Acquired Magnetic Resonance Images Research Articles

Motion Artifact Reduction Using U-Net Model with Three-Dimensional Simulation-Based Datasets for Brain Magnetic Resonance Images.

This study aimed to remove motion artifacts from brain magnetic resonance (MR) images using a U-Net model. In addition, a simulation method was proposed to increase the size of the dataset required to train the U-Net model while avoiding the overfitting problem. The volume data were rotated and translated with random intensity and frequency, in three dimensions, and were iterated as the number of slices in the volume data. Then, for every slice, a portion of the motion-free k-space data was replaced with motion k-space data, respectively. In addition, based on the transposed k-space data, we acquired MR images with motion artifacts and residual maps and constructed datasets. For a quantitative evaluation, the root mean square error (RMSE), peak signal-to-noise ratio (PSNR), coefficient of correlation (CC), and universal image quality index (UQI) were measured. The U-Net models for motion artifact reduction with the residual map-based dataset showed the best performance across all evaluation factors. In particular, the RMSE, PSNR, CC, and UQI improved by approximately 5.35×, 1.51×, 1.12×, and 1.01×, respectively, and the U-Net model with the residual map-based dataset was compared with the direct images. In conclusion, our simulation-based dataset demonstrates that U-Net models can be effectively trained for motion artifact reduction.

Pulmonary magnetic resonance-guided online adaptive radiotherapy of locally advanced non-small-cell lung cancer: the PUMA trial

BackgroundPatients with locally-advanced non-small-cell lung cancer (LA-NSCLC) are often ineligible for surgery, so that definitive chemoradiotherapy (CRT) represents the treatment of choice. Nevertheless, long-term tumor control is often not achieved. Intensification of radiotherapy (RT) to improve locoregional tumor control is limited by the detrimental effect of higher radiation exposure of thoracic organs-at-risk (OAR). This narrow therapeutic ratio may be expanded by exploiting the advantages of magnetic resonance (MR) linear accelerators, mainly the online adaptation of the treatment plan to the current anatomy based on daily acquired MR images. However, MR-guidance is both labor-intensive and increases treatment times, which raises the question of its clinical feasibility to treat LA-NSCLC. Therefore, the PUMA trial was designed as a prospective, multicenter phase I trial to demonstrate the clinical feasibility of MR-guided online adaptive RT in LA-NSCLC.MethodsThirty patients with LA-NSCLC in stage III A-C will be accrued at three German university hospitals to receive MR-guided online adaptive RT at two different MR-linac systems (MRIdian Linac®, View Ray Inc. and Elekta Unity®, Elekta AB) with concurrent chemotherapy. Conventionally fractioned RT with isotoxic dose escalation up to 70 Gy is applied. Online plan adaptation is performed once weekly or in case of major anatomical changes. Patients are followed-up by thoracic CT- and MR-imaging for 24 months after treatment. The primary endpoint is twofold: (1) successfully completed online adapted fractions, (2) on-table time. Main secondary endpoints include adaptation frequency, toxicity, local tumor control, progression-free and overall survival.DiscussionPUMA aims to demonstrate the clinical feasibility of MR-guided online adaptive RT of LA-NSCLC. If successful, PUMA will be followed by a clinical phase II trial that further investigates the clinical benefits of this approach. Moreover, PUMA is part of a large multidisciplinary project to develop MR-guidance techniques.Trial registrationClinicalTrials.gov: NCT05237453.

Applications of a deep learning method for anti-aliasing and super-resolution in MRI

Magnetic resonance (MR) images with both high resolutions and high signal-to-noise ratios (SNRs) are desired in many clinical and research applications. However, acquiring such images takes a long time, which is both costly and susceptible to motion artifacts. Acquiring MR images with good in-plane resolution and poor through-plane resolution is a common strategy that saves imaging time, preserves SNR, and provides one viewpoint with good resolution in two directions. Unfortunately, this strategy also creates orthogonal viewpoints that have poor resolution in one direction and, for 2D MR acquisition protocols, also creates aliasing artifacts. A deep learning approach called SMORE that carries out both anti-aliasing and super-resolution on these types of acquisitions using no external atlas or exemplars has been previously reported but not extensively validated. This paper reviews the SMORE algorithm and then demonstrates its performance in four applications with the goal to demonstrate its potential for use in both research and clinical scenarios. It is first shown to improve the visualization of brain white matter lesions in FLAIR images acquired from multiple sclerosis patients. Then it is shown to improve the visualization of scarring in cardiac left ventricular remodeling after myocardial infarction. Third, its performance on multi-view images of the tongue is demonstrated and finally it is shown to improve performance in parcellation of the brain ventricular system. Both visual and selected quantitative metrics of resolution enhancement are demonstrated.

Generative Adversarial Networks for the Creation of Realistic Artificial Brain Magnetic Resonance Images.

Even as medical data sets become more publicly accessible, most are restricted to specific medical conditions. Thus, data collection for machine learning approaches remains challenging, and synthetic data augmentation, such as generative adversarial networks (GAN), may overcome this hurdle. In the present quality control study, deep convolutional GAN (DCGAN)–based human brain magnetic resonance (MR) images were validated by blinded radiologists. In total, 96 T1-weighted brain images from 30 healthy individuals and 33 patients with cerebrovascular accident were included. A training data set was generated from the T1-weighted images and DCGAN was applied to generate additional artificial brain images. The likelihood that images were DCGAN-created versus acquired was evaluated by 5 radiologists (2 neuroradiologists [NRs], vs 3 non-neuroradiologists [NNRs]) in a binary fashion to identify real vs created images. Images were selected randomly from the data set (variation of created images, 40%–60%). None of the investigated images was rated as unknown. Of the created images, the NRs rated 45% and 71% as real magnetic resonance imaging images (NNRs, 24%, 40%, and 44%). In contradistinction, 44% and 70% of the real images were rated as generated images by NRs (NNRs, 10%, 17%, and 27%). The accuracy for the NRs was 0.55 and 0.30 (NNRs, 0.83, 0.72, and 0.64). DCGAN-created brain MR images are similar enough to acquired MR images so as to be indistinguishable in some cases. Such an artificial intelligence algorithm may contribute to synthetic data augmentation for “data-hungry” technologies, such as supervised machine learning approaches, in various clinical applications.

Development and Testing of a Magnetic Resonance (MR) Conditional Afterloader for Source Tracking in Magnetic Resonance Imaging-Guided High-Dose-Rate (HDR) Brachytherapy

For the purpose of magnetic resonance imaging (MRI)-guided high-dose-rate (HDR) brachytherapy, a prototype magnetic resonance (MR) conditional afterloader was developed. This study demonstrates the development and testing of the prototype, while operating simultaneously with MRI. In combination with an MR-based method for HDR source localization, this development enables treatment verification of HDR brachytherapy. Additionally, this allows a direct reconstruction of the source dwell positions after catheter insertion (when using a dummy source) and introduction of a clinical workflow where the patient remains in the same position during dwell position reconstruction, treatment planning and irradiation. A prototype MR conditional afterloader was developed by providing radiofrequency (RF) shielding and a plastic source cable containing a dummy source. Simultaneous functioning of the afterloader and MRI acquisition was tested in an experimental setting where the afterloader was placed next to the scanner and programmed to send the source to predefined positions within a phantom, while acquiring MR images. The HDR source positions were determined using MR artifact simulation and matching of the MR images to the simulated artifact. Additionally, the impact of the presence and use of the afterloader on the MRI performance was investigated by assessment of RF interference, signal-to-noise ratio (SNR), and B0 field homogeneity. The experiments demonstrated that the prototype MR conditional afterloader and the MRI scanner fully functioned while operating simultaneously, without influencing the other system. The step sizes between the source positions obtained from the MR images corresponded with the afterloader settings. Besides, the MRI performance tests demonstrated no deterioration due to the presence or functioning of the afterloader next to the scanner. This research has demonstrated the feasibility of simultaneous MR acquisition and employment of an MR conditional afterloader. This development enables real-time HDR source localization for treatment verification of MRI-guided HDR brachytherapy using an MR conditional afterloader.

Left Ventricular Wall Stress Is Sensitive Marker of Hypertrophic Cardiomyopathy With Preserved Ejection Fraction

Hypertrophic cardiomyopathy (HCM) patients present altered myocardial mechanics due to the hypertrophied ventricular wall and are typically diagnosed by the increase in myocardium wall thickness. This study aimed to quantify regional left ventricular (LV) shape, wall stress and deformation from cardiac magnetic resonance (MR) images in HCM patients and controls, in order to establish superior measures to differentiate HCM from controls. A total of 19 HCM patients and 19 controls underwent cardiac MR scans. The acquired MR images were used to reconstruct 3D LV geometrical models and compute the regional parameters (i.e., wall thickness, curvedness, wall stress, area strain and ejection fraction) based on the standard 16 segment model using our in-house software. HCM patients were further classified into four quartiles based on wall thickness at end diastole (ED) to assess the impact of wall thickness on these regional parameters. There was a significant difference between the HCM patients and controls for all regional parameters (P < 0.001). Wall thickness was greater in HCM patients at the end-diastolic and end-systolic phases, and thickness was most pronounced in segments at the septal regions. A multivariate stepwise selection algorithm identified wall stress index at ED (σi,ED) as the single best independent predictor of HCM (AUC = 0.947). At the cutoff value σi,ED < 1.64, both sensitivity and specificity were 94.7%. This suggests that the end-diastolic wall stress index incorporating regional wall curvature—an index based on mechanical principle—is a sensitive biomarker for HCM diagnosis with potential utility in diagnostic and therapeutic assessment.

Application of Modified Spin-Echo–based Sequences for Hepatic MR Elastography: Evaluation, Comparison with the Conventional Gradient-Echo Sequence, and Preliminary Clinical Experience

Purpose To (a) evaluate modified spin-echo (SE) magnetic resonance (MR) elastographic sequences for acquiring MR images with improved signal-to-noise ratio (SNR) in patients in whom the standard gradient-echo (GRE) MR elastographic sequence yields low hepatic signal intensity and (b) compare the stiffness values obtained with these sequences with those obtained with the conventional GRE sequence. Materials and Methods This HIPAA-compliant retrospective study was approved by the institutional review board; the requirement to obtain informed consent was waived. Data obtained with modified SE and SE echo-planar imaging (EPI) MR elastographic pulse sequences with short echo times were compared with those obtained with the conventional GRE MR elastographic sequence in two patient cohorts, one that exhibited adequate liver signal intensity and one that exhibited low liver signal intensity. Shear stiffness values obtained with the three sequences in 130 patients with successful GRE-based examinations were retrospectively tested for statistical equivalence by using a 5% margin. In 47 patients in whom GRE examinations were considered to have failed because of low SNR, the SNR and confidence level with the SE-based sequences were compared with those with the GRE sequence. Results The results of this study helped confirm the equivalence of SE MR elastography and SE-EPI MR elastography to GRE MR elastography (P = .0212 and P = .0001, respectively). The SE and SE-EPI MR elastographic sequences provided substantially improved SNR and stiffness inversion confidence level in 47 patients in whom GRE MR elastography had failed. Conclusion Modified SE-based MR elastographic sequences provide higher SNR MR elastographic data and reliable stiffness measurements; thus, they enable quantification of stiffness in patients in whom the conventional GRE MR elastographic sequence failed owing to low signal intensity. The equivalence of the three sequences indicates that the current diagnostic thresholds are applicable to SE MR elastographic sequences for assessing liver fibrosis. © RSNA, 2016.

Magnetic resonance imaging of ultrasmall superparamagnetic iron oxide-labeled exosomes from stem cells: a new method to obtain labeled exosomes.

PurposeRecent findings indicate that the beneficial effects of adipose stem cells (ASCs), reported in several neurodegenerative experimental models, could be due to their paracrine activity mediated by the release of exosomes. The aim of this study was the development and validation of an innovative exosome-labeling protocol that allows to visualize them with magnetic resonance imaging (MRI).Materials and methodsAt first, ASCs were labeled using ultrasmall superparamagnetic iron oxide nanoparticles (USPIO, 4–6 nm), and optimal parameters to label ASCs in terms of cell viability, labeling efficiency, iron content, and magnetic resonance (MR) image contrast were investigated. Exosomes were then isolated from labeled ASCs using a standard isolation protocol. The efficiency of exosome labeling was assessed by acquiring MR images in vitro and in vivo as well as by determining their iron content. Transmission electron microscopy images and histological analysis were performed to validate the results obtained.ResultsBy using optimized experimental parameters for ASC labeling (200 µg Fe/mL of USPIO and 72 hours of incubation), it was possible to label 100% of the cells, while their viability remained comparable to unlabeled cells; the detection limit of MR images was of 102 and 2.5×103 ASCs in vitro and in vivo, respectively. Exosomes isolated from previously labeled ASCs retain nanoparticles, as demonstrated by transmission electron microscopy images. The detection limit by MRI was 3 µg and 5 µg of exosomes in vitro and in vivo, respectively.ConclusionWe report a new approach for labeling of exosomes by USPIO that allows detection by MRI while preserving their morphology and physiological characteristics.

Magnetic Resonance Imaging Guided Vacuum Assisted and Core Needle Biopsies.

The purpose of this study to present the results of Magnetic resonance imaging (MRI) guided cutting needle biopsy procedures of suspicious breast lesions that can be solely detected on Magnetic resonance (MR) examination. The study included 48 patients with 48 lesions which were solely be observed in breast MRI, indistinguishable in ultrasonography and mammography, for MR guided vacuum-assisted cutting needle biopsy and 42 patients with 42 lesions for MR guided cutting needle biopsy for the lesions of the same nature. MR imaging was performed using a 1.5-Tesla MRI device. Acquired MR images were determined and biopsy protocol was performed using computer-aided diagnosis system on the workstation. Vacuum biopsies were performed using 10 G or 12 G automatic biopsy systems, cutting needle biopsy procedures were performed using fully automated 12 G biopsy needle. All biopsy procedures were finalized successfully without major complications. The lesions were 54 mass (60%), 28 were non-mass contrast enhancement (31%) and 8 were foci (9%) in the MR examination. Histopathological evaluation revealed 18 malignant (invasive, in-situ ductal carcinoma and lobular carcinoma), 66 benign (apocrine metaplasia, fibrosis, fibroadenomatoid lesion, sclerosing adenosis, fibrocystic disease and mild-to-severe epithelial proliferation) and 6 high-risk (atypical ductal hyperplasia, intraductal papilloma, radial scar) lesions. Magnetic resonance guided vacuum and cutting needle biopsy methods are successful methods fort he evaluation of solely MRI detected suspicious breast lesions. There are several advantages relative to each other in both methods.

Magnetic resonance image denoising using nonlocal maximum likelihood paradigm in DCT-framework

The data acquired by magnetic resonance (MR) imaging system are inherently degraded by noise that has its origin in the thermal Brownian motion of electrons. Denoising can enhance the quality (by improving the SNR) of the acquired MR image, which is important for both visual analysis and other post processing operations. Recent works on maximum likelihood (ML) based denoising shows that ML methods are very effective in denoising MR images and has an edge over the other state-of-the-art methods for MRI denoising. Among the ML based approaches, the Nonlocal maximum likelihood (NLML) method is commonly used. In the conventional NLML method, the samples for the ML estimation of the unknown true pixel are chosen in a nonlocal fashion based on the intensity similarity of the pixel neighborhoods. Euclidean distance is generally used to measure this similarity. It has been recently shown that computing similarity measure is more robust in discrete cosine transform (DCT) subspace, compared with Euclidean image subspace. Motivated by this observation, we integrated DCT into NLML to produce an improved MRI filtration process. Other than improving the SNR, the time complexity of the conventional NLML can also be significantly reduced through the proposed approach. On synthetic MR brain image, an average improvement of 5% in PSNR and 86%reduction in execution time is achieved with a search window size of 91 × 91 after incorporating the improvements in the existing NLML method. On an experimental kiwi fruit image an improvement of 10% in PSNR is achieved. We did experiments on both simulated and real data sets to validate and to demonstrate the effectiveness of the proposed method. © 2015 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 25, 256–264, 2015

Design of a Dedicated Five Degree-of-Freedom Magnetic Resonance Imaging Compatible Robot for Image Guided Prostate Biopsy

In order to improve the current clinical application of magnetic resonance (MR)-guided prostate biopsies, a new, fully magnetic resonance imaging (MRI)-compatible solution has been developed. This solution consists of a five degree-of-freedom (5DOF) pneumatic robot, a programmable logic controller (PLC), and a software application for visualization and robot control. The robot can be freely positioned on the MR table. For the calibration of the robot and MR coordinate system, the robot’s needle guide (NG) is used. The software application supports the calibration with image segmentation and graphic overlays and guides the user through the interventional planning process. After selecting a target point, the application calculates the needed movements via solving the kinematics of the robot and translating the adjustment into commands for the PLC driving the step motors of the robot. In case further adjustments are required, the software also allows for manual control of the robot, to position the NG according to the acquired MR images.

A prototype MR insertable brain PET using tileable GAPD arrays

The aim of this study was to develop a prototype magnetic resonance (MR)-compatible positron emission tomography (PET) that can be inserted into a MR imager and that allows simultaneous PET and MR imaging of the human brain. This paper reports the initial results of the authors' prototype brain PET system operating within a 3-T magnetic resonance imaging (MRI) system using newly developed Geiger-mode avalanche photodiode (GAPD)-based PET detectors, long flexible flat cables, position decoder circuit with high multiplexing ratio, and digital signal processing with field programmable gate array-based analog to digital converter boards. A brain PET with 72 detector modules arranged in a ring was constructed and mounted in a 3-T MRI. Each PET module was composed of cerium-doped lutetium yttrium orthosilicate (LYSO) crystals coupled to a tileable GAPD. The GAPD output charge signals were transferred to preamplifiers using 3 m long flat cables. The LYSO and GAPD were located inside the MR bore and all electronics were positioned outside the MR bore. The PET detector performance was investigated both outside and inside the MRI, and MR image quality was evaluated with and without the PET system. The performance of the PET detector when operated inside the MRI during MR image acquisition showed no significant change in energy resolution and count rates, except for a slight degradation in timing resolution with an increase from 4.2 to 4.6 ns. Simultaneous PET/MR images of a hot-rod and Hoffman brain phantom were acquired in a 3-T MRI. Rods down to a diameter of 3.5 mm were resolved in the hot-rod PET image. The activity distribution patterns between the white and gray matter in the Hoffman brain phantom were well imaged. The hot-rod and Hoffman brain phantoms on the simultaneously acquired MR images obtained with standard sequences were observed without any noticeable artifacts, although MR image quality requires some improvement. These results demonstrate that the simultaneous acquisition of PET and MR images is feasible using the MR insertable PET developed in this study.

A modified Rician LMMSE estimator for the restoration of magnitude MR images

In this article two modified versions of a simple and recently proposed method which is known as linear minimum mean square error (LMMSE) estimator for denoising of magnetic resonance (MR) images are presented. In the introduced approaches the self-similarity and natural redundancy of the acquired MR image are considered to achieve the higher performance of unknown signal estimation. Since, the MR data are in a large majority 3D, the proposed methods are developed to deal with 3D volumes. The introduced methods are compared with related state-of-the-art schemes over both clinical and simulated MR data. The quantitative and qualitative results show their superior denoising ability.

Nonlocal maximum likelihood estimation method for denoising multiple-coil magnetic resonance images

Effective denoising is vital for proper analysis and accurate quantitative measurements from magnetic resonance (MR) images. Even though many methods were proposed to denoise MR images, only few deal with the estimation of true signal from MR images acquired with phased-array coils. If the magnitude data from phased array coils are reconstructed as the root sum of squares, in the absence of noise correlations and subsampling, the data is assumed to follow a non central-χ distribution. However, when the k-space is subsampled to increase the acquisition speed (as in GRAPPA like methods), noise becomes spatially varying. In this note, we propose a method to denoise multiple-coil acquired MR images. Both the non central-χ distribution and the spatially varying nature of the noise is taken into account in the proposed method. Experiments were conducted on both simulated and real data sets to validate and to demonstrate the effectiveness of the proposed method.

Imaging-guided Prostate Biopsy: Conventional and Emerging Techniques

Transrectal ultrasonography (US)-guided biopsy is the standard approach for histopathologic diagnosis of prostate cancer. However, this technique has multiple limitations owing to the operator's inability in most cases to directly visualize and target prostate lesions. Magnetic resonance (MR) imaging of the prostate overcomes many of these limitations by directly depicting areas of abnormality and allowing targeted biopsies. Accuracy in the detection of prostate cancer is improved by the combined use of standard T2-weighted MR imaging and advanced MR imaging techniques such as diffusion-weighted imaging, dynamic contrast-enhanced imaging, and MR spectroscopy. Suspicious-appearing regions of the prostate seen on MR images can be targeted at real-time transrectal US-guided biopsy to improve the diagnostic yield. MR imaging also can be performed for real-time guidance of transrectal prostate biopsy. Studies among patients who underwent at least one transrectal US-guided biopsy with a negative result before undergoing an MR imaging-guided biopsy showed improved detection rates with MR imaging-guided biopsy in comparison with the detection rates achieved with a repeat transrectal US-guided biopsy; however, MR imaging-guided biopsy is a more time-consuming procedure. A technique known as fused MR imaging- and transrectal US-guided biopsy, which relies on the coregistration of previously acquired MR images with real-time transrectal US images acquired during the procedure, shows promise but is limited by deformation of the prostate; this limitation is the subject of ongoing investigation. Another technique that is currently under investigation, MR imaging-guided prostate biopsy with robotic assistance, may one day help improve the accuracy of biopsy needle placement.

DESCRIPTION AND VALIDATION OF A MAGNETIC RESONANCE IMAGING‐GUIDED STEREOTACTIC BRAIN BIOPSY DEVICE IN THE DOG

A stereotactic brain biopsy system that is magnetic resonance (MR) imaging-guided has not been validated in dogs. Our purpose was to determine the mean needle placement error in the caudate nucleus, thalamus, and midbrain of a canine cadaver brain using the modified Brainsight stereotactic system. Relocatable reference markers (fiducial markers) were attached to the cadaver head using a dental bite block. A T1-weighted gradient echo three-dimensional (3D) sequence was acquired using set parameters. Fiducial markers were used to register the head to the acquired MR images in reference to a 3D position sensor. This allowed the planning of trajectory path to brain targets in real time. Coordinates (X, Y, Z) were established for each target and 0.5 microl of diluted gadolinium was injected at each target using a 26-gauge needle to create a lesion. The center of the gadolinium deposition was identified on the postoperative MR images and coordinates (X', Y', Z') were established. The precision of this system in bringing the needle to target (needle placement error) was calculated. Seventeen sites were targeted in the brain. The mean needle placement error for all target sites was 1.79 +/- 0.87 mm. The upper bound of error for this stereotactic system was 3.31 mm. There was no statistically significant relationship between needle placement error and target depth (P = 0.23). The ease of use and precision of this stereotactic system support its development for clinical use in dogs with brain lesions > 3.31 mm.

Connectivity of the human pedunculopontine nucleus region and diffusion tensor imaging in surgical targeting

The pedunculopontine nucleus (PPN) region of the brainstem has become a new stimulation target for the treatment of gait freezing, akinesia, and postural instability in advanced Parkinson disease (PD). Because PD locomotor symptoms are probably caused by excessive gamma-aminobutyric acidergic inhibition of the PPN, low-frequency stimulation of the PPN may overcome this inhibition and improve the symptoms. However, the anatomical connections of this region in humans are not known in any detail. Diffusion weighted magnetic resonance (MR) images were acquired at 1.5 teslas, and probabilistic tractography was used to trace the connections of the PPN region in eight healthy volunteers. A single seed voxel (2 x 2 x 2 mm) was chosen in the PPN just lateral to the decussation of the superior cerebellar peduncle, and the Diffusion Toolbox of the Oxford Centre for Functional Magnetic Resonance Imaging of the Brain was used to process the acquired MR images. The connections of each volunteer's PPN region were analyzed using a human brain MR imaging atlas. The PPN region was connected with the cerebellum and spinal cord below and to the thalamus, pallidum, subthalamic nucleus, and motor cortex above. The regions of the primary motor cortex that control the trunk and upper and lower extremities had the highest connectivity compared with other parts of motor cortex. These findings suggest that connections of the PPN region with the primary motor cortex, basal ganglia, thalamus, cerebellum, and spinal cord may play important roles in the regulation of movement by the PPN region. Diffusion tensor imaging tractography of the PPN region may be used preoperatively to optimize placement of stimulation electrodes and postoperatively it may also be useful to reassess electrode positions.

Automated detection of prostatic adenocarcinoma from high-resolution ex vivo MRI

Prostatic adenocarcinoma is the most commonly occurring cancer among men in the United States, second only to skin cancer. Currently, the only definitive method to ascertain the presence of prostatic cancer is by trans-rectal ultrasound (TRUS) directed biopsy. Owing to the poor image quality of ultrasound, the accuracy of TRUS is only 20%-25%. High-resolution magnetic resonance imaging (MRI) has been shown to have a higher accuracy of prostate cancer detection compared to ultrasound. Consequently, several researchers have been exploring the use of high resolution MRI in performing prostate biopsies. Visual detection of prostate cancer, however, continues to be difficult owing to its apparent lack of shape, and the fact that several malignant and benign structures have overlapping intensity and texture characteristics. In this paper, we present a fully automated computer-aided detection (CAD) system for detecting prostatic adenocarcinoma from 4 Tesla ex vivo magnetic resonance (MR) imagery of the prostate. After the acquired MR images have been corrected for background inhomogeneity and nonstandardness, novel three-dimensional (3-D) texture features are extracted from the 3-D MRI scene. A Bayesian classifier then assigns each image voxel a "likelihood" of malignancy for each feature independently. The "likelihood" images generated in this fashion are then combined using an optimally weighted feature combination scheme. Quantitative evaluation was performed by comparing the CAD results with the manually ascertained ground truth for the tumor on the MRI. The tumor labels on the MR slices were determined manually by an expert by visually registering the MR slices with the corresponding regions on the histology slices. We evaluated our CAD system on a total of 33 two-dimensional (2-D) MR slices from five different 3-D MR prostate studies. Five slices from two different glands were used for training. Our feature combination scheme was found to outperform the individual texture features, and also other popularly used feature combination methods, including AdaBoost, ensemble averaging, and majority voting. Further, in several instances our CAD system performed better than the experts in terms of accuracy, the expert segmentations being determined solely from visual inspection of the MRI data. In addition, the intrasystem variability (changes in CAD accuracy with changes in values of system parameters) was significantly lower than the corresponding intraobserver and interobserver variability. CAD performance was found to be very similar for different training sets. Future work will focus on extending the methodology to guide high-resolution MRI-assisted in vivo prostate biopsies.

Image-matching as a medical diagnostic support tool (DST) for brain diseases in children

Imaging-matching is an important research area in imaging informatics. We have developed and evaluated a novel diagnostic support tool (DST) based on medical image matching using MR brain images. The approach consists of two steps, database generation and image matching. The database contains pre-diagnosed MR brain images. As the images are added to the database, they are registered to the 3D Talairach coordinate system. In addition, regions of interests (ROI) are generated, and image-processing techniques are used to extract relevant image parameters related to the brain and diseases from the ROIs and from the entire MR image. The second step is to retrieve relevant information from the database by performing image matching. In this step, the physician first submits a query image. The DST computes the similarity between the query image and each of the images in the database, and then presents the most similar images to the user. Since the database contains pre-diagnosed images, the retrieved cases tend to contain relevant diagnostic information. To evaluate the usefulness of the DST in a clinical setting, pediatric brain diseases were used. The database contains 2500 pediatric patients between ages 0 and 18 with brain Magnetic Resonance (MR) images of known brain lesions. A testbed was established at the Childrens Hospital Los Angeles (CHLA) for acquiring MR images from the PACS server of patients with known lesions. These images were matched against those in the DST pediatric brain MR database. An expert pediatric neuroradiologist evaluated the matched results. We found that in most cases, the image-matching method was able to quickly retrieve images with relevant diagnostic content. The evaluation method and results are given.

Three-Dimensional Velocity Field Reconstruction

The problem of inter-slice magnetic resonance (MR) image reconstruction is encountered often in medical imaging applications, in such scenarios, there is a need to approximate information not captured in contiguously acquired MR images due to hardware sampling limitations. In the context of velocity field reconstruction, these data are required for visualization and computational analyses of flow fields to be effective. To provide more complete velocity information, a method has been developed for the reconstruction of flow fields based on adaptive control grid interpolation (ACGI). In this study, data for reconstruction were acquired via MRJ from in vitro models of surgically corrected pediatric cardiac vasculatures. Reconstructed velocity fields showed strong qualitative agreement with those obtained via other acquisition techniques. Quantitatively reconstruction was shown to produce data of comparable quality to accepted velocity data acquisition methods. Results indicate that ACGI-based velocity field reconstruction is capable of producing information suitable for a variety of applications demanding three-dimensional in vivo velocity data.