependent (BOLD)-based techniques for the quantification of brain hemodynamic and metabolic properties - theoretical models and experimental approaches

A fully synthetic three-dimensional human cerebrovascular model based on histological characteristics to investigate the hemodynamic fingerprint of the layer BOLD fMRI signal formation

Estimation of Tissue Mixture Perfusions Using Arterial Spin Labeled Magnetic Resonance Imaging

Multiparametric estimation of brain hemodynamics with MR fingerprinting ASL

Statistical Intensity- and Shape-Modeling to Automate Cerebrovascular Segmentation from TOF-MRA Data