Multispectral analysis of magnetic resonance images

Automatic change detection in multimodal serial MRI: application to multiple sclerosis lesion evolution

The automatic analysis of subtle changes between MRI scans is an important tool for assessing disease evolution over time. Manual labeling of evolutions in 3D data sets is tedious and error prone. Automatic change detection, however, remains a challenging image processing problem. A variety of MRI artifacts introduce a wide range of unrepresentative changes between images, making standard change detection methods unreliable. In this study we describe an automatic image processing system that addresses these issues. Registration errors and undesired anatomical deformations are compensated using a versatile multiresolution deformable image matching method that preserves significant changes at a given scale. A nonlinear intensity normalization method is associated with statistical hypothesis test methods to provide reliable change detection. Multimodal data is optionally exploited to reduce the false detection rate. The performance of the system was evaluated on a large database of 3D multimodal, MR images of patients suffering from relapsing remitting multiple sclerosis (MS). The method was assessed using receiver operating characteristics (ROC) analysis, and validated in a protocol involving two neurologists. The automatic system outperforms the human expert, detecting many lesion evolutions that are missed by the expert, including small, subtle changes.

Detection of prostate cancer by integration of line-scan diffusion, T2-mapping and T2-weighted magnetic resonance imaging; a multichannel statistical classifier

A multichannel statistical classifier for detecting prostate cancer was developed and validated by combining information from three different magnetic resonance (MR) methodologies: T2-weighted, T2-mapping, and line scan diffusion imaging (LSDI). From these MR sequences, four different sets of image intensities were obtained: T2-weighted (T2W) from T2-weighted imaging, Apparent Diffusion Coefficient (ADC) from LSDI, and proton density (PD) and T2 (T2 Map) from T2-mapping imaging. Manually segmented tumor labels from a radiologist, which were validated by biopsy results, served as tumor "ground truth." Textural features were extracted from the images using co-occurrence matrix (CM) and discrete cosine transform (DCT). Anatomical location of voxels was described by a cylindrical coordinate system. A statistical jack-knife approach was used to evaluate our classifiers. Single-channel maximum likelihood (ML) classifiers were based on 1 of the 4 basic image intensities. Our multichannel classifiers: support vector machine (SVM) and Fisher linear discriminant (FLD), utilized five different sets of derived features. Each classifier generated a summary statistical map that indicated tumor likelihood in the peripheral zone (PZ) of the prostate gland. To assess classifier accuracy, the average areas under the receiver operator characteristic (ROC) curves over all subjects were compared. Our best FLD classifier achieved an average ROC area of 0.839(+/-0.064), and our best SVM classifier achieved an average ROC area of 0.761(+/-0.043). The T2W ML classifier, our best single-channel classifier, only achieved an average ROC area of 0.599(+/-0.146). Compared to the best single-channel ML classifier, our best multichannel FLD and SVM classifiers have statistically superior ROC performance (P=0.0003 and 0.0017, respectively) from pairwise two-sided t-test. By integrating the information from multiple images and capturing the textural and anatomical features in tumor areas, summary statistical maps can potentially aid in image-guided prostate biopsy and assist in guiding and controlling delivery of localized therapy under image guidance.

Improved detection of metastases on magnetic resonance images by digital tissue recognition: validation using VX-2 tumor in the rabbit

PURPOSE

To evaluate the ability of a prototype digital tissue recognition (DTR) system to improve the accuracy of detection of metastases on magnetic resonance (MR) images in the rabbit VX-2 tumor model.

MATERIALS AND METHODS

Multiple MR imaging (MRI) sequences, including pre-contrast and post-contrast enhanced T1-weighted, T2-weighted, proton-density, and fast short inversion time inversion recovery (FSTIR), were acquired for six rabbits implanted with VX-2 adenocarcinoma. For each rabbit, DTR used the MR intensity characteristics of a known tumor site to highlight other areas suspicious for tumor. Three independent veterinary radiologists with extensive experience in animal MRI interpreted the images for tumor both without and with the results of DTR. The conventional and DTR-assisted interpretations were compared to pathology.

RESULTS

Using DTR, the radiologists found an average of 13.2% more true positive sites with a 10.3% reduction in false positives compared to unassisted interpretation. The improvement for the radiologists was statistically significant (McNemar's test, P = 0.0004). The agreement between radiologists using DTR was consistently higher than for their conventional interpretations (kappa statistic).

CONCLUSION

Compared with conventional interpretation of MR images, the use of DTR provided a statistically significant improvement in the accuracy of locating more and smaller sites of tumor. This improvement was achieved without the benefit of post-contrast images.

A variational framework for integrating segmentation and registration through active contours

Traditionally, segmentation and registration have been solved as two independent problems, even though it is often the case that the solution to one impacts the solution to the other. In this paper, we introduce a geometric, variational framework that uses active contours to simultaneously segment and register features from multiple images. The key observation is that multiple images may be segmented by evolving a single contour as well as the mappings of that contour into each image.

MRI tissue characterization of experimental cerebral ischemia in rat

PURPOSE

To extend the ISODATA image segmentation method to characterize tissue damage in stroke, by generating an MRI score for each tissue that corresponds to its histological damage.

MATERIALS AND METHODS

After preprocessing and segmentation (using ISODATA clustering), the proposed method scores tissue regions between 1 and 100. Score 1 is assigned to normal brain matter (white or gray matter), and score 100 to cerebrospinal fluid (CSF). Lesion zones are assigned a score based on their relative levels of similarities to normal brain matter and CSF. To evaluate the method, 15 rats were imaged by a 7T MRI system at one of three time points (acute, subacute, chronic) after MCA occlusion. Then they were killed and their brains were sliced and prepared for histological studies. MRI of two or three slices of each rat brain (using two DWI (b = 400, b = 800), one PDWI, one T2WI, and one T1WI) was performed, and an MRI score between 1 and 100 was determined for each region. Segmented regions were mapped onto the histology images and scored on a scale of 1-10 by an experienced pathologist. The MRI scores were validated by comparison with histology scores. To this end, correlation coefficients between the two scores (MRI and histology) were determined.

RESULTS

Experimental results showed excellent correlations between MRI and histology scores at different time points. Depending on the reference tissue (gray matter or white matter) used in the standardization, the correlation coefficients ranged from 0.73 (P < 0.0001) to 0.78 (P < 0.0001) using the entire dataset, including acute, subacute, and chronic time points. This suggests that the proposed multiparametric approach accurately identified and characterized ischemic tissue in a rat model of cerebral ischemia at different stages of stroke evolution.

CONCLUSION

The proposed approach scores tissue regions and characterizes them using unsupervised clustering and multiparametric image analysis techniques. The method can be used for a variety of applications in the field of computer-aided diagnosis and treatment, including evaluation of response to treatment. For example, volume changes for different zones of the lesion over time (e.g., tissue recovery) can be evaluated.

Automated segmentation of multispectral brain MR images

This work presents a robust and comprehensive approach for the in vivo automated segmentation and quantitative tissue volume measurement of normal brain composition from multispectral magnetic resonance imaging (MRI) data. Statistical pattern recognition methods based on a finite mixture model are used to partition the intracranial volume into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) spaces. A masking algorithm initially extracts the brain volume from surrounding extrameningeal tissue. Radio frequency (RF) field inhomogeneity effects in the images are then removed using a recursive method that adapts to the intrinsic local tissue contrast. Our technique supports heterogeneous data with multispectral MR images of different contrast and intensity weighting acquired at varying spatial resolution and orientation. The proposed image segmentation methods have been tested using multispectral T1-, proton density-, and T2-weighted MRI data from young and aged non-human primates as well as from human subjects.

Feature-based, automated segmentation of cerebral infarct patterns using T2- and diffusion-weighted imaging

Diffusion-weighted imaging enables the diagnosis of cerebral ischemias very early, thus supporting therapies such as thrombolysis. However, morphology and tissue-characterizing parameters (e.g. relaxation times or water diffusion) may vary strongly in ischemic regions, indicating different underlying pathologic processes. As the determination of the parameters by a supervised segmentation is very time consuming, we evaluated whether different infarct patterns may be segmented by an automated, multidimensional feature-based method using a unified segmentation procedure. Ischemias were classified into 5 characteristic patterns. For each class, a 3D histogram based on T(2)- and diffusion-weighted images as well as calculated apparent diffusion coefficients (ADC) was generated from a representative data set. Healthy and pathologic tissue classes were segmented in the histogram as separate, local density maxima with freely shaped borders. Segmentation control parameters were optimized in a 3-step procedure. The method was evaluated using synthetic images as well as results of a supervised segmentation. For the analysis of cerebral ischemias, the optimal control parameter set led to sensitivities and specificities between 1.0 and 0.9.

Segmentation of MR images for computer-assisted surgery of the lumbar spine

This paper describes a segmentation algorithm designed to separate bone from soft tissue in magnetic resonance (MR) images developed for computer-assisted surgery of the spine. The algorithm was applied to MR images of the spine of healthy volunteers. Registration experiments were carried out on a physical model of a spine generated from computed tomography (CT) data of a surgical patient. Segmented CT, manually segmented MR and MR images segmented using the developed algorithm were compared. The algorithm performed well at segmenting bone from soft tissue on images taken of healthy volunteers. Registration experiments showed similar results between the CT and MR data. The MR data, which were manually segmented, performed worse on visual verification experiments than both the CT and semi-automatic segmented data. The algorithm developed performs well at segmenting bone from soft tissue in MR images of the spine as measured using registration experiments.

Current methods in medical image segmentation

Image segmentation plays a crucial role in many medical-imaging applications, by automating or facilitating the delineation of anatomical structures and other regions of interest. We present a critical appraisal of the current status of semi-automated and automated methods for the segmentation of anatomical medical images. Terminology and important issues in image segmentation are first presented. Current segmentation approaches are then reviewed with an emphasis on the advantages and disadvantages of these methods for medical imaging applications. We conclude with a discussion on the future of image segmentation methods in biomedical research.

Feature space analysis: effects of MRI protocols

We present a method for exploring the relationship between the image segmentation results obtained by an optimal feature space method and the MRI protocols used. The steps of the work accomplished are as follows. (1) Patients with brain tumors were imaged on a 1.5 T General Electric Signa MRI System using multiple protocols (T1 and T2-weighted fast spin-echo and FLAIR). T1-weighted images were acquired before and after gadolinium injection. (2) Image volumes were co-registered, and images of a slice through the center of the tumor were selected for processing. (3) For each patient, several image sets were defined by selecting certain MR images (e.g., 4T2's+ IT1, 4T2's+FLAIR, 2T2's+ 1T1). (4) Using each image set, the optimal feature space was generated and images were segmented into normal tissues and different tumor zones. (5) Segmentation results obtained using the different MRI sets were compared. Based on the analysis results from 27 image sets, we found that the locations of the clusters for the tumor zones and their corresponding regions in the image domain changed as a function of the MR images (MRI protocols) used. However, the segmentation results for the total lesion and normal tissues remained relatively unchanged.

Color enhancement of multispectral MR images: improving the visualization of subcortical structures

PURPOSE

The current investigation was undertaken to evaluate a new method for creating MR multispectral color images, which we call "Superimages." They were developed to improve the delineation of small brain structures composed of mixed tissue types, such as the basal ganglia.

METHOD

To qualitatively validate the method, visual comparisons were made of six unimodal and multispectral images, including the Superimage. Quantitative validation was undertaken by comparing the reliability values for parcellation of the globus pallidus (GP) using either a gray scale (T1-weighted) image or the Superimage.

RESULTS

Qualitative assessment of the Superimage revealed enhanced visualization of the GP, caudate, and putamen. Quantitative assessment resulted in good reliability for Superimage traces.

CONCLUSION

The Superimage significantly improves both the visualization and the parcellation of structures visualized by MRI.

Novel approaches to imaging brain tumors

Brain imaging techniques are assuming a greater range of roles in neuro-oncology. New techniques promise earlier recognition of the spread of tumors to the brain, which is useful in staging of disseminated disease, as well as better definition of small lesions associated with presentations of epilepsy. There is the promise that entirely noninvasive, specific diagnosis of brain tumors may become possible. Imaging methods are being used increasingly to direct and monitor therapy. Preoperative and intraoperative imaging are being used for guiding tumor surgery. An exciting potential goal for greater use of imaging is in the individualization of medical therapies either by analysis of in vitro responses or by visualization of drug responses on the tumor in situ. An important focus for technical development is in the robust integration of complementary information to allow optimization of the sensitivity and specificity of multimodal examinations.

A model for multiparametric mri tissue characterization in experimental cerebral ischemia with histological validation in rat: part 1

BACKGROUND AND PURPOSE

After stroke, brain tissue undergoes time-dependent heterogeneous histopathological change. These tissue alterations have MRI characteristics that allow segmentation of ischemic from nonischemic tissue. Moreover, MRI segmentation generates different zones within the lesion that may reflect heterogeneity of tissue damage.

METHODS

A vector tissue signature model is presented that uses multiparametric MRI for segmentation and characterization of tissue. An objective (unsupervised) computer segmentation algorithm was incorporated into this model with the use of a modified version of the Iterative Self-Organizing Data Analysis Technique (ISODATA). The ability of the model to characterize ischemic tissue after permanent middle cerebral ischemia occlusion in the rat was tested. Multiparametric ISODATA measurements of the ischemic tissue were compared with quantitative histological characterization of the tissue from 4 hours to 1 week after stroke.

RESULTS

The ISODATA segmentation of tissue identified a gradation of cerebral tissue damage at all time points after stroke. The histological scoring of ischemic tissue from 4 hours to 1 week after stroke on all the animals was significantly correlated with ISODATA segmentation (r=0.78, P<0.001; n=20) when a multiparametric (T2-, T1-, diffusion-weighted imaging) data set was used, less correlated (r=0.70, P<0.01; n=20) when a T2- and T1-weighted data set was used, and not correlated (r=-0.12, P>0.47; n=20) when only a diffusion-weighted imaging data set was used.

CONCLUSIONS

Our data indicate that an integrated set of MRI parameters can distinguish and stage ischemic tissue damage in an objective manner.

Mathematical segmentation of grey matter, white matter and cerebral spinal fluid from MR image pairs

The aims of this study were (1) to design a mathematical segmentation technique to allow extraction of grey matter, white matter and cerebral spinal fluid volumes from paired high resolution MR images and (2) to document the statistical accuracy of the method with different image combinations. A series of linear equations were derived that describe proportional tissue volumes in individual image voxels. The equations use estimates of pure tissue values to derive the proportion of each tissue within a single voxel. Repeatability of manual estimations of pure tissue values was assessed both using regions of interest and thresholding techniques. Statistical accuracy of tissue estimations for a variety of image pairs was assessed from measurements of root-mean-square noise and mean grey level intensity. The technique was used to produce parametric images of grey and white matter distribution. The segmentation technique showed greatest statistical accuracy when the first image has high grey/white matter contrast and the second image has little contrast or the rank order of the signal intensities from pure tissue is reversed. A combination of inversion recovery fast spin echo and fast FLAIR images produced a statistical error of 11% for grey matter and 10% for white matter for any given voxel. The effect of increasing sample size improves both of these figures to give a 1% statistical error on a 100 pixel sample.

Automated segmentation of MR images of brain tumors

An automated brain tumor segmentation method was developed and validated against manual segmentation with three-dimensional magnetic resonance images in 20 patients with meningiomas and low-grade gliomas. The automated method (operator time, 5-10 minutes) allowed rapid identification of brain and tumor tissue with an accuracy and reproducibility comparable to those of manual segmentation (operator time, 3-5 hours), making automated segmentation practical for low-grade gliomas and meningiomas.

Automatic segmentation of non-enhancing brain tumors in magnetic resonance images

Tumor segmentation from magnetic resonance (MR) images may aid in tumor treatment by tracking the progress of tumor growth and/or shrinkage. In this paper we present the first automatic segmentation method which separates non-enhancing brain tumors from healthy tissues in MR images to aid in the task of tracking tumor size over time. The MR feature images used for the segmentation consist of three weighted images (T1, T2 and proton density (PD)) for each axial slice through the head. An initial segmentation is computed using an unsupervised fuzzy clustering algorithm. Then, integrated domain knowledge and image processing techniques contribute to the final tumor segmentation. They are applied under the control of a knowledge-based system. The system knowledge was acquired by training on two patient volumes (14 images). Testing has shown successful tumor segmentations on four patient volumes (31 images). Our results show that we detected all six non-enhancing brain tumors, located tumor tissue in 35 of the 36 ground truth (radiologist labeled) slices containing tumor and successfully separated tumor regions from physically connected CSF regions in all the nine slices. Quantitative measurements are promising as correspondence ratios between ground truth and segmented tumor regions ranged between 0.368 and 0.871 per volume, with percent match ranging between 0.530 and 0.909 per volume.

Validation of partial tissue segmentation of single-channel magnetic resonance images of the brain

We describe and evaluate a practical, automated algorithm based on local statistical mixture modeling for segmenting single-channel, T1-weighted volumetric magnetic resonance images of the brain into gray matter, white matter, and cerebrospinal fluid. We employed a stereological sampling method to assess, prospectively, the performance of the method with respect to human experts on 10 normal T1-weighted brain scans acquired with a three-dimensional gradient echo pulse sequence. The overall kappa statistic for the concordance of the algorithm with the human experts was 0.806, while that among raters, excluding the algorithm, was 0.802. The algorithm had better agreement with the modal expert decision (kappa = 0.878). The algorithm could not be distinguished from the experts by this measure. We also validated the algorithm on a simulated MR scan of a digital brain phantom with known tissue composition. Global gray matter and white matter errors were 1% and <1%, respectively, and correlation coefficients with the underlying tissue model were 0.95 for gray matter, 0.98 for white matter, and 0.95 for cerebrospinal fluid. In both approaches to validation, we evaluated both local and global performance of the algorithm. Human experts generated slightly higher global gray matter proportion estimates on the test brain scans relative to the algorithm (3.7%) and on the simulated MR scan relative to the true tissue model (4.4%). The algorithm underestimated gray in some subcortical nuclei which contain admixed gray and white matter. We demonstrate the reliability of the method on individual 1 NEX data sets of the test subjects, and its insensitivity to the precise values of initial model parameters. The output of this algorithm is suitable for quantifying cerebral cortical tissue, using a commonly performed commercial pulse sequence.

Normal brain development and aging: quantitative analysis at in vivo MR imaging in healthy volunteers

PURPOSE

To quantitate neuroanatomic parameters in healthy volunteers and to compare the values with normative values from postmortem studies.

MATERIALS AND METHODS

Magnetic resonance (MR) images of 116 volunteers aged 19 months to 80 years were analyzed with semiautomated procedures validated by means of comparison with manual tracings. Volumes measured included intracranial space, whole brain, gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). Results were compared with values from previous postmortem studies.

RESULTS

Whole brain and intracranial space grew by 25%-27% between early childhood (mean age, 26 months; age range, 19-33 months) and adolescence (mean age, 14 years; age range, 12-15 years); thereafter, whole-brain volume decreased such that volunteers (age range, 71-80 years) had volumes similar to those of young children. GM increased 13% from early to later (6-9 years) childhood. Thereafter, GM increased more slowly and reached a plateau in the 4th decade; it decreased by 13% in the oldest volunteers. The GM-WM ratio decreased exponentially from early childhood through the 4th decade; thereafter, it gradually declined. In vivo patterns of change in the intracranial space, whole brain, and GM-WM ratio agreed with published postmortem data.

CONCLUSION

MR images accurately depict normal patterns of age-related change in intracranial space, whole brain, GM, WM, and CSF. These quantitative MR imaging data can be used in research studies and clinical settings for the detection of abnormalities in fundamental neuroanatomic parameters.

Affine adaptive filtering of CT data

A novel method for resampling and enhancing image data using multidimensional adaptive filters is presented. The underlying issue that this paper addresses is segmentation of image structures that are close in size to the voxel geometry. Adaptive filtering is used to reduce both the effects of partial volume averaging by resampling the data to a lattice with higher sample density and to reduce the image noise level. Resampling is achieved by constructing filter sets that have subpixel offsets relative to the original sampling lattice. The filters are also frequency corrected for ansisotropic voxel dimensions. The shift and the voxel dimensions are described by an affine transform and provides a model for tuning the filter frequency functions. The method has been evaluated on CT data where the voxels are in general non cubic. The in-plane resolution in CT image volumes is often higher by a factor of 3-10 than the through-plane resolution. The method clearly shows an improvement over conventional resampling techniques such as cubic spline interpolation and sinc interpolation.

Adaptive, template moderated, spatially varying statistical classification

A novel image segmentation algorithm was developed to allow the automatic segmentation of both normal and abnormal anatomy from medical images. The new algorithm is a form of spatially varying statistical classification, in which an explicit anatomical template is used to moderate the segmentation obtained by statistical classification. The algorithm consists of an iterated sequence of spatially varying classification and nonlinear registration, which forms an adaptive, template moderated (ATM), spatially varying statistical classification (SVC). Classification methods and nonlinear registration methods are often complementary, both in the tasks where they succeed and in the tasks where they fail. By integrating these approaches the new algorithm avoids many of the disadvantages of each approach alone while exploiting the combination. The ATM SVC algorithm was applied to several segmentation problems, involving different image contrast mechanisms and different locations in the body. Segmentation and validation experiments were carried out for problems involving the quantification of normal anatomy (MRI of brains of neonates) and pathology of various types (MRI of patients with multiple sclerosis, MRI of patients with brain tumors, MRI of patients with damaged knee cartilage). In each case, the ATM SVC algorithm provided a better segmentation than statistical classification or elastic matching alone.

Multi-level adaptive segmentation of multi-parameter MR brain images

MR brain image segmentation into several tissue classes is of significant interest to visualize and quantify individual anatomical structures. Traditionally, the segmentation is performed manually in a clinical environment that is operator dependent and may be difficult to reproduce. Though several algorithms have been investigated in the literature for computerized automatic segmentation of MR brain images, they are usually targeted to classify image into a limited number of classes such as white matter, gray matter, cerebrospinal fluid and specific lesions. We present a novel model-based method for the automatic segmentation and classification of multi-parameter MR brain images into a larger number of tissue classes of interest. Our model employs 15 brain tissue classes instead of the commonly used set of four classes, which were of clinical interest to neuroradiologists for following-up with patients suffering from cerebrovascular deficiency (CVD) and/or stroke. The model approximates the spatial distribution of tissue classes by a Gauss Markov random field and uses the maximum likelihood method to estimate the class probabilities and transitional probabilities for each pixel of the image. Multi-parameter MR brain images with T(1), T(2), proton density, Gd+T(1), and perfusion imaging were used in segmentation and classification. In the development of the segmentation model, true class-membership of measured parameters was determined from manual segmentation of a set of normal and pathologic brain images by a team of neuroradiologists. The manual segmentation was performed using a human-computer interface specifically designed for pixel-by-pixel segmentation of brain images. The registration of corresponding images from different brains was accomplished using an elastic transformation. The presented segmentation method uses the multi-parameter model in adaptive segmentation of brain images on a pixel-by-pixel basis. The method was evaluated on a set of multi-parameter MR brain images of a twelve-year old patient 48h after suffering a stroke. The results of classification as compared to the manual segmentation of the same data show the efficacy and accuracy of the presented methods as well as its capability to create and learn new tissue classes.

Intimate combination of low- and high-resolution image data: I. Real-space PET and (1)H(2)O MRI, PETAMRI

Two different types of (co-registered) images of the same slice of tissue will generally have different spatial resolutions. The judicious pixel-by-pixel combination of their data can be accomplished to yield a single image exhibiting properties of both. Here, axial (18)FDG PET and (1)H(2)O MR images of the human brain are used as the low- and high-resolution members of the pair. A color scale is necessary in order to provide for separate intensity parameters from the two image types. However, not all color scales can accommodate this separability. The HSV color model allows one to choose a color scale in which the intensity of the low-resolution image type is coded as hue, while that of the high-resolution type is coded as value, a reasonably independent parameter. Furthermore, the high-resolution image must have high contrast and be quantitative in the same sense as the low-resolution image almost always is. Here, relaxographic MR images (naturally segmented quantitative (1)H(2)O spin-density components) are used. Their essentially complete contrast serves to effect an apparent editing function when encoded as the value of the color scale. Thus, the combination of (18)FDG PET images with gray-matter (GM) relaxographic (1)H(2)O images produces visually "GM-edited" (18)FDG PETAMR (positron emission tomography and magnetic resonance) images. These exhibit the high sensitivity to tracer amounts characteristic of PET along with the high spatial resolution of (1)H(2)O MRI. At the same time, however, they retain the complete quantitative measures of each of their basis images. Magn Reson Med 42:345-360, 1999. Published 1999 Wiley-Liss, Inc.

Spatial attention deficits in patients with acquired or developmental cerebellar abnormality

Recent imaging and clinical studies have challenged the concept that the functional role of the cerebellum is exclusively in the motor domain. We present evidence of slowed covert orienting of visuospatial attention in patients with developmental cerebellar abnormality (patients with autism, a disorder in which at least 90% of all postmortem cases reported to date have Purkinje neuron loss), and in patients with cerebellar damage acquired from tumor or stroke. In spatial cuing tasks, normal control subjects across a wide age range were able to orient attention within 100 msec of an attention-directing cue. Patients with cerebellar damage showed little evidence of having oriented attention after 100 msec but did show the effects of attention orienting after 800-1200 msec. These effects were demonstrated in a task in which results were independent of the motor response. In this task, smaller cerebellar vermal lobules VI-VII (from magnetic resonance imaging) were associated with greater attention-orienting deficits. Although eye movements may also be disrupted in patients with cerebellar damage, abnormal gaze shifting cannot explain the timing and nature of the attention-orienting deficits reported here. These data may be consistent with evidence from animal models that suggest damage to the cerebellum disrupts both the spatial encoding of a location for an attentional shift and the subsequent gaze shift. These data are also consistent with a model of cerebellar function in which the cerebellum supports a broad spectrum of brain systems involved in both nonmotor and motor function.

Non-invasive assessment of axonal fiber connectivity in the human brain via diffusion tensor MRI

A technique for assessing in vivo fiber connectivity in the human brain is presented. The method utilizes a novel connectivity algorithm that operates in three spatial dimensions and uses estimates of fiber tract orientation and tissue anisotropy, obtained from diffusion tensor magnetic resonance imaging, to establish the pathways of fiber tracts. Sample in vivo connectivity images from healthy human brain are presented that demonstrate connections in the white matter tracts. White matter connectivity information is potentially of interest in the study of a range of neurological, psychiatric, and developmental disorders and shows promise for following the natural history of disease.

Three-dimensional visualization and analysis methodologies: a current perspective

Three-dimensional (3D) imaging was developed to provide both qualitative and quantitative information about an object or object system from images obtained with multiple modalities including digital radiography, computed tomography, magnetic resonance imaging, positron emission tomography, single photon emission computed tomography, and ultrasonography. Three-dimensional imaging operations may be classified under four basic headings: preprocessing, visualization, manipulation, and analysis. Preprocessing operations (volume of interest, filtering, interpolation, registration, segmentation) are aimed at extracting or improving the extraction of object information in given images. Visualization operations facilitate seeing and comprehending objects in their full dimensionality and may be either scene-based or object-based. Manipulation may be either rigid or deformable and allows alteration of object structures and of relationships between objects. Analysis operations, like visualization operations, may be either scene-based or object-based and deal with methods of quantifying object information. There are many challenges involving matters of precision, accuracy, and efficiency in 3D imaging. Nevertheless, 3D imaging is an exciting technology that promises to offer an expanding number and variety of applications.

Quantitative follow-up of patients with multiple sclerosis using MRI: technical aspects

A highly reproducible automated procedure for quantitative analysis of serial brain magnetic resonance (MR) images was developed for use in patients with multiple sclerosis (MS). The intracranial cavity (ICC) was identified on standard dual-echo spin-echo brain MR images using a supervised automated procedure. MR images obtained from one MS patient at 24 time points in the course of a 1-year follow-up were aligned with the images of one of the time points. Next, the contents of the ICC in each MR exam were segmented into four tissues, using a self-adaptive statistical algorithm. Misclassifications due to partial voluming were corrected using a combination of morphologic operators and connectivity criteria. Finally, a connectivity detection algorithm was used to separate the tissue classified as lesions into individual entities. Registration, classification of the contents of the ICC, and identification of individual lesions are fully automatic. Only identification of the ICC requires operator interaction. In each MR exam, the program estimated volumes for the ICC, gray matter (GM), white matter (WM), white matter lesions (WML), and cerebrospinal fluid (CSF). The reproducibility of the system was superior to that of supervised segmentation, as evidenced by the coefficient of variation: CSF supervised 45.9% vs. automated 7.7%, GM 16.0% vs. 1.4%, WM 15.7% vs. 1.3%, and WML 39.5% vs 52.0%. Our results demonstrate that this computerized procedure allows routine reproducible quantitative analysis of large serial MRI data sets.

Quantification of white matter and gray matter volumes from three-dimensional magnetic resonance volume studies using fuzzy classifiers

We accurately measured white matter (WM) and gray matter (GM) from three-dimensional (3D) volume studies, using a fuzzy classification technique. The new segmentation method is a modification of a recently published method developed for T1 parametric images. 3D MR images were transformed into pseudo forms of T1 parametric images and segmented into WM and GM voxel fraction images with a set of standardized fuzzy classifiers. This segmentation method was validated with synthesized 3D MR images as phantoms. These phantoms were developed from cryosectioned human brain images located in the superior, middle, and inferior regions of the cerebrum. Phantom volume measurements revealed that, generally, the difference between measured and actual volumes was less than 3% for 1.5-mm simulated brain slices. The average cerebral GM/WM ratio calculated from 3D MR studies in four subjects was 1.77, which compared favorably with the estimate of 1.67 derived from anatomical data. Results indicate that this is an accurate and rapid method for quantifying WM and GM from Tl-weighted 3D volume studies.

Feature space analysis of MRI

This paper presents an MRI feature-space image-analysis method and its application to brain tumor studies. The proposed method generates a transformed feature space in which the normal tissues (white matter, gray matter, and CSF) become orthonormal. As such, the method is expected to have site-to-site and patient-to-patient consistency, and is useful for identification of tissue types, segmentation of tissues, and quantitative measurements on tissues. The steps of the work accomplished are as follows: (1) Four T2-weighted and two T1-weighted images (before and after injection of gadolinium) were acquired for 10 tumor patients. (2) Images were analyzed by an image analyst according to the proposed algorithm. (3) Biopsy samples were extracted from each patient and were subsequently analyzed by the pathology laboratory. (4) Image-analysis results were compared with the biopsy results. Pre- and postsurgery feature spaces were also compared. The proposed method made it possible to visualize the MRI feature space and to segment the image. In all cases, the operators were able to find clusters for normal and abnormal tissues. Also, clusters for different zones of the tumor were found. The method successfully segmented the image into normal tissues (white matter, gray matter, and CSF) and different zones of the lesion (tumor, cyst, edema, radiation necrosis, necrotic core, and infiltrated tumor). The results agreed with those obtained from the biopsy samples. Comparison of pre- with postsurgery and radiation feature spaces illustrated that the original solid tumor was not present in the second study, but a new tissue component appeared in a different location of the feature space. This tissue could be radiation necrosis generated as a result of radiation.

Application of the extremum stack to neurological MRI

The extremum stack, as proposed by Koenderink, is a multiresolution image description and segmentation scheme which examines intensity extrema (minima and maxima) as they move and merge through a series of progressively isotropically diffused images known as scale space. Such a data-driven approach is attractive because it is claimed to be a generally applicable and natural method of image segmentation. The performance of the extremum stack is evaluated here using the case of neurological magnetic resonance imaging data as a specific example, and means of improving its performance proposed. It is confirmed experimentally that the extremum stack has the desirable property of being shift-, scale-, and rotation-invariant, and produces natural results for many compact regions of anatomy. It handles elongated objects poorly, however, and subsections of regions may merge prematurely before each region is represented as a single node. It is shown that this premature merging can often be avoided by the application of either a variable conductance-diffusing preprocessing step, or more effectively, the use of an adaptive variable conductance diffusion method within the extremum stack itself in place of the isotropic Gaussian diffusion proposed by Koenderink.

Brain tumor segmentation and characterization by pattern analysis of multispectral NMR images

A major problem in tumor treatment planning and evaluation is determination of the tumor extent. This paper presents a pattern analysis methodology for segmentation and characterization of brain tumors from multispectral NMR images. The proposed approach has been used in 15 clinical studies of cerebral tumor patients who have been scheduled for surgical biopsy and resection. The tissue biopsy results, obtained at specific spatial coordinates determined in the analysis, have been utilized to validate the methodology. It was found that in all cases the lesion had extended into normal tissue, at least to the location where the sample was taken. In most cases, the proposed method suggested that the lesion had extended several millimetres beyond the point from where the biopsy sample was taken. In some cases, the extent of the lesion into normal tissue was well beyond the boundary seen on T1- or T2-weighted images. It is concluded that the proposed approach indicates brain tumor infiltration more precisely than what is visualized in the original NMR images and therefore its utilization facilitates proper treatment planning for the cerebral tumor patients.

Robust parameter estimation of intensity distributions for brain magnetic resonance images

This paper presents two new methods for robust parameter estimation of mixtures in the context of magnetic resonance (MR) data segmentation. The head is constituted of different types of tissue that can be modeled by a finite mixture of multivariate Gaussian distributions. Our goal is to estimate accurately the statistics of desired tissues in presence of other ones of lesser interest. These latter can be considered as outliers and can severely bias the estimates of the former. For this purpose, we introduce a first method, which is an extension of the expectation-maximization (EM) algorithm, that estimates parameters of Gaussian mixtures but incorporates an outlier rejection scheme which allows to compute the properties of the desired tissues in presence of atypical data. The second method is based on genetic algorithms and is well suited for estimating the parameters of mixtures of different kind of distributions. We use this property by adding a uniform distribution to the Gaussian mixture for modeling the outliers. The proposed genetic algorithm can efficiently estimate the parameters of this extended mixture for various initial settings. Also, by changing the minimization criterion, estimates of the parameters can be obtained by histogram fitting which considerably reduces the computational cost. Experiments on synthetic and real MR data show that accurate estimates of the gray and white matters parameters are computed.

Automatic tumor segmentation using knowledge-based techniques

A system that automatically segments and labels glioblastoma-multiforme tumors in magnetic resonance images (MRI's) of the human brain is presented. The MRI's consist of T1-weighted, proton density, and T2-weighted feature images and are processed by a system which integrates knowledge-based (KB) techniques with multispectral analysis. Initial segmentation is performed by an unsupervised clustering algorithm. The segmented image, along with cluster centers for each class are provided to a rule-based expert system which extracts the intracranial region. Multispectral histogram analysis separates suspected tumor from the rest of the intracranial region, with region analysis used in performing the final tumor labeling. This system has been trained on three volume data sets and tested on thirteen unseen volume data sets acquired from a single MRI system. The KB tumor segmentation was compared with supervised, radiologist-labeled "ground truth" tumor volumes and supervised k-nearest neighbors tumor segmentations. The results of this system generally correspond well to ground truth, both on a per slice basis and more importantly in tracking total tumor volume during treatment over time.

Partial-volume Bayesian classification of material mixtures in MR volume data using voxel histograms

We present a new algorithm for identifying the distribution of different material types in volumetric datasets such as those produced with magnetic resonance imaging (MRI) or computed tomography (CT). Because we allow for mixtures of materials and treat voxels as regions, our technique reduces errors that other classification techniques can create along boundaries between materials and is particularly useful for creating accurate geometric models and renderings from volume data. It also has the potential to make volume measurements more accurately and classifies noisy, low-resolution data well. There are two unusual aspects to our approach. First, we assume that, due to partial-volume effects, or blurring, voxels can contain more than one material, e.g., both muscle and fat; we compute the relative proportion of each material in the voxels. Second, we incorporate information from neighboring voxels into the classification process by reconstructing a continuous function, rho(x), from the samples and then looking at the distribution of values that rho(x) takes on within the region of a voxel. This distribution of values is represented by a histogram taken over the region of the voxel; the mixture of materials that those values measure is identified within the voxel using a probabilistic Bayesian approach that matches the histogram by finding the mixture of materials within each voxel most likely to have created the histogram. The size of regions that we classify is chosen to match the spacing of the samples because the spacing is intrinsically related to the minimum feature size that the reconstructed continuous function can represent.

A phase I trial of human corticotropin-releasing factor (hCRF) in patients with peritumoral brain edema

BACKGROUND

Human corticotropin-releasing factor (hCRF) is an endogenous peptide responsible for the secretion and synthesis of corticosteroids. In animal models of peritumoral brain edema, hCRF has significant anti-edematous action. This effect, which appears to be independent of the release of adrenal steroids, appears mediated by a direct effect on endothelial cells. We conducted a feasibility and phase I study with hCRF given by continuous infusion to patients with brain metastasis.

PATIENTS AND METHODS

Peritumoral brain edema documented by MRI and the use of either no steroids or stable steroid doses for more than a week were required. MRIs were repeated at completion of infusion and estimations by dual echo-image sequence (Proton density and T2-weighted images) of the amount of peritumoral edema were performed. The study was performed in two stages. In the feasibility part, patients were randomized to receive either 0.66 or 1 microgram/kg/h of hCRF or placebo over 24 hours. The second part was a dose finding study of hCRF over 72 hours at escalating doses.

RESULTS

Seventeen patients were enrolled; only one was receiving steroids (stable doses) at study entrance; dose-limiting toxicity (hypotension) was observed at 4 micrograms/kg/h x 72 hours in two out of four patients, while zero of five patients treated at 2 micrograms/kg/h developed dose-limiting toxicities. Flushing and hot flashes were also observed. Improvement of neurological symptoms and/or exam were seen in 10 patients. Only small changes were detected by MRI. Improvement in symptoms did not correlate with changes in cortisol levels, and changes in cortisol levels were not correlated with changes in peritumoral edema.

CONCLUSIONS

hCRF is well tolerated in doses up to 2 micrograms/kg/h by continuous infusion x 72 hours. Hypotension limits administration of higher doses. The observation of clinical benefit in the absence of corticosteroids suggests hCRF may be an alternative to steroids for the treatment of patients with peritumoral brain edema. Further exploration of this agent in efficacy studies is warranted.

MR multispectral analysis of multiple sclerosis lesions

Although quantification of the lesion burden from serial MR examinations of patients with multiple sclerosis (MS) is a common technique to assess disease activity in clinical trials, pathologic change may occur within a lesion without a corresponding change in volume. Therefore, measures of lesion volume and composition may improve the sensitivity of detecting disease activity. A new technique has been developed that provides information about the intensity composition of MS lesions in standard spin-echo MR examinations. The new technique is based on the multispectral "feature space" intensity distributions of the lesions and normal tissues. Analysis of MR examinations of materials with known T1 and T2 times showed that feature space position from spin-echo examinations is largely determined from proton density (rho), T2, and the interecho delay. Information about intensity composition was obtained by reducing the multidimensional intensity distribution to one dimension while minimizing the loss of information. This technique was used to analyze eight lesions in standard spin-echo MR examinations of three patients with MS. Lesion distributions were compared between examinations by first calibrating the examinations based on the intensity distributions of cerebrospinal fluid (CSF), an internal reference tissue. Many of the lesion distributions had a distinctive peak at low intensity, corresponding to normal-appearing white matter (WM). Within the lesion distributions, increases in high intensity peaks generally were accompanied by reductions in the WM peak. Serial analysis of the lesion distributions revealed some dramatic fluctuations, even when lesion volume remained constant.

Statistical approach to segmentation of single-channel cerebral MR images

A statistical model is presented that represents the distributions of major tissue classes in single-channel magnetic resonance (MR) cerebral images. Using the model, cerebral images are segmented into gray matter, white matter, and cerebrospinal fluid (CSF). The model accounts for random noise, magnetic field inhomogeneities, and biological variations of the tissues. Intensity measurements are modeled by a finite Gaussian mixture. Smoothness and piecewise contiguous nature of the tissue regions are modeled by a three-dimensional (3-D) Markov random field (MRF). A segmentation algorithm, based on the statistical model, approximately finds the maximum a posteriori (MAP) estimation of the segmentation and estimates the model parameters from the image data. The proposed scheme for segmentation is based on the iterative conditional modes (ICM) algorithm in which measurement model parameters are estimated using local information at each site, and the prior model parameters are estimated using the segmentation after each cycle of iterations. Application of the algorithm to a sample of clinical MR brain scans, comparisons of the algorithm with other statistical methods, and a validation study with a phantom are presented. The algorithm constitutes a significant step toward a complete data driven unsupervised approach to segmentation of MR images in the presence of the random noise and intensity inhomogeneities.

Unsupervised, automated segmentation of the normal brain using a multispectral relaxometric magnetic resonance approach

The purpose of this study was the development and testing of a method for unsupervised, automated brain segmentation. Two spin-echo sequences were used to obtain relaxation rates and proton-density maps from 1.5 T MR studies, with two axial data sets including the entire brain. Fifty normal subjects (age range, 16 to 76 years) were studied. A Three-dimensional (3D) spectrum of the tissue voxels was used for automatic segmentation of gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) and for calculation of their volumes. Accuracy and reproducibility were tested with a three-compartment phantom simulating GM, WM, and CSF. In the normal subjects, a significant decrease of GM fractional volume and increased CSF volume with age were observed (P < 0.0001), with no significant changes in WM. This multispectral segmentation method permits reproducible, operator-independent volumetric measurements.

A new computer-assisted method for the quantification of enhancing lesions in multiple sclerosis

PURPOSE

Our goal is to describe a new computerized method for the detection and quantification of enhanced multiple sclerosis (MS) lesions.

METHOD

Gd-DTPA-enhanced, thin section, T1-weighted images of seven patients (involving 336 slice images) with definite MS were analyzed using a new method based on the theory of "fuzzy connected components," developed and implemented on the 3DVIEWNIX software system. Four neuroradiologists selected "true" lesions from the computer-detected potential lesions with a yes/no response to the program query on 2 different days. The enhanced lesion volume and number of enhancing lesions for each image and each observer were subsequently computed. Additional studies involving 720 slices were conducted to determine lesions that were missed by the system.

RESULTS

The intra- and interobserver variability in the system was 0%. It took approximately 1 min of operator time per 3D study. The system output has no false positives and a mean false-negative volume of 1.3%.

CONCLUSION

The novel system calculates enhancing lesion volume and the number of enhancing lesions with very little operator time, inter- and intraoperator variability, or false-positive and false-negative volumes. Computer-based quantification of enhancing lesion volume is an important objective measure of the activity of MS. The system is now in routine use in clinical investigations that study the role of enhancing lesions in the MS disease.

Tumour volume determination from MR images by morphological segmentation

Accurate tumour volume measurement from MR images requires some form of objective image segmentation, and therefore a certain degree of automation. Manual methods of separating data according to the various tissue types which they are thought to represent are inherently prone to operator subjectivity and can be very time consuming. A segmentation procedure based on morphological edge detection and region growing has been implemented and tested on a phantom of known adjustable volume. Comparisons have been made with a traditional data thresholding procedure for the determination of tumour volumes on a set of patients with intracerebral glioma. The two methods are shown to give similar results, with the morphological segmentation procedure having the advantages of being automated and faster.

Multispectral magnetic resonance images segmentation using fuzzy Hopfield neural network

This paper demonstrates a fuzzy Hopfield neural network for segmenting multispectral MR brain images. The proposed approach is a new unsupervised 2-D Hopfield neural network based upon the fuzzy clustering technique. Its implementation consists of the combination of 2-D Hopfield neural network and fuzzy c-means clustering algorithm in order to make parallel implementation for segmenting multispectral MR brain images feasible. For generating feasible results, a fuzzy c-means clustering strategy is included in the Hopfield neural network to eliminate the need for finding weighting factors in the energy function which is formulated and based on a basic concept commonly used in pattern classification, called the 'within-class scatter matrix' principle. The suggested fuzzy c-means clustering strategy has also been proven to be convergent and to allow the network to learn more effectively than the conventional Hopfield neural network. The experimental results show that a near optimal solution can be obtained using the fuzzy Hopfield neural network based on the within-class scatter matrix.

Quantification of white matter and gray matter volumes from T1 parametric images using fuzzy classifiers

White matter (WM) and gray matter (GM) were accurately measured using a technique based on a single standardized fuzzy classifier (FC) for each tissue. Fuzzy classifier development was based on experts' visual assessments of WM and GM boundaries from a set of T1 parametric MR images. The fuzzy classifier method's accuracy was validated and optimized by a set of T1 phantom images that were based on hand-detailed human brain cryosection images. Nine sets of axial T1 images of varying thickness equally distributed throughout the brain were simulated. All T1 data sets were mapped to the standardized FCs and rapidly segmented into WM and GM voxel fraction images. Resulting volumes revealed that, in most cases, the difference between measured and actual volumes was less than 5%. This was consistent throughout most of the brain, and as expected, the accuracy improved to generally less than 2% for the 1-mm simulated brain slices.

Cerebral magnetic resonance image segmentation using data fusion

OBJECTIVE

A semiautomated method is described for segmenting dual echo MR head scans into gray and white matter and CSF. The method is applied to brain scans of 80 healthy children and adolescents.

MATERIALS AND METHODS

A probabilistic data fusion equation was used to combine simultaneously acquired T2-weighted and proton density head scans for tissue segmentation. The fusion equation optimizes the probability of a voxel being a particular tissue type, given the corresponding probabilities from both images. The algorithm accounts for the intensity inhomogeneities present in the images by fusion of local regions of the images.

RESULTS

The method was validated using a phantom (agarose gel with iron oxide particles) and hand-segmented images. Gray and white matter volumes for subjects aged 20-30 years were close to those previously published. White matter and CSF volume increased and gray matter volume decreased significantly across ages 4-18 years. White matter, gray matter, and CSF volumes were larger for males than for females. Males and females showed similar change of gray and white matter volumes with age.

CONCLUSION

This simple, reliable, and valid method can be employed in clinical research for quantification of gray and white matter and CSF volumes in MR head scans. Increase in white matter volume may reflect ongoing axonal growth and myelination, and gray matter reductions may reflect synaptic pruning or cell death in the age span of 4-18 years.

Simulation of MRI cluster plots and application to neurological segmentation

The advent of magnetic resonance imaging has provided new opportunities for volume measurement of tissues, with applications increasing dramatically in recent years. Cluster classification techniques have proved the most popular for volume measurement, yet little attention has been paid to how the choice of images for analysis affects the quality and ease of segmentation. To address this issue, we have developed a system to simulate MRI cluster plots using multicompartmental anthropomorphic software models of anatomy, and components for image contrast, signal-to-noise ratio, image nonuniformity, tissue heterogeneity, imager field strength, the partial volume effect, correlation between proton density, T1 and T2, and a variety of data preprocessing techniques. The effect of these components on tissue cluster size, shape, orientation, and separation is demonstrated. The simulation allows an informed choice of pulse sequence, acquisition parameters, and data preprocessing for cluster classification to be made as well as providing an aid to interpretation of acquired data cluster plots and a valuable educational tool. The system has been used to choose suitable images for neurological segmentation of grey matter, white matter, CSF, and multiple sclerosis lesions using spin-echo, inversion recovery, and gradient-echo pulse sequences. Constraints on image selection are discussed.

Optimal linear transformation for MRI feature extraction

This paper presents development and application of a feature extraction method for magnetic resonance imaging (MRI), without explicit calculation of tissue parameters. A three-dimensional (3-D) feature space representation of the data is generated in which normal tissues are clustered around prespecified target positions and abnormalities are clustered elsewhere. This is accomplished by a linear minimum mean square error transformation of categorical data to target positions. From the 3-D histogram (cluster plot) of the transformed data, clusters are identified and regions of interest (ROI's) for normal and abnormal tissues are defined. These ROI's are used to estimate signature (prototype) vectors for each tissue type which in turn are used to segment the MRI scene. The proposed feature space is compared to those generated by tissue-parameter-weighted images, principal component images, and angle images, demonstrating its superiority for feature extraction and scene segmentation. Its relationship with discriminant analysis is discussed. The method and its performance are illustrated using a computer simulation and MRI images of an egg phantom and a human brain.

MR image segmentation using vector decomposition and probability techniques: a general model and its application to dual-echo images

A general model is developed for segmenting magnetic resonance images using vector decomposition and probability techniques. Each voxel is assigned fractional volumes of q tissues from p differently weighted images (q < or = p + 1) in the presence of partial-volume mixing, random noise, and other tissues. Compared with the eigenimage method, fewer differently weighted images are needed for segmenting the q tissues, and the contrast-to-noise ratio in the calculated fractional volumes is improved. The model can produce composite tissue-type images similar to that of the probability methods, by comparing the fractional volumes assigned to different tissues on each voxel. A three-tissue (p = 2, q = 3) model is illustrated for segmenting three tissues from dual-echo images. It provides statistical analysis to the algebraic method. A three-compartment phantom is segmented for validation. Two clinical examples are presented.