Renal malacoplakia: demonstration by MR imaging

We report herein a case of histologically verified bilateral renal malacoplakia. MRI features were distinctive and include multiple nodules 1-2 cm in diameter that were low in signal on T1, T2 and early and late post gadolinium images with intervening fibrous stroma. Demonstration of renal malacoplakia on MR images may obviate the need for major surgery and rapidly direct patients to appropriate antimicrobial therapy for treatment.     

Bowel-related abscesses: MR demonstration preliminary results

The purpose of this study is to describe the appearance of bowel-related abscesses on magnetic resonance (MR) images. Sixteen consecutive patients who had bowel-related abscesses underwent MR examination at 1.5T. MR sequences included T₁-weighted fat-suppressed imaging pre- and post-intravenous gadolinium chelate administration (all patients) and breathing-independent single-shot T₂-weighted half Fourier turbo (fast) spin echo (6 patients). Patients with pelvic abscesses also underwent sagittal imaging with post-gadolinium T₁-weighted images (9 patients) and T₂-weighted turbo (fast) spin echo (8 patients). Abscesses were confirmed by open surgery or surgical drainage (6 patients), percutaneous drainage (8 patients), or combined physical examination, fluoroscopic fistulogram, and clinical follow-up (2 patients). Oval-shaped fluid collections were identified in all of the patients, which ranged in diameter from 2 cm to 18 cm, mean: 8 cm. Abscesses were low to intermediate in signal on T₁-weighted images, heterogenous and moderately high signal on T₂-weighted images, and low signal on post-gadolinium images. A layering effect of lower signal material in the dependent portion of the abscess was noted in abscesses in 6 of 14 patients on T₂-weighted images. Post-gadolinium images demonstrated a definable 3- to 7-mm thick abscess wall, which enhanced substantially with contrast. Definition of the wall was best shown on fat-suppressed images post-gadolinium. Substantial enhancement of surrounding periabscess tissues was demonstrated in all cases and was most clearly defined on fat-suppressed images. Image acquisition in two orthogonal planes was of value to demonstrate that fluid collections were oval, and separate from bowel. Image acquisition in the sagittal plane was useful in the evaluation of pelvic abscesses. The results from this preliminary study show that bowel-related abscesses are demonstrable on MR images using gadolinium-enhanced fat-suppressed T₁-weighted and turbo (fast) spin-echo T₂-weighted sequences. The presence of a thickened, enhancing lesion wall and enhancement of perilesional tissues on T₁-weighted fat-suppressed images were observed in all abscesses. A layering effect of low signal intensity material in the dependent portion of the abscess was an important ancillary feature.     

Hepatic angiomyolipoma with minimal fat content: MR demonstration

Hepatic angiomyolipoma is a rare entity. In this report, we describe the unusual MR findings of an angiomyolipoma with minimal fat content with histopathology correlation. This tumor had signal intensity features on various MR sequences including intense heterogeneous enhancement on immediate postgadolinium spoiled gradient echo images that resembled the appearance of a well-differentiated primary hepatocellular carcinoma. Angiomyolipoma with minimal fat content should be considered in the differential diagnosis of tumors with early diffuse heterogeneous enhancement on contrast enhanced images.     

Observations on maximum entropy processing of MR images

A maximum entropy (MAXENT) criteria for MR image processing optimizations has previously shown poor performance, but this note observes that there are two entirely different kinds of "data transmission" applications which appear to have been intermixed. In the two cases, "image entropy" actually refers to different kinds of data variables. The previous literature formulations are for transfer of data in which pixel-locations are the transmitted variable, and these pixels may be neither uniform nor constant. The second application concerns the MRI data set for display. Its data variables are image pixel-values of magnetization intensity, and the data transfer mode has the sense of visual display. When MAXENT criteria are modified to address an array of pixel-value intensities, and use a pixel-value information entropy rather than pixel-locations entropy, then successful data processing results. Restoring display visualization from highly nonuniform surface coils for lumbar spine scans are demonstrated, as an example of MAXENT usefulness.     

MR diagnosis of lymphangioleiomyomatosis: Visibility of pulmonary cysts on spin-echo images

The signs and symptoms of pulmonary lymphangioleiomyomatosis are nonspecific, but the radiographic findings, particularly those on computed tomographic (CT) scans, often suggest the diagnosis. Although evaluation of the lungs on magnetic resonance (MR) scans is limited, MR findings in some diffuse lung diseases have been described. This report illustrates a case of lymphangioleiomyomatosis in which the diagnosis was suggested by the appearance of the lungs on spin-echo MR images.     

Deriving new soft tissue contrasts from conventional MR images using deep learning

Versatile soft tissue contrast in magnetic resonance imaging is a unique advantage of the imaging modality. However, the versatility is not fully exploited. In this study, we propose a deep learning-based strategy to derive more soft tissue contrasts from conventional MR images obtained in standard clinical MRI. Two types of experiments are performed. First, MR images corresponding to different pulse sequences are predicted from one or more images already acquired. As an example, we predict T_1ρ weighted knee image from T₂ weighted image and/or T₁ weighted image. Furthermore, we estimate images corresponding to alternative imaging parameter values. In a representative case, variable flip angle images are predicted from a single T₁ weighted image, whose accuracy is further validated in quantitative T₁ map subsequently derived. To accomplish these tasks, images are retrospectively collected from 56 subjects, and self-attention convolutional neural network models are trained using 1104 knee images from 46 subjects and tested using 240 images from 10 other subjects. High accuracy has been achieved in resultant qualitative images as well as quantitative T₁ maps. The proposed deep learning method can be broadly applied to obtain more versatile soft tissue contrasts without additional scans or used to normalize MR data that were inconsistently acquired for quantitative analysis.     

Spinal MR imaging in suspected metastases: correlation with skeletal scintigraphy

Bone scintigraphy (RN) and magnetic resonance imaging (MR) were prospectively and retrospectively correlated in 64 patients with suspected spinal metastatic disease and possible spinal cord compression. Images were retrospectively interpreted and compared with the prospective official RN and MR reports to help decide relative prospective lesion conspicuity. Spinal lesions were confirmed by radiography, computed tomography, myelography or MR and RN follow-up in 56 patients (88%). Prospectively, MR detected 11 lesions not reported on RN while RN detected two lesions not reported on MR. Retrospective review of RN detected six lesions previously not reported. Retrospectively MR showed all lesions. Those lesions seen only in retrospect by RN were rather subtle and would be difficult to detect prospectively. In general, lesions not well seen on RN had relatively more bone marrow abnormality and less cortical bone involvement. In some cases, MR imaging shows spinal marrow lesions not well seen on planar RN.     

Differentiation of hepatic cavernous hemangioma from metastases by rare sequence MR imaging.

In this study, in order to differentiate cavernous hemangioma and hepatic metastases, rapid acquisition relaxation enhanced (RARE) sequence was used. First, in vivo measurements of T1, T2 relaxation times and proton density were obtained using T1, T2 calculation protocol (TOMIKON S50, 0.5T) and multipoint techniques. These measurements were made from regions of interest placed over the liver, spleen (because of similarity of relaxation time values between hepatic metastases and spleen) and cavernous hemangioma (HCH). Based on these intrinsic parameters, T2 curves signal intensity of three different tissues were constructed. At TE = 500 ms, the signal intensity of the liver and spleen has been near zero whereas in HCH, the signal intensity remained. As RARE sequence is very similar to spin echo (SE), by replacing effective TE(ETE) = 500 ms in the RARE equation, two dimensional contrast-to-noise ratio (CNR) contour plots were constructed demonstrating signal intensity contrast between liver-spleen, liver-Hemangioma for two different scan times (3 min, 7.5 s) and pulse timing. Then, optimal RARE factor and inter echo times were obtained in order to have maximum CNR between liver-Hemangioma and minimum CNR between liver-spleen. These optimal parameters were performed on ten normal and five persons with known HCH. Images showed that in both scan times (3 min, 7.5 s); the liver and spleen were suppressed whereas the HCH was enhanced. The image quality in the scan time of 3 min was better than the scan time of 7.5 s. Moreover, in this study, two different sequences were compared: i) Multi-slice single echo (MSSE) for T1 weighted image ii) RARE (ETE = 80 ms) for T2-weighted image. This comparison was done to show maximum CNR between liver-spleen (metastases) and to choose a better sequence for detecting metastases. CNR in the RARE sequence was more than in the MSSE sequence.     

Model-based electron microscopy: From images toward precise numbers for unknown structure parameters

Statistical parameter estimation theory is proposed as a method to quantify electron microscopy images. It aims at obtaining precise and accurate values for the unknown structure parameters including, for example, atomic column positions and types. In this theory, observations are purely considered as data planes, from which structure parameters have to be determined using a parametric model describing the images. The method enables us to measure positions of atomic columns with a precision of the order of a few picometers even though the resolution of the electron microscope is one or two orders of magnitude larger. Moreover, small differences in averaged atomic number, which cannot be distinguished visually, can be quantified using high-angle annular dark field scanning transmission electron microscopy images. Finally, it is shown how to optimize the experimental design so as to attain the highest precision. As an example, the optimization of the probe size for nanoparticle radius measurements is considered. It is also shown how to quantitatively balance signal-to-noise ratio and resolution by adjusting the probe size.     

Acute renal failure: common occurrence of preservation of corticomedullary differentiation on MR images

The purpose of this study was to evaluate the relationship between renal corticomedullary differentiation (CMD) on MR imaging and serum creatinine (sCr) level in patients with acute renal failure (ARF). Twenty-one patients with ARF were retrospectively investigated. In all 21 patients, sCr levels were obtained on the same date as the MR study, and within 8 days before and after the MR study. CMD was assessed on non-contrast T(1)-weighted images and immediate post-gadolinium spoiled gradient echo (Gd-SGE) images. Presence of CMD was graded into 3 groups as 'preserved', 'intermediate', or 'loss'. On non-contrast T(1)-weighted images, 12/21 (57%) showed loss of CMD and 9/21 (43%) showed preserved CMD. On immediate Gd-SGE images, 5/21 (24%) showed loss of CMD, 12/21 (57%) preserved CMD, and 4/21 (19%) intermediate CMD. The sCr levels of 9 patients with preserved CMD on non-contrast T(1)-weighted images ranged from 1.4 to 10.5 mg/dl (mean 4.6 mg/dl), while those of 12 patients with loss of CMD ranged from 1.6 to 7.6 mg/dl (mean 4.8 mg/dl), which was not statistically significant (p > 0.2). Renal CMD can remain preserved on non-contrast T(1)-weighted or immediate Gd-SGE images in patients with acute presentation of ARF, independent of sCr level.     

Experimental demonstration of quantitation errors in MR spectroscopy resulting from saturation corrections under changing conditions

Metabolite concentration measurements in in vivo NMR are generally performed under partially saturated conditions, with correction for partial saturation performed after data collection using a measured saturation factor. Here, we present an experimental test of the hypothesis that quantitation errors can occur due to application of such saturation factor corrections in changing systems. Thus, this extends our previous theoretical work on quantitation errors due to varying saturation factors. We obtained results for two systems frequently studied by 31P NMR, the ischemic rat heart and the electrically stimulated rat gastrocnemius muscle. The results are interpreted in light of previous theoretical work which defined the degree of saturation occurring in a one-pulse experiment for a system with given spin-lattice relaxation times, T(1)s, equilibrium magnetizations, M(0)s, and reaction rates. We found that (i) the assumption of constancy of saturation factors leads to quantitation errors on the order of 40% in inorganic phosphate; (ii) the dominant contributor to the quantitation errors in inorganic phosphate is most likely changes in T(1); (iii) T(1) and M(0) changes between control and intervention periods, and chemical exchange contribute to different extents to quantitation errors in phosphocreatine and gamma-ATP; (iv) relatively small increases in interpulse delay substantially decreased quantitation errors for metabolites in ischemic rat hearts; (v) random error due to finite SNR led to approximately 4% error in quantitation, and hence was a substantially smaller contributor than were changes in saturation factors.     

Automatic phasing of MR images. Part I: linearly varying phase

In spin-echo and well shimmed gradient-echo images, the phase of the complex image often varies linearly in both the readout and phase-encode directions. Thus, in principle, it is possible to display an image in absorption mode. However, manually determining the two first-order and one zero-order phase parameters needed to display an absorption-mode image is a formidable task. In this paper, the Bayesian calculations needed to automatically determine these parameters are presented, and the calculations are illustrated using spin-echo images.     

Automatic phasing of MR images. Part II: voxel-wise phase estimation

Magnetic resonance images are typically displayed as the absolute value of the discrete Fourier transform of the k-space data. However, absorption-mode images, the real part of the discrete Fourier transform of the data after applying an appropriate phase correction, have significant advantages over absolute-value images. In a companion paper, the problem of estimating the phase parameters needed to produce an absorption-mode image when the phase of the complex image varies linearly as a function of position, a situation common in magnetic resonance images, was addressed. However, some magnetic resonance images have phases that can vary in a complicated, nonlinear, positionally dependent fashion. To produce an absorption-mode image from these data, one must first estimate the positionally dependent phase, and then use that phase estimate to produce an absorption-mode image. This paper addresses both of these problems by first using Bayesian probability theory to estimate the constant or zero-order phase as a function of image position, and then the calculations are illustrated by using them to generate absorption-mode images from data where the phase of the image is a nonlinear function of position.     

Three-dimensional segmentation of anatomical structures in MR images on large data bases

In this paper an image-based method founded on mathematical morphology is presented in order to facilitate the segmentation of cerebral structures over large data bases of 3D magnetic resonance images (MRIs). The segmentation is described as an immersion simulation, applied to the modified gradient image, modeled by a generated 3D-region adjacency graph (RAG). The segmentation relies on two main processes: homotopy modification and contour decision. The first one is achieved by a marker extraction stage where homogeneous 3D-regions are identified. This stage uses contrasted regions from morphological reconstruction and labeled flat regions constrained by the RAG. Then, the decision stage intends to precisely locate the contours of regions detected by the marker extraction. This decision is performed by a 3D extension of the watershed transform. The method has been applied on a data base of 3D brain MRIs composed of fifty patients. Results are illustrated by segmenting the ventricles, corpus callosum, cerebellum, hippocampus, pons, medulla and midbrain on our data base and the approach is validated on two phantom 3D MRIs.     

A statistical method for characterizing the noise in nonlinearly reconstructed images from undersampled MR data: The POCS example

The projection-onto-convex-sets (POCS) algorithm is a powerful tool for reconstructing high-resolution images from undersampled k-space data. It is a nonlinear iterative method that attempts to estimate values for missing data. The convergence of the algorithm and its other deterministic properties are well established, but relatively little is known about how noise in the source data influences noise in the final reconstructed image. In this paper, we present an experimental treatment of the statistical properties in POCS and investigate 12 stochastic models for its noise distribution beside its nonlinear point spread functions. Statistical results show that as the ratio of the missing k-space data increases, the noise distribution in POCS images is no longer Rayleigh as with conventional linear Fourier reconstruction. Instead, the probability density function for the noise is well approximated by a lognormal distribution. For small missing data ratios, however, the noise remains Rayleigh distributed. Preliminary results show that in the presence of noise, POCS images are often dominated by POCS-enhanced noise rather than POCS-induced artifacts. Implicit in this work is the presentation of a general statistical method that can be used to assess the noise properties in other nonlinear reconstruction algorithms.     

Thrombophlebitis of the inferior vena cava involving the retroperitoneum with Crohn's disease: MR demonstration

We present a case of thrombophlebitis of the inferior vena cava (IVC) with Crohn's disease after intestinal perforation and prolonged indwelling of a catheter in the IVC. Magnetic resonance imaging demonstrated abnormal thickening and enhancement of the IVC wall. In addition, IVC thrombus formation was shown. The abnormal enhancement extended from the wall in the pericaval tissue and into the retroperitoneum, and regressed as the Crohn's disease subsided.     

Understanding the Discrepancies Between 31P MR Spectroscopy Assessed Liver Metabolite Concentrations from Different Institutions

The high divergence between the liver metabolite concentrations and pH values reported in previous quantitative ³¹P magnetic resonance studies, for instance phosphomonoester (0.7–3.8 mM) and phosphodiester (3.5–9.7 mM), has not been addressed in the literature. To assess what level of discrepancy can be caused by processing and metabolite integration, in this study chemical shift imaging localized ³¹P magnetic resonance spectra of human liver were quantitated by three methods currently applied in clinical practice: peak areas defined manually by placement of two cursors vs. frequency domain curve fitting with the assumption of either Gaussian or Lorentzian line shapes. Large reproducible differences were found in liver metabolite peak areas but not in pH, indicating that processing and peak integration methods can only explain part of the discrepancies between the results from different institutions.     

Automatic knee cartilage segmentation from multi-contrast MR images using support vector machine classification with spatial dependencies

Accurate segmentation of knee cartilage is required to obtain quantitative cartilage measurements, which is crucial for the assessment of knee pathology caused by musculoskeletal diseases or sudden injuries. This paper presents an automatic knee cartilage segmentation technique which exploits a rich set of image features from multi-contrast magnetic resonance (MR) images and the spatial dependencies between neighbouring voxels. The image features and the spatial dependencies are modelled into a support vector machine (SVM)-based association potential and a discriminative random field (DRF)-based interaction potential. Subsequently, both potentials are incorporated into an inference graphical model such that the knee cartilage segmentation is cast into an optimal labelling problem which can be efficiently solved by loopy belief propagation. The effectiveness of the proposed technique is validated on a database of multi-contrast MR images. The experimental results show that using diverse forms of image and anatomical structure information as the features are helpful in improving the segmentation, and the joint SVM-DRF model is superior to the classification models based solely on DRF or SVM in terms of accuracy when the same features are used. The developed segmentation technique achieves good performance compared with gold standard segmentations and obtained higher average DSC values than the state-of-the-art automatic cartilage segmentation studies.     

Evaluation of cystic ovarian lesions using apparent diffusion coefficient calculated from reordered turboflash MR images.

Reordered snapshot fast low-angle shot images with, and without, diffusion-perfusion gradients were used for the evaluation of contents of cystic ovarian lesions. Sonographically detected 51 cystic ovarian lesions (13 endometrial cysts, 17 ovarian cysts, 7 serous cystadenomas, 6 mucinous cystadenomas, 8 malignant cystic ovarian tumors) were studied. T1- and T2-weighted images, reordered snapshot fast low-angle shot images with and without diffusion-perfusion gradients (b = 106 and 0 s/mm2, respectively) were obtained. Using these images, apparent diffusion coefficients (ADCs) were calculated in the cystic contents of these lesions. Endometrial cysts and malignant cystic ovarian tumors showed lower ADC values than ovarian cysts, serous cystadenomas and mucinous cystadenomas (p < 0.02). There was no distinct ADC difference among ovarian cysts, serous cystadenomas, mucinous cystadenomas (p > 0.2). In conclusion, diffusion-weighted magnetic resonance imaging is possible to be useful to evaluate cystic contents of ovarian lesions.