Diffusion-tensor-based method for robust and practical estimation of axial and radial diffusional kurtosis

Quantitative measurement of diffusion-weighted imaging signal using expression-controlled aquaporin-4 cells: Comparative study of 2-compartment and diffusion kurtosis imaging models

The purpose of this study was to compare parameter estimates for the 2-compartment and diffusion kurtosis imaging models obtained from diffusion-weighted imaging (DWI) of aquaporin-4 (AQP4) expression-controlled cells, and to look for biomarkers that indicate differences in the cell membrane water permeability. DWI was performed on AQP4-expressing and non-expressing cells and the signal was analyzed with the 2-compartment and diffusion kurtosis imaging models. For the 2-compartment model, the diffusion coefficients (Df, Ds) and volume fractions (Ff, Fs, Ff = 1-Fs) of the fast and slow compartments were estimated. For the diffusion kurtosis imaging model, estimates of the diffusion kurtosis (K) and corrected diffusion coefficient (D) were obtained. For the 2-compartment model, Ds and Fs showed clear differences between AQP4-expressing and non-expressing cells. Fs was also sensitive to cell density. There was no clear relationship with the cell type for the diffusion kurtosis imaging model parameters. Changes to cell membrane water permeability due to AQP4 expression affected DWI of cell suspensions. For the 2-compartment and diffusion kurtosis imaging models, Ds was the parameter most sensitive to differences in AQP4 expression.

The Utility of a Convolutional Neural Network for Generating a Myelin Volume Index Map from Rapid Simultaneous Relaxometry Imaging

Purpose:

A current algorithm to obtain a synthetic myelin volume fraction map (SyMVF) from rapid simultaneous relaxometry imaging (RSRI) has a potential problem, that it does not incorporate information from surrounding pixels. The purpose of this study was to develop a method that utilizes a convolutional neural network (CNN) to overcome this problem.

Methods:

RSRI and magnetization transfer images from 20 healthy volunteers were included. A CNN was trained to reconstruct RSRI-related metric maps into a myelin volume-related index (generated myelin volume index: GenMVI) map using the MVI map calculated from magnetization transfer images (MTMVI) as reference. The SyMVF and GenMVI maps were statistically compared by testing how well they correlated with the MTMVI map. The correlations were evaluated based on: (i) averaged values obtained from 164 atlas-based ROIs, and (ii) pixel-based comparison for ROIs defined in four different tissue types (cortical and subcortical gray matter, white matter, and whole brain).

Results:

For atlas-based ROIs, the overall correlation with the MTMVI map was higher for the GenMVI map than for the SyMVF map. In the pixel-based comparison, correlation with the MTMVI map was stronger for the GenMVI map than for the SyMVF map, and the difference in the distribution for the volunteers was significant (Wilcoxon sign-rank test, P < 0.001) in all tissue types.

Conclusion:

The proposed method is useful, as it can incorporate more specific information about local tissue properties than the existing method. However, clinical validation is necessary.

The Utility of Applying Various Image Preprocessing Strategies to Reduce the Ambiguity in Deep Learning-based Clinical Image Diagnosis

PURPOSE

A general problem of machine-learning algorithms based on the convolutional neural network (CNN) technique is that the reason for the output judgement is unclear. The purpose of this study was to introduce a strategy that may facilitate better understanding of how and why a specific judgement was made by the algorithm. The strategy is to preprocess the input image data in different ways to highlight the most important aspects of the images for reaching the output judgement.

MATERIALS AND METHODS

T2-weighted brain image series falling into two age-ranges were used. Classifying each series into one of the two age-ranges was the given task for the CNN model. The images from each series were preprocessed in five different ways to generate five different image sets: (1) subimages from the inner area of the brain, (2) subimages from the periphery of the brain, (3-5) subimages of brain parenchyma, gray matter area, and white matter area, respectively, extracted from the subimages of (2). The CNN model was trained and tested in five different ways using one of these image sets. The network architecture and all the parameters for training and testing remained unchanged.

RESULTS

The judgement accuracy achieved by training was different when the image set used for training was different. Some of the differences was statistically significant. The judgement accuracy decreased significantly when either extra-parenchymal or gray matter area was removed from the periphery of the brain (P < 0.05).

CONCLUSION

The proposed strategy may help visualize what features of the images were important for the algorithm to reach correct judgement, helping humans to understand how and why a particular judgement was made by a CNN.

Whole brain analyses of age-related microstructural changes quantified using different diffusional magnetic resonance imaging methods

PURPOSE

The new diffusional magnetic resonance imaging (dMRI) techniques, diffusional kurtosis imaging (DKI) and neurite orientation dispersion and density imaging (NODDI) have been developed to clarify the microstructural changes. To our knowledge, however, there is little information on the similarities and differences of these metrics evaluated by the image-by-image paired t test.

MATERIALS AND METHODS

Twenty-three healthy subjects underwent dMRI. We estimated the relationships of these metrics evaluated by the image-by-image paired t-test and compared aging effects on each metric.

RESULTS

We found that fractional anisotropy (FA), mean kurtosis (MK) derived from DKI and neurite density index (NDI) values derived from NODDI correlated with each other positively, and mean diffusivity (MD) and orientation dispersion index (ODI) values from NODDI correlated negatively with the FA value. There were no significant relationships of age with FA or MD values, while MK, ODI and NDI values showed significant correlations with age.

CONCLUSION

These results may indicate not only the similar tendency among the metrics, but also the higher sensitivity of NODDI and DKI to the changes in microstructural tissue organization with advancing age. These techniques could shed light on both normal and degenerated brain changes.

Fast imaging of mean, axial and radial diffusion kurtosis

Diffusion kurtosis imaging (DKI) is being increasingly reported to provide sensitive biomarkers of subtle changes in tissue microstructure. However, DKI also imposes larger data requirements than diffusion tensor imaging (DTI), hence, the widespread adaptation and exploration of DKI would benefit from more efficient acquisition and computational methods. To meet this demand, we recently developed a method capable of estimating mean kurtosis with only 13 diffusion weighted images. This approach was later shown to provide very accurate mean kurtosis estimates and to be more efficient in terms of contrast to noise per unit time. However, insofar, the computation of two other critical DKI parameters, radial and axial kurtosis, has required the estimation of all 22 variables parameterizing the full DKI signal expression. Here, we present two strategies for estimating all of DKI's principal parameters - mean kurtosis, radial kurtosis, and axial kurtosis - using only 19 diffusion weighted images, compared to the current state-of-the-art acquisitions typically requiring about 60 images. The first approach is based on axially symmetric diffusion and kurtosis tensors, presented here for the first time, and referred to as axially symmetric DKI. The second approach is applicable in tissues with a priori known principal diffusion direction, and does not require fitting of any kind. The approaches are evaluated in human brain in vivo as well as in fixed rat spinal cord, and are demonstrated to provide metrics in good agreement with their full DKI counterparts estimated with nonlinear least squares. For small data sets and in white matter, axially symmetric DKI provides more accurate and robust estimates than unconstrained DKI. The significant acceleration achieved further paves the way to routine application of the technique.