Markov random field segmentation of brain MR images

Intensity correction with a pair of spoiled gradient recalled echo images

Intensity inhomogeneities in magnetic resonance images (MRI) are a frequently occurring artefact, and result in the same tissue class to have vastly different intensities within an image. These inhomogeneities can be modelled by a slowly varying field, which is also called the bias field. Previous phantom-, image- or sequence based approaches suffer from long scan times, post-processing times or do not sufficiently remove the intensity variations. These intensity variations cause problems for quantitative image analysis algorithms (segmentation, registration) as well as clinicians (e.g. by complicating the visual assessment). This paper presents a novel technique (COIN, correction of intensity inhomogeneities) that uses two calibration images (fast spoiled gradient echo) to map a parameter containing the bias field, which is specific to the patient during a particular exam. This parametric map can then be used to correct any other images acquired during the same exam, regardless of the sequence employed. By using a short repetition time (less than 5 ms) for the calibration scans, the additional scan time is reduced to 60 s (max). The subsequent post-processing time is approximately 60 s per 20 slices. We successfully validate our approach on simulated brain MRI as well as real liver and spinal images. These images were acquired with a number of different coils, sequences and weightings. A comparison of our method with an existing, commercially available algorithm by radiologists shows that COIN is superior.

Real-time myocardial segmentation using coupled active geometric functions

Myocardial segmentation is essential for quantitative evaluation of cardiac functional images. As imaging techniques advance, 3D and 4D image data have become available. These data can provide clinically important cardiac dynamic information at high spatial or temporal resolution. However, the enormous amount of information contained in these data has also raised a challenge for traditional image analysis algorithms in terms of efficiency and clinical workflow. In this context, an automated real-time myocardial segmentation framework based on coupled Active Geometric Functions was proposed and tested on 414 frames of Phase Train Imaging data, a real-time cardiac MR imaging technique, with an average temporal resolution of 2 ms. The performance of myocardial segmentation was visually and quantitatively validated. Implemented in Matlab(c), the current method takes less than 1.2 ms per cardiac phase, allowing realization of true real-time online segmentation.

Towards an accurate and robust method based on fuzzy logic principles for the reconstruction and quantification of large volumes from MR and CT images

The authors have previously evaluated a new method of volume reconstruction and quantification from MR images, based on fuzzy logic (FL) principles. The technique is evaluated here for larger and more complex structures by investigating its accuracy and robustness using MR and CT images. Four large (50-71 cm(3)) and complex (e.g. mimicking a prostate) structures were created and imaged on MR and CT scanners, both with increasing slice thickness. Contours were delineated to generate 112 volumes. MR and CT images were processed using the FL method and a "classical" method of reconstruction on research software. In addition, the CT images were also processed on commercial virtual simulation software. Calculated volumes were compared with actual volumes. The mean +/- standard deviation of the relative variations in calculated target volume using the FL method was found to be 4.4%+/-2.8%, whereas with the "classical" method it was 23.7%+/-6% from axial MR images and 23.3%+/-9.8% from CT images. With the "classical" method, the relative variations in calculated volumes rise with increasing slice thickness, and the displayed volumes show deformations in the longitudinal direction. With the FL method, the volume calculation is not sensitive to the slice thickness and so the deformations are minimal. When used with MR images, our FL method of volume reconstruction is accurate and robust with respect to changes in slice thickness. For CT images, the results are encouraging but some work is still needed to improve the accuracy of the FL method.

Application of hidden Markov random field approach for quantification of perfusion/diffusion mismatch in acute ischemic stroke

The perfusion/diffusion 'mismatch model' in acute ischemic stroke provides the potential to more accurately understand the consequences of thrombolytic therapy on an individual patient basis. Few methods exist to quantify mismatch extent (ischemic penumbra) and none have shown a robust ability to predict infarcted tissue outcome. Hidden Markov random field (HMRF) approaches have been used successfully in many other applications. The aim of the study was to develop a method for rapid and reliable identification and quantification of perfusion/diffusion mismatch using an HMRF approach. An HMRF model was used in combination with automated contralateral identification to segment normal tissue from non-infarcted tissue with perfusion abnormality. The infarct was used as a seed point to initialize segmentation, along with the contralateral mirror tissue. The two seeds were then allowed to compete for ownership of all unclassified tissue. In addition, a novel method was presented for quantifying tissue salvageability by weighting the volume with the degree of hypoperfusion, allowing the penumbra voxels to contribute unequal potential damage estimates. Simulated and in vivo datasets were processed and compared with results from a conventional thresholding approach. Both simulated and in vivo experiments demonstrated a dramatic improvement in accuracy with the proposed technique. For the simulated dataset, the mean absolute error decreased from 171.9% with conventional thresholding to 2.9% for the delay-weighted HMRF approach. For the in vivo dataset, the mean absolute error decreased from 564.6% for thresholding to 34.2% for the delay-weighted HMRF approach. The described method represents a significant improvement over thresholding techniques.

Graph-partitioned spatial priors for functional magnetic resonance images

Spatial models of functional magnetic resonance imaging (fMRI) data allow one to estimate the spatial smoothness of general linear model (GLM) parameters and eschew pre-process smoothing of data entailed by conventional mass-univariate analyses. Recently diffusion-based spatial priors [Harrison, L.M., Penny, W., Daunizeau, J., and Friston, K.J. (2008). Diffusion-based spatial priors for functional magnetic resonance images. NeuroImage.] were proposed, which provide a way to formulate an adaptive spatial basis, where the diffusion kernel of a weighted graph-Laplacian (WGL) is used as the prior covariance matrix over GLM parameters. An advantage of these is that they can be used to relax the assumption of isotropy and stationarity implicit in smoothing data with a fixed Gaussian kernel. The limitation of diffusion-based models is purely computational, due to the large number of voxels in a brain volume. One solution is to partition a brain volume into slices, using a spatial model for each slice. This reduces computational burden by approximating the full WGL with a block diagonal form, where each block can be analysed separately. While fMRI data are collected in slices, the functional structures exhibiting spatial coherence and continuity are generally three-dimensional, calling for a more informed partition. We address this using the graph-Laplacian to divide a brain volume into sub-graphs, whose shape can be arbitrary. Their shape depends crucially on edge weights of the graph, which can be based on the Euclidean distance between voxels (isotropic) or on GLM parameters (anisotropic) encoding functional responses. The result is an approximation the full WGL that retains its 3D form and also has potential for parallelism. We applied the method to high-resolution (1 mm(3)) fMRI data and compared models where a volume was divided into either slices or graph-partitions. Models were optimized using Expectation-Maximization and the approximate log-evidence computed to compare these different ways to partition a spatial prior. The high-resolution fMRI data presented here had greatest evidence for the graph partitioned anisotropic model, which was best able to preserve fine functional detail.

A modified fuzzy C-means classification method using a multiscale diffusion filtering scheme

A fully automatic, multiscale fuzzy C-means (MsFCM) classification method for MR images is presented in this paper. We use a diffusion filter to process MR images and to construct a multiscale image series. A multiscale fuzzy C-means classification method is applied along the scales from the coarse to fine levels. The objective function of the conventional fuzzy C-means (FCM) method is modified to allow multiscale classification processing where the result from a coarse scale supervises the classification in the next fine scale. The method is robust for noise and low-contrast MR images because of its multiscale diffusion filtering scheme. The new method was compared with the conventional FCM method and a modified FCM (MFCM) method. Validation studies were performed on synthesized images with various contrasts and on the McGill brain MR image database. Our MsFCM method consistently performed better than the conventional FCM and MFCM methods. The MsFCM method achieved an overlap ratio of greater than 90% as validated by the ground truth. Experiments results on real MR images were given to demonstrate the effectiveness of the proposed method. Our multiscale fuzzy C-means classification method is accurate and robust for various MR images. It can provide a quantitative tool for neuroimaging and other applications.

Restoration of MRI data for intensity non-uniformities using local high order intensity statistics

MRI at high magnetic fields (>3.0 T) is complicated by strong inhomogeneous radio-frequency fields, sometimes termed the "bias field". These lead to non-biological intensity non-uniformities across the image. They can complicate further image analysis such as registration and tissue segmentation. Existing methods for intensity uniformity restoration have been optimized for 1.5 T, but they are less effective for 3.0 T MRI, and not at all satisfactory for higher fields. Also, many of the existing restoration algorithms require a brain template or use a prior atlas, which can restrict their practicalities. In this study an effective intensity uniformity restoration algorithm has been developed based on non-parametric statistics of high order local intensity co-occurrences. These statistics are restored with a non-stationary Wiener filter. The algorithm also assumes a smooth non-uniformity and is stable. It does not require a prior atlas and is robust to variations in anatomy. In geriatric brain imaging it is robust to variations such as enlarged ventricles and low contrast to noise ratio. The co-occurrence statistics improve robustness to whole head images with pronounced non-uniformities present in high field acquisitions. Its significantly improved performance and lower time requirements have been demonstrated by comparing it to the very commonly used N3 algorithm on BrainWeb MR simulator images as well as on real 4 T human head images.

Fast nosological imaging using canonical correlation analysis of brain data obtained by two-dimensional turbo spectroscopic imaging

A new fast and accurate tissue typing technique has recently been successfully applied to prostate MR spectroscopic imaging (MRSI) data. This technique is based on canonical correlation analysis (CCA), a statistical method able to simultaneously exploit the spectral and spatial information characterizing the MRSI data. Here, the performance of CCA is further investigated by using brain data obtained by two-dimensional turbo spectroscopic imaging (2DTSI) from patients affected by glioblastoma. The purpose of this study is to investigate the applicability of CCA when typing tissues of heterogeneous tumors. The performance of CCA is also compared with that of ordinary correlation analysis on simulated as well as in vivo data. The results show that CCA outperforms ordinary correlation analysis in terms of robustness and accuracy.

Fully automatic segmentation of multiple sclerosis lesions in brain MR FLAIR images using adaptive mixtures method and Markov random field model

In this paper, an approach is proposed for fully automatic segmentation of MS lesions in fluid attenuated inversion recovery (FLAIR) Magnetic Resonance (MR) images. The proposed approach, based on a Bayesian classifier, utilizes the adaptive mixtures method (AMM) and Markov random field (MRF) model to obtain and upgrade the class conditional probability density function (CCPDF) and the a priori probability of each class. To compare the performance of the proposed approach with those of previous approaches including manual segmentation, the similarity criteria of different slices related to 20 MS patients were calculated. Also, volumetric comparison of lesions volume between the fully automated segmentation and the gold standard was performed using correlation coefficient (CC). The results showed a better performance for the proposed approach, compared to those of previous works.

Null point imaging: a joint acquisition/analysis paradigm for MR classification

Automatic classification of neurological tissues is a first step to many structural analysis pipelines. Most computational approaches are designed to extract the best possible classification results out of MR data acquired with standard clinical protocols. We observe that the characteristics of the latter owe more to the historical circumstances under which they were developed and the visual appreciation of the radiographer who acquires the images than to the optimality with which they can be classified with an automatic algorithm. We submit that better performances could be obtained by considering the acquisition and analysis processes conjointly rather than optimising them independently. Here, we propose such a joint approach to MR tissue classification in the form of a fast MR sequence, which nulls the magnitude and changes the sign of the phase at the boundary between tissue types. A simple phase-based thresholding algorithm then suffices to segment the tissues. Preliminary results show promises to simplify and shorten the overall classification process.

An artificial immune-activated neural network applied to brain 3D MRI segmentation

In this paper, a new neural network model inspired by the biological immune system functions is presented. The model, termed Artificial Immune-Activated Neural Network (AIANN), extracts classification knowledge from a training data set, which is then used to classify input patterns or vectors. The AIANN is based on a neuron activation function whose behavior is conceptually modeled after the chemical bonds between the receptors and epitopes in the biological immune system. The bonding is controlled through an energy measure to ensure accurate recognition. The AIANN model was applied to the segmentation of 3-dimensional magnetic resonance imaging (MRI) data of the brain and a contextual basis was developed for the segmentation problem. Evaluation of the segmentation results was performed using both real MRI data obtained from the Center for Morphometric Analysis at Massachusetts General Hospital and simulated MRI data generated using the McGill University BrainWeb MRI simulator. Experimental results demonstrated that the AIANN model attained higher average results than those obtained using published methods for real MRI data and simulated MRI data, especially at low levels of noise.

A hierarchical algorithm for MR brain image parcellation

We introduce an algorithm for segmenting brain magnetic resonance (MR) images into anatomical compartments such as the major tissue classes and neuro-anatomical structures of the gray matter. The algorithm is guided by prior information represented within a tree structure. The tree mirrors the hierarchy of anatomical structures and the subtrees correspond to limited segmentation problems. The solution to each problem is estimated via a conventional classifier. Our algorithm can be adapted to a wide range of segmentation problems by modifying the tree structure or replacing the classifier. We evaluate the performance of our new segmentation approach by revisiting a previously published statistical group comparison between first-episode schizophrenia patients, first-episode affective psychosis patients, and comparison subjects. The original study is based on 50 MR volumes in which an expert identified the brain tissue classes as well as the superior temporal gyrus, amygdala, and hippocampus. We generate analogous segmentations using our new method and repeat the statistical group comparison. The results of our analysis are similar to the original findings, except for one structure (the left superior temporal gyrus) in which a trend-level statistical significance (p = 0.07) was observed instead of statistical significance.

A Dirichlet process mixture model for brain MRI tissue classification

Accurate classification of magnetic resonance images according to tissue type or region of interest has become a critical requirement in diagnosis, treatment planning, and cognitive neuroscience. Several authors have shown that finite mixture models give excellent results in the automated segmentation of MR images of the human normal brain. However, performance and robustness of finite mixture models deteriorate when the models have to deal with a variety of anatomical structures. In this paper, we propose a nonparametric Bayesian model for tissue classification of MR images of the brain. The model, known as Dirichlet process mixture model, uses Dirichlet process priors to overcome the limitations of current parametric finite mixture models. To validate the accuracy and robustness of our method we present the results of experiments carried out on simulated MR brain scans, as well as on real MR image data. The results are compared with similar results from other well-known MRI segmentation methods.

Restoration of MRI Data for Field Nonuniformities using High Order Neighborhood Statistics

MRI at high magnetic fields (> 3.0 T ) is complicated by strong inhomogeneous radio-frequency fields, sometimes termed the "bias field". These lead to nonuniformity of image intensity, greatly complicating further analysis such as registration and segmentation. Existing methods for bias field correction are effective for 1.5 T or 3.0 T MRI, but are not completely satisfactory for higher field data. This paper develops an effective bias field correction for high field MRI based on the assumption that the nonuniformity is smoothly varying in space. Also, nonuniformity is quantified and unmixed using high order neighborhood statistics of intensity cooccurrences. They are computed within spherical windows of limited size over the entire image. The restoration is iterative and makes use of a novel stable stopping criterion that depends on the scaled entropy of the cooccurrence statistics, which is a non monotonic function of the iterations; the Shannon entropy of the cooccurrence statistics normalized to the effective dynamic range of the image. The algorithm restores whole head data, is robust to intense nonuniformities present in high field acquisitions, and is robust to variations in anatomy. This algorithm significantly improves bias field correction in comparison to N3 on phantom 1.5 T head data and high field 4 T human head data.

Fully automated segmentation of oncological PET volumes using a combined multiscale and statistical model

The widespread application of positron emission tomography (PET) in clinical oncology has driven this imaging technology into a number of new research and clinical arenas. Increasing numbers of patient scans have led to an urgent need for efficient data handling and the development of new image analysis techniques to aid clinicians in the diagnosis of disease and planning of treatment. Automatic quantitative assessment of metabolic PET data is attractive and will certainly revolutionize the practice of functional imaging since it can lower variability across institutions and may enhance the consistency of image interpretation independent of reader experience. In this paper, a novel automated system for the segmentation of oncological PET data aiming at providing an accurate quantitative analysis tool is proposed. The initial step involves expectation maximization (EM)-based mixture modeling using a k-means clustering procedure, which varies voxel order for initialization. A multiscale Markov model is then used to refine this segmentation by modeling spatial correlations between neighboring image voxels. An experimental study using an anthropomorphic thorax phantom was conducted for quantitative evaluation of the performance of the proposed segmentation algorithm. The comparison of actual tumor volumes to the volumes calculated using different segmentation methodologies including standard k-means, spatial domain Markov Random Field Model (MRFM), and the new multiscale MRFM proposed in this paper showed that the latter dramatically reduces the relative error to less than 8% for small lesions (7 mm radii) and less than 3.5% for larger lesions (9 mm radii). The analysis of the resulting segmentations of clinical oncologic PET data seems to confirm that this methodology shows promise and can successfully segment patient lesions. For problematic images, this technique enables the identification of tumors situated very close to nearby high normal physiologic uptake. The use of this technique to estimate tumor volumes for assessment of response to therapy and to delineate treatment volumes for the purpose of combined PET/CT-based radiation therapy treatment planning is also discussed.

A segmentation model using compound Markov random fields based on a boundary model

Markov random field (MRF) theory has been widely applied to the challenging problem of image segmentation. In this paper, we propose a new nontexture segmentation model using compound MRFs, in which the original label MRF is coupled with a new boundary MRF to help improve the segmentation performance. The boundary model is relatively general and does not need prior training on boundary patterns. Unlike some existing related work, the proposed method offers a more compact interaction between label and boundary MRFs. Furthermore, our boundary model systematically takes into account all the possible scenarios of a single edge existing in a 3 x 3 neighborhood and, thus, incorporates sophisticated prior information about the relation between label and boundary. It is experimentally shown that the proposed model can segment objects with complex boundaries and at the same time is able to work under noise corruption. The new method has been applied to medical image segmentation. Experiments on synthetic images and real clinical datasets show that the proposed model is able to produce more accurate segmentation results and satisfactorily keep the delicate boundary. It is also less sensitive to noise in both high and low signal-to-noise ratio regions than some of the existing models in common use.

Variational Bayes inference of spatial mixture models for segmentation

Mixture models are commonly used in the statistical segmentation of images. For example, they can be used for the segmentation of structural medical images into different matter types, or of statistical parametric maps into activating and nonactivating brain regions in functional imaging. Spatial mixture models have been developed to augment histogram information with spatial regularization using Markov random fields (MRFs). In previous work, an approximate model was developed to allow adaptive determination of the parameter controlling the strength of spatial regularization. Inference was performed using Markov Chain Monte Carlo (MCMC) sampling. However, this approach is prohibitively slow for large datasets. In this work, a more efficient inference approach is presented. This combines a variational Bayes approximation with a second-order Taylor expansion of the components of the posterior distribution, which would otherwise be intractable to Variational Bayes. This provides inference on fully adaptive spatial mixture models an order of magnitude faster than MCMC. We examine the behavior of this approach when applied to artificial data with different spatial characteristics, and to functional magnetic resonance imaging statistical parametric maps.

Constrained Gaussian mixture model framework for automatic segmentation of MR brain images

An automated algorithm for tissue segmentation of noisy, low-contrast magnetic resonance (MR) images of the brain is presented. A mixture model composed of a large number of Gaussians is used to represent the brain image. Each tissue is represented by a large number of Gaussian components to capture the complex tissue spatial layout. The intensity of a tissue is considered a global feature and is incorporated into the model through tying of all the related Gaussian parameters. The expectation-maximization (EM) algorithm is utilized to learn the parameter-tied, constrained Gaussian mixture model. An elaborate initialization scheme is suggested to link the set of Gaussians per tissue type, such that each Gaussian in the set has similar intensity characteristics with minimal overlapping spatial supports. Segmentation of the brain image is achieved by the affiliation of each voxel to the component of the model that maximized the a posteriori probability. The presented algorithm is used to segment three-dimensional, T1-weighted, simulated and real MR images of the brain into three different tissues, under varying noise conditions. Results are compared with state-of-the-art algorithms in the literature. The algorithm does not use an atlas for initialization or parameter learning. Registration processes are therefore not required and the applicability of the framework can be extended to diseased brains and neonatal brains.

Model-based region-of-interest selection in dynamic breast MRI

Magnetic resonance imaging (MRI) is emerging as a powerful tool for the diagnosis of breast abnormalities. Dynamic analysis of the temporal pattern of contrast uptake has been applied in differential diagnosis of benign and malignant lesions to improve specificity. Selecting a region of interest (ROI) is an almost universal step in the process of examining the contrast uptake characteristics of a breast lesion. We propose an ROI selection method that combines model-based clustering of the pixels with Bayesian morphology, a new statistical image segmentation method. We then investigate tools for subsequent analysis of signal intensity time course data in the selected region. Results on a database of 19 patients indicate that the method provides informative segmentations and good detection rates.

A Review on MR Image Intensity Inhomogeneity Correction

Intensity inhomogeneity (IIH) is often encountered in MR imaging, and a number of techniques have been devised to correct this artifact. This paper attempts to review some of the recent developments in the mathematical modeling of IIH field. Low-frequency models are widely used, but they tend to corrupt the low-frequency components of the tissue. Hypersurface models and statistical models can be adaptive to the image and generally more stable, but they are also generally more complex and consume more computer memory and CPU time. They are often formulated together with image segmentation within one framework and the overall performance is highly dependent on the segmentation process. Beside these three popular models, this paper also summarizes other techniques based on different principles. In addition, the issue of quantitative evaluation and comparative study are discussed.

MRI tissue classification with neighborhood statistics: a nonparametric, entropy-minimizing approach

We introduce a novel approach for magnetic resonance image (MRI) brain tissue classification by learning image neighborhood statistics from noisy input data using nonparametric density estimation. The method models images as random fields and relies on minimizing an entropy-based metric defined on high dimensional probability density functions. Combined with an atlas-based initialization, it is completely automatic. Experiments on real and simulated data demonstrate the advantages of the method in comparison to other approaches.

Unbiased atlas formation via large deformations metric mapping

The construction of population atlases is a key issue in medical image analysis, and particularly in brain mapping. Large sets of images are mapped into a common coordinate system to study intra-population variability and inter-population differences, to provide voxel-wise mapping of functional sites, and to facilitate tissue and object segmentation via registration of anatomical labels. We formulate the unbiased atlas construction problem as a Fréchet mean estimation in the space of diffeomorphisms via large deformations metric mapping. A novel method for computing constant speed velocity fields and an analysis of atlas stability and robustness using entropy are presented. We address the question: how many images are required to build a stable brain atlas?

Method to correct intensity inhomogeneity in MR images for atherosclerosis characterization

We are developing methods to characterize atherosclerotic disease in human carotid arteries using multiple MR images having different contrast mechanisms (T1W, T2W, PDW). To enable the use of voxel gray values for interpretation of disease, we created a new method, local entropy minimization with a bicubic spline model (LEMS), to correct the severe (approximately 80%) intensity inhomogeneity that arises from the surface coil array. This entropy-based method does not require classification and robustly addresses some problems that are more severe than those found in brain imaging, including noise, steep bias field, sensitivity of artery wall voxels to edge artifacts, and signal voids near the artery wall. Validation studies were performed on a synthetic digital phantom with realistic intensity inhomogeneity, a physical phantom roughly mimicking the neck, and patient carotid artery images. We compared LEMS to a modified fuzzy c-means segmentation based method (mAFCM), and a linear filtering method (LINF). Following LEMS correction, skeletal muscles in patient images were relatively isointense across the field of view. In the physical phantom, LEMS reduced the variation in the image to 1.9% and across the vessel wall region to 2.5%, a value which should be sufficient to distinguish plaque tissue types, based on literature measurements. In conclusion, we believe that the correction method shows promise for aiding human and computerized tissue classification from MR signal intensities.

A versatile segmentation procedure

In this paper, a new method for the segmentation of natural images is proposed. Original images g(x, y) are first regularized by using a self-adaptive implementation of the Mumford-Shah functional so that the two parameters alpha and gamma controlling the smoothness and fidelity, automatically adapt to the local scale and contrast of g(x, y). From the regularized image u(x, y) which is piecewise smooth, it is possible to obtain a piecewise constant image sN(x, y) representing a segmentation of the original image g(x, y). Indeed, sN(X, y) is the union of N closed regions, having a constant grey level, preserving thin bars and trihedral junctions present in the original image g(x, y). If the number N of closed regions is too high, closed regions can be merged by minimizing a functional which depends on a parameter n. When n is set equal to 1, a coarse segmentation is obtained with a few tens of distinct regions. With larger values of n, finer segmentations are obtained with about a hundred distinct regions. Therefore, by selecting the value of n it is possible to obtain segmentations at different resolutions. The proposed method for image segmentation was evaluated in two cases where a ground truth segmentation is available. The proposed procedure for image segmentation is rather versatile and depends on only one parameter n and seems suitable for higher level processing, such as categorization, recognition, and scene understanding.

3D Cryo-Section/Imaging of Blood Vessel Lesions for Validation of MRI Data

Vascular disease is a leading cause of death and disability in the western world. Diagnosis and staging of atherosclerosis is a challenge, especially with regards to the identification of plaque vulnerability. We are developing imaging methods based upon MRI and intravascular microcoils. In order to rigorously validate our MRI imaging methods and algorithms, we have developed a new cryo-imaging system that allows one to alternately section and image the block face of tissue. We obtain 3D pathology of vessel segments excised from cadaver and we characterize the tissues of atheroma using episcopic autofluorescence and bright field microscopy images. After embedding the vessel, the block is frozen, and block face microscopic images are taken every 200µm with an image resolution of 30µm×30µm. The series of images is then corrected for uneven illumination, serially registered to one another, and the 3D vessel segment is reconstructed. Some sections are recovered and processed with histological staining for validation. Seven tissue types can be readily identified from the cryo-images: necrotic core, calcification, lipid pool, media, adventitia, fibrosis, thrombus, and normal intima. Since the whole vessel segment is available, we could register 3D data to images from MR, or other modalities, for validation. In addition, visualization tools such as multi-planar reformatting 3D rendering can be used to study 3D plaque morphology, in microscopic detail.

Segmentation of skull and scalp in 3-D human MRI using mathematical morphology

We present a new technique for segmentation of skull and scalp in T(1)-weighted magnetic resonance images (MRIs) of the human head. Our method uses mathematical morphological operations to generate realistic models of the skull, scalp, and brain that are suitable for electroencephalography (EEG) and magnetoencephalography (MEG) source modeling. We first segment the brain using our Brain Surface Extractor algorithm; using this, we can ensure that the brain does not intersect our skull segmentation. We next generate a scalp mask using a combination of thresholding and mathematical morphology. We use the scalp mask in our skull segmentation procedure, as it allows us to automatically exclude background voxels with intensities similar to those of the skull. We find the inner and outer skull boundaries using thresholding and morphological operations. Finally, we mask the results with the scalp and brain volumes to ensure closed and nonintersecting skull boundaries. Visual evaluation indicated accurate segmentations of the cranium at a gross anatomical level (other than small holes in the zygomatic bone in eight subjects) in all 44 MRI volumes processed when run using default settings. In a quantitative comparison with coregistered CT images as a gold standard, MRI skull segmentation accuracy, as measured using the Dice coefficient, was found to be similar to that which would be obtained using CT imagery with a registration error of 2-3 mm.

Unwrapping of MR phase images using a Markov random field model

Phase unwrapping is an important problem in many magnetic resonance imaging applications, such as field mapping and flow imaging. The challenge in two-dimensional phase unwrapping lies in distinguishing jumps due to phase wrapping from those due to noise and/or abrupt variations in the actual function. This paper addresses this problem using a Markov random field to model the true phase function, whose parameters are determined by maximizing the a posteriori probability. To reduce the computational complexity of the optimization procedure, an efficient algorithm is also proposed for parameter estimation using a series of dynamic programming connected by the iterated conditional modes. The proposed method has been tested with both simulated and experimental data, yielding better results than some of the state-of-the-art method (e.g., the popular least-squares method) in handling noisy phase images with rapid phase variations.

A multiresolution diffused expectation-maximization algorithm for medical image segmentation

In this paper a new method for segmenting medical images is presented, the multiresolution diffused expectation-maximization (MDEM) algorithm. The algorithm operates within a multiscale framework, thus taking advantage of the fact that objects/regions to be segmented usually reside at different scales. At each scale segmentation is carried out via the expectation-maximization algorithm, coupled with anisotropic diffusion on classes, in order to account for the spatial dependencies among pixels. This new approach is validated via experiments on a variety of medical images and its performance is compared with more standard methods.

Comparison and validation of tissue modelization and statistical classification methods in T1-weighted MR brain images

This paper presents a validation study on statistical nonsupervised brain tissue classification techniques in magnetic resonance (MR) images. Several image models assuming different hypotheses regarding the intensity distribution model, the spatial model and the number of classes are assessed. The methods are tested on simulated data for which the classification ground truth is known. Different noise and intensity nonuniformities are added to simulate real imaging conditions. No enhancement of the image quality is considered either before or during the classification process. This way, the accuracy of the methods and their robustness against image artifacts are tested. Classification is also performed on real data where a quantitative validation compares the methods' results with an estimated ground truth from manual segmentations by experts. Validity of the various classification methods in the labeling of the image as well as in the tissue volume is estimated with different local and global measures. Results demonstrate that methods relying on both intensity and spatial information are more robust to noise and field inhomogeneities. We also demonstrate that partial volume is not perfectly modeled, even though methods that account for mixture classes outperform methods that only consider pure Gaussian classes. Finally, we show that simulated data results can also be extended to real data.

Partial volume segmentation of brain magnetic resonance images based on maximum a posteriori probability

Noise, partial volume (PV) effect, and image-intensity inhomogeneity render a challenging task for segmentation of brain magnetic resonance (MR) images. Most of the current MR image segmentation methods focus on only one or two of the above-mentioned effects. The objective of this paper is to propose a unified framework, based on the maximum a posteriori probability principle, by taking all these effects into account simultaneously in order to improve image segmentation performance. Instead of labeling each image voxel with a unique tissue type, the percentage of each voxel belonging to different tissues, which we call a mixture, is considered to address the PV effect. A Markov random field model is used to describe the noise effect by considering the nearby spatial information of the tissue mixture. The inhomogeneity effect is modeled as a bias field characterized by a zero mean Gaussian prior probability. The well-known fuzzy C-mean model is extended to define the likelihood function of the observed image. This framework reduces theoretically, under some assumptions, to the adaptive fuzzy C-mean (AFCM) algorithm proposed by Pham and Prince. Digital phantom and real clinical MR images were used to test the proposed framework. Improved performance over the AFCM algorithm was observed in a clinical environment where the inhomogeneity, noise level, and PV effect are commonly encountered.

Automated segmentation of neuroanatomical structures in multispectral MR microscopy of the mouse brain

We present the automated segmentation of magnetic resonance microscopy (MRM) images of the C57BL/6J mouse brain into 21 neuroanatomical structures, including the ventricular system, corpus callosum, hippocampus, caudate putamen, inferior colliculus, internal capsule, globus pallidus, and substantia nigra. The segmentation algorithm operates on multispectral, three-dimensional (3D) MR data acquired at 90-microm isotropic resolution. Probabilistic information used in the segmentation is extracted from training datasets of T2-weighted, proton density-weighted, and diffusion-weighted acquisitions. Spatial information is employed in the form of prior probabilities of occurrence of a structure at a location (location priors) and the pairwise probabilities between structures (contextual priors). Validation using standard morphometry indices shows good consistency between automatically segmented and manually traced data. Results achieved in the mouse brain are comparable with those achieved in human brain studies using similar techniques. The segmentation algorithm shows excellent potential for routine morphological phenotyping of mouse models.

Segmentation of cDNA microarray spots using markov random field modeling

MOTIVATION

Spot segmentation is a critical step in microarray gene expression data analysis. Therefore, the performance of segmentation may substantially affect the results of subsequent stages of the analysis, such as the detection of differentially expressed genes. Several methods have been developed to segment microarray spots from the surrounding background. In this study, we have proposed a new approach based on Markov random field (MRF) modeling and tested its performance on simulated and real microarray images against a widely used segmentation method based on Mann-Whitney test adopted by QuantArray software (Boston, MA). Spot addressing was performed using QuantArray. We have also devised a simulation method to generate microarray images with realistic features. Such images can be used as gold standards for the purposes of testing and comparing different segmentation methods, and optimizing segmentation parameters.

RESULTS

Experiments on simulated and 14 actual microarray image sets show that the proposed MRF-based segmentation method can detect spot areas and estimate spot intensities with higher accuracy.

An advanced boundary element method (BEM) implementation for the forward problem of electromagnetic source imaging

The forward problem of electromagnetic source imaging has two components: a numerical model to solve the related integral equations and a model of the head geometry. This study is on the boundary element method (BEM) implementation for numerical solutions and realistic head modelling. The use of second-order (quadratic) isoparametric elements and the recursive integration technique increase the accuracy in the solutions. Two new formulations are developed for the calculation of the transfer matrices to obtain the potential and magnetic field patterns using realistic head models. The formulations incorporate the use of the isolated problem approach for increased accuracy in solutions. If a personal computer is used for computations, each transfer matrix is calculated in 2.2 h. After this pre-computation period, solutions for arbitrary source configurations can be obtained in milliseconds for a realistic head model. A hybrid algorithm that uses snakes, morphological operations, region growing and thresholding is used for segmentation. The scalp, skull, grey matter, white matter and eyes are segmented from the multimodal magnetic resonance images and meshes for the corresponding surfaces are created. A mesh generation algorithm is developed for modelling the intersecting tissue compartments, such as eyes. To obtain more accurate results quadratic elements are used in the realistic meshes. The resultant BEM implementation provides more accurate forward problem solutions and more efficient calculations. Thus it can be the firm basis of the future inverse problem solutions.

Tissue segmentation and classification of MRSI data using canonical correlation analysis

In this article an accurate and efficient technique for tissue typing is presented. The proposed technique is based on Canonical Correlation Analysis, a statistical method able to simultaneously exploit the spectral and spatial information characterizing the Magnetic Resonance Spectroscopic Imaging (MRSI) data. Recently, Canonical Correlation Analysis has been successfully applied to other types of biomedical data, such as functional MRI data. Here, Canonical Correlation Analysis is adapted for MRSI data processing in order to retrieve in an accurate and efficient way the possible tissue types that characterize the organ under investigation. The potential and limitations of the new technique have been investigated by using simulated as well as in vivo prostate MRSI data, and extensive studies demonstrate a high accuracy, robustness, and efficiency. Moreover, the performance of Canonical Correlation Analysis has been compared to that of ordinary correlation analysis. The test results show that Canonical Correlation Analysis performs best in terms of accuracy and robustness.

Vascular segmentation of phase contrast magnetic resonance angiograms based on statistical mixture modeling and local phase coherence

In this paper, we present an approach to segmenting the brain vasculature in phase contrast magnetic resonance angiography (PC-MRA). According to our prior work, we can describe the overall probability density function of a PC-MRA speed image as either a Maxwell-uniform (MU) or Maxwell-Gaussian-uniform (MGU) mixture model. An automatic mechanism based on Kullback-Leibler divergence is proposed for selecting between the MGU and MU models given a speed image volume. A coherence measure, namely local phase coherence (LPC), which incorporates information about the spatial relationships between neighboring flow vectors, is defined and shown to be more robust to noise than previously described coherence measures. A statistical measure from the speed images and the LPC measure from the phase images are combined in a probabilistic framework, based on the maximum a posteriori method and Markov random fields, to estimate the posterior probabilities of vessel and background for classification. It is shown that segmentation based on both measures gives a more accurate segmentation than using either speed or flow coherence information alone. The proposed method is tested on synthetic, flow phantom and clinical datasets. The results show that the method can segment normal vessels and vascular regions with relatively low flow rate and low signal-to-noise ratio, e.g., aneurysms and veins.

A meta-algorithm for brain extraction in MRI

Accurate identification of brain tissue and cerebrospinal fluid (CSF) in a whole-head MRI is a critical first step in many neuroimaging studies. Automating this procedure can eliminate intra- and interrater variance and greatly increase throughput for a labor-intensive step. Many available procedures perform differently across anatomy and under different acquisition protocols. We developed the Brain Extraction Meta-Algorithm (BEMA) to address these concerns. It executes many extraction algorithms and a registration procedure in parallel to combine the results in an intelligent fashion and obtain improved results over any of the individual algorithms. Using an atlas space, BEMA performs a voxelwise analysis of training data to determine the optimal Boolean combination of extraction algorithms to produce the most accurate result for a given voxel. This allows the provided extractors to be used differentially across anatomy, increasing both the accuracy and robustness of the procedure. We tested BEMA using modified forms of BrainSuite's Brain Surface Extractor (BSE), FSL's Brain Extraction Tool (BET), AFNI's 3dIntracranial, and FreeSurfer's MRI Watershed as well as FSL's FLIRT for the registration procedure. Training was performed on T1-weighted scans of 136 subjects from five separate data sets with different acquisition parameters on separate scanners. Testing was performed on 135 separate subjects from the same data sets. BEMA outperformed the individual algorithms, as well as interrater results from a subset of the scans, when compared for the mean Dice coefficient, a rating of the similarity of output masks to the manually defined gold standards.

A method for the automatic segmentation of autoradiographic image stacks and spatial normalization of functional cortical activity patterns

This paper introduces two new methods for the automatic anatomical and functional analysis of neurobiological autoradiographic image stacks, such as 2-fluoro-deoxyglucose (2FDG) images. The difficulty in the evaluation of these "2(1/2)D" datasets is that they do not inherently represent a continuous 3D data volume (as generated by MRI or CT), but consist of a stack of images from single tissue slices, suffering from unavoidable preparation artifacts. In the first part of the paper, a semi-automatic segmentation method is presented which generates a 3D surface model of certain brain structures and which is robust against different cutting directions with respect to the brain coordinate system. The method saves man-hours compared to manual segmentation and the results are highly reproducible. In the second part, a fully automatic method for the extraction, analysis and 3D visualization of functional information is described, which allows not only a more accurate localization of activation sites, but also greatly enhances the comparability of different individuals. Results are shown for 2FDG autoradiographs from rat brains under acoustical stimulation.

Automated segmentation of human brain MR images using a multi-agent approach

Image interpretation consists in finding a correspondence between radiometric information and symbolic labeling with respect to specific spatial constraints. It is intrinsically a distributed process in terms of goals to be reached, zones in the image to be processed and methods to be applied. To cope the the difficulty inherent in this process, several information processing steps are required to gradually extract information. In this paper we advocate the use of situated cooperative agents as a framework for managing such steps. Dedicated agent behaviors are dynamically adapted depending on their position in the image, of their topographic relationships and of the radiometric information available. The information collected by the agents is gathered, shared via qualitative maps, or used as soon as available by acquaintances. Incremental refinement of interpretation is obtained through a coarse to fine strategy. Our work is essentially focused on radiometry-based tissue interpretation where knowledge is introduced or extracted at several levels to estimate models for tissue-intensity distribution and to cope with noise, intensity non-uniformity and partial volume effect. Several experiments on phantom and real images were performed. A complete volume can be segmented in less than 5 min with about 0.84% accuracy of the segmented reference. Comparison with other techniques demonstrates the potential interest of our approach for magnetic resonance imaging (MRI) brain scan interpretation.

A fully automatic and robust brain MRI tissue classification method

A novel, fully automatic, adaptive, robust procedure for brain tissue classification from 3D magnetic resonance head images (MRI) is described in this paper. The procedure is adaptive in that it customizes a training set, by using a 'pruning' strategy, such that the classification is robust against anatomical variability and pathology. Starting from a set of samples generated from prior tissue probability maps (a 'model') in a standard, brain-based coordinate system ('stereotaxic space'), the method first reduces the fraction of incorrectly labeled samples in this set by using a minimum spanning tree graph-theoretic approach. Then, the corrected set of samples is used by a supervised kNN classifier for classifying the entire 3D image. The classification procedure is robust against variability in the image quality through a non-parametric implementation: no assumptions are made about the tissue intensity distributions. The performance of this brain tissue classification procedure is demonstrated through quantitative and qualitative validation experiments on both simulated MRI data (10 subjects) and real MRI data (43 subjects). A significant improvement in output quality was observed on subjects who exhibit morphological deviations from the model due to aging and pathology.

Automatic segmentation of the brain and intracranial cerebrospinal fluid in T1-weighted volume MRI scans of the head, and its application to serial cerebral and intracranial volumetry

A new fully automatic algorithm for the segmentation of the brain and total intracranial cerebrospinal fluid (CSF) from T(1)-weighted volume MRI scans of the head, called Exbrain v.2, is described. The algorithm was developed in the context of serial intracranial volumetry. A brain mask obtained using a previous version of the algorithm forms the basis of the CSF segmentation. Improved brain segmentation is then obtained by iterative tracking of the brain-CSF interface. Gray matter (GM), white matter (WM), and intracranial CSF volumes and probability maps are calculated based on a model of intensity probability distribution (IPD) that includes two partial volume classes: GM-CSF and GM-WM. Accuracy was assessed using the Montreal Neurological Institute's (MNI) digital phantom scan. Reproducibility was assessed using scan pairs from 24 controls and 10 patients with epilepsy. Segmentation overlap with the gold standard was 98% for the brain and 95%, 96%, and 97% for the GM, WM, and total intracranial contents, respectively; CSF overlap was 86%. In the controls, the Bland and Altman coefficient of reliability (CR) was 35.2 cm(3) for the total brain volume (TBV) and 29.0 cm(3) for the intracranial volume (ICV). Scan-matching reduced CR to 25.2 cm(3) and 17.1 cm(3) for the TBV and ICV, respectively. For the patients, similar CR values were obtained for the ICV.

A unifying framework for partial volume segmentation of brain MR images

Accurate brain tissue segmentation by intensity-based voxel classification of magnetic resonance (MR) images is complicated by partial volume (PV) voxels that contain a mixture of two or more tissue types. In this paper, we present a statistical framework for PV segmentation that encompasses and extends existing techniques. We start from a commonly used parametric statistical image model in which each voxel belongs to one single tissue type, and introduce an additional downsampling step that causes partial voluming along the borders between tissues. An expectation-maximization approach is used to simultaneously estimate the parameters of the resulting model and perform a PV classification. We present results on well-chosen simulated images and on real MR images of the brain, and demonstrate that the use of appropriate spatial prior knowledge not only improves the classifications, but is often indispensable for robust parameter estimation as well. We conclude that general robust PV segmentation of MR brain images requires statistical models that describe the spatial distribution of brain tissues more accurately than currently available models.

A robust method for extraction and automatic segmentation of brain images

A new protocol is introduced for brain extraction and automatic tissue segmentation of MR images. For the brain extraction algorithm, proton density and T2-weighted images are used to generate a brain mask encompassing the full intracranial cavity. Segmentation of brain tissues into gray matter (GM), white matter (WM), and cerebral spinal fluid (CSF) is accomplished on a T1-weighted image after applying the brain mask. The fully automatic segmentation algorithm is histogram-based and uses the Expectation Maximization algorithm to model a four-Gaussian mixture for both global and local histograms. The means of the local Gaussians for GM, WM, and CSF are used to set local thresholds for tissue classification. Reproducibility of the extraction procedure was excellent, with average variation in intracranial capacity (TIC) of 0.13 and 0.66% TIC in 12 healthy normal and 33 Alzheimer brains, respectively. Repeatability of the segmentation algorithm, tested on healthy normal images, indicated scan-rescan differences in global tissue volumes of less than 0.30% TIC. Reproducibility at the regional level was established by comparing segmentation results within the 12 major Talairach subdivisions. Accuracy of the algorithm was tested on a digital brain phantom, and errors were less than 1% of the phantom volume. Maximal Type I and Type II classification errors were low, ranging between 2.2 and 4.3% of phantom volume. The algorithm was also insensitive to variation in parameter initialization values. The protocol is robust, fast, and its success in segmenting normal as well as diseased brains makes it an attractive clinical application.