Multispectral analysis of magnetic resonance images

A technique for single-channel MR brain tissue segmentation: application to a pediatric sample

A segmentation method is presented for gray matter, white matter, and cerebrospinal fluid (CSF) in thin-sliced single-channel brain magnetic resonance (MR) scans. The method is based on probabilistic modeling of intensity distributions and on a region growing technique. Interrater and intrarater reliabilities for the method were high, and comparison with phantom studies and hand-traced results from an experienced rater indicated good validity. The method was designed to account for spatially dependent image intensity inhomogeneities. Segmentation of MR brain scans of 105 (56 male and 49 female) healthy children and adolescents showed that although the total brain volume was stable over age 4-18, white matter increased and gray matter decreased significantly. There were no sex differences in total gray and white matter growth after correction for total brain volume. White matter volume increased the most in superior and posterior regions and laterality effects were seen in hemisphere tissue volumes. These findings are consistent with other reports, and further validate the segmentation technique.

Neuroradiologic MR applications with multiparametric color composite display

The purpose of this article is to demonstrate the application of a PC-based multiparameter full color composite display technique of MR images of 14 selected patients with neuropathology while assessing the ability of this technique to display clinically important neuroanatomic and neuropathologic tissues. Using a PC with a 386 microprocessor and full color 24-bit graphics display capabilities, custom and commercially available image-processing softwares were applied to spatially aligned multiparameter proton density, T1-weighted (with and/or without gadolinium-DTPA) and T2-weighted MR image sets obtained from 14 patients with known neuropathology to generate intensity-based color composites. Quantitative color channel applications were used to assess the ability of this technique to differentiate anatomically and pathologically confirmed tissue types into unique color regions within the full color spectrum display in each patient case. Based on the results of pathologic correlation and quantitative color imaging analysis, the application of full color composite generation techniques to multiple MR images of selected neuropathology cases represents a viable technique for displaying diagnostically relevant tissue contrast information in one color image. With this technique, it is possible to generate composites that simultaneously display uniquely color-coded neuroanatomic and neuropathologic tissue information within the context of partially natural-appearing images.

Adaptive segmentation of MRI data

Intensity-based classification of MR images has proven problematic, even when advanced techniques are used. Intrascan and interscan intensity inhomogeneities are a common source of difficulty. While reported methods have had some success in correcting intrascan inhomogeneities, such methods require supervision for the individual scan. This paper describes a new method called adaptive segmentation that uses knowledge of tissue intensity properties and intensity inhomogeneities to correct and segment MR images. Use of the expectation-maximization (EM) algorithm leads to a method that allows for more accurate segmentation of tissue types as well as better visualization of magnetic resonance imaging (MRI) data, that has proven to be effective in a study that includes more than 1000 brain scans. Implementation and results are described for segmenting the brain in the following types of images: axial (dual-echo spin-echo), coronal [three dimensional Fourier transform (3-DFT) gradient-echo T1-weighted] all using a conventional head coil, and a sagittal section acquired using a surface coil. The accuracy of adaptive segmentation was found to be comparable with manual segmentation, and closer to manual segmentation than supervised multivariant classification while segmenting gray and white matter.

Unsupervised measurement of brain tumor volume on MR images

We examined unsupervised methods of segmentation of MR images of the brain for measuring tumor volume in response to treatment. Two clustering methods were used: fuzzy c-means and a nonfuzzy clustering algorithm. Results were compared with volume segmentations by two supervised methods, k-nearest neighbors and region growing, and all results were compared with manual labelings. Results of individual segmentations are presented as well as comparisons on the application of the different methods with 10 data sets of patients with brain tumors. Unsupervised segmentation is preferred for measuring tumor volumes in response to treatment, as it eliminates operator dependency and may be adequate for delineation of the target volume in radiation therapy. Some obstacles need to be overcome, in particular regarding the detection of anatomically relevant tissue classes. This study shows that these improvements are possible.

Selection of MR images for automated segmentation

MR images show a large range of contrast for various tissues in the body and are ideal for multispectral segmentation. Typically, only two MR images (dual-echo series) are used for segmentation; however, other images are often available. We evaluated MR images from 40 patients to determine the optimal type and number of images required for segmentation of tissues associated with brain tumors (normal brain, edema, necrosis, and active tumor). Pattern recognition methods indicated that three MR images from the same slice location were adequate for segmentation, as defined by feature selection and feature extraction measures based on training fields. This result was also confirmed by visually examining segmented images for all 40 patients. This work demonstrates that by using existing image/statistical analysis techniques (feature selection and feature extraction), one can systematically determine the optimal type and number of MR images for tissue segmentation.

Automatic magnetic resonance tissue characterization for three-dimensional magnetic resonance imaging of the brain

Computer-assisted diagnostic systems enhance the information available from magnetic resonance imaging. Segmentations are the basis on which three-dimensional volume renderings are made. The application of a raw data-based, operator-independent (automatic), magnetic resonance segmentation technique for tissue differentiation is demonstrated. Segmentation images of vasogenic edema with gross and histopathological correlation are presented for demonstration of the technique. A pixel was classified into a tissue class based on a feature vector using unsupervised fuzzy clustering techniques as the pattern recognition method. Correlation of fuzzy segmentations and gross and histopathology were successfully performed. Based on the results of neuropathological correlation, the application of fuzzy magnetic resonance image segmentation to a patient with a brain tumor and extensive edema represents a viable technique for automatically displaying clinically important tissue differentiation. With this pattern recognition technique, it is possible to generate automatic segmentation images that display diagnostically relevant neuroanatomical and neuropathological tissue contrast information from raw magnetic resonance data for use in three-dimensional volume reconstructions.

MRI segmentation: methods and applications

The current literature on MRI segmentation methods is reviewed. Particular emphasis is placed on the relative merits of single image versus multispectral segmentation, and supervised versus unsupervised segmentation methods. Image pre-processing and registration are discussed, as well as methods of validation. The application of MRI segmentation for tumor volume measurements during the course of therapy is presented here as an example, illustrating problems associated with inter- and intra-observer variations inherent to supervised methods.

Feature selection in the pattern classification problem of digital chest radiograph segmentation

In pattern classification problems, the choice of variables to include in the feature vector is a difficult one. The authors have investigated the use of stepwise discriminant analysis as a feature selection step in the problem of segmenting digital chest radiographs. In this problem, locally calculated features are used to classify pixels into one of several anatomic classes. The feature selection step was used to choose a subset of features which gave performance equivalent to the entire set of candidate features, while utilizing less computational resources. The impact of using the reduced/selected feature set on classifier performance is evaluated for two classifiers: a linear discriminator and a neural network. The results from the reduced/selected feature set were compared to that of the full feature set as well as a randomly selected reduced feature set. The results of the different feature sets were also compared after applying an additional postprocessing step which used a rule-based spatial information heuristic to improve the classification results. This work shows that, in the authors' pattern classification problem, using a feature selection step reduced the number of features used, reduced the processing time requirements, and gave results comparable to the full set of features.

Comparison of supervised MRI segmentation methods for tumor volume determination during therapy

Two different multispectral pattern recognition methods are used to segment magnetic resonance images (MRI) of the brain for quantitative estimation of tumor volume and volume changes with therapy. A supervised k-nearest neighbor (kNN) rule and a semi-supervised fuzzy c-means (SFCM) method are used to segment MRI slice data. Tumor volumes as determined by the kNN and SFCM segmentation methods are compared with two reference methods, based on image grey scale, as a basis for an estimation of ground truth, namely: (a) a commonly used seed growing method that is applied to the contrast enhanced T1-weighted image, and (b) a manual segmentation method using a custom-designed graphical user interface applied to the same raw image (T1-weighted) dataset. Emphasis is placed on measurement of intra and inter observer reproducibility using the proposed methods. Intra- and interobserver variation for the kNN method was 9% and 5%, respectively. The results for the SFCM method was a little better at 6% and 4%, respectively. For the seed growing method, the intra-observer variation was 6% and the interobserver variation was 17%, significantly larger when compared with the multispectral methods. The absolute tumor volume determined by the multispectral segmentation methods was consistently smaller than that observed for the reference methods. The results of this study are found to be very patient case-dependent. The results for SFCM suggest that it should be useful for relative measurements of tumor volume during therapy, but further studies are required. This work demonstrates the need for minimally supervised or unsupervised methods for tumor volume measurements.

Pure phase-encoded MRI and classification of solids

Here, the authors combine a pure phase-encoded magnetic resonance imaging (MRI) method with a new tissue-classification technique to make geometric models of a human tooth. They demonstrate the feasibility of three-dimensional imaging of solids using a conventional 11.7-T NMR spectrometer. In solid-state imaging, confounding line-broadening effects are typically eliminated using coherent averaging methods. Instead, the authors circumvent them by detecting the proton signal at a fixed phase-encode time following the radio-frequency excitation. By a judicious choice of the phase-encode time in the MRI protocol, the authors differentiate enamel and dentine sufficiently to successfully apply a new classification algorithm. This tissue-classification algorithm identifies the distribution of different material types, such as enamel and dentine, in volumetric data. In this algorithm, the authors treat a voxel as a volume, not as a single point, and assume that each voxel may contain more than one material. They use the distribution of MR image intensities within each voxel-sized volume to estimate the relative proportion of each material using a probabilistic approach. This combined approach, involving MRI and data classification, is directly applicable to bone imaging and hard-tissue contrast-based modeling of biological solids.

Application of fuzzy c-means segmentation technique for tissue differentiation in MR images of a hemorrhagic glioblastoma multiforme

The application of a raw data-based, operator-independent MR segmentation technique to differentiate boundaries of tumor from edema or hemorrhage is demonstrated. A case of a glioblastoma multiforme with gross and histopathologic correlation is presented. The MR image data set was segmented into tissue classes based on three different MR weighted image parameters (T1-, proton density-, and T2-weighted) using unsupervised fuzzy c-means (FCM) clustering algorithm technique for pattern recognition. A radiological examination of the MR images and correlation with fuzzy clustering segmentations was performed. Results were confirmed by gross and histopathology which, to the best of our knowledge, reports the first application of this demanding approach. Based on the results of neuropathologic correlation, the application of FCM MR image segmentation to several MR images of a glioblastoma multiforme represents a viable technique for displaying diagnostically relevant tissue contrast information used in 3D volume reconstruction. With this technique, it is possible to generate segmentation images that display clinically important neuroanatomic and neuropathologic tissue contrast information from raw MR image data.

Dual-echo MRI segmentation using vector decomposition and probability techniques: a two-tissue model

We combined a vector decomposition technique with Gaussian probability thresholding in feature space to segment normal brain tissues, tumors, or other abnormalities on dual-echo MR images. The vector decomposition technique assigns to each voxel a fractional volume for each of two tissues. A probability threshold, based on an assumed Gaussian probability density function describing random noise, isolates a region in feature space for fractional volume calculation that minimizes contamination from other tissues. The calculated fractional volumes are unbiased estimates of the true fractional volumes. The contrast-to-noise ratio (CNR) between tissues on the segmented images is the same as the Euclidean norm of CNRs in the original images. The method is capable of segmenting more than two tissues from a set of dual-echo images by sequentially analyzing different pairs of tissues. The model is analyzed mathematically and in experiments with a phantom. Two clinical examples are presented.

Computer-assisted identification and quantification of multiple sclerosis lesions in MR imaging volumes in the brain

Magnetic resonance (MR) imaging is the principal imaging technique for the diagnosis of multiple sclerosis (MS). However, quantifying the number and extent of lesions on MR images manually is arduous. The authors have developed a computerized three-dimensional (3D) quantitative system to assist in the identification and analysis of MS lesions in proton-density (PD)- and T2-weighted volumes of the head. The system provides intuitive, interactive operations that allow flexible extraction of information from the data. Use of the system to analyze MR examinations of a phantom containing regular "lesions" showed that accurate (average error, < 0.21 cm3) and precise (10% or better for lesions > 1 cm3) measurements of objects less than 7 cm3 is possible, and that an estimate of the quantization error predicted the uncertainty in the volume. Analysis of four MR examinations of a chronic-progressive MS patient conducted over an 18-month period was performed. A two-dimensional histogram showing the frequency of voxels with particular PD- and T2-weighted intensities revealed a distinct cluster only in histograms of sections that contained lesions. Measurements and 3D volume rendering of lesions clearly showed changes in lesion shape, position, and size.

Multispectral analysis of the brain using magnetic resonance imaging

The authors demonstrate an improved differentiation of the most common tissue types in the human brain and surrounding structures by quantitative validation using multispectral analysis of magnetic resonance images. This is made possible by a combination of a special training technique and an increase in the number of magnetic resonance channel images with different pulse acquisition parameters. The authors give a description of the tissue-specific multivariate statistical distributions of the pixel intensity values and discuss how their properties may be explored to improve the statistical modeling further. A statistical method to estimate the tissue-specific longitudinal and transverse relaxation times is also given. It is concluded that multispectral analysis of magnetic resonance images is a valuable tool to recognize the most common normal tissue types in the brain and surrounding structures.

Parameter estimation and tissue segmentation from multispectral MR images

A statistical method is developed to classify tissue types and to segment the corresponding tissue regions from relaxation time T(1 ), T(2), and proton density P(D) weighted magnetic resonance images. The method assumes that the distribution of image intensities associated with each tissue type can be expressed as a multivariate likelihood function of three weighted signal intensity values (T(1), T(2), P(D)) at each location within that tissue regions. The method further assumes that the underlying tissue regions are piecewise contiguous and can be characterized by a Markov random field prior. In classifying the tissue types, the method models the likelihood of realizing the images as a finite multivariate-mixture function. The class parameters associated with the tissue types (i.e. the weighted intensity means, variances and correlation coefficients of the multivariate function, as well as the number of voxels within regions of the tissue types of are estimated by maximum likelihood. The estimation fits the class parameters to the image data via the expectation-maximization algorithm. The number of classes associated with the tissue types is determined by the information criterion of minimum description length. The method segments the tissue regions, given the estimated class parameters, by maximum a posteriori probability. The prior is constructed by the tissue-region membership of the first- and second-order neighborhood. The method is tested by a few sets of T(1), T(2), and P(D) weighted images of the brain acquired with a 1.5 Tesla whole body scanner. The number of classes and the associated class parameters are automatically estimated. The regions of different brain tissues are satisfactorily segmented.

Improvements to the quality of MRI cluster analysis

Cluster analysis techniques are gaining widespread use for segmentation of MRI data, especially for volume measurement and 3-D display purposes. This paper describes four improvements to such techniques: (1) The use of intensity simulations to model cluster plots; (2) Correction of image nonuniformity; (3) Anisotropic smoothing of data; and (4) Automatic isolation of tissues of interest. Simulation of cluster plots allows an informed choice of pulse sequence(s) and acquisition parameters to be made. Correction of image nonuniformity and anisotropic smoothing reduce the spread of signal intensity from a single tissue thus producing significantly more compact clusters, whilst the isolation of tissues of interest prevents overlap of clusters from the tissues of interest with those not under consideration. These techniques may be used to improve the results of cluster analysis or traded off, for example to allow lower signal-to-noise images, shorter repetition time images, or fewer images to be used for segmentation.

3D display: a survey from theory to applications

We survey some of the literature on three-dimensional medical imaging. We report both on technical developments and on medical applications, with a concentration on material that has been published within the years 1990-1992.

A multispectral analysis of brain tissues

With the increasing use of three-dimensional MRI techniques it is becoming necessary to explore automated techniques for locating pathology in the volume images. The suitability of a specific technique to locate and identify healthy tissues of the brain was examined as a first step toward eventually identifying pathology in images. This technique, called multispectral image segmentation, is based on the classification of tissue types in an image according to their characteristics in various spectral regions. The spectral regions chosen for this study were the hydrogen spin-lattice relaxation time T1, spin-spin relaxation time T2, and spin density, rho. Single-echo, spin-echo magnetic resonance images of axial slices through the brain at the level of the lateral ventricles were recorded on a 1.5 Tesla imager from 20 volunteers ranging in age from 17 to 72 years. These images were used to calculate the T1, T2, and rho images used for the classification. Tissue classification was performed by locating clusters of pixels in a three-dimensional T1(-1)-T2(-1)-rho histogram. Gray matter, white matter, cerebrospinal fluid, meninges, muscle, and adipose tissues were readily classified in magnetic resonance images of the volunteers with a single set of T1, T2, and rho values. Cluster characteristics, such as size, shape, and location, provided information on the imaging procedure and tissue characteristics.

Three-dimensional MR reconstruction images of skull base tumors

Forty-eight patients with skull base tumors were evaluated prospectively with T1-weighted spin-echo two-dimensional (2D) magnetic resonance (MR) sequences, a three-dimensional (3D) MR TurboFLASH (fast low-angle shot) sequence, and a 3D reconstruction window technique. All patients underwent surgery with histopathologic correlation, and the three MR imaging techniques were compared to assess representation of tumor margins and the topographic relationship of tumor to surrounding tissue and adjacent vasculature. The best results were obtained with standard 2D spin-echo sequences after administration of the paramagnetic contrast agent gadopentetate dimeglumine. The 2D MR sequences gave the highest contrast-to-noise ratios, with decreasing values for 3D sequences and 3D reconstructions, respectively. Nevertheless, 3D MR imaging, by virtue of its good representation of adjacent structures, aided surgeons in planning surgical intervention. This study presents the technical features of 3D imaging of the skull base, the choices involved in its implementation, and its potential clinical applications.

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.

MRI: stability of three supervised segmentation techniques

Supervised segmentation methods from three families of pattern recognition techniques were used to segment multispectral MRI data. Studied were the maximum likelihood method (MLM), k-nearest neighbors (k-NN), and a back-propagation artificial neural net (ANN). Performance was measured in terms of execution speed, and stability for the selection of training data, namely, region of interest (ROI) selection, and interslice and interpatient classifications. MLM proved to have the smallest execution times, but demonstrated the least stability. k-NN showed the best stability for training data selection. To evaluate the segmentation techniques, multispectral images were used of normal volunteers and patients with gliomas, the latter with and without MR contrast material. All measures applied indicated that k-NN provides the best results.

Neural-network-based segmentation of multi-modal medical images: a comparative and prospective study

This work presents an investigation of the potential of artificial neural networks for classification of registered magnetic resonance and X-ray computer tomography images of the human brain. First, topological and learning parameters are established experimentally. Second, the learning and generalization properties of the neural networks are compared to those of a classical maximum likelihood classifier and the superiority of the neural network approach is demonstrated when small training sets are utilized. Third, the generalization properties of the neural networks are utilized to develop an adaptive learning scheme able to overcome interslice intensity variations typical of MR images. This approach permits the segmentation of image volumes based on training sets selected on a single slice. Finally, the segmentation results obtained both with the artificial neural network and the maximum likelihood classifiers are compared to contours drawn manually.

Routine quantitative analysis of brain and cerebrospinal fluid spaces with MR imaging

A computerized system for processing spin-echo magnetic resonance (MR) imaging data was implemented to estimate whole brain (gray and white matter) and cerebrospinal fluid volumes and to display three-dimensional surface reconstructions of specified tissue classes. The techniques were evaluated by assessing the radiometric variability of MR volume data and by comparing automated and manual procedures for measuring tissue volumes. Results showed (a) the homogeneity of the MR data and (b) that automated techniques were consistently superior to manual techniques. Both techniques, however, were affected by the complexity of the structure, with simpler structures (eg, the intracranial cavity) showing less variability and better spatial correlation of segmentation results between raters. Moreover, the automated techniques were completed for whole brain in a fraction of the time required to complete the equivalent segmentation manually. Additional evaluations included interrater reliability and an evaluation that included longitudinal measurement, in which one subject was imaged sequentially 24 times, with reliability computed from data collected by three raters over 1 year. Results showed good reliability for the automated segmentation procedures.

Applications of hierarchical image segmentation techniques: aorta segmentation

The segmentation of objects from complex images is difficult due to indistinct boundaries between objects and similarity of objects. We have used a hierarchical segmentation approach to accurately distinguish between objects and identify the corresponding boundaries. This approach has been used successfully to extract the aorta from transverse magnetic resonance (MR) images of the abdomen. The procedure to segment the abdominal aorta involves three progressive steps: aorta detection, aorta extraction, and estimation of the aorta wall boundary. Comparison of hierarchical segmentation techniques with single-step segmentation methods (e.g., region-growing, edge-detection) shows that hierarchical segmentation yields more reliable results.

Composite and classified color display in MR imaging of the female pelvis

Because of its superior soft-tissue-imaging capabilities, MRI has proved to be an excellent modality for visualizing the contents of the female pelvis. In an effort to potentially improve gynecological MRI studies, we have applied color composite techniques to sets of spin-echo and gradient-echo gray-tone MR images obtained from various individuals. For composite generation, based on tissue region of interest calculated mean pixel intensity values, various colors were applied to spatially aligned images using a DEC MicroVAX II computer with interactive digital language (IDL) so that tissue contrast patterns could be optimized in the final image. The IDL procedures, which are similar to those used in NASA's LANDSAT image processing system, allowed the generation of single composite images displaying the combined information present in a series of spatially aligned images acquired using different pulse sequences. With our composite generation techniques, it was possible to generate seminatural-appearing color images of the female pelvis that possessed enhanced conspicuity of specific tissues and fluids. For comparison with color composites, classified images were also generated based on computer recognition and statistical separation of distinct tissue intensity patterns in an image set using the maximum likelihood processing algorithm.

A comparison of neural network and fuzzy clustering techniques in segmenting magnetic resonance images of the brain

Magnetic resonance (MR) brain section images are segmented and then synthetically colored to give visual representations of the original data with three approaches: the literal and approximate fuzzy c-means unsupervised clustering algorithms, and a supervised computational neural network. Initial clinical results are presented on normal volunteers and selected patients with brain tumors surrounded by edema. Supervised and unsupervised segmentation techniques provide broadly similar results. Unsupervised fuzzy algorithms were visually observed to show better segmentation when compared with raw image data for volunteer studies. For a more complex segmentation problem with tumor/edema or cerebrospinal fluid boundary, where the tissues have similar MR relaxation behavior, inconsistency in rating among experts was observed, with fuzz-c-means approaches being slightly preferred over feedforward cascade correlation results. Various facets of both approaches, such as supervised versus unsupervised learning, time complexity, and utility for the diagnostic process, are compared.

Generation of color composites for enhanced tissue differentiation in magnetic resonance imaging of the brain

Currently, the diagnostic interpretation of magnetic resonance (MR) images requires that radiologists integrate specific tissue contrast information from several different images obtained at the same anatomic slice position. Each of these images has its own unique tissue contrast patterns which are based on the image acquisition parameters (pulse sequence) selected. The complex contrast patterns observable in these images reflect the inherent biophysical characteristics of the tissues and fluids present in the imaged section. In an effort to increase the diagnostic accuracy and efficiency of MR image interpretation, we have generated color composite images from quantitatively analyzed achromatic MR images of the brain, obtained while utilizing different pulse sequences. By using a DEC MicroVAX II computer with Interactive Digital Language (IDL), this color display method has been applied to images obtained from General Electric Signa and Siemens Magnatom imagers. For this study, our image sets included T1-weighted, T2-weighted, and proton density spin echo sequences as well as both high and low flip angle gradient echo sequences. Advantages of our color composite methods, in contrast to many other image processing techniques that have been described, are that minimal information is lost, computer misclassification of tissues is avoided, and the conspicuity of specific tissues is enhanced. Furthermore, with this method it is possible to produce composite images whose color renditions approach a natural anatomic tissue appearance. Availability of these color composites to radiologists may improve the efficiency and accuracy of the diagnostic interpretation of MR images.

Validation of magnetic resonance imaging (MRI) multispectral tissue classification

The application of NASA multispectral image processing technology for analysis of Magnetic Resonance Imaging (MRI) scans has been studied. Software and hardware capability has been developed, and a statistical evaluation of multispectral analysis application to MRI scans of the head has been performed.

Improvement of 3D acquisition and visualization in MRI

Three-dimensional (3D) visualization techniques are becoming an ever more important aid in the interpretation of tomographic data. Up to now, however, they have not received widespread use in MRI, because both acquisition and visualization techniques have been inadequate. In this paper we describe new 3D acquisition techniques which can acquire up to 128 slices with a resolution of 256 x 256 pixels in from 8 to 20 min. These techniques produce 3D data sets with excellent contrast and few motion artifacts, which are very well suited for 3D visualization techniques. For the visualization we investigate several rendering techniques, describe some improvements and compare their results. We found that there is no single method which renders all objects equally well. We show which shading method is best suited for different objects and why the other methods fail. Our studies suggest that in a 3D view with several objects each object should be rendered with a separate shading method. In so doing, 3D views can be generated which look like the real human anatomy.

Computer-human interface issues in the design of an intelligent workstation for scientific visualization

The long-range goal of our research is to create an intelligent assistant for interactive scientific data visualization via both sight and sound. There are a variety of computer-human interface (CHI) issues that are unique to our approach to interactive visualization. It is upon these issues that we focus here. In this paper, we: (1) describe the approach to interactive visualization taken by the project which is the context of our work; (2) specify the CHI issues that are peculiar to this approach; (3) summarize the current capabilities of our workstation for performing human factors experiments; (4) describe the research plan we have developed for learning how to provide a user with intelligent assistance for dealing with those issues; (5) present a representative pilot study that has contributed useful information; (6) summarize the results of our pilot studies; and (7) discuss the direction of our future work. We do not claim to be solving the general case of how to provide intelligent assistance for scientific visualization. We do, however, expect that the progress we make in one visualization environment will contribute to understanding of the general case.

Three-dimensional magnetic resonance images of the brain: application to neurosurgical planning

Data from single 10-minute magnetic resonance scans were used to create three-dimensional (3-D) views of the surfaces of the brain and skin of 12 patients. In each case, these views were used to make a preoperative assessment of the relationship of lesions to brain surface structures associated with movement, sensation, hearing, and speech. Interactive software was written so that the user could "slice" through the 3-D computer model and inspect cross-sectional images at any level. A surgery simulation program was written so that surgeons were able to "rehearse" craniotomies on 3-D computer models before performing the actual operations. In each case, the qualitative accuracy of the 3-D views was confirmed by intraoperative inspection of the brain surface and by intraoperative electrophysiological mapping, when available.

Three-dimensional imaging in craniofacial surgery

Over the past decade, three-dimensional (3-D) imaging has been developed to such a stage of perfection and to such a level of interactive selective imaging of specific anatomic and pathologic structures that craniofacial surgeons can now use this technique effectively in the planning of complicated reconstructive surgery. In addition, modeling techniques have been devised that can be used in surgical simulation and in the manufacture of implants and prosthetic devices. The technical aspects of 3-D imaging are discussed in relation to their applications in craniofacial surgery, and reference is made to the literature describing these techniques in full detail. The results are illustrated with cases that the authors have processed by means of: (a) a clinical research program that was developed on a general purpose computer which provided full flexibility in changing and improving the reconstruction algorithms (Lobregt algorithms and DEC VAX 750 computer), (b) a system under development (Pixar PICS 2000), and (c) a commercial system (Cemax 1500X). Finally, a number of emerging techniques are discussed such as surgical stimulation (electronic sculpting), and trends such as multimodality imaging.

Unified multiple-feature color display for MR images

A display method is proposed in which the spin-lattice relaxation time T1, the spin-spin relaxation time T2, and the proton density rho of each pixel in a MR image are simultaneously expressed in color features in a unified way that allows international standardization. MR images were made from a phantom, a healthy volunteer, and patients in such a way that T1 and T2 and proton density images could be derived. T1 and T2 data were compared with accurate relaxation time measurements of the phantom content. Color images were computed from the acquired T1 and T2 images using matrix multiplication on a pixel base. In this way the color combination in each pixel represents the properties of that particular pixel by a unique mixing of the elementary colors red, green, and blue. Color resolution could be modified using different choices of the reference triangle in which the color combinations were defined. This method of representation offers a means for displaying multiple features as T1 and T2 in one directly interpretable image, independent of instrumental settings.

High-resolution 1H NMR spectral signature from human atheroma

Coronary artery disease due to atherosclerosis takes the lives of approximately 550,000 Americans each year--an enormous toll. Put in economic terms, the cost to the United States alone has been estimated to exceed 60 billion dollars annually. We have found that well-resolved proton (1H) NMR spectra can be obtained from human atheroma (fatty plaque), despite its macroscopic solid appearance. The fraction of the total spectral intensity corresponding to the sharp 1H NMR signals is temperature dependent and approaches unity at body temperature (37 degrees C). Studies of the total lipids extracted from atheroma and cholesteryl esters were conducted to identify the chemical and physical origin of the spectral signature. The samples were characterized through assignment of their chemical shifts and by measurement of their T1 and T2 relaxation times as a function of magnetic field strength. The results suggest that the relatively sharp 1H NMR signals from human atheroma (excluding water) are due to a mixture of cholesteryl esters, whose liquid-crystalline to isotropic fluid phase transition is near body temperature. Preliminary applications to NMR imaging of human atheroma are reported, which demonstrate early fatty plaque formation within the wall of the aorta. These findings offer a basis for noninvasive imaging by NMR to monitor early and potentially reversible stages of human atherogenesis.

Automated classification of periodontal disease using bitewing radiographs

The feasibility of applying a prototype, computer-based pattern recognition system to the objective classification of periodontal disease using dental radiographs was tested. Twenty-nine observer-classified bitewing radiographs, representing seven individuals with varying grades of periodontal disease, were selected. The radiographs were digitized using a computer-controlled TV camera. Mathematical features of these radiographs were interactively extracted using a digital image processing system (International Imaging Systems Model 75 and System/575). The features extracted from these radiographs included the brightness levels of cortical and trabecular bone and ratios of bone-loss to linear-crown height. Twenty-eight mathematically defined features (variables) were determined for each radiograph. Stepwise linear discriminant analysis used these features to classify subjects based on the presence and extent of periodontal disease. This pattern recognition system was able to grade periodontal disease in our test series with percentages of correct classifications ranging from 78.8% to 91%. This technology is particularly applicable to the development of morbidity and activity indices for periodontal diseases.

Identification and 3-D quantification of atherosclerosis using magnetic resonance imaging

Cardiovascular disease due to atherosclerosis is a leading cause of death in the United States as well as other developed countries. This paper describes the development of image processing, pattern recognition, and graphical display techniques to non-invasively quantify the atherosclerotic disease process using magnetic resonance imaging (MRI). We have demonstrated the ability to identify the soft tissue classes of (1) normal, smooth muscle wall, (2) fatty plaque, (3) complex, fibrous plaque, and (4) calcified plaque. The objective of this work has been to combine functional information, such as plaque tissue type, with structural information, represented by 3-D display of vessel structure, into a single composite display. The results of this work provide a "high information content" display which will aid in the diagnosis and analysis of the atherosclerotic disease process, and permit detailed and quantitative studies to assess the effectiveness of therapies (e.g. changes in diet, exercise and drug administration).

The computer and stereotactic surgery in neurological surgery

The technical aspects, current uses, and future clinical applications of stereotactic surgery and three-dimensional imaging in neurological surgery are reviewed.

Oscillating intensity display of soft tissue lesions in MRI

A new postprocessing method for improving visualization of soft tissue lesions in MR images is described. Abnormal tissues are detected by a computerized tissue characterization algorithm which is based on measurements of intensity in a spatially matched pair of T1- and T2-weighted images. Simultaneous display of information from this pair of static images is achieved by using a temporal parameter (amplitude or frequency of intensity oscillation) to encode abnormal pixels. Specifically, a movie is created in which pixel intensities of abnormal tissues are made to oscillate so that the amplitude (or frequency) of oscillation is proportional to an abnormality index which depends on the difference between intensities of normal and abnormal tissues in the original image pair. The visual effect is that of a churning motion within the lesion, while surrounding normal tissues are displayed as stable structures. This technique increases the conspicuity of the lesion by exploiting the eye's great sensitivity to motion.