3D Magnetic Resonance Imaging Datasets Research Articles

EVOLVE: Learning volume-adaptive phases for fast 3D magnetic resonance scan and image reconstruction

Compared with 2D Magnetic Resonance Imaging (MRI), 3D MRI is more powerful for generating high resolution images and visualizing small anatomical structures. However, 3D MRI acquisition is much more time-consuming due to the significantly larger number of phase encoding steps, which is directly proportional to the acquisition time. This paper proposes to select a volume-adaptive small subset of phases to accelerate 3D MRI scans and accurately reconstruct 3D images from the corresponding undersampled 3D k-space data. To avoid the delays caused by computationally expensive yet high-performance volume-adaptive phase selection, we propose a strategy of selecting multiple phases based on sampled slices from the volume during idle time within the repetition time (TR). To enhance the performance of phase selection, we propose a novel three-directional cross-attention phase selection network. Additionally, to improve the reconstruction performance, we introduce a three-directional slice-wise volume reconstruction. To the best of our knowledge, the proposed method, which we called EVOLVE (learning volume-adaptive phases), is the first work that learns volume-adaptive phases for fast 3D MRI. The extensive experimental results on a large-scale 3D MRI dataset at various acceleration factors demonstrate the substantial performance improvement in terms of image reconstruction achieved by using the EVOLVE method for phase selection compared to traditional learning free 3D MRI phase selection methods.

Fet-Net Algorithm for Automatic Detection of Fetal Orientation in Fetal MRI

Identifying fetal orientation is essential for determining the mode of delivery and for sequence planning in fetal magnetic resonance imaging (MRI). This manuscript describes a deep learning algorithm named Fet-Net, composed of convolutional neural networks (CNNs), which allows for the automatic detection of fetal orientation from a two-dimensional (2D) MRI slice. The architecture consists of four convolutional layers, which feed into a simple artificial neural network. Compared with eleven other prominent CNNs (different versions of ResNet, VGG, Xception, and Inception), Fet-Net has fewer architectural layers and parameters. From 144 3D MRI datasets indicative of vertex, breech, oblique and transverse fetal orientations, 6120 2D MRI slices were extracted to train, validate and test Fet-Net. Despite its simpler architecture, Fet-Net demonstrated an average accuracy and F1 score of 97.68% and a loss of 0.06828 on the 6120 2D MRI slices during a 5-fold cross-validation experiment. This architecture outperformed all eleven prominent architectures (p < 0.05). An ablation study proved each component's statistical significance and contribution to Fet-Net's performance. Fet-Net demonstrated robustness in classification accuracy even when noise was introduced to the images, outperforming eight of the 11 prominent architectures. Fet-Net's ability to automatically detect fetal orientation can profoundly decrease the time required for fetal MRI acquisition.

CAN3D: Fast 3D medical image segmentation via compact context aggregation.

Direct automatic segmentation of objects in 3D medical imaging, such as magnetic resonance (MR) imaging, is challenging as it often involves accurately identifying multiple individual structures with complex geometries within a large volume under investigation. Most deep learning approaches address these challenges by enhancing their learning capability through a substantial increase in trainable parameters within their models. An increased model complexity will incur high computational costs and large memory requirements unsuitable for real-time implementation on standard clinical workstations, as clinical imaging systems typically have low-end computer hardware with limited memory and CPU resources only. This paper presents a compact convolutional neural network (CAN3D) designed specifically for clinical workstations and allows the segmentation of large 3D Magnetic Resonance (MR) images in real-time. The proposed CAN3D has a shallow memory footprint to reduce the number of model parameters and computer memory required for state-of-the-art performance and maintain data integrity by directly processing large full-size 3D image input volumes with no patches required. The proposed architecture significantly reduces computational costs, especially for inference using the CPU. We also develop a novel loss function with extra shape constraints to improve segmentation accuracy for imbalanced classes in 3D MR images. Compared to state-of-the-art approaches (U-Net3D, improved U-Net3D and V-Net), CAN3D reduced the number of parameters up to two orders of magnitude and achieved much faster inference, up to 5 times when predicting with a standard commercial CPU (instead of GPU). For the open-access OAI-ZIB knee MR dataset, in comparison with manual segmentation, CAN3D achieved Dice coefficient values of (mean = 0.87±0.02 and 0.85±0.04) with mean surface distance errors (mean = 0.36±0.32mm and 0.29±0.10mm) for imbalanced classes such as (femoral and tibial) cartilage volumes respectively when training volume-wise under only 12G video memory. Similarly, CAN3D demonstrated high accuracy and efficiency on a pelvis 3D MR imaging dataset for prostate cancer consisting of 211 examinations with expert manual semantic labels (bladder, body, bone, rectum, prostate) now released publicly for scientific use as part of this work.

Multi-mask self-supervised learning for physics-guided neural networks in highly accelerated magnetic resonance imaging.

Self-supervised learning has shown great promise because of its ability to train deep learning (DL) magnetic resonance imaging (MRI) reconstruction methods without fully sampled data. Current self-supervised learning methods for physics-guided reconstruction networks split acquired undersampled data into two disjoint sets, where one is used for data consistency (DC) in the unrolled network, while the other is used to define the training loss. In this study, we propose an improved self-supervised learning strategy that more efficiently uses the acquired data to train a physics-guided reconstruction network without a database of fully sampled data. The proposed multi-mask self-supervised learning via data undersampling (SSDU) applies a holdout masking operation on the acquired measurements to split them into multiple pairs of disjoint sets for each training sample, while using one of these pairs for DC units and the other for defining loss, thereby more efficiently using the undersampled data. Multi-mask SSDU is applied on fully sampled 3D knee and prospectively undersampled 3D brain MRI datasets, for various acceleration rates and patterns, and compared with the parallel imaging method, CG-SENSE, and single-mask SSDU DL-MRI, as well as supervised DL-MRI when fully sampled data are available. The results on knee MRI show that the proposed multi-mask SSDU outperforms SSDU and performs as well as supervised DL-MRI. A clinical reader study further ranks the multi-mask SSDU higher than supervised DL-MRI in terms of signal-to-noise ratio and aliasing artifacts. Results on brain MRI show that multi-mask SSDU achieves better reconstruction quality compared with SSDU. The reader study demonstrates that multi-mask SSDU at R = 8 significantly improves reconstruction compared with single-mask SSDU at R = 8, as well as CG-SENSE at R = 2.

Future activity prediction of multiple sclerosis with 3D MRI using 3D discrete wavelet transform

Multiple Sclerosis (MS) is a chronic and autoimmune neurological disease that is frequently seen especially in young people. MS lesions that can be seen with magnetic resonance imaging (MRI) findings are important biomarkers that provide information about the clinical prognosis and activity of the disease. The presence of new MS lesions is associated with future disease activity. This study aims to predict the future activity of MS using the 3D discrete wavelet transform (DWT) as a feature extraction method from 3D MRI. The 3D-DWT can be used as it provides spatial and spectral location features of MS lesions without losing their relationship between MRI slices. Ten different wavelet families of DWT are used individually, each of them is classified by six machine learning algorithms, and their feature extraction performances are compared. The highest F1-score, Precision, and Recall of 95.0% are obtained by the support vector machine algorithm on the SYM4, SYM8, and Haar wavelet families in the 3D MRI dataset consisting of 40 patients based on 5-fold cross validation. The results show that the 3D-DWT method is an effective method for feature extraction in predicting the future activity of MS.

Automated Brain Metastases Detection Framework for T1-Weighted Contrast-Enhanced 3D MRI.

Brain Metastases (BM) complicate 20-40% of cancer cases. BM lesions can present as punctate (1mm) foci, requiring high-precision Magnetic Resonance Imaging (MRI) in order to prevent inadequate or delayed BM treatment. However, BM lesion detection remains challenging partly due to their structural similarities to normal structures (e.g., vasculature).We propose a BM-detection framework using a single-sequence gadolinium-enhanced T1-weighted 3D MRI dataset. The framework focuses on the detection of smaller (<15mm) BM lesions and consists of: (1) candidate-selection stage, using Laplacian of Gaussian approach for highlighting parts of an MRI volume holding higher BM occurrence probabilities, and (2) detection stage that iteratively processes cropped region-of-interest volumes centered by candidates using a custom-built 3D convolutional neural network ("CropNet"). Data is augmented extensively during training via a pipeline consisting of random gamma correction and elastic deformation stages; the framework thereby maintains its invariance for a plausible range of BM shape and intensity representations.This approach is tested using five-fold cross-validation on 217 datasets from 158 patients, with training and testing groups randomized per patient to eliminate learning bias. The BM database included lesions with a mean diameter of ∼5.4mm and a mean volume of ∼160mm3. For 90% BM-detection sensitivity, the framework produced on average 9.12 false-positive BM detections per patient (standard deviation of 3.49); for 85% sensitivity, the average number of false-positives declined to 5.85. Comparative analysis showed that the framework produces comparable BM-detection accuracy with the state-of-art approaches validated for significantly larger lesions.

ADAPTIVE REGION GROWING FOR SKULL, BRAIN, AND SCALP SEGMENTATION FROM 3D MRI

Many of the diseases that are related to the nervous system are serious and can lead to pain, headaches, and many other symptoms. Magnetic resonance imaging (MRI), which produces detailed medical images for the diagnosis of nervous system-related diseases, offers an advanced approach to the study of these diseases. In this paper, a method for the automatic segmentation of the skull, scalp, and brain of 3D MRI imagery is proposed to help doctors diagnose numerous injuries and diseases. For the proposed method, the 3D MRI dataset was first transformed to 2D axial slices, which were grouped based on the skull shape. Second, the histogram method was used to compute the fitting rectangle boundary around the object. Third, the segmentation of the skull was performed using the proposed adaptive region growing method, depending on the slice groups. Finally, the brain and the scalp were segmented. The proposed method was tested and validated on 20 subjects from the BrainWeb database and seven adults from the Neurodevelopmental MRI database with the use of the tissues for the ground truth data. The performance of the proposed method was evaluated using the Dice percentage and a comparison with the other methods.

Technical Note: Real-time 3D MRI in the presence of motion for MRI-guided radiotherapy: 3D Dynamic keyhole imaging with super-resolution.

The purpose of this study was to present real-time three-dimensional (3D) magnetic resonance imaging (MRI) in the presence of motion for MRI-guided radiotherapy (MRgRT) using dynamic keyhole imaging for high-temporal acquisition and super-resolution generative (SRG) model for high-spatial reconstruction. We propose a unique real-time 3D MRI technique by combining a data sharing technique (3D dynamic keyhole imaging) with a SRG model using cascaded deep learning technique. 3D dynamic keyhole imaging utilizes the data sharing mechanism by combining keyhole central k-space data acquired in real-time with high-spatial, high-temporal resolution prior peripheral k-space data at various motion positions prepared by the SRG model. The efficacy of the 3D dynamic keyhole imaging with super-resolution (SR_dKeyhole) was compared to the ground-truth super-resolution images with the original full k-space data. It was also compared with the zero-filling reconstruction (zero-filling), conventional keyhole reconstruction with low-spatial high-temporal prior data (LR_cKeyhole), and conventional keyhole reconstruction with super-resolution prior data (SR_cKeyhole). High-spatial, high-temporal resolution 3D MRI datasets (1.5×1.5×6mm3 ) were generated from low-spatial, high-temporal resolution 3D MRI datasets (6×6×6mm3 ) using the cascaded deep learning SRG framework (<100ms/volume). 3D dynamic keyhole imaging with the SRG model provided high-spatial, high-temporal resolution images (1.5×1.5×6mm3 , 455ms) with the highest similarity to the ground-truth SR images without any noticeable artifacts. Structural similarity indices (SSIM) of the reconstructed 3D MRI to the original SR 3D MRI were 0.65, 0.66, 0.86, and 0.89 for zero-filling, LR_cKeyhole, SR_cKeyhole, and SR_dKeyhole, respectively (1 for identical image). In addition, average value of image relative error (IRE) of the reconstructed 3D MRI to the original SR 3D MRI were 0.169, 0.191, 0.079, and 0.067 for zero-filling, LR_cKeyhole, SR_cKeyhole, and SR_dKeyhole, respectively (0 for identical image). We demonstrated that high-spatial, high-temporal resolution 3D MRI was feasible by combing 3D dynamic keyhole imaging with a SRG model in terms of image quality and imaging time. The proposed technique can be utilized for real-time 3D MRgRT.

Electrode placement accuracy in robot-assisted epilepsy surgery: A comparison of different referencing techniques including frame-based CT versus facial laser scan based on CT or MRI

BackgroundPrecise robotic or stereotactic implantation of stereoelectroencephalography (sEEG) electrodes relies on the exact referencing of the planning images in order to match the patient's anatomy to the stereotactic device or robot. We compared the accuracy of sEEG electrode implantation with stereotactic frame versus laser scanning of the face based on computed tomography (CT) or magnetic resonance imaging (MRI) datasets for referencing. MethodsThe accuracy was determined by calculating the Euclidian distance between the planned trajectory and the postoperative position of the sEEG electrode, defining the entry point error (EPE) and the target point error (TPE). The sEEG electrodes (n = 171) were implanted with the robotic surgery assistant (ROSA) in 19 patients. Preoperative trajectory planning was performed on three-dimensional (3D) MRI datasets. Referencing was accomplished either by performing (A) 1.25-mm slice CT with the patient's head fixed in a Leksell stereotactic frame (CT-frame, n = 49), fused with a 3D-T1-weighted, contrast enhanced- and T2-weighted 1.5 Tesla (T) MRI; (B) 1.25 mm CT (CT-laser, n = 60), fused with 3D-3.0-T MRI; (C) 3.0-T MRI T1-based laser scan (3.0-T MRI-laser, n = 56) or (D) in one single patient, because of a pacemaker, 3D-1.5-T MRI T1-based laser scan (1.5-T MRI-laser, n = 6). ResultsIn (A) CT-frame referencing, the mean EPE amounted to 0.86 mm and the mean TPE amounted to 2.28 mm (n = 49). In (B) CT-laser referencing, the EPE amounted to 1.85 mm and the TPE to 2.41 mm (n = 60). In (C) 3.0-T MRI-laser referencing, the mean EPE amounted to 3.02 mm and the mean TPE to 3.51 mm (n = 56). In (D) 1.5-T MRI, surprisingly the mean EPE amounted only to 0.97 mm and the TPE to 1.71 mm (n = 6). In 3 cases using CT-laser and 1 case using 3.0 T MRI-laser for referencing, small asymptomatic intracerebral hemorrhages were detected. No further complications were observed. ConclusionRobot-guided sEEG electrode implantation using CT-frame referencing and CT-laser-based referencing is most accurate and can serve for high precision placement of electrodes. In contrast, 3.0-T MRI-laser-based referencing is less accurate, but saves radiation. Most trajectories can be reached if alternative routes over less vascularized brain areas are used.This article is part of the Special Issue "Individualized Epilepsy Management: Medicines, Surgery and Beyond".

Comparison of MRI Under-Sampling Techniques for Compressed Sensing with Translation Invariant Wavelets Using FastTestCS: A Flexible Simulation Tool

A sparsifying transform for use in Compressed Sensing (CS) is a vital piece of image reconstruction for Magnetic Resonance Imaging (MRI). Previously, Translation Invariant Wavelet Transforms (TIWT) have been shown to perform exceedingly well in CS by reducing repetitive line pattern image artifacts that may be observed when using orthogonal wavelets. To further establish its validity as a good sparsifying transform, the TIWT is comprehensively investigated and compared with Total Variation (TV), using six under-sampling patterns through simulation. Both trajectory and random mask based under-sampling of MRI data are reconstructed to demonstrate a comprehensive coverage of tests. Notably, the TIWT in CS reconstruction performs well for all varieties of under-sampling patterns tested, even for cases where TV does not improve the mean squared error. This improved Image Quality (IQ) gives confidence in applying this transform to more CS applications which will contribute to an even greater speed-up of a CS MRI scan. High vs low resolution time of flight MRI CS re-constructions are also analyzed showing how partial Fourier acquisitions must be carefully addressed in CS to prevent loss of IQ. In the spirit of reproducible research, novel software is introduced here as FastTestCS. It is a helpful tool to quickly develop and perform tests with many CS customizations. Easy integration and testing for the TIWT and TV minimization are exemplified. Simulations of 3D MRI datasets are shown to be efficiently distributed as a scalable solution for large studies. Comparisons in reconstruction computation time are made between the Wavelab toolbox and Gnu Scientific Library in FastTestCS that show a significant time savings factor of 60×. The addition of FastTestCS is proven to be a fast, flexible, portable and reproducible simulation aid for CS research.

Benchmark for Algorithms Segmenting the Left Atrium From 3D CT and MRI Datasets.

Knowledge of left atrial (LA) anatomy is important for atrial fibrillation ablation guidance, fibrosis quantification and biophysical modelling. Segmentation of the LA from Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) images is a complex problem. This manuscript presents a benchmark to evaluate algorithms that address LA segmentation. The datasets, ground truth and evaluation code have been made publicly available through the http://www.cardiacatlas.org website. This manuscript also reports the results of the Left Atrial Segmentation Challenge (LASC) carried out at the STACOM'13 workshop, in conjunction with MICCAI'13. Thirty CT and 30 MRI datasets were provided to participants for segmentation. Each participant segmented the LA including a short part of the LA appendage trunk and proximal sections of the pulmonary veins (PVs). We present results for nine algorithms for CT and eight algorithms for MRI. Results showed that methodologies combining statistical models with region growing approaches were the most appropriate to handle the proposed task. The ground truth and automatic segmentations were standardised to reduce the influence of inconsistently defined regions (e.g., mitral plane, PVs end points, LA appendage). This standardisation framework, which is a contribution of this work, can be used to label and further analyse anatomical regions of the LA. By performing the standardisation directly on the left atrial surface, we can process multiple input data, including meshes exported from different electroanatomical mapping systems.

Nonlocal denoising using anisotropic structure tensor for 3D MRI

Noise in magnetic resonance imaging (MRI) data is widely recognized to be harmful to image processing and subsequent quantitative analysis. To ameliorate the effects of image noise, the authors present a structure-tensor based nonlocal mean (NLM) denoising technique that can effectively reduce noise in MRI data and improve tissue characterization. The proposed 3D NLM algorithm uses a structure tensor to characterize information around tissue boundaries. The similarity weight of a pixel (or patch), which determines its contribution to the denoising process, is determined by the intensity and structure tensor simultaneously. Meanwhile, similarity of structure tensors is computed using an affine-invariant Riemannian metrics, which compares tensor properties more comprehensively and avoids orientation inaccuracy of structure subsequently. The proposed method is further extended for denoising high dimensional MRI data such as diffusion weighted MRI. It is also extended to handle Rician noise corruption so that denoising effects are further enhanced. The proposed method was implemented in both simulated datasets and multiply modalities of real 3D MRI datasets. Comparisons with related state-of-the-art algorithms demonstrated that this method improves denoising performance qualitatively and quantitatively. In this paper, high order structure information of 3D MRI was characterized by 3D structure tensor and compared for NLM denoising in a Riemannian space. Experiments with simulated and real human MRI data demonstrate a great potential of the proposed technique for routine clinical use.

Normal ascent of the conus medullaris: a post-mortem foetal MRI study

Objectives: The position of the conus medullaris is considered abnormal if it ends below lumbar vertebrae three (L3) at birth. We used magnetic resonance imaging (MRI) to measure the position of the conus in post-mortem foetuses, to investigate the timing of normal ascent.Methods: The position of the conus in 84 post-mortem foetuses (mean gestation 26.3 weeks; range 14–41 weeks) was identified using 3D MRI datasets. A numerical scale was used for vertebral levels, from 1 (S2) to 15 (T12).Results: There was significant ascent of the conus medullaris with increasing gestation. At 20 weeks gestation, an estimated 84.2% (95% confidence interval (CI): 72.9, 93.2%) of foetuses have a conus position of L4/5 or higher, but only 22.8% (95% CI 11.7, 34.9%) at L3 or higher. By 26 weeks, an estimated 50.7% (95% CI 34.1, 67.5) will have reached L3, and 94.8% (95% CI 87.0, 98.5%) reach L3 by 40 weeks.Conclusion: There is regular ascent of the conus throughout foetal life. Although growth for each individual foetus may be non-linear, most foetuses have a conus level within the normal adult range by 33 weeks gestation.

Virtual bronchoscopy in the fetus

The term ‘virtual bronchoscopy’ describes the creation and evaluation of representations of the bronchial tree and surrounding structures using spatial information derived from imaging sources other than the bronchoscope itself. Initially, this referred to two-dimensional and later three-dimensional (3D) bronchial tree representations generated from computed tomography images. However, virtual bronchoscopy can now also be performed using data from magnetic resonance imaging (MRI). Several software options are available that can be used to generate accurate and realistic renderings of the bronchial tree 1–3 . Advances in imaging technology have led to vast improvements in fetal evaluation. Ultrasound examination is the primary method of fetal assessment because it is patient-friendly, effective, cost-efficient and considered to be safe 4 . MRI is generally used when ultrasound cannot provide sufficiently high-quality images. It offers high-resolution fetal imaging with excellent contrast that allows visualization of internal tissues 5 .I n this article we describe virtual bronchoscopy performed in a normal fetus that underwent MRI at 28 weeks’ gestation due to placenta previa with suspicion of placenta accreta. MRI examination was performed using a 1.5-T scanner (Siemens, Erlangen, Germany). The protocol involved a T2-weighted sequence in the three planes of the fetal body (HASTE; repetition time (TR), shortest; echo time (TE), 140 ms; field of view, 300–200 mm; matrix, 256 × 256; slice thickness, 4 mm; acquisition time, 17 s; 40 slices). In addition, we applied a 3D, T2-weighted TrueFISP sequence in the sagittal plane (TRUFI; TR/TE = 3.02/1.34; voxel size 1.6 × 1.6 × 1. 6m m 3 ; flip angle = 70 ◦ ; parallel acquisition techniques 2; acquisition time, 0.26 s) (Figure 1). The entire examination time did not exceed 20 min. The 3D MRI datasets were reconstructed to produce an interactive surface model of the fetal respiratory tract that could be viewed from any angle. Firstly, a 3D model of the fetal airway was created from the overlapping image layers generated by MRI using the software Mimics (Materialise, Leuven, Belgium), which allowed the surface of the airway to be delineated using contrast detection in the relevant areas of interest. The 3D model that was generated was exported using the standard triangular language file format and then converted into an OBJ file using the MAYA 3D modeling software (Autodesk, San Rafael, CA, USA) 6,7 . This program allows the virtual positioning of observation cameras while working with multiple on-screen windows. After the development of the 3D model, the software allows the user to determine the best positioned viewpoints for visualization of the 3D model (Figure 2) and also facilitates the adjustment of lighting parameters to improve contrast resolution. Using the navigation mode it is also possible to perform virtual bronchoscopy to visualize the upper respiratory tract from the pharynx downwards through the tracheobronchial tree (Videoclip S1) with a quality similar to that which could be obtained by videotaped bronchoscopy. We have demonstrated that MRI data can be used to create a 3D model of the respiratory tract in a normal fetus. We believe that this technique could become a useful tool for the assessment of fetal airway patency.

COREGISTRATION OF DIGITAL PHOTOGRAPHY OF THE HUMAN CORTEX AND CRANIAL MAGNETIC RESONANCE IMAGING FOR VISUALIZATION OF SUBDURAL ELECTRODES IN EPILEPSY SURGERY

To develop a method for the coregistration of digital photographs of the human cortex with head magnetic resonance imaging (MRI) scans for invasive diagnostics and resective neocortical epilepsy surgery. Six chronically epileptic patients (two women, four men; mean age, 34 yr; age range, 20-43 yr) underwent preoperative three-dimensional (3D) T1-weighted MRI scans. Digital photographs of the exposed cortex were taken during implantation of subdural grid electrodes. Rendering software (Analyze 3.1; Biomedical Imaging Resource, Mayo Foundation, Rochester, MN) was used to create an MRI-based 3D model of the brain surface. Digital photographs were manually coregistered with the brain surface MRI model using the registration tool in the Analyze software. By matching the digital photograph and the brain surface model, the position of the subdural electrodes was integrated into the coordinate system of the preoperatively acquired 3D MRI dataset. In all patients, the position of the labeled electrode contacts in relation to the cortical anatomy could be visualized on the 3D models of the cortical surface. At the time of resection, the resulting image of the coregistration process provides a realistic view of the cortex and the position of the subdural electrode. The coregistration of digital photographs of the brain cortex with the results of 3D MRI data sets is possible. This allows for identification of anatomic details underlying the subdural grid electrodes and enhances the orientation of the surgeon.

Cardiovascular MRI: A Literature Review

Substrate-based catheter ablation of post-myocardial infarction ventricular tachycardia requires delineation of the scarred myocardium, typically using an electroanatomic mapping (EAM) system. Cardiac magnetic resonance imaging (MRI) might facilitate this procedure by localizing this myocardial scar. In a series of in vitro and in vivo experiments, the authors evaluated the feasibility of integrating 3D MRI and EAM data to guide real-time left ventricular (LV) catheter manipulation. A custom program was employed to integrate 3D MRI datasets with real-time EAM. Initially, a plastic model of the LV was used to determine the optimal alignment/registration strategy. To determine the in vivo accuracy of the registration process, ablation lesions were directed at iatrogenic MRI-visible ‘‘targets’’ (iron oxide injections) within normal porcine LVs. Finally, this image integration strategy was assessed in a porcine infarction model by targeting ablation lesions to the scar border. Results: The in vitro experiments revealed that registration of the LV alone results in inaccurate alignment due primarily to rotation along the chamber’s long axis. Inclusion of the aorta in the registration process rectified this error. In the iron oxide injection experiments, the ablation lesions were 1.8 ± 0.5 from the targets. In the porcine infarct model, the catheter could be reliably navigated to the mitral valve annulus, and the ablation lesions were uniformly situated at the scar borders.

Atlas-based automatic segmentation of MR images: Validation study on the brainstem in radiotherapy context

Brain tumor radiotherapy requires the volume measurements and the localization of several individual brain structures. Any tool that can assist the physician to perform the delineation would then be of great help. Among segmentation methods, those that are atlas-based are appealing because they are able to segment several structures simultaneously, while preserving the anatomy topology. This study aims to evaluate such a method in a clinical context. The brain atlas is made of two three-dimensional (3D) volumes: the first is an artificial 3D magnetic resonance imaging (MRI); the second consists of the segmented structures in this artificial MRI. The elastic registration of the artificial 3D MRI against a patient 3D MRI dataset yields an elastic transformation that can be applied to the labeled image. The elastic transformation is obtained by minimizing the sum of the square differences of the image intensities and derived from the optical flow principle. This automatic delineation (AD) enables the mapping of the segmented structures onto the patient MRI. Parameters of the AD have been optimized on a set of 20 patients. Results are obtained on a series of 6 patients' MRI. A comprehensive validation of the AD has been conducted on performance of atlas-based segmentation in a clinical context with volume, position, sensitivity, and specificity that are compared by a panel of seven experimented physicians for the brain tumor treatments. Expert interobserver volume variability ranged from 16.70 cm(3) to 41.26 cm(3). For patients, the ratio of minimal to maximal volume ranged from 48% to 70%. Median volume varied from 19.47 cm(3) to 27.66 cm(3) and volume of the brainstem calculated by AD varied from 17.75 cm(3) to 24.54 cm(3). Medians of experts ranged, respectively, for sensitivity and specificity, from 0.75 to 0.98 and from 0.85 to 0.99. Median of AD were, respectively, 0.77 and 0.97. Mean of experts ranged, respectively, from 0.78 to 0.97 and from 0.86 to 0.99. Mean of AD were, respectively, 0.76 and 0.97. Results demonstrate that the method is repeatable, provides a good trade-off between accuracy and robustness, and leads to reproducible segmentation and labeling. These results can be improved by enriching the atlas with the rough information of tumor or by using different laws of deformation for the different structures. Qualitative results also suggest that this method can be used for automatic segmentation of other organs such as neck, thorax, abdomen, pelvis, and limbs.

Integration of cardiac magnetic resonance imaging with three-dimensional electroanatomic mapping to guide left ventricular catheter manipulation: Feasibility in a porcine model of healed myocardial infarction

In a series of in vitro and in vivo experiments, we evaluated the feasibility of integrating three-dimensional (3D) magnetic resonance imaging (MRI) and electroanatomic mapping (EAM) data to guide real-time left ventricular (LV) catheter manipulation. Substrate-based catheter ablation of post-myocardial infarction ventricular tachycardia requires delineation of the scarred myocardium, typically using an EAM system. Cardiac MRI might facilitate this procedure by localizing this myocardial scar. A custom program was employed to integrate 3D MRI datasets with real-time EAM. Initially, a plastic model of the LV was used to determine the optimal alignment/registration strategy. To determine the in vivo accuracy of the registration process, ablation lesions were directed at iatrogenic MRI-visible "targets" (iron oxide injections) within normal porcine LVs (n = 5). Finally, this image integration strategy was assessed in a porcine infarction model (n = 6) by targeting ablation lesions to the scar border. The in vitro experiments revealed that registration of the LV alone results in inaccurate alignment due primarily to rotation along the chamber's long axis. Inclusion of the aorta in the registration process rectified this error. In the iron oxide injection experiments, the ablation lesions were 1.8 +/- 0.5 mm from the targets. In the porcine infarct model, the catheter could be reliably navigated to the mitral valve annulus, and the ablation lesions were uniformly situated at the scar borders. These data suggest that registration of pre-acquired magnetic resonance images with real-time mapping is sufficiently accurate to guide LV catheter manipulation in a reliable and clinically relevant manner.