Recommendations to improve imaging and analysis of brain lesion load and atrophy in longitudinal studies of multiple sclerosis.

Focal lesions and brain atrophy are the most extensively studied aspects of multiple sclerosis (MS), but the image acquisition and analysis techniques used can be further improved, especially those for studying within-patient changes of lesion load and atrophy longitudinally. Improved accuracy and sensitivity will reduce the numbers of patients required to detect a given treatment effect in a trial, and ultimately, will allow reliable characterization of individual patients for personalized treatment. Based on open issues in the field of MS research, and the current state of the art in magnetic resonance image analysis methods for assessing brain lesion load and atrophy, this paper makes recommendations to improve these measures for longitudinal studies of MS. Briefly, they are (1) images should be acquired using 3D pulse sequences, with near-isotropic spatial resolution and multiple image contrasts to allow more comprehensive analyses of lesion load and atrophy, across timepoints. Image artifacts need special attention given their effects on image analysis results. (2) Automated image segmentation methods integrating the assessment of lesion load and atrophy are desirable. (3) A standard dataset with benchmark results should be set up to facilitate development, calibration, and objective evaluation of image analysis methods for MS.

Structural asymmetries in the human brain: a voxel-based statistical analysis of 142 MRI scans.

The use of computational approaches in the analysis of high resolution magnetic resonance images (MRI) of the human brain provides a powerful tool for in vivo studies of brain anatomy. Here, we report results obtained with a voxel-wise statistical analysis of hemispheric asymmetries in regional 'amounts' of gray matter, based on MRI scans obtained in 142 healthy young adults. Firstly, the voxel-wise analysis detected the well-known frontal (right > left) and occipital (left > right) petalias. Secondly, our analysis confirmed the presence of left-greater-than-right asymmetries in several posterior language areas, including the planum temporale and the angular gyrus; no significant asymmetry was detected in the anterior language regions. We also found previously described asymmetries in the cingulate sulcus (right > left) and the caudate nucleus (right > left). Finally, in some brain regions we observed highly significant asymmetries that were not reported before, such as in the anterior insular cortex (right > left). The above asymmetries were observed in men and women. Our results thus provide confirmation of the known structural asymmetries in the human brain as well as new findings that may stimulate further research of hemispheric specialization.

Maturation of white matter in the human brain: a review of magnetic resonance studies.

This review focuses on the maturation of brain white-matter, as revealed by magnetic resonance (MR) imaging carried out in healthy subjects. The review begins with a brief description of the nature of the MR signal and its possible biological underpinnings, and proceeds with a description of MR findings obtained in newborns, infants, children and adolescents. On MR images, a significant decrease in water content leads to a decrease of longitudinal relaxation times (T1) and transverse relaxation times (T2) and consequent "adult-like" appearance of T1-weighted and T2-weighted images becomes evident towards the end of the first year of life. Owing to the onset of myelination and the related increase of lipid content, MR images gradually acquire an exquisite grey-white matter contrast in a temporal sequence reflecting the time course of myelination. Albeit less pronounced, age-related changes in white matter continue during childhood and adolescence; white matter increases its overall volume and becomes more myelinated in a region-specific fashion. Detection of more subtle changes during this "late" phase of brain development is greatly aided by computational analyses of MR images. The review also briefly outlines future directions, including the use of novel MR techniques such as diffusion tensor imaging and magnetization transfer, as well as the suggestion for the concurrent use of experimental behavioral test-batteries, with structural MR imaging, to study developmental changes in structure-function relationships.

Quantitative brain magnetic resonance imaging in girls with attention-deficit/hyperactivity disorder.

BACKGROUND: Anatomic studies of boys with attention-deficit/hyperactivity disorder (ADHD) have detected decreased volumes in total and frontal brain, basal ganglia, and cerebellar vermis. We tested these findings in a sample of girls with ADHD. METHODS: Anatomic brain magnetic resonance images from 50 girls with ADHD, of severity comparable with that in previously studied boys, and 50 healthy female control subjects, aged 5 to 15 years, were obtained with a 1.5-T scanner with contiguous 2-mm coronal slices and 1.5-mm axial slices. We measured volumes of total cerebrum, frontal lobes, caudate nucleus, globus pallidus, cerebellum, and cerebellar vermis. Behavioral measures included structured psychiatric interviews, parent and teacher ratings, and the Wechsler vocabulary and block design subtests. RESULTS: Total brain volume was smaller in girls with ADHD than in control subjects (effect size, 0.40; P =.05). As in our previous study in boys with ADHD, girls with ADHD had significantly smaller volumes in the posterior-inferior cerebellar vermis (lobules VIII-X; effect size, 0.54; P =.04), even when adjusted for total cerebral volume and vocabulary score. Patients and controls did not differ in asymmetry in any region. Morphometric differences correlated significantly with several ratings of ADHD severity and were not predicted by past or present stimulant drug exposure. CONCLUSIONS: These results confirm previous findings for boys in the posterior-inferior lobules of the cerebellar vermis. The influence of the cerebellar vermis on prefrontal and striatal circuitry should be explored.

Progressive cortical change during adolescence in childhood-onset schizophrenia. A longitudinal magnetic resonance imaging study.

BACKGROUND: Adolescence provides a window to examine regional and disease-specific late abnormal brain development in schizophrenia. Because previous data showed progressive brain ventricular enlargement for a group of adolescents with childhood-onset schizophrenia at 2-year follow-up, with no significant changes for healthy controls, we hypothesized that there would be a progressive decrease in volume in other brain tissue in these patients during adolescence. METHODS: To examine cortical change, we used anatomical brain magnetic resonance imaging scans for 15 patients with childhood-onset schizophrenia (defined as onset of psychosis by age 12 years) and 34 temporally yoked, healthy adolescents at a mean (SD) age of 13.17 (2.73) years at initial baseline scan and 17.46 (2.96) years at follow-up scan. Cortical gray and white matter volumes were obtained with an automated analysis system that classifies brain tissue into gray matter, white matter, and cerebrospinal fluid and separates the cortex into anatomically defined lobar regions. RESULTS: A significant decrease in cortical gray matter volume was seen for healthy controls in the frontal (2.6%) and parietal (4.1%) regions. For the childhood-onset schizophrenia group, there was a decrease in volume in these regions (10.9% and 8.5%, respectively) as well as a 7% decrease in volume in the temporal gray matter. Thus, the childhood-onset schizophrenia group showed a distinctive disease-specific pattern (multivariate analysis of variance for change X region X diagnosis: F, 3.68; P = .004), with the frontal and temporal regions showing the greatest between-group differences. Changes in white matter volume did not differ significantly between the 2 groups. CONCLUSIONS: Patients with very early-onset schizophrenia had both a 4-fold greater decrease in cortical gray matter volume during adolescence and a disease-specific pattern of change. Etiologic models for these patients' illness, which seem clinically and neurobiologically continuous with later-onset schizophrenia, must take into account both early and late disruptions of brain development.

Structural maturation of neural pathways in children and adolescents: in vivo study.

Structural maturation of fiber tracts in the human brain, including an increase in the diameter and myelination of axons, may play a role in cognitive development during childhood and adolescence. A computational analysis of structural magnetic resonance images obtained in 111 children and adolescents revealed age-related increases in white matter density in fiber tracts constituting putative corticospinal and frontotemporal pathways. The maturation of the corticospinal tract was bilateral, whereas that of the frontotemporal pathway was found predominantly in the left (speech-dominant) hemisphere. These findings provide evidence for a gradual maturation, during late childhood and adolescence, of fiber pathways presumably supporting motor and speech functions.

Design and construction of a realistic digital brain phantom.

After conception and implementation of any new medical image processing algorithm, validation is an important step to ensure that the procedure fulfills all requirements set forth at the initial design stage. Although the algorithm must be evaluated on real data, a comprehensive validation requires the additional use of simulated data since it is impossible to establish ground truth with in vivo data. Experiments with simulated data permit controlled evaluation over a wide range of conditions (e.g., different levels of noise, contrast, intensity artefacts, or geometric distortion). Such considerations have become increasingly important with the rapid growth of neuroimaging, i.e., computational analysis of brain structure and function using brain scanning methods such as positron emission tomography and magnetic resonance imaging. Since simple objects such as ellipsoids or parallelepipedes do not reflect the complexity of natural brain anatomy, we present the design and creation of a realistic, high-resolution, digital, volumetric phantom of the human brain. This three-dimensional digital brain phantom is made up of ten volumetric data sets that define the spatial distribution for different tissues (e.g., grey matter, white matter, muscle, skin, etc.), where voxel intensity is proportional to the fraction of tissue within the voxel. The digital brain phantom can be used to simulate tomographic images of the head. Since the contribution of each tissue type to each voxel in the brain phantom is known, it can be used as the gold standard to test analysis algorithms such as classification procedures which seek to identify the tissue "type" of each image voxel. Furthermore, since the same anatomical phantom may be used to drive simulators for different modalities, it is the ideal tool to test intermodality registration algorithms. The brain phantom and simulated MR images have been made publicly available on the Internet (http://www.bic.mni.mcgill.ca/brainweb).

Improved correlation between scores on the expanded disability status scale and cerebral lesion load in relapsing-remitting multiple sclerosis. Results of the application of new imaging methods.

We hypothesized that a better correlation between MRI and clinical measures of neurological disability using the expanded disability status scale (EDSS) in multiple sclerosis could be obtained by assessing lesion load only in and around the corticospinal tracts, since the EDSS is weighted towards motor and ambulatory deficits. Multiple sclerosis lesions in cerebral MRIs from 39 patients with relapsing-remitting multiple sclerosis were manually painted using a three-dimensional computer display tool and mapped into a standardized three-dimensional coordinate space. Total lesion load was then measured. A mask to expose only the corticospinal tract was extracted from an MRI atlas and used to measure lesion load in the corticospinal tract. To account for the residual anatomical variability among the different MRI volumes after stereotaxic transformation, the corticospinal tract mask was dilated to various degrees and the lesion load remeasured. Spearman's rank correlation coefficient was used to calculate the correlation between the EDSS and total lesion load and corticospinal tract lesion load and between the EDSS subscores and total lesion load and corticospinal tract lesion load. Spearman's rank correlation coefficient between the EDSS and total lesion load was 0.6, probably reflecting the rather broad EDSS range represented in the study. The highest correlation of 0.67 was between the EDSS and corticospinal tract lesion load, dilated with a blurring kernel of 8-10 mm. The pyramidal subscore alone showed a weaker correlation with total lesion load, and with corticospinal tract lesion load, than did the overall EDSS, possibly reflecting the narrow range of disability in these subscores in patients with EDSS scores of 1-6.5. The imperfect correlation between the EDSS and corticospinal tract lesion load suggests that factors other than cerebral T2-weighted lesion volume are important determinants of disability.

Quantitative assessment of magnetic resonance imaging lesion load in multiple sclerosis.

Changes of lesion load on yearly conventional spin echo (CSE) T2-weighted scans of the brain from patients with multiple sclerosis, measured using computer assisted techniques, are used to monitor long term disease evolution, either natural or modified by treatment. Although lesion load measurements have several advantages over clinical measures of outcome (they provide a more objective and sensitive measure of disease evolution, which has a linear distribution and a more strict relation with the underlying pathology), the poor correlation between changes of lesion load and changes of disability is of concern when using such an approach for monitoring multiple sclerosis trials. In this review, the main sources of variation in T2 lesion load from brain MRI of patients with multiple sclerosis will be considered, along with possible strategies to, at least partially, overcome them. Also, some of the newer fully automated techniques to segment multiple sclerosis lesions, which have been validated against manual outlining, and a recently developed coregistration technique are presented. It is hoped that a more reliable and standardised approach to lesion load measurements in multiple sclerosis will lead to better correlation with clinical disease course, to a higher confidence in the results of trials, and to reduced numbers of scans needed to conduct the trials, thus improving cost efficiency and reducing discomfort of the patients.

A nonparametric method for automatic correction of intensity nonuniformity in MRI data.

A novel approach to correcting for intensity nonuniformity in magnetic resonance (MR) data is described that achieves high performance without requiring a model of the tissue classes present. The method has the advantage that it can be applied at an early stage in an automated data analysis, before a tissue model is available. Described as nonparametric nonuniform intensity normalization (N3), the method is independent of pulse sequence and insensitive to pathological data that might otherwise violate model assumptions. To eliminate the dependence of the field estimate on anatomy, an iterative approach is employed to estimate both the multiplicative bias field and the distribution of the true tissue intensities. The performance of this method is evaluated using both real and simulated MR data.

In vivo morphometry of the intrasulcal gray matter in the human cingulate, paracingulate, and superior-rostral sulci: hemispheric asymmetries, gender differences and probability maps.

Volumes of the intrasulcal gray matter were measured in three cerebral sulci located on the medial wall of the human frontal lobe: cingulate sulcus (CS), paracingulate sulcus (PCS), and superior-rostral sulcus (SRS). The measurements were carried out on T1-weighted 3-D high-resolution magnetic-resonance (MR) images acquired in 105 young right-handed volunteers (42 female and 63 male). Before the measurement, the images were transformed into a standardized stereotaxic space (Talairach and Tournoux [1988] Human Brain: 3-Dimensional Proportional System. An Approach to Cerebral Imaging. Stuttgart, New York: Georg Thieme Verlag), thus removing inter-individual differences in brain size. The intrasulcal gray matter was segmented in a semi-automatic manner. Significant gender differences were found in the volume of the CS (female > male) and the PCS (male > female). Hemispheric asymmetries were observed between the left and right volumes of the intrasulcal gray matter in the anterior (right > left) and posterior (left > right) segments of the CS, as well as between the left and right volumes of the PCS (left > right). There was no interaction between the asymmetries and gender. In addition, significant positive correlations were found between the left and right gray-matter volumes in the anterior (r = 0.43) and posterior (r = 0.66) segments of the CS, whereas significant negative correlations were observed between the gray-matter volumes of the anterior segment of the CS and those of the PCS (left hemisphere: r = -0.48; right hemisphere: r = -0.42). The observed hemispheric asymmetries in the CS and PCS gray-matter volumes are consistent with the proposed role of these structures in the integration of emotions with cognition (CS) and in the control of speech/vocalization (PCS). The pattern of inter-hemispheric correlations in the sulcal gray-matter points to an increasing asynchrony in the foetal development of primary (CS), secondary (SRS), and tertiary (PCS) sulci, respectively. The presence of negative correlations between the two neighbouring sulci (CS and PCS) suggests that a process of compensation could underlie interactions between adjacent primary and tertiary sulci. Besides the above volumetric analysis, we also provide average (probability) maps of the three sulci; the use of such maps for the parcellation of the medial frontal lobe and localization of "peaks" obtained in blood-flow activation studies is discussed.

Morphometric analysis of white matter lesions in MR images: method and validation.

The analysis of MR images is evolving from qualitative to quantitative. More and more, the question asked by clinicians is how much and where, rather than a simple statement on the presence or absence of abnormalities. The authors present a study in which the results obtained with a semiautomatic, multispectral segmentation technique are quantitatively compared to manually delineated regions. The core of the semiautomatic image analysis system is a supervised artificial neural network classifier augmented with dedicated preand postprocessing algorithms, including anisotropic noise filtering and a surface-fitting method for the correction of spatial intensity variations. The study was focused on the quantitation of white matter lesions in the human brain. A total of 36 images from six brain volumes was analyzed twice by each of two operators, under supervision of a neuroradiologist. Both the intra- and interrater variability of the methods were studied in terms of the average tissue area detected per slice, the correlation coefficients between area measurements, and a measure of similarity derived from the kappa statistic. The results indicate that, compared to a manual method, the use of the semiautomatic technique not only facilitates the analysis of the images, but also has similar or lower intra- and interrater variabilities.

Brain segmentation and white matter lesion detection in MR images.

This paper presents a review of methods and techniques that have been proposed for the segmentation of magnetic resonance (MR) images of the brain, with a special emphasis on the segmentation of white matter lesions. First, artifacts affecting MR images (noise, partial volume effect, and shading artifact) are reviewed and methods that have been proposed to correct for these artifacts are discussed. Next, a taxonomy of generic segmentation algorithms is presented, categorized as region-based, edge-based, and classification algorithms. For each category, the applications proposed in the literature are subdivided into 2-D, 3-D, or multimodal approaches. In each case, tables listing authors, bibliographic references, and methods used have been compiled and are presented. This description of segmentation algorithms is followed by a section on techniques proposed specifically for the analysis of white matter lesions. Finally, a section is dedicated to a review and a comparison of validation methods proposed to assess the accuracy and the reliability of the results obtained with various segmentation algorithms.

Automatic detection of intracranial contours in MR images.

Segmentation of the intracranial cavity in medical images is valuable in several research areas such as the quantitative analysis of normal and abnormal brain tissues, the registration of different imaging modalities (MRI, PET, CT) based on surface models of the brain, and the rendering of volume data. Because the manual delineation of the brain contour in the images can be demanding and error prone, an automatic procedure to perform this task is desirable. We have developed and tested a robust method that permits the automatic detection of the intracranial contour in transverse MR images. The method is described and its performance evaluated.

Correction of intensity variations in MR images for computer-aided tissue classification.

A number of supervised and unsupervised pattern recognition techniques have been proposed in recent years for the segmentation and the quantitative analysis of MR images. However, the efficacy of these techniques is affected by acquisition artifacts such as inter-slice, intra-slice, and inter-patient intensity variations. Here a new approach to the correction of intra-slice intensity variations is presented. Results demonstrate that the correction process enhances the performance of backpropagation neural network classifiers designed for the segmentation of the images. Two slightly different versions of the method are presented. The first version fits an intensity correction surface directly to reference points selected by the user in the images. The second version fits the surface to reference points obtained by an intermediate classification operation. Qualitative and quantitative evaluation of both methods reveals that the first one leads to a better correction of the images than the second but that it is more sensitive to operator errors.

Developments towards the slice-wise three-dimensional reconstruction of the distribution of the contrast perfusion in the myocardial muscle from biplane angiographic views.

In theory, radiographic myocardial perfusion imaging allows a quantitative assessment of the functional significance of a coronary stenosis. However, in the conventional two-dimensional projection images there does not exist a one-two-one relationship between a selected myocardial region of interest (ROI) and one particular coronary segment perfusing that area due to over-projection of myocardial regions in front of and behind the selected ROI perfused by other arterial segments, which may result in measurements which are difficult to interpret or even unreliable. To overcome these problems, we have developed two algorithms to determine the spatial distribution of perfusion levels in slices of the heart, selected approximately perpendicular to the left ventricular long axis, from two orthogonal angiographic views: the Segmental Reconstruction Technique (SRT) and the Network Programming Reconstruction Technique (NPRT). Both techniques require a priori geometric information about the myocardium, which can be obtained from the epicardial coronary tree (epicardial boundaries) and the left ventricular lumen (endocardial boundaries). Using the SRT approach, pie-shaped segments are defined for each slice within the myocardial geometric constraints such that superimposition of these segments when projected in orthogonal biplane views is minimal. The reconstruction process uses a model with identical myocardial geometry and definition of segments. Each segment of the model is assigned a relative perfusion level with unit one if no other a priori information is available. In this case, the model contains geometric information only. In case a priori information about expected segmental perfusion levels is available, a level between zero and one is assigned to each segment. The a priori information on the myocardial perfusion levels can be extracted from either anatomic information about the location and severity of existing coronary arterial obstructions, or from a slice adjacent to the one under reconstruction. Using the NPRT approach perfusion levels are computed for each volume picture element of a slice within the reconstructed myocardial geometry, thus resulting in a much higher spatial resolution than the SRT approach. A priori information of perfusion levels must be included in this approach, again based upon anatomical information, or upon the slice adjacent to the one under reconstruction. The very first slice of a myocardial study will be reconstructed by the SRT approach. Extensive computer simulations for the SRT have proved that the mean difference between the actual and reconstructed segmental perfusion levels, on a scale from 0 to 1, is smaller than 0.45 (SEE = 0.0033, REE = 1.80) for various coronary artery disease states without the use of a priori information on expected perfusion levels. This error becomes smaller than 0.36 (SEE = 0.0026, REE = 1.42), if a priori information in the reconstruction technique is included. Similar computer simulations for the NPRT have proved that these mean differences in geometric segments equal to those defined for the SRT, are smaller than 2.94 (SEE = 0.0308, REE = 0.77) on a scale from 0 to 16, without the use of a priori information on expected perfusion levels, and smaller than 1.72 (SEE = 0.0304, REE = 1.10) on the same scale when a priori information is included. Therefore, it may be concluded that slice-wise three-dimensional reconstruction of perfusion levels is feasible from biplane computer-simulated data, and that a similarity exists for mean perfusion levels in corresponding regions in the simulated and reconstructed slices, for various states of single coronary artery disease.