Data-driven curvilinear reconstructions of 3D MR images: application to cryptogenic extratemporal epilepsy

This paper presents a data-driven method for the reconstruction and visualisation of curvilinear slices from three-dimensional (3D) magnetic resonance (MR) scans of the head. Visualisation of curvilinear slices, rather than standard planar slices, produces symmetrical views of the cortex and allows small abnormalities to be detected by comparing the two hemispheres of the brain. In our method, the surface defined by the upper half of the brain is used as a reference shape for curvilinear reconstructions. The brain is first segmented from the 3D scan using a 3D region growing method associated to an unsupervised threshold selection technique. The upper half of the segmented brain is then extracted and fitted by a deformable surface model. This surface is finally interactively moved by the operator in the 3D scan, to visualise the desired curvilinear slice, which is projected on the screen as a two-dimensional image. We show an application of this visualisation technique to the localisation of cerebral epileptogenic lesions. The procedure has proven efficient and handy in clinical use.     

Application of fuzzy c-means segmentation technique for tissue differentiation in MR images of a hemorrhagic glioblastoma multiforme

The application of a raw data-based, operator-independent MR segmentation technique to differentiate boundaries of tumor from edema or hemorrhage is demonstrated. A case of a glioblastoma multiforme with gross and histopathologic correlation is presented. The MR image data set was segmented into tissue classes based on three different MR weighted image parameters (T₁-, proton density-, and T₂-weighted) using unsupervised fuzzy c-means (FCM) clustering algorithm technique for pattern recognition. A radiological examination of the MR images and correlation with fuzzy clustering segmentations was performed. Results were confirmed by gross and histopathology which, to the best of our knowledge, reports the first application of this demanding approach. Based on the results of neuropathologic correlation, the application of FCM MR image segmentation to several MR images of a glioblastoma multiforme represents a viable technique for displaying diagnostically relevant tissue contrast information used in 3D volume reconstruction. With this technique, it is possible to generate segmentation images that display clinically important neuroanatomic and neuropathologic tissue contrast information from raw MR image data.     

Optimization of tissue segmentation of brain MR images based on multispectral 3D feature maps

The purpose of this work was to optimize and increase the accuracy of tissue segmentation of the brain magnetic resonance (MR) images based on multispectral 3D feature maps. We used three sets of MR images as input to the in-house developed semi-automated 3D tissue segmentation algorithm: proton density (PD) and T2-weighted fast spin echo and, T1-weighted spin echo. First, to eliminate the random noise, non-linear anisotropic diffusion type filtering was applied to all the images. Second, to reduce the nonuniformity of the images, we devised and applied a correction algorithm based on uniform phantoms. Following these steps, the qualified observer "seeded" (identified training points) the tissue of interest. To reduce the operator dependent errors, cluster optimization was also used; this clustering algorithm identifies the densest clusters pertaining to the tissues. Finally, the images were segmented using k-NN (k-Nearest Neighborhood) algorithm and a stack of color-coded segmented images were created along with the connectivity algorithm to generate the entire surface of the brain. The application of pre-processing optimization steps substantially improved the 3D tissue segmentation methodology.     

Preoperative fast MRI of brain tumors using three-dimensional segmented echo planar imaging compared to three-dimensional gradient echo technique

The purpose of this study was to compare the diagnostic efficacy of a newly developed T(1)-weighted three-dimensional segmented echo planar imaging (3D EPI) sequence versus a conventional T(1)-weighted three dimensional spoiled gradient echo (3D GRE) sequence in the evaluation of brain tumors. Forty-four patients with cerebral tumors and infections were examined on a 1.0 T MR unit with 23 mT/m gradient strength. The total scan time for the T(1) 3D EPI sequence was 2 min 12 s, and for a conventional 3D GRE sequence it was 4 min 59 s. Both sequences were performed after administration of a contrast agent. The images were analyzed by three radiologists. Image assessment criteria included lesion conspicuity, contrast between different types of normal tissue, and image artifacts. In addition, signal-to-noise and contrast-to-noise-ratio (C/N) were calculated. The gray-white differentiation and C/N ratio of 3D EPI were found to be inferior to conventional 3D GRE images, but the difference was not statistically significant. In the qualitative comparison, lesion detection and conspicuity of 3D EPI images and conventional 3D GRE images were similar, but a tow-fold reduction of the scanning time was obtained. With the 3D EPI technique, a 50% scan time reduction could be achieved with acceptable image quality compared to conventional 3D GRE. Thus, the 3D EPI technique could replace conventional 3D GRE in the preoperative imaging of brain.     

Transfer learning in deep neural network based under-sampled MR image reconstruction

In Magnetic Resonance Imaging (MRI), the success of deep learning-based under-sampled MR image reconstruction depends on: (i) size of the training dataset, (ii) generalization capabilities of the trained neural network. Whenever there is a mismatch between the training and testing data, there is a need to retrain the neural network from scratch with thousands of MR images obtained using the same protocol. This may not be possible in MRI as it is costly and time consuming to acquire data. In this research, a transfer learning approach i.e. end-to-end fine tuning is proposed for U-Net to address the data scarcity and generalization problems of deep learning-based MR image reconstruction. First the generalization capabilities of a pre-trained U-Net (initially trained on the human brain images of 1.5 T scanner) are assessed for: (a) MR images acquired from MRI scanners of different magnetic field strengths, (b) MR images of different anatomies and (c) MR images under-sampled by different acceleration factors. Later, end-to-end fine tuning of the pre-trained U-Net is proposed for the reconstruction of the above-mentioned MR images (i.e. (a), (b) and (c)). The results show successful reconstructions obtained from the proposed method as reflected by the Structural SIMilarity index, Root Mean Square Error, Peak Signal-to-Noise Ratio and central line profile of the reconstructed images.     

Removal of extracerebral tissues in dual-echo magnetic resonance images via linear scale-space features

The location and masking of the brain and surrounding cerebrospinal fluid (CSF) in two-dimensional (2D) dual-echo fast spin-echo (FSE) magnetic resonance (MR) images of the head is achieved by an automated procedure with a voxel-based computational algorithm. Linear scale-space features are derived from the short-echo, proton-density (PD)-weighted images. The second-order Gaussian derivative (the Laplacian) operator is applied at three different spatial scales as a measure of image convexity/concavity with a first-order Gaussian derivative measure (the squared gradient) at a single scale used to circumscribe cortical regions. A mask obtained from the long-echo, T2-weighted image is used to remove extracerebral components of the visual system. A three-dimensional (3D) connectivity analysis then identifies the largest connected volume as the brain. Five dual-echo fast spin-echo images acquired by repeated scanning of the same normal volunteer were used to verify reproducibility; and coronal and axial acquisitions from another normal volunteer to demonstrate the method's robustness to data collected with non-cubic voxels. Images acquired from five individuals with Alzheimer's disease are also presented to show that the algorithm can be used in cases of non-normative anatomy. Validity is affirmed by demonstrating that cerebral volumes estimated by this method for all 12 images are highly correlated (R = 0.98) with estimates obtained by an expert human operator.     

Singular Value Decomposition Using Jacobi Algorithm in pMRI and CS

Parallel magnetic resonance imaging (pMRI) and compressed sensing (CS) have been recently used to accelerate data acquisition process in MRI. Matrix inversion (for rectangular matrices) is required to reconstruct images from the acquired under-sampled data in various pMRI algorithms (e.g., SENSE, GRAPPA) and CS. Singular value decomposition (SVD) provides a mechanism to accurately estimate pseudo-inverse of a rectangular matrix. This work proposes the use of Jacobi SVD algorithm to reconstruct MR images from the acquired under-sampled data both in pMRI and in CS. The use of Jacobi SVD algorithm is proposed in advance MRI reconstruction algorithms, including SENSE, GRAPPA, and low-rank matrix estimation in L + S model for matrix inversion and estimation of singular values. Experiments are performed on 1.5T human head MRI data and 3T cardiac perfusion MRI data for different acceleration factors. The reconstructed images are analyzed using artifact power and central line profiles. The results show that the Jacobi SVD algorithm successfully reconstructs the images in SENSE, GRAPPA, and L + S algorithms. The benefit of using Jacobi SVD algorithm for MRI image reconstruction is its suitability for parallel computation on GPUs, which may be a great help in reducing the image reconstruction time.     

Automatic measurement of changes in brain volume on consecutive 3D MR images by segmentation propagation

This article presents a technique to automatically measure changes in the volume of a structure of interest in successive 3D magnetic resonance (MR) images and its application in the study of the brain and lateral cerebral ventricles. The only manual step is a segmentation of the structure of interest in the first image. The analysis comprises, first, precise rigid co-registration of the time series of images; second, computation of residual deformations between pairs of images; third, automatic quantification of the volume change, obtained by propagation of the segmentation of the structure of interest through the series of MR images. This approach has been applied to monitor changes in the volume of the brain and lateral cerebral ventricles in a healthy subject and a patient with primary progressive aphasia (PPA). Results are consistent with those obtained by application of the boundary shift integral (BSI) and by stereology in the same subjects.     

磁共振成象诊断出血性胆囊炎——离体及活体研究

在离体研究的基础上,对三个出血性胆囊炎的病人术前做出诊断,出血性胆囊炎可分为混合性及非混合性.在离体实验中,如果血液未与胆汁混合,T₁加权象可发现加于10mL胆汁中的0.2mL的血液表现为高信号区;质子密度加权象可发现加于10mL,胆汁中的0.4mL血液表现为稍高信号区;T₂加权象对此不敏感.如果血液与胆汁完全混合,在所有采用的磁共振成象上均使胆汁信号增高.非混合性出血性胆囊炎的磁共振成象具有特征性:在T₁加权象及质子密度加权象胆囊内有高信号区,T₂加权象此区为等信号、低信号或其中心为低信号周围与胆汁相接的为高信号。混合性出血性胆囊炎在所采用的磁共振成象图象中,相对于肝脏,胆囊内容物表现为均匀高信号,临床资料及胆囊壁、胆囊周围渗出有助于这种出血性胆囊炎的诊断.     

MR image reconstruction of sparsely sampled 3D k-space data by projection-onto-convex sets

In many rapid three-dimensional (3D) magnetic resonance (MR) imaging applications, such as when following a contrast bolus in the vasculature using a moving table technique, the desired k-space data cannot be fully acquired due to scan time limitations. One solution to this problem is to sparsely sample the data space. Typically, the central zone of k-space is fully sampled, but the peripheral zone is partially sampled. We have experimentally evaluated the application of the projection-onto-convex sets (POCS) and zero-filling (ZF) algorithms for the reconstruction of sparsely sampled 3D k-space data. Both a subjective assessment (by direct image visualization) and an objective analysis [using standard image quality parameters such as global and local performance error and signal-to-noise ratio (SNR)] were employed. Compared to ZF, the POCS algorithm was found to be a powerful and robust method for reconstructing images from sparsely sampled 3D k-space data, a practical strategy for greatly reducing scan time. The POCS algorithm reconstructed a faithful representation of the true image and improved image quality with regard to global and local performance error, with respect to the ZF images. SNR, however, was superior to ZF only when more than 20% of the data were sparsely sampled. POCS-based methods show potential for reconstructing fast 3D MR images obtained by sparse sampling.     

Time-course magnetic resonance imaging of rat pancreatic cyst after experimental pancreatitis

Fast magnetic resonance (MR) imaging of the rat pancreas was carried out using a snapshot method to observe three-dimensional (3D) and temporal development of the pancreatic cyst after experimental pancreatitis. Acute pancreatitis was induced by a retrograde infusion of the trypsin-taurocholate solution into the pancreatic duct in 23 rats, of which seven survived for one month. Under 2% enflurane anesthesia, (1)H images of the rat abdomen were taken by a 4.7 T magnetic resonance spectrometer under spontaneous breathing. 3D images of the pancreas and cyst were reconstructed from the axial, sagittal and coronal images taken before, 24 h, 7 days, 14 days, 21 days and 28 days after the induction of pancreatitis. The 3D images reconstructed from different slice orientations at each time point showed good agreement with each other. The calculated volumes of the cyst on 7th, 14th, 21st, and 28th day were 0.3 +/- 0.1, 0.8 +/- 0.3, 2.1 +/- 0.6, 6.5 +/- 1.3 mL, respectively. The cystic fluid volume on 28th day was 6.4 +/- 1.4 mL, which confirmed reliability of volume measurement by MR imaging. Fast MR imaging (snapshot) together with 3D reconstruction allows us to understand the detailed chronological and spatial development of pancreatic cyst after acute pancreatitis in rats.     

The detection limit of a Gd3+-based T1 agent is substantially reduced when targeted to a protein microdomain

Simple low molecular weight (MW) chelates of Gd(3+) such as those currently used in clinical MRI are considered too insensitive for most molecular imaging applications. Here, we evaluated the detection limit (DL) of a molecularly targeted low MW Gd(3+)-based T(1) agent in a model where the receptor concentration was precisely known. The data demonstrate that receptors clustered together to form a microdomain of high local concentration can be imaged successfully even when the bulk concentration of the receptor is quite low. A GdDO3A-peptide identified by phage display to target the anti-FLAG antibody was synthesized, purified and characterized. T(1-)weighted MR images were compared with the agent bound to antibody in bulk solution and with the agent bound to the antibody localized on agarose beads. Fluorescence competition binding assays show that the agent has a high binding affinity (K(D)=150 nM) for the antibody, while the fully bound relaxivity of the GdDO3A-peptide/anti-FLAG antibody in solution was a relatively modest 17 mM(-1) s(-1). The agent/antibody complex was MR silent at concentrations below approximately 9 microM but was detectable down to 4 microM bulk concentrations when presented to antibody clustered together on the surface of agarose beads. These results provided an estimate of the DLs for other T(1)-based agents with higher fully bound relaxivities or multimeric structures bound to clustered receptor molecules. The results demonstrate that the sensitivity of molecularly targeted contrast agents depends on the local microdomain concentration of the target protein and the molecular relaxivity of the bound complex. A model is presented, which predicts that for a molecularly targeted agent consisting of a single Gd(3+) complex with bound relaxivity of 100 mM(-1) s(-1) or, more reasonably, four tethered Gd(3+) complexes each having a bound relaxivity of 25 mM(-1) s(-1), the DL of a protein microdomain is approximately 690 nM at 9.4 T. These experimental and extrapolated DLs are both well below current literature estimates and suggests that detection of low MW molecularly targeted T(1) agents is not an unrealistic goal.     

Magnetic resonance imaging in hemophilic children: value of gradient echo and contrast-enhanced imaging

The purpose of this study was to determine the value of Gradient Echo imaging for the evaluation of cartilage (3D fatsat) and blood products (2D Hemoflash), and the use of contrast enhanced SE imaging for the evaluation of synovial changes, in comparison to the clinical evaluation of children with hemophilia A. We investigated 21 joints in 16 patients with evidence of hemophilia A (mean age 11.3+/-2.1 years). In all patients, clinical examination, plain film radiographs, and MR evaluation were performed magnetic resonance imaging (MRI) was performed by using sagittal T1 SE and T2 SE images, as well as 3D fatsat GE and 2D GE images. Axial and sagittal T1 weighted SE images were obtained before and after contrast application. Findings from the clinical examination and MR imaging, regarding the evaluation of blood, synovia, and cartilage were compared. Clinical examination revealed evidence of a bleeding episode in 12 joints (57.1%), whereas MRI revealed evidence of blood or blood products in 15 joints (71.4%). Clinical investigations, including bleeding scores, pain scores, and physical examination scores did not correlate with MR findings. Due to the MR findings in 6 of 16 patients, therapeutic management was changed from on demand to prophylactic therapy. MR imaging with gradient echo and contrast-enhanced sequences is more sensitive than clinical examination for the detection of blood products in children with hemophilia. Its ability to demonstrate potentially early stages of cartilage or synovial alterations might assist in therapy planning. Clinical scores might underestimate effects of hemophilia.     

Joint reconstruction of multiecho MR images using correlated sparsity

This works addresses the problem of reconstructing multiple T1- or T2-weighted images of the same anatomical cross section from partially sampled K-space data. Previous studies in reconstructing magnetic resonance (MR) images from partial samples of the K-space used compressed sensing (CS) techniques to exploit the spatial correlation of the images (leading to sparsity in wavelet domain). Such techniques can be employed to reconstruct the individual T1- or T2-weighted images. However, in the current context, the different images are not really independent; they are images of the same cross section and, hence, are highly correlated. We exploit the correlation between the images, along with the spatial correlation within the images to achieve better reconstruction results than exploiting spatial correlation only.For individual MR images, CS-based techniques lead to a sparsity-promoting optimization problem in the wavelet domain. In this article, we show that the same framework can be extended to incorporate correlation between images leading to group/row sparsity-promoting optimization. Algorithms for solving such optimization problems have already been developed in the CS literature. We show that significant improvement in reconstruction accuracy can be achieved by considering the correlation between different T1- and T2-weighted images. If the reconstruction accuracy is considered to be constant, our proposed group sparse formulation can yield the same result with 33% less K-space samples compared with simple sparsity-promoting reconstruction. Moreover, the reconstruction time by our proposed method is about two to four times less than the previous method.     

Brain MRI tissue classification based on local Markov random fields

A new method for tissue classification of brain magnetic resonance images (MRI) of the brain is proposed. The method is based on local image models where each models the image content in a subset of the image domain. With this local modeling approach, the assumption that tissue types have the same characteristics over the brain needs not to be evoked. This is important because tissue type characteristics, such as T1 and T2 relaxation times and proton density, vary across the individual brain and the proposed method offers improved protection against intensity non-uniformity artifacts that can hamper automatic tissue classification methods in brain MRI. A framework in which local models for tissue intensities and Markov Random Field (MRF) priors are combined into a global probabilistic image model is introduced. This global model will be an inhomogeneous MRF and it can be solved by standard algorithms such as iterative conditional modes. The division of the whole image domain into local brain regions possibly having different intensity statistics is realized via sub-volume probabilistic atlases. Finally, the parameters for the local intensity models are obtained without supervision by maximizing the weighted likelihood of a certain finite mixture model. For the maximization task, a novel genetic algorithm almost free of initialization dependency is applied. The algorithm is tested on both simulated and real brain MR images. The experiments confirm that the new method offers a useful improvement of the tissue classification accuracy when the basic tissue characteristics vary across the brain and the noise level of the images is reasonable. The method also offers better protection against intensity non-uniformity artifact than the corresponding method based on a global (whole image) modeling scheme.     

Graded image segmentation of brain tissue in the presence of inhomogeneous radio frequency fields

Image segmentation is used increasingly to interpret MR spectroscopic data of the brain, using image contrast to identify gray matter (GM), white matter (WM), and cerebral spinal fluid (CSF). T(1)- or T(2)-weighted images are typically used, but poor shimming, susceptibility effects, and small variations in B(1) and receptivity cause difficulties in tissue identification. Quantitative imaging of T(1) can reduce many of these difficulties but is still subject to complications when B(1) has large variations like those observed with the surface coils often used for spectroscopy. In this study, B(1) imaging was implemented to support quantitative imaging of T(1) with either a surface coil or a volume coil. The T(1) observed by this method is a continuous function across mixtures of WM/GM and GM/CSF, and this function was measured and used to convert the images of T(1) to maps of percent GM, WM, and CSF.     

Increased differentiation of intracranial white matter lesions by multispectral 3D-tissue segmentation: preliminary results

MRI is a very sensitive imaging modality, however with relatively low specificity. The aim of this work was to determine the potential of image post-processing using 3D-tissue segmentation technique for identification and quantitative characterization of intracranial lesions primarily in the white matter. Forty subjects participated in this study: 28 patients with brain multiple sclerosis (MS), 6 patients with subcortical ischemic vascular dementia (SIVD), and 6 patients with lacunar white matter infarcts (LI). In routine MR imaging these pathologies may be almost indistinguishable. The 3D-tissue segmentation technique used in this study was based on three input MR images (T(1), T(2)-weighted, and proton density). A modified k-Nearest-Neighbor (k-NN) algorithm optimized for maximum computation speed and high quality segmentation was utilized. In MS lesions, two very distinct subsets were classified using this procedure. Based on the results of segmentation one subset probably represent gliosis, and the other edema and demyelination. In SIVD, the segmented images demonstrated homogeneity, which differentiates SIVD from the heterogeneity observed in MS. This homogeneity was in agreement with the general histological findings. The LI changes pathophysiologically from subacute to chronic. The segmented images closely correlated with these changes, showing a central area of necrosis with cyst formation surrounded by an area that appears like reactive gliosis. In the chronic state, the cyst intensity was similar to that of CSF, while in the subacute stage, the peripheral rim was more prominent. Regional brain lesion load were also obtained on one MS patient to demonstrate the potential use of this technique for lesion load measurements. The majority of lesions were identified in the parietal and occipital lobes. The follow-up study showed qualitatively and quantitatively that the calculated MS load increase was associated with brain atrophy represented by an increase in CSF volume as well as decrease in "normal" brain tissue volumes. Importantly, these results were consistent with the patient's clinical evolution of the disease after a six-month period. In conclusion, these results show there is a potential application for a 3D tissue segmentation technique to characterize white matter lesions with similar intensities on T(2)-weighted MR images. The proposed methodology warrants further clinical investigation and evaluation in a large patient population.     

Optimizing brain MRI protocols in the follow-up of patients with multiple sclerosis T2-weighted MRI of the brain after the administration of gadopentetate dimeglumine

The aim of our study was to determine whether T2-weighted (T2w) MRI of the brain could be performed immediately after the administration of gadopentetate dimeglumine (gadolinium DTPA) in patients with multiple sclerosis (MS) without a loss in image quality or diagnostic reliability. Sixteen patients with clinically diagnosed MS were included in the study. Twenty-four patients with various cerebral pathologies (14 patients with multiple lacunar lesions) were examined in order to exclude masking of T2 hyperintense lesions other than MS lesions. Images of 10 patients without pathological changes served as a control condition for the qualitative analysis. In these 50 patients, T1w and T2w MRI was performed before and after the administration of gadolinium DTPA. Signal intensities were measured within T2 hyperintense cerebral lesions, in T1-enhancing lesions and in normal appearing brain tissue on T2w turbo spin-echo (TSE) sequences. Both quantitative and qualitative analysis did not show significant differences between T2w pre- and postcontrast series. T2w MRI performed prior to and after the administration of gadolinium DTPA provides similar information in patients with MS. With a TR of 3.2 s, not a single lesion was obscured on T2w postcontrast series. Acquisition of T2w MR images immediately after the administration of gadolinium DTPA allows for shorter examination time and assures sufficient time for contrast enhancement in cerebral lesions with a disrupted blood-brain barrier.     

Initial in vivo rodent sodium and proton MR imaging at 21.1 T

The first in vivo sodium and proton magnetic resonance (MR) images and localized spectra of rodents were attained using the wide bore (105 mm) high resolution 21.1-T magnet, built and operated at the National High Magnetic Field Laboratory (Tallahassee, FL, USA). Head images of normal mice (C57BL/6J) and Fisher rats (∼250 g) were acquired with custom designed radiofrequency probes at frequencies of 237/900 MHz for sodium and proton, respectively. Sodium MR imaging resolutions of ∼0.125 μl for mouse and rat heads were achieved by using a 3D back-projection pulse sequence. A gain in SNR of ∼3 for sodium and ∼2 times for proton were found relative to corresponding MR images acquired at 9.4 T. 3D Fast Low Angle Shot (FLASH) proton mouse images (50×50×50 μm³) were acquired in 90 min and corresponding rat images (100×100×100 μm³) within a total time of 120 min. Both in vivo large rodent MR imaging and localized spectroscopy at the extremely high field of 21.1 T are feasible and demonstrate improved resolution and sensitivity valuable for structural and functional brain analysis.     

Statistical analysis of fractal-based brain tumor detection algorithms

Fractals are geometric objects that have a noninteger fractal dimension (FD). The FD has been exploited for various biomedical recognition applications such as breast tumor and lung tumor detection. Our previous work shows that the FD is useful in the detection of brain tumors when a reference nontumor image is available. In this work, we extend our previous work by statistically validating the results of FD analysis on a set of 80 real MR and CT images. Our half-image technique requires that the tumor is located in one half of the brain whereas our whole-image technique does not. Furthermore, we alleviate the need for a reference (control) nontumor image to compute the tumor FD, which was necessary in our previous work. We also compare the brain tumor detection performance of our algorithms with other fractal-based algorithms in the literature and statistically validate our results against manually segmented tumor images. We find that the tumor region offers a statistically significant lower FD compared with that of the nontumor area for most of the FD algorithms studied in this work. Thus, our statistical analysis suggests that these FD algorithms may be exploited successfully to determine the possible presence and location of brain tumors in MR and CT images.