Adaptive fuzzy segmentation of magnetic resonance images

Segmentation of brain images using adaptive atlases with application to ventriculomegaly

Segmentation of brain images often requires a statistical atlas for providing prior information about the spatial position of different structures. A major limitation of atlas-based segmentation algorithms is their deficiency in analyzing brains that have a large deviation from the population used in the construction of the atlas. We present an expectation-maximization framework based on a Dirichlet distribution to adapt a statistical atlas to the underlying subject. Our model combines anatomical priors with the subject's own anatomy, resulting in a subject specific atlas which we call an "adaptive atlas". The generation of this adaptive atlas does not require the subject to have an anatomy similar to that of the atlas population, nor does it rely on the availability of an ensemble of similar images. The proposed method shows a significant improvement over current segmentation approaches when applied to subjects with severe ventriculomegaly, where the anatomy deviates significantly from the atlas population. Furthermore, high levels of accuracy are maintained when the method is applied to subjects with healthy anatomy.

White matter atlas generation using HARDI based automated parcellation

Most diffusion imaging studies have used subject registration to an atlas space for enhanced quantification of anatomy. However, standard diffusion tensor atlases lack information in regions of fiber crossing and are based on adult anatomy. The degree of error associated with applying these atlases to studies of children for example has not yet been estimated but may lead to suboptimal results. This paper describes a novel technique for generating population-specific high angular resolution diffusion imaging (HARDI)-based atlases consisting of labeled regions of homogenous white matter. Our approach uses a fiber orientation distribution (FOD) diffusion model and a data driven clustering algorithm. White matter regional labeling is achieved by our automated data driven clustering algorithm that has the potential to delineate white matter regions based on fiber complexity and orientation. The advantage of such an atlas is that it is study specific and more comprehensive in describing regions of white matter homogeneity as compared to standard anatomical atlases. We have applied this state of the art technique to a dataset consisting of adolescent and preadolescent children, creating one of the first examples of a HARDI-based atlas, thereby establishing the feasibility of the atlas creation framework. The white matter regions generated by our automated clustering algorithm have lower FOD variance than when compared to the regions created from a standard anatomical atlas.

A multiscale and multiblock fuzzy C-means classification method for brain MR images

PURPOSE

Classification of magnetic resonance (MR) images has many clinical and research applications. Because of multiple factors such as noise, intensity inhomogeneity, and partial volume effects, MR image classification can be challenging. Noise in MRI can cause the classified regions to become disconnected. Partial volume effects make the assignment of a single class to one region difficult. Because of intensity inhomogeneity, the intensity of the same tissue can vary with respect to the location of the tissue within the same image. The conventional "hard" classification method restricts each pixel exclusively to one class and often results in crisp results. Fuzzy C-mean (FCM) classification or "soft" segmentation has been extensively applied to MR images, in which pixels are partially classified into multiple classes using varying memberships to the classes. Standard FCM, however, is sensitive to noise and cannot effectively compensate for intensity inhomogeneities. This paper presents a method to obtain accurate MR brain classification using a modified multiscale and multiblock FCM.

METHODS

An automatic, multiscale and multiblock fuzzy C-means (MsbFCM) classification method with MR intensity correction is presented in this paper. We use a bilateral filter to process MR images and to build a multiscale image series by increasing the standard deviation of spatial function and by reducing the standard deviation of range function. At each scale, we separate the image into multiple blocks and for every block a multiscale fuzzy C-means classification method is applied along the scales from the coarse to fine levels in order to overcome the effect of intensity inhomogeneity. The result from a coarse scale supervises the classification in the next fine scale. The classification method is tested with noisy MR images with intensity inhomogeneity.

RESULTS

Our method was compared with the conventional FCM, a modified FCM (MFCM) and multiscale FCM (MsFCM) method. Validation studies were performed on synthesized images with various contrasts, on the simulated brain MR database, and on real MR images. Our MsbFCM method consistently performed better than the conventional FCM, MFCM, and MsFCM methods. The MsbFCM method achieved an overlap ratio of 91% or higher. Experimental results using real MR images demonstrate the effectiveness of the proposed method. Our MsbFCM classification method is accurate and robust for various MR images.

CONCLUSIONS

As our classification method did not assume a Gaussian distribution of tissue intensity, it could be used on other image data for tissue classification and quantification. The automatic classification method can provide a useful quantification tool in neuroimaging and other applications.

ODVBA: optimally-discriminative voxel-based analysis

Gaussian smoothing of images prior to applying voxel-based statistics is an important step in voxel-based analysis and statistical parametric mapping (VBA-SPM) and is used to account for registration errors, to Gaussianize the data and to integrate imaging signals from a region around each voxel. However, it has also become a limitation of VBA-SPM based methods, since it is often chosen empirically and lacks spatial adaptivity to the shape and spatial extent of the region of interest, such as a region of atrophy or functional activity. In this paper, we propose a new framework, named optimally-discriminative voxel-based analysis (ODVBA), for determining the optimal spatially adaptive smoothing of images, followed by applying voxel-based group analysis. In ODVBA, nonnegative discriminative projection is applied regionally to get the direction that best discriminates between two groups, e.g., patients and controls; this direction is equivalent to local filtering by an optimal kernel whose coefficients define the optimally discriminative direction. By considering all the neighborhoods that contain a given voxel, we then compose this information to produce the statistic for each voxel. Finally, permutation tests are used to obtain a statistical parametric map of group differences. ODVBA has been evaluated using simulated data in which the ground truth is known and with data from an Alzheimer's disease (AD) study. The experimental results have shown that the proposed ODVBA can precisely describe the shape and location of structural abnormality.

Compensation of RF field and receiver coil induced inhomogeneity effects in abdominal MR images by a priori knowledge on the human adipose tissue distribution

PURPOSE

To reliably compensate bias field effects in abdominal areas to accurately quantify visceral adipose tissue using standard T1-weighted sequences on MR scanners with up to 3 Tesla (T) field strength.

MATERIALS AND METHODS

Compensation is achieved in two steps: The bias field is first estimated by picking and fitting sampling points from the subcutaneous adipose tissue, using active contours and a thin plate fitting spline. Then, additional sampling points from visceral adipose tissue compartments are detected by thresholding and the bias field estimation is refined. It was compared with an established method using a simulated abdominal image and real 3T data.

RESULTS

At low bias field amplitudes (40-50%), the simulation study showed a good reduction of the mean coefficients of variance (CV) for both approaches (>80%). At higher amplitudes, the CV reduction was significantly higher for our approach (83.6%), compared with LEMS (54.3%). In the real data study, our approach showed reliable reduction of the inhomogeneities, while the LEMS algorithm sometimes even amplified the inhomogeneities.

CONCLUSION

The proposed method enables accurate and reliable segmentation of abdominal adipose tissue using simple thresholding techniques, even in severely corrupted images slices, obtained when using high field strengths and/or phased-array coils.

Molecular image segmentation based on improved fuzzy clustering

Segmentation of molecular images is a difficult task due to the low signal-to-noise ratio of images. A novel two-dimensional fuzzy C-means (2DFCM) algorithm is proposed for the molecular image segmentation. The 2DFCM algorithm is composed of three stages. The first stage is the noise suppression by utilizing a method combining a Gaussian noise filter and anisotropic diffusion techniques. The second stage is the texture energy characterization using a Gabor wavelet method. The third stage is introducing spatial constraints provided by the denoising data and the textural information into the two-dimensional fuzzy clustering. The incorporation of intensity and textural information allows the 2DFCM algorithm to produce satisfactory segmentation results for images corrupted by noise (outliers) and intensity variations. The 2DFCM can achieve 0.96 +/- 0.03 segmentation accuracy for synthetic images under different imaging conditions. Experimental results on a real molecular image also show the effectiveness of the proposed algorithm.

Automatic segmentation of neonatal images using convex optimization and coupled level sets

Accurate segmentation of neonatal brain MR images remains challenging mainly due to their poor spatial resolution, inverted contrast between white matter and gray matter, and high intensity inhomogeneity. Most existing methods for neonatal brain segmentation are atlas-based and voxel-wise. Although active contour/surface models with geometric information constraint have been successfully applied to adult brain segmentation, they are not fully explored in the neonatal image segmentation. In this paper, we propose a novel neonatal image segmentation method by combining local intensity information, atlas spatial prior, and cortical thickness constraint in a single level-set framework. Besides, we also provide a robust and reliable tissue surface initialization for the proposed method by using a convex optimization technique. Thus, tissue segmentation, as well as inner and outer cortical surface reconstruction, can be obtained simultaneously. The proposed method has been tested on a large neonatal dataset, and the validation on 10 neonatal brain images (with manual segmentations) shows very promising results.

Image segmentation using fuzzy region competition and spatial/frequency information

This paper presents a multiphase fuzzy region competition model that takes into account spatial and frequency information for image segmentation. In the proposed energy functional, each region is represented by a fuzzy membership function and a data fidelity term that measures the conformity of spatial and frequency data within each region to (generalized) gaussian densities whose parameters are determined jointly with the segmentation process. Compared with the classical region competition model, our approach gives soft segmentation results via the fuzzy membership functions, and moreover, the use of frequency data provides additional region information that can improve the overall segmentation result. To efficiently solve the minimization of the energy functional, we adopt an alternate minimization procedure and make use of Chambolle's fast duality projection algorithm. We apply the proposed method to synthetic and natural textures as well as real-world natural images. Experimental results show that our proposed method has very promising segmentation performance compared with the current state-of-the-art approaches.

Pattern Analysis in Neuroimaging: Beyond Two-Class Categorization

One of the many advantages of multivariate pattern recognition approaches over conventional mass-univariate group analysis using voxel-wise statistical tests is their potential to provide highly sensitive and specific markers of diseases on an individual basis. However, a vast majority of imaging problems addressed by pattern recognition are viewed from the perspective of a two-class classification. In this article, we provide a summary of selected works that propose solutions to biomedical problems where the widely-accepted classification paradigm is not appropriate. These pattern recognition approaches address common challenges in many imaging studies: high heterogeneity of populations and continuous progression of diseases. We focus on diseases associated with aging and propose that clustering-based approaches may be more suitable for disentanglement of the underlying heterogeneity, while high-dimensional pattern regression methodology is appropriate for prediction of continuous and gradual clinical progression from magnetic resonance brain images.

Segmentation of multiple sclerosis lesions in MR images: a review

INTRODUCTION

Multiple sclerosis (MS) is an inflammatory demyelinating disease that the parts of the nervous system through the lesions generated in the white matter of the brain. It brings about disabilities in different organs of the body such as eyes and muscles. Early detection of MS and estimation of its progression are critical for optimal treatment of the disease.

METHODS

For diagnosis and treatment evaluation of MS lesions, they may be detected and segmented in Magnetic Resonance Imaging (MRI) scans of the brain. However, due to the large amount of MRI data to be analyzed, manual segmentation of the lesions by clinical experts translates into a very cumbersome and time consuming task. In addition, manual segmentation is subjective and prone to human errors. Several groups have developed computerized methods to detect and segment MS lesions. These methods are not categorized and compared in the past.

RESULTS

This paper reviews and compares various MS lesion segmentation methods proposed in recent years. It covers conventional methods like multilevel thresholding and region growing, as well as more recent Bayesian methods that require parameter estimation algorithms. It also covers parameter estimation methods like expectation maximization and adaptive mixture model which are among unsupervised techniques as well as kNN and Parzen window methods that are among supervised techniques.

CONCLUSIONS

Integration of knowledge-based methods such as atlas-based approaches with Bayesian methods increases segmentation accuracy. In addition, employing intelligent classifiers like Fuzzy C-Means, Fuzzy Inference Systems, and Artificial Neural Networks reduces misclassified voxels.

Segmentation of multiple sclerosis lesions in MR images: a review

INTRODUCTION

Multiple sclerosis (MS) is an inflammatory demyelinating disease that the parts of the nervous system through the lesions generated in the white matter of the brain. It brings about disabilities in different organs of the body such as eyes and muscles. Early detection of MS and estimation of its progression are critical for optimal treatment of the disease.

METHODS

For diagnosis and treatment evaluation of MS lesions, they may be detected and segmented in Magnetic Resonance Imaging (MRI) scans of the brain. However, due to the large amount of MRI data to be analyzed, manual segmentation of the lesions by clinical experts translates into a very cumbersome and time consuming task. In addition, manual segmentation is subjective and prone to human errors. Several groups have developed computerized methods to detect and segment MS lesions. These methods are not categorized and compared in the past.

RESULTS

This paper reviews and compares various MS lesion segmentation methods proposed in recent years. It covers conventional methods like multilevel thresholding and region growing, as well as more recent Bayesian methods that require parameter estimation algorithms. It also covers parameter estimation methods like expectation maximization and adaptive mixture model which are among unsupervised techniques as well as kNN and Parzen window methods that are among supervised techniques.

CONCLUSIONS

Integration of knowledge-based methods such as atlas-based approaches with Bayesian methods increases segmentation accuracy. In addition, employing intelligent classifiers like Fuzzy C-Means, Fuzzy Inference Systems, and Artificial Neural Networks reduces misclassified voxels.

Intelligence and cortical thickness in children with complex partial seizures

Prior studies on healthy children have demonstrated regional variations and a complex and dynamic relationship between intelligence and cerebral tissue. Yet, there is little information regarding the neuroanatomical correlates of general intelligence in children with epilepsy compared to healthy controls. In vivo imaging techniques, combined with methods for advanced image processing and analysis, offer the potential to examine quantitative mapping of brain development and its abnormalities in childhood epilepsy. A surface-based, computational high resolution 3-D magnetic resonance image analytic technique was used to compare the relationship of cortical thickness with age and intelligence quotient (IQ) in 65 children and adolescents with complex partial seizures (CPS) and 58 healthy controls, aged 6-18 years. Children were grouped according to health status (epilepsy; controls) and IQ level (average and above; below average) and compared on age-related patterns of cortical thickness. Our cross-sectional findings suggest that disruption in normal age-related cortical thickness expression is associated with intelligence in pediatric CPS patients both with average and below average IQ scores.

A MR Brain Classification Method Based on Multiscale and Multiblock Fuzzy C-means

A fully automatic, multiscale and multiblock fuzzy C-means (MsbFCM) classification method with intensity correction for MR images is presented in this paper. We use a bilateral filter to process MR images and to build a multiscale image series by increasing the standard deviation of spatial function and reducing the standard deviation of range function. We separate every scale image into multiple blocks and for every block a multiscale fuzzy C-means classification method is applied along the scales from the coarse to fine levels to overcome the effect of intensity inhomogeneity. The method is robust for noise MR images with intensity inhomogeneity because of its multiscale and multiblock bilateral filtering scheme. Our method was compared with the conventional FCM, a modified FCM (MFCM) and multiscale FCM (MsFCM) method on synthesized images, simulated brain MR images, and real MR images. The MsbFCM method achieved an overlap ratio of greater than 91% as validated by the ground truth even if original images have 9% noise and 40% intensity inhomogeneity. Experimental results using real MR images demonstrate the effectiveness of the proposed method. Our MsbFCM classification method is accurate and robust for various MR images.

Intensity Inhomogeneity Correction of Magnetic Resonance Images using Patches

This paper presents a patch-based non-parametric approach to the correction of intensity inhomogeneity from magnetic resonance (MR) images of the human brain. During image acquisition, the inhomogeneity present in the radio-frequency coil, is usually manifested on the reconstructed MR image as a smooth shading effect. This artifact can significantly deteriorate the performance of any kind of image processing algorithm that uses intensities as a feature. Most of the current inhomogeneity correction techniques use explicit smoothness assumptions on the inhomogeneity field, which sometimes limit their performance if the actual inhomogeneity is not smooth, a problem that becomes prevalent in high fields. The proposed patch-based inhomogeneity correction method does not assume any parametric smoothness model, instead, it uses patches from an atlas of an inhomogeneity-free image to do the correction. Preliminary results show that the proposed method is comparable to N3, a current state of the art method, when the inhomogeneity is smooth, and outperforms N3 when the inhomogeneity contains non-smooth elements.

TOPOLOGY PRESERVING AUTOMATIC SEGMENTATION OF THE SPINAL CORD IN MAGNETIC RESONANCE IMAGES

Magnetic resonance images of the spinal cord play an important role in studying neurological diseases, particularly multiple sclerosis, where spinal cord atrophy can provide a measure of disease progression and disability. Current practices involve segmenting the spinal cord manually, which can be an inconsistent and time-consuming process. We present an automatic segmentation method for the spinal cord using a combination of deformable atlas based registration and topology preserving classification. Using real MR data, our method is shown to be highly accurate when compared to segmentations by manual raters. In addition, our results always maintain the correct topology of the spinal cord, therefore providing segmentations more consistent with the known anatomy.

A modified possibilistic fuzzy c-means clustering algorithm for bias field estimation and segmentation of brain MR image

A modified possibilistic fuzzy c-means clustering algorithm is presented for fuzzy segmentation of magnetic resonance (MR) images that have been corrupted by intensity inhomogeneities and noise. By introducing a novel adaptive method to compute the weights of local spatial in the objective function, the new adaptive fuzzy clustering algorithm is capable of utilizing local contextual information to impose local spatial continuity, thus allowing the suppression of noise and helping to resolve classification ambiguity. To estimate the intensity inhomogeneity, the global intensity is introduced into the coherent local intensity clustering algorithm and takes the local and global intensity information into account. The segmentation target therefore is driven by two forces to smooth the derived optimal bias field and improve the accuracy of the segmentation task. The proposed method has been successfully applied to 3 T, 7 T, synthetic and real MR images with desirable results. Comparisons with other approaches demonstrate the superior performance of the proposed algorithm. Moreover, the proposed algorithm is robust to initialization, thereby allowing fully automatic applications.

Multi-Kernel Classification for Integration of Clinical and Imaging Data: Application to Prediction of Cognitive Decline in Older Adults

Diagnosis of neurologic and neuropsychiatric disorders typically involves considerable assessment including clinical observation, neuroimaging, and biological and neuropsychological measurements. While it is reasonable to expect that the integration of neuroimaging data and complementary non-imaging measures is likely to improve early diagnosis on individual basis, due to technical challenges associated with the task of combining different data types, medical image pattern recognition analysis has been largely focusing solely on neuroimaging evaluations. In this paper, we explore the potential of integrating neuroimaging and clinical information within a pattern classification framework, and propose that the multi-kernel learning (MKL) paradigm may be suitable for building a multimodal classifier of a disorder, as well as for automatic identification of the relevance of each information type. We apply our approach to the problem of detecting cognitive decline in healthy older adults from single-visit evaluations, and show that the performance of a classifier can be improved when nouroimaging and clinical evaluations are used simultaneously within a MKL-based classification framework.

UNDERSTANDING HETEROGENEITY IN NORMAL OLDER ADULT POPULATIONS VIA CLUSTERING OF LONGITUDINAL DATA

Populations of healthy older individuals are often highly heterogeneous, as prevalence of various underlying pathologies increases with age. Finding coherent groups of normal older adults may allow to identify subpopulations that are at risk of developing Alzheimer's disease (AD). In this paper, we propose an approach that utilizes longitudinal magnetic resonance imaging (MRI) data to obtain natural groupings of older adult subjects via an unsupervised (i.e., clustering) technique. We develop a k-medoids-like clustering algorithm that simultaneously finds clusters of longitudinal images, as well as weights brain regions in such a way that the obtained clusters are maximally coherent. We propose a cluster-based measure that reflects the individual subject's cognitive decline. The proposed method is unsupervised and is suitable for analyzing AD at its very early stages.

Semi-supervised pattern classification of medical images: application to mild cognitive impairment (MCI)

Many progressive disorders are characterized by unclear or transient diagnoses for specific subgroups of patients. Commonly used supervised pattern recognition methodology may not be the most suitable approach to deriving image-based biomarkers in such cases, as it relies on the availability of categorically labeled data (e.g., patients and controls). In this paper, we explore the potential of semi-supervised pattern classification to provide image-based biomarkers in the absence of precise diagnostic information for some individuals. We employ semi-supervised support vector machines (SVM) and apply them to the problem of classifying MR brain images of patients with uncertain diagnoses. We examine patterns in serial scans of ADNI participants with mild cognitive impairment (MCI), and propose that in the absence of sufficient follow-up evaluations of individuals with MCI, semi-supervised strategy is potentially more appropriate than the fully-supervised paradigm employed up to date.

Segmentation priors from local image properties: without using bias field correction, location-based templates, or registration

We present a novel approach for generating information about a voxel's tissue class membership based on its signature--a collection of local image textures estimated over a range of neighborhood sizes. The approach produces a form of tissue class priors that can be used to initialize and regularize image segmentation. The signature-based approach is a departure from current location-based methods, which derive tissue class likelihoods based on a voxel's location in standard template space. To use location-based priors, one needs to register the volume in question to the template space, and estimate the image intensity bias field. Two optimizations, over more than a thousand parameters, are needed when high order nonlinear registration is employed. In contrast, the signature-based approach is independent of volume orientation, voxel position, and largely insensitive to bias fields. For these reasons, the approach does not require the use of population derived templates. The prior information is generated from variations in image texture statistics as a function of spatial scale, and an SVM approach is used to associate signatures with tissue types. With the signature-based approach, optimization is needed only during the training phase for the parameter estimation stages of the SVM hyperplanes, and associated PDFs; a training process separate from the segmentation step. We found that signature-based priors were superior to location-based ones aligned under favorable conditions, and that signature-based priors result in improved segmentation when replacing location-based ones in FAST (Zhang et al., 2001), a widely used segmentation program. The software implementation of this work is freely available as part of AFNI http://afni.nimh.nih.gov.

Integrating spatial fuzzy clustering with level set methods for automated medical image segmentation

The performance of the level set segmentation is subject to appropriate initialization and optimal configuration of controlling parameters, which require substantial manual intervention. A new fuzzy level set algorithm is proposed in this paper to facilitate medical image segmentation. It is able to directly evolve from the initial segmentation by spatial fuzzy clustering. The controlling parameters of level set evolution are also estimated from the results of fuzzy clustering. Moreover the fuzzy level set algorithm is enhanced with locally regularized evolution. Such improvements facilitate level set manipulation and lead to more robust segmentation. Performance evaluation of the proposed algorithm was carried on medical images from different modalities. The results confirm its effectiveness for medical image segmentation.

Fast algorithm for calculation of inhomogeneity gradient in magnetic resonance imaging data

PURPOSE

To develop and implement a new approach for correcting the intensity inhomogeneity in magnetic resonance imaging (MRI) data.

MATERIALS AND METHODS

The algorithm is based on the assumption that intensity inhomogeneity in MR data is multiplicative and smoothly varying. Using a statistically stable method, the algorithm first calculates the partial derivative of the inhomogeneity gradient across the data. The algorithm then solves for the gradient field and fits it to a parametric surface. It was tested on both simulated and real human and animal MRI data.

RESULTS

The algorithm is shown to restore the homogeneity in all images that were tested. On real human brain images the algorithm demonstrated superior or comparable performance relative to some of the commonly used intensity inhomogeneity correction methods such as SPM, BrainSuite, and N3.

CONCLUSION

The proposed algorithm provides an alternative method for correcting the intensity inhomogeneity in MR images. It is shown to be fast and its performance is superior or comparable to algorithms described in the published literature. Due to its generality, this algorithm is applicable to MR images of both humans and animals.

Semi-supervised cluster analysis of imaging data

In this paper, we present a semi-supervised clustering-based framework for discovering coherent subpopulations in heterogeneous image sets. Our approach involves limited supervision in the form of labeled instances from two distributions that reflect a rough guess about subspace of features that are relevant for cluster analysis. By assuming that images are defined in a common space via registration to a common template, we propose a segmentation-based method for detecting locations that signify local regional differences in the two labeled sets. A PCA model of local image appearance is then estimated at each location of interest, and ranked with respect to its relevance for clustering. We develop an incremental k-means-like algorithm that discovers novel meaningful categories in a test image set. The application of our approach in this paper is in analysis of populations of healthy older adults. We validate our approach on a synthetic dataset, as well as on a dataset of brain images of older adults. We assess our method's performance on the problem of discovering clusters of MR images of human brain, and present a cluster-based measure of pathology that reflects the deviation of a subject's MR image from normal (i.e. cognitively stable) state. We analyze the clusters' structure, and show that clustering results obtained using our approach correlate well with clinical data.

Development of a virtual reality guided diagnostic tool based on magnetic resonance imaging

Computed tomography (CT) and virtual reality (VR) made it possible to create internal views of the human body without actual penetration. During the last two decades, several endoscopic diagnosis procedures have received virtual counter candidates. This paper presents an own concept of a virtual reality guided diagnostic tool, based on magnetic resonance images representing parallel cross-sections of the investigated organ. A series of image processing methods are proposed for image quality enhancement, accurate segmentation in two dimensions, and three-dimensional reconstruction of detected surfaces. These techniques provide improved accuracy in image segmentation, and thus they represent excellent support for three dimensional imaging. The implemented software system allows interactive navigation within the investigated volume, and provides several facilities to quantify important physical properties including distances, areas, and volumes.

A novel fuzzy C-means algorithm for unsupervised heterogeneous tumor quantification in PET

PURPOSE

Accurate and robust image segmentation was identified as one of the most challenging issues facing PET quantification in oncological imaging. This difficulty is compounded by the low spatial resolution and high noise characteristics of PET images. The fuzzy C-means (FCM) clustering algorithm was largely used in various medical image segmentation approaches. However, the algorithm is sensitive to both noise and intensity heterogeneity since it does not take into account spatial contextual information.

METHODS

To overcome this limitation, a new fuzzy segmentation technique adapted to typical noisy and low resolution oncological PET data is proposed. PET images smoothed using a nonlinear anisotropic diffusion filter are added as a second input to the proposed FCM algorithm to incorporate spatial information (FCM-S). In addition, a methodology was developed to integrate the a trous wavelet transform in the standard FCM algorithm (FCM-SW) to allow handling of heterogeneous lesions' uptake. The algorithm was applied to the simulated data of the NCAT phantom, incorporating heterogeneous lesions in the lung and clinical PET/CT images of 21 patients presenting with histologically proven nonsmall-cell lung cancer (NSCLC) and 7 patients presenting with laryngeal squamous cell carcinoma (LSCC) to assess its performance for segmenting tumors with arbitrary size, shape, and tracer uptake. For NSCLC patients, the maximal tumor diameters measured from the macroscopic examination of the surgical specimen served as the ground truth for comparison with the maximum diameter estimated by the segmentation technique, whereas for LSCC patients, the 3D macroscopic tumor volume was considered as the ground truth for comparison with the corresponding PET-based volume. The proposed algorithm was also compared to the classical FCM segmentation technique.

RESULTS

There is a good correlation (R2 = 0.942) between the actual maximal diameter of primary NSCLC tumors estimated using the proposed PET segmentation procedure and those measured from the macroscopic examination, and the regression line agreed well with the line of identity (slope = 1.08) for the group analysis of the clinical data. The standard FCM algorithm seems to underestimate actual maximal diameters of the clinical data, resulting in a mean error of -4.6 mm (relative error of -10.8 +/- 23.1%) for all data sets. The mean error of maximal diameter estimation was reduced to 0.1 mm (0.9 +/- 14.4%) using the proposed FCM-SW algorithm. Likewise, the mean relative error on the estimated volume for LSCC patients was reduced from 21.7 +/- 22.0% for FCM to 8.6 +/- 28.3% using the proposed FCM-SW technique.

CONCLUSIONS

A novel unsupervised PET image segmentation technique that allows the quantification of lesions in the presence of heterogeneity of tracer uptake was developed and evaluated. The technique is being further refined and assessed in clinical setting to delineate treatment volumes for the purpose of PET-guided radiation therapy treatment planning but could find other applications in clinical oncology such as the assessment of response to treatment.

Positional brain deformation visualized with magnetic resonance morphometry

OBJECTIVE

To assess and visualize gravitational effects on brain morphology and the position of the brain within the skull by magnetic resonance (MR) morphometry in order to identify confounding effects and possible sources of error for accurate planning of neurosurgical interventions.

METHODS

Three-dimensional MR imaging data sets of 13 healthy adults were acquired in different positions in the scanner. With a morphometric approach, data sets were evaluated by deformation field analysis and the brain boundary shift integral. Distortions of the brain were assessed comparing right versus left and prone versus supine positioning, respectively.

RESULTS

Two effects could be differentiated: 1) greatest brain deformation of up to 1.7 mm predominantly located around central brain structures in the lateral direction and a less pronounced change after position changes in posterior-anterior direction, and 2) the brain boundary shift integral depicted position-dependent brain shift relative to the inner skull.

CONCLUSION

Position-dependent effects on brain structure may undermine the accuracy of neuronavigational and other neurosurgical procedures. Furthermore, in longitudinal MR volumetric studies, gravitational effects should be kept in mind and the scanning position should be rigidly controlled for.

Estimating Myocardial Motion by 4D Image Warping

A method for spatio-temporally smooth and consistent estimation of cardiac motion from MR cine sequences is proposed. Myocardial motion is estimated within a 4-dimensional (4D) registration framework, in which all 3D images obtained at different cardiac phases are simultaneously registered. This facilitates spatio-temporally consistent estimation of motion as opposed to other registration-based algorithms which estimate the motion by sequentially registering one frame to another. To facilitate image matching, an attribute vector (AV) is constructed for each point in the image, and is intended to serve as a "morphological signature" of that point. The AV includes intensity, boundary, and geometric moment invariants (GMIs). Hierarchical registration of two image sequences is achieved by using the most distinctive points for initial registration of two sequences and gradually adding less-distinctive points to refine the registration. Experimental results on real data demonstrate good performance of the proposed method for cardiac image registration and motion estimation. The motion estimation is validated via comparisons with motion estimates obtained from MR images with myocardial tagging.

Fast and robust multi-atlas segmentation of brain magnetic resonance images

We introduce an optimised pipeline for multi-atlas brain MRI segmentation. Both accuracy and speed of segmentation are considered. We study different similarity measures used in non-rigid registration. We show that intensity differences for intensity normalised images can be used instead of standard normalised mutual information in registration without compromising the accuracy but leading to threefold decrease in the computation time. We study and validate also different methods for atlas selection. Finally, we propose two new approaches for combining multi-atlas segmentation and intensity modelling based on segmentation using expectation maximisation (EM) and optimisation via graph cuts. The segmentation pipeline is evaluated with two data cohorts: IBSR data (N=18, six subcortial structures: thalamus, caudate, putamen, pallidum, hippocampus, amygdala) and ADNI data (N=60, hippocampus). The average similarity index between automatically and manually generated volumes was 0.849 (IBSR, six subcortical structures) and 0.880 (ADNI, hippocampus). The correlation coefficient for hippocampal volumes was 0.95 with the ADNI data. The computation time using a standard multicore PC computer was about 3-4 min. Our results compare favourably with other recently published results.

Using cerebral white matter for estimation of nondisplaceable binding of 5-HT1A receptors in temporal lobe epilepsy

The estimation of nondisplaceable binding from cerebellar white matter, rather than from whole cerebellum, was proposed for the PET tracer carbonyl-(11)C-WAY-100635 (N-{2-[4-(2-methoxyphenyl)-1-piperazinyl]ethyl}-N-(2-pyridyl)cyclohexanecarboxamidel]) because of the heterogeneity of total ligand binding in this region. For the 5-hydroxytryptamine receptor 1A (5-HT(1A)) antagonist (18)F-N-{2-[4-(2-methoxyphenyl)piperazin-1-yl]ethyl}-N-2-pyridyl)trans-4-fluorocyclohexanecarboxamide ((18)F-FCWAY), the estimation of nondisplaceable binding from cerebellum (V(ND)) may be additionally biased by spillover of (18)F-fluoride activity from skull. We aimed to assess the effect of using cerebral white matter as reference region on detection of group differences in 5-HT(1A) binding with PET and (18)F-FCWAY.

METHODS

In 22 temporal lobe epilepsy patients (TLE) and 10 healthy controls, (18)F-FCWAY distribution volume in cerebral white matter (V(WM)) was computed using an extrapolation method as part of a partial-volume correction (PVC) algorithm. To assess the feasibility of applying this method to clinical studies in which PVC is not performed, V(WM) was also calculated by placing circular, 6-mm-diameter regions of interest (ROIs) in the centrum semiovalis on parametric images. Binding potentials were BP(F) = (V(T) - V(ND))/f(P) and BP(F-WM) = (V(T) - V(WM))/f(P), where V(T) is total distribution volume and f(P) = (18)F-FCWAY plasma free fraction. Statistical analysis was performed using t tests and linear regression.

RESULTS

In the whole group, V(WM) was 14% +/- 19% lower than V(ND) (P < 0.05). V(WM)/f(P) was significantly (P < 0.05) lower in patients than in controls. All significant (P < 0.05) reductions of 5-HT(1A) receptor availability in TLE patients detected by BP(F) were also detected using BP(F-WM). Significant (P < 0.05) reductions of 5-HT(1A) specific binding were detected by BP(F-WM), but not BP(F), in ipsilateral inferior temporal cortex, contralateral fusiform gyrus, and contralateral amygdala. However, effect sizes were similar for BP(F-WM) and BP(F). The value of V(WM) calculated with the ROI approach did not significantly (P > 0.05) differ from that calculated with the extrapolation approach (0.67 +/- 0.32 mL/mL and 0.72 +/- 0.34 mL/mL, respectively).

CONCLUSION

Cerebral white matter can be used for the quantification of nondisplaceable binding of 5-HT(1A) without loss of statistical power for detection of regional group differences. The ROI approach is a good compromise between computational complexity and sensitivity to spillover of activity, and it appears suitable to studies in which PVC is not performed. For (18)F-FCWAY, this approach has the advantage of avoiding spillover of (18)F-fluoride activity onto the reference region.

Intensity Standardization Simplifies Brain MR Image Segmentation

Typically, brain MR images present significant intensity variation across patients and scanners. Consequently, training a classifier on a set of images and using it subsequently for brain segmentation may yield poor results. Adaptive iterative methods usually need to be employed to account for the variations of the particular scan. These methods are complicated, difficult to implement and often involve significant computational costs. In this paper, a simple, non-iterative method is proposed for brain MR image segmentation. Two preprocessing techniques, namely intensity inhomogeneity correction, and more importantly MR image intensity standardization, used prior to segmentation, play a vital role in making the MR image intensities have a tissue-specific numeric meaning, which leads us to a very simple brain tissue segmentation strategy.Vectorial scale-based fuzzy connectedness and certain morphological operations are utilized first to generate the brain intracranial mask. The fuzzy membership value of each voxel within the intracranial mask for each brain tissue is then estimated. Finally, a maximum likelihood criterion with spatial constraints taken into account is utilized in classifying all voxels in the intracranial mask into different brain tissue groups. A set of inhomogeneity corrected and intensity standardized images is utilized as a training data set. We introduce two methods to estimate fuzzy membership values. In the first method, called SMG (for simple membership based on a gaussian model), the fuzzy membership value is estimated by fitting a multivariate Gaussian model to the intensity distribution of each brain tissue whose mean intensity vector and covariance matrix are estimated and fixed from the training data sets. The second method, called SMH (for simple membership based on a histogram), estimates fuzzy membership value directly via the intensity distribution of each brain tissue obtained from the training data sets. We present several studies to evaluate the performance of these two methods based on 10 clinical MR images of normal subjects and 10 clinical MR images of Multiple Sclerosis (MS) patients. A quantitative comparison indicates that both methods have overall better accuracy than the k-nearest neighbors (kNN) method, and have much better efficiency than the Finite Mixture (FM) model based Expectation-Maximization (EM) method. Accuracy is similar for our methods and EM method for the normal subject data sets, but much better for our methods for the patient data sets.

A clinically feasible method to estimate pharmacokinetic parameters in breast cancer

Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) is the MRI technique of choice for detecting breast cancer, which can be roughly classified as either quantitative or semiquantitative. The major advantage of quantitative DCE-MRI is its ability to provide pharmacokinetic parameters such as volume transfer constant (Ktrans) and extravascular extracellular volume fraction (ve). However, semiquantitative DCE-MRI is still the clinical MRI technique of choice for breast cancer diagnosis due to several major practical difficulties in the implementation of quantitative DCE-MRI in a clinical setting, including (1) long acquisition necessary to acquire 3D T1(0) map, (2) challenges in obtaining accurate artery input function (AIF), (3) long computation time required by conventional nonlinear least square (NLS) fitting, and (4) many illogical values often generated by conventional NLS method. The authors developed a new analysis method to estimate pharmacokinetic parameters Ktrans and ve from clinical DCE-MRI data, including fixed T1(0) to eliminate the long acquisition for T1(0) map and "reference region" model to remove the requirement of measuring AIF. Other techniques used in our analysis method are (1) an improved formula to calculate contrast agent (CA) concentration based on signal intensity of SPGR data, (2) FCM clustering-based techniques for automatic segmentation and generation of a clustered concentration data set (3) an empirical formula for CA time course to fit the clustered data sets, and (4) linear regression for the estimation of pharmacokinetic parameters. Preliminary results from computer simulation and clinical study of 39 patients have demonstrated (1) the feasibility of their analysis method for estimating Ktrans and ve from clinical DCE-MRI data, (2) significantly less illogical values compared to NLS method (typically less than 1% versus more than 7%), (3) relative insensitivity to the noise in DCE-MRI data; (4) reduction in computation time by a factor of more than 30 times compared to NLS method on average, (5) high statistic correlation between the method used and NLS method (correlation coefficients: 0.941 for Ktrans and 0.881 for ve), and (6) the potential clinical usefulness of the new method.

Joint brain parametric T1-map segmentation and RF inhomogeneity calibration

We propose a constrained version of Mumford and Shah's (1989) segmentation model with an information-theoretic point of view in order to devise a systematic procedure to segment brain magnetic resonance imaging (MRI) data for parametric T(1)-Map and T(1)-weighted images, in both 2-D and 3D settings. Incorporation of a tuning weight in particular adds a probabilistic flavor to our segmentation method, and makes the 3-tissue segmentation possible. Moreover, we proposed a novel method to jointly segment the T(1)-Map and calibrate RF Inhomogeneity (JSRIC). This method assumes the average T(1) value of white matter is the same across transverse slices in the central brain region, and JSRIC is able to rectify the flip angles to generate calibrated T(1)-Maps. In order to generate an accurate T(1)-Map, the determination of optimal flip-angles and the registration of flip-angle images are examined. Our JSRIC method is validated on two human subjects in the 2D T(1)-Map modality and our segmentation method is validated by two public databases, BrainWeb and IBSR, of T(1)-weighted modality in the 3D setting.

Three-dimensional segmentation of tumors from CT image data using an adaptive fuzzy system

A new segmentation method using a fuzzy rule based system to segment tumors in a three-dimensional CT data was developed. To initialize the segmentation process, the user selects a region of interest (ROI) within the tumor in the first image of the CT study set. Using the ROI's spatial and intensity properties, fuzzy inputs are generated for use in the fuzzy rules inference system. With a set of predefined fuzzy rules, the system generates a defuzzified output for every pixel in terms of similarity to the object. Pixels with the highest similarity values are selected as tumor. This process is automatically repeated for every subsequent slice in the CT set without further user input, as the segmented region from the previous slice is used as the ROI for the current slice. This creates a propagation of information from the previous slices, used to segment the current slice. The membership functions used during the fuzzification and defuzzification processes are adaptive to the changes in the size and pixel intensities of the current ROI. The method is highly customizable to suit different needs of a user, requiring information from only a single two-dimensional image. Test cases success in segmenting the tumor from seven of the 10 CT datasets with <10% false positive errors and five test cases with <10% false negative errors. The consistency of the segmentation results statistics also showed a high repeatability factor, with low values of inter- and intra-user variability for both methods.

Imaging incorporation of circulating docosahexaenoic acid into the human brain using positron emission tomography

Docosahexaenoic acid (DHA; 22:6n-3) is a critical constituent of the brain, but its metabolism has not been measured in the human brain in vivo. In monkeys, using positron emission tomography (PET), we first showed that intravenously injected [1-(11)C]DHA mostly entered nonbrain organs, with approximately 0.5% entering the brain. Then, using PET and intravenous [1-(11)C]DHA in 14 healthy adult humans, we quantitatively imaged regional rates of incorporation (K*) of DHA. We also imaged regional cerebral blood flow (rCBF) using PET and intravenous [(15)O]water. Values of K* for DHA were higher in gray than white matter regions and correlated significantly with values of rCBF in 12 of 14 subjects despite evidence that rCBF does not directly influence K*. For the entire human brain, the net DHA incorporation rate J(in), the product of K*, and the unesterified plasma DHA concentration equaled 3.8 +/- 1.7 mg/day. This net rate is equivalent to the net rate of DHA consumption by brain and, considering the reported amount of DHA in brain, indicates that the half-life of DHA in the human brain approximates 2.5 years. Thus, PET with [1-(11)C]DHA can be used to quantify regional and global human brain DHA metabolism in relation to health and disease.

Improvement of brain segmentation accuracy by optimizing non-uniformity correction using N3

Smoothly varying and multiplicative intensity variations within MR images that are artifactual, can reduce the accuracy of automated brain segmentation. Fortunately, these can be corrected. Among existing correction approaches, the nonparametric non-uniformity intensity normalization method N3 (Sled, J.G., Zijdenbos, A.P., Evans, A.C., 1998. Nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans. Med. Imag. 17, 87-97.) is one of the most frequently used. However, at least one recent study (Boyes, R.G., Gunter, J.L., Frost, C., Janke, A.L., Yeatman, T., Hill, D.L.G., Bernstein, M.A., Thompson, P.M., Weiner, M.W., Schuff, N., Alexander, G.E., Killiany, R.J., DeCarli, C., Jack, C.R., Fox, N.C., 2008. Intensity non-uniformity correction using N3 on 3-T scanners with multichannel phased array coils. NeuroImage 39, 1752-1762.) suggests that its performance on 3 T scanners with multichannel phased-array receiver coils can be improved by optimizing a parameter that controls the smoothness of the estimated bias field. The present study not only confirms this finding, but additionally demonstrates the benefit of reducing the relevant parameter values to 30-50 mm (default value is 200 mm), on white matter surface estimation as well as the measurement of cortical and subcortical structures using FreeSurfer (Martinos Imaging Centre, Boston, MA). This finding can help enhance precision in studies where estimation of cerebral cortex thickness is critical for making inferences.