Abolfazl Mehranian^{1}, Andrew J Reader^{1}, and Enrico De Vita^{1}

A MOtion-Corrected and High-resolution Anatomically-assisted (MOCHA) reconstruction framework is proposed for ASL MRI. The method simultaneously accounts and corrects for rigid motion and partial volume effects (PVE), and reduces noise by guided high-resolution anatomical MR images without any need for segmentation. The proposed method was compared with standard methods and a 3D linear regression (3DLR) correction method using realistic simulations and in-vivo data. Results show that MOCHA outperforms 3DLR not only in preservation of structural and local details, including simulated lesions, but also in PVE correction of deep grey matter structures, often subject to segmentation errors in conventional methods.

In the proposed MOtion-Corrected and High-resolution Anatomically-assisted (MOCHA) ASL reconstruction framework, all $$$\small{N}$$$ low-resolution control-label ASL pairs ($$$\small\boldsymbol{C},\boldsymbol{L}$$$) are simultaneously used to reconstruct a single perfusion-weighted image ($$$\small\boldsymbol{x}$$$) using the following penalized weighted least squares minimization: $$$\small\widehat{\boldsymbol{x}}\mathrm{=}{\mathop{\mathrm{argmin}}_{\boldsymbol{x}} \left\{\frac{\mathrm{1}}{\mathrm{2}N}\sum^{\ }_i{{\left\|\boldsymbol{E}{\boldsymbol{M}}_i\boldsymbol{Bx}\mathrm{-}{\left(\boldsymbol{C}\boldsymbol{-}\boldsymbol{L}\right)}_i\right\|}^{\mathrm{2}}_{\boldsymbol{\ }}}\mathrm{+}\beta \sum^{\ }_j{{\left\|{\left(\boldsymbol{Dx}\right)}_j\right\|}^{\mathrm{2}}_{\boldsymbol{\omega }}}\right\}\ }$$$.This method takes all data acquisition processes into account including point-spread-function ($$$\small\boldsymbol{B}$$$), motion ($$$\small\boldsymbol{M}$$$) and MR Fourier encoding matrix ($$$\small\boldsymbol{E}$$$, composed of Fourier bases, coil sensitivity profiles and k-space undersampling masks). Additionally, a quadratic smoothness prior, weighted ($$$\boldsymbol{\omega }$$$) by an anatomical image, is utilized to guide reconstruction of the target high-resolution perfusion image. $$$\small\boldsymbol{D}$$$ is a high-order finite differences, and $$$\beta$$$ is a regularization parameter. MOCHA therefore removes the need for segmentation and downsampling of the anatomical images into ASL space. Using realistic simulations and in-vivo datasets MOCHA was compared with standard reconstruction and 3DLR PVE correction.

Our results suggest that the proposed MOCHA ASL reconstruction method improves spatial resolution and anatomical fidelity of CBF maps taking into account motion and PVE. MOCHA has potential to improve the diagnostic confidence and quantification of current ASL protocols in clinical practice and further validation with clinical dataset is planned.

Fig 1. Results for reconstruction of simulated data for a motion
corrected CBF map obtained from the standard method, corrected for partial
volume effect of GM and WM using the 3DLR method and reconstructed using the
MOCHA method.
Note 3DLR produces separate GM and WM CBF maps.

Fig 2. Normalized root mean squared error
(NRMSE) of the studied methods in different regions of the simulated brain
phantom with (left) and without (right) motion correction.

Fig 3. CBF results for an exemplary in-vivo
dataset, calculated using the standard, 3DLR and proposed reconstruction
methods. Arrows indicate areas of interest demonstrating the resolution
enhancements achieved by MOCHA reconstruction.

Fig 4. The CBF results of the standard,
3DLR and MOCHA methods for different regions averaged over all in-vivo datasets
(n=5). The error bars show the standard deviation of the mean CBF values over
the subjects in each region. NB: Grey matter here represents ‘cortical’ grey
matter only.

Fig 5. Comparison of CBF maps from higher-resolution (4×4×2 mm3) and standard-resolution
(4×4×4 mm3) acquisitions.