Bong-Soo Han

Regularization of DT-MR images using a successive Fermat median filtering method

Tractography using diffusion tensor magnetic resonance imaging (DT-MRI) is a method to determine the architecture of axonal fibers in the central nervous system by computing the direction of greatest diffusion in the white matter of the brain. To reduce the noise in DT-MRI measurements, a tensor-valued median filter, which is reported to be denoising and structure preserving in the tractography, is applied. In this paper, we proposed the successive Fermat (SF) method, successively using Fermat point theory for a triangle contained in the two-dimensional plane, as a median filtering method. We discussed the error analysis and numerical study about the SF method for phantom and experimental data. By considering the computing time and the image quality aspects of the numerical study simultaneously, we showed that the SF method is much more efficient than the simple median (SM) and gradient descents (GD) methods.

Symmetrical Location Characteristics of Corticospinal Tract Associated With Hand Movement in the Human Brain: A Probabilistic Diffusion Tensor Tractography

AbstractThe purpose of this study is to elucidate the symmetrical characteristics of corticospinal tract (CST) related with hand movement in bilateral hemispheres using probabilistic fiber tracking method.Seventeen subjects were participated in this study. Fiber tracking was performed with 2 regions of interest, hand activated functional magnetic resonance imaging (fMRI) results and pontomedullary junction in each cerebral hemisphere. Each subjects extracted fiber tract was normalized with a brain template. To measure the symmetrical distributions of the CST related with hand movement, the laterality and anteriority indices were defined in upper corona radiata (CR), lower CR, and posterior limb of internal capsule.The measured laterality and anteriority indices between the hemispheres in each different brain location showed no significant differences with P < 0.05. There were significant differences in the measured indices among 3 different brain locations in each cerebral hemisphere with P < 0.001. Our results clearly showed that the hand CST had symmetric structures in bilateral hemispheres.The probabilistic fiber tracking with fMRI approach demonstrated that the hand CST can be successfully extracted regardless of crossing fiber problem. Our analytical approaches and results seem to be helpful for providing the database of CST somatotopy to neurologists and clinical researches.

Diffusion-Tensor Magnetic Resonance Imaging for Hand and Foot Fibers Location at the Corona Radiata: Comparison with Two Lesion Studies

The corticospinal tract is the motor pathway in the human brain, and corona radiata (CR) is an important location to diagnose stroke. We detected hand and foot motor fiber tracts in the CR to investigate accurate locations using diffusion-tensor imaging (DTI) and functional imaging. Ten right-handed normal volunteers participated in this study. We used a probabilistic tracking algorithm, a brain normalization method, and functional imaging results to set out region of interests. Moreover, our results were compared to previous results of lesion studies to confirm their accuracy and usefulness. The location measurements were performed in two index types; anteriority index on the basis of the anterior and posterior location of lateral ventricle and laterality index on the basis of the left and right location. The anteriority indices were 56.40/43.2 (hand/foot) at the upper CR and lower CR 40.72/30.90 at the lower CR. The measurements of anteriority and laterality of motor fibers were represented as anteriority index 0.40/0.31 and laterality index 0.60/0.47 (hand/foot). Our results showed that the hand and foot fibers were in good agreements with previous lesion studies. This study and approaches can be used as a standard for DTI combined with lesion location studies in patients who need rehabilitation or follow-up.

Possibility Study of Scale Invariant Feature Transform (SIFT) Algorithm Application to Spine Magnetic Resonance Imaging.

The purpose of this study is an application of scale invariant feature transform (SIFT) algorithm to stitch the cervical-thoracic-lumbar (C-T-L) spine magnetic resonance (MR) images to provide a view of the entire spine in a single image. All MR images were acquired with fast spin echo (FSE) pulse sequence using two MR scanners (1.5 T and 3.0 T). The stitching procedures for each part of spine MR image were performed and implemented on a graphic user interface (GUI) configuration. Moreover, the stitching process is performed in two categories; manual point-to-point (mPTP) selection that performed by user specified corresponding matching points, and automated point-to-point (aPTP) selection that performed by SIFT algorithm. The stitched images using SIFT algorithm showed fine registered results and quantitatively acquired values also indicated little errors compared with commercially mounted stitching algorithm in MRI systems. Our study presented a preliminary validation of the SIFT algorithm application to MRI spine images, and the results indicated that the proposed approach can be performed well for the improvement of diagnosis. We believe that our approach can be helpful for the clinical application and extension of other medical imaging modalities for image stitching.

Brodmann’s Area Template Based Region of Interest Setting and Probabilistic Pathway Map Generation in Diffusion Tensor Tractography: Application to the Arcuate Fasciculus Fiber Tract in the Human Brain

The purpose of this study is to acquire accurate diffusion tensor tractography (DTT) results for arcuate fasciculus (AF) fiber tract using Brodmann’s area (BA) template for region of interest (ROI) setting. Thirteen healthy subjects were participated in this study. Fractional anisotropy (FA) map of each subject was calculated using diffusion tensor data, and T1w template was co-registered to FA map. The BA template was also co-registered using the transformation matrix. The ROIs were drawn in the co-registered BA template, and AF fiber tract was extracted. To generate the probabilistic pathway map, a binary mask image was generated based on the fiber tract image and co-registered to T1w template image. We also measured relative location of the AF fiber tract. The location of the probabilistic pathway map of each subject’s AF fiber tract was well defined in the brain. By using this probabilistic map, the mediolateral position ratio of AF was measured 18%, and the anteroposterior position ratio of AF was measured 35%, respectively. This study demonstrated that the AF fiber tract can be extracted using BA template for ROI setting and probabilistic pathway of fiber tract. Our results and analytical approaches can helpful for accurate fiber tracking and application of perspective clinical researches.

Simple image intensity compensation SIMIC method prior to application of distortion correction algorithms in brain diffusion Tensor Magnetic Resonance Imaging: Validation test for two cost functions of distortion correction algorithms

The purpose of this study is to design a simple image intensity compensation (SIMIC) method prior to the application of a variety of cost functions for distortion correction in diffusion tensor imaging (DTI). The synthetic dataset consists of each direction of diffusion weighted imaging (DWI) made by multiplication of nondiffusion weighted image (b = 0 image) and tensor matrices. We added the effects of patient motion and eddy current distortion using translation, rotation, scaling and shearing matrices. We calculated the b = 0 image of each direction from original DTI, inversely. A co‐registration method was applied to the extracted b = 0 images of each direction based on the original b = 0 image and then, the transformation matrices were generated and the original DTI were transformed using this transformation matrix. For the DTI distortion correction, two kinds of cost functions, normalized mutual information (NMI) and normalized cross‐correlation (NCC), were used. Visual assessments and quantitative measurements were used to evaluate the results. When using the NMI as a cost function, the quantitative results showed no significant differences between NMI and NMI with SIMIC method. However, there are significant differences compared with using the NCC as a cost function. Our study showed cost function for image distortion correction with SIMIC method improved the results both quantitatively and in terms of qualitative accuracy. This method may be helpful for DTI analysis and helpful for increasing accuracy.

A simple auto prescan calibration method for multislice fast spin echo MRI

The image quality of fast spin echo (FSE) is more sensitive than the typical spin echo pulse sequence caused by the eddy current effect. Microsecond‐scale misalignment of primary spin echoes produces a large spatial variation in image signal intensity. In this study, we describe an auto prescan calibration method that can improve the FSE image quality and minimize the eddy current effect on the image. We used a 0.32 T MRI system and obtained phantom and lumbar images. For FSE image correction, the optimal ranges and steps were determined to find the appropriate values, which were added to or subtracted from the gradient area values for each slice. The appropriate value of each slice could be found using the maximum signal intensity when the refocusing gradient area was changed by a number of steps in the optimal range. The determined value of each slice was applied before each slice image acquisition. The determined optimal step numbers and ranges were applied to in vivo image acquisition, and confirmed the reconstructed image quality. Based on our results, the obtained phantom and lumbar images were shown to be well corrected. The corrected images represented the image quality improvement and elimination of ghosting and blurring artifacts. In conclusion, we have proposed an FSE correction technique that automatically adjusts slice selection for the refocusing gradient balance during prescan, and confirmed that the calibration technique is very reliable even within complex in vivo images. We believe that our proposed technique will greatly benefit in MRI systems.

Sparse sampling MR image reconstruction using bregman iteration: A feasibility study at low tesla MRI system

MR images reconstruction need many samples that are sequentially sampled by phase encoding gradients in MRI system. MRI takes long scan time, therefore, many researchers have been studied to reduce scan time. Especially, the Compressed Sensing (CS) that is used sparse images and reconstruction from fewer sampling data which the k-space is not fully sampled. Recently, an iterative technique based on Bregman method is developed for denoising. The Bregman iteration method improves on the Total Variation (TV) regularization by gradually recovering the fine scale structures that are usually lost in the TV regularization. In this study, we studied sparse sampling image reconstruction using Bregman iteration at low tesla MRI system for improving the temporal resolution and validated the usefulness. The image was obtained at 0.32T MRI scanner (Magfinder II, Genpia, Korea) using 2D T1-weighed spin-echo pulse sequence with phantom and in-vivo human brain in the head coil. We applied the random k-space sampling and sampling ratios are determined by half of fully sampled k-space. The Bregman iteration was used to generate the final images based on the reduced data. The number of Bregman iterations used for the reconstruction was minimum 1 to maximum 100. We also calculated Root Mean Square Error (RMSE) values from error images that were performed according to number of bregman iterations. The results which are reconstructed images using the bregman iteration to sparse sampling image shown well reconstruction images compared with original images. Moreover, the RMSE values can be seen that sparse reconstructed phantom image and human images are converge to the original image. We confirmed the feasibility of sparse sampling image reconstruction methods using Bregman iteration at low tesla MRI system and obtained good results. Although our results used half of sampling ratio, this method will helpful to increase the temporal resolution at low tesla MRI system.

Minimisation of Signal Intensity Differences in Distortion Correction Approaches of Brain Magnetic Resonance Diffusion Tensor Imaging

ObjectivesTo evaluate the effects of signal intensity differences between the b0 image and diffusion tensor imaging (DTI) in the image registration process.MethodsTo correct signal intensity differences between the b0 image and DTI data, a simple image intensity compensation (SIMIC) method, which is a b0 image re-calculation process from DTI data, was applied before the image registration. The re-calculated b0 image (b0ext) from each diffusion direction was registered to the b0 image acquired through the MR scanning (b0nd) with two types of cost functions and their transformation matrices were acquired. These transformation matrices were then used to register the DTI data. For quantifications, the dice similarity coefficient (DSC) values, diffusion scalar matrix, and quantified fibre numbers and lengths were calculated.ResultsThe combined SIMIC method with two cost functions showed the highest DSC value (0.802 ± 0.007). Regarding diffusion scalar values and numbers and lengths of fibres from the corpus callosum, superior longitudinal fasciculus, and cortico-spinal tract, only using normalised cross correlation (NCC) showed a specific tendency toward lower values in the brain regions.ConclusionImage-based distortion correction with SIMIC for DTI data would help in image analysis by accounting for signal intensity differences as one additional option for DTI analysis.Key points• We evaluated the effects of signal intensity differences at DTI registration.• The non-diffusion-weighted image re-calculation process from DTI data was applied.• SIMIC can minimise the signal intensity differences at DTI registration.

Signal intensity correction for multichannel MR images using radon transformation

The purpose of this study is to correct signal intensity at low‐field MRI system with multichannel receiver coils using Radon transformation and filtered backprojection (FBP) method. An open‐type 0.32 T MRI system and a body size phantom were used to acquire the MR images. We used various types of coils from 2‐channels to 4‐channels, which minimized the loss of signal. In the intensity correction process, Radon transform was used for the images of each channel and low‐pass filtering was applied to reduce noise. After that FBP was used for the space transform again from the Radon space to the image space. We also made changes to the projection ranges and their intervals, and then confirmed them to evaluate the optimal parameters. All the intensity corrected results were compared with its original sum‐of‐square (SOS) images, and the corrected images showed more uniform and homogeneous intensities than the images without correction. In addition, these results were also shown in the quantitative values through the signal intensity variations according to the cut view along the horizontal lines of the images. The feasibility of our approach and results for signal intensity correction may be useful and helpful for the researchers of low‐field system with multichannel coils.