Noninvasive three-dimensional computer imaging of matrix-filled fossil skulls by high-resolution computed tomography.

A noninvasive computer imaging technique allows three-dimensional images of fossil skulls to be generated from two-dimensional serial computerized tomographic scan data. The computer programs can "dissect" the skull in different planes by making portions of it and any obstructing matrix transparent in order to reveal intracranial morphology. The computer image is geometrically precise so that linear distances, angles, areas, volumes, and evaluations of symmetry can be determined.

Endocranial features of Australopithecus africanus revealed by 2- and 3-D computed tomography.

The earliest hominid from South Africa, Australopithecus africanus, is known from only six specimens in which accurate assessment of endocranial capacity and cranial venous outflow pattern can be obtained. This places a severe limit on a number of hypotheses concerning early hominid evolution, particularly those involving brain-body size relationships and adaptations of the circulatory system to evolving upright posture. Advances in high-resolution two- and three-dimensional computed tomography (CT) now allow the inclusion of another important specimen to this list, MLD 37/38 from Makapansgat. A new computer imaging technique is described that "reconstructs" the missing portions of the endocranial cavity in order to determine endocranial capacity. In addition, CT evaluation allows assessment of cranial venous outflow pattern even in cases where the endocranial cavity is completely filled with stone matrix. Results show that endocranial capacity in this specimen is less than originally proposed and also support the view that gracile and robust australopithecines evolved different cranial venous outflow patterns in response to upright postures.

A method to analyze 2-dimensional daily radiotherapy portal images from an on-line fiber-optic imaging system.

On-line radiotherapy imaging systems allow convenient treatment verification and generate a wealth of data. Quantitative analysis of data will provide important information about the nature of treatment variations. Using an inhouse fiber-optic imaging system to acquire daily portal images for five patients, we have developed a method to analyze the cumulative positional variation of blocks in the 2-dimensional images. For each beam arrangement used to treat a particular patient, a reference portal image was established. All other images for that patient were registered with respect to the anatomical landmarks visible on the reference image. Two-dimensional frequency distributions describing the overlap of the blocks during the course of treatment were then calculated and superimposed on the reference image. Results of the analysis show positional and quantitative information about the daily variation in block placement, and appeared to be site-dependent. Long term verification studies using on-line imaging systems will be important in the understanding of treatment uncertainties.

Portal dose images. I: Quantitative treatment plan verification.

The comparison of a predicted portal dose image, calculated during treatment planning, with the measured image obtained during treatment is proposed as an approach to verify the correct implementation of a patient treatment plan. The comparison inherently verifies both the geometric alignment and the dose delivered. Feasibility studies were conducted with 60Co irradiation of a modular plastic phantom, an anthropomorphic phantom and a patient with lung cancer. Calculations were made with the 3-dimensional scatter ray-trace Delta Volume method. Calculated distributions and/or selected points of transmitted dose correction factors were compared with measurements made with TLD, scanning ionization chamber and film. For the two phantom studies, excellent agreement, usually to within 3%, was achieved when positioning of the phantoms were accurate. The patient study showed that selected point comparisons were inadequate in identifying the cause of errors when disagreement occurred. Simple subtraction of the calculated and measured images showed a 4 mm translational misalignment. The results are encouraging and demonstrate that portal dose images can be used to detect large geometric and dosimetric discrepancies between treatment plan calculations and measurements. The results also show that perfect verification is virtually impossible in the clinical situation. More work is required to use the verification information for improving the estimation of dose to the patient.

Image-based dose planning of intracavitary brachytherapy: registration of serial-imaging studies using deformable anatomic templates.

PURPOSE: To demonstrate that high-dimensional voxel-to-voxel transformations, derived from continuum mechanics models of the underlying pelvic tissues, can be used to register computed tomography (CT) serial examinations into a single anatomic frame of reference for cumulative dose calculations. METHODS AND MATERIALS: Three patients with locally advanced cervix cancer were treated with CT-compatible intracavitary (ICT) applicators. Each patient underwent five volumetric CT examinations: before initiating treatment, and immediately before and after the first and second ICT insertions, respectively. Each serial examination was rigidly registered to the patient's first ICT examination by aligning the bony anatomy. Detailed nonrigid alignment for organs (or targets) of interest was subsequently achieved by deforming the CT exams as a viscous-fluid, described by the Navier-Stokes equation, until the coincidence with the corresponding targets on CT image was maximized. In cases where ICT insertion induced very large and topologically complex rearrangements of pelvic organs, e.g., extreme uterine canal reorientation following tandem insertion, a viscous-fluid-landmark transformation was used to produce an initial registration. RESULTS: For all three patients, reasonable registrations for organs (or targets) of interest were achieved. Fluid-landmark initialization was required in 4 of the 11 registrations. Relative to the best rigid bony landmark alignment, the viscous-fluid registration resulted in average soft-tissue displacements from 2.8 to 28.1 mm, and improved organ coincidence from the range of 5.2% to 72.2% to the range of 90.6% to 100%. Compared to the viscous-fluid transformation, global registration of bony anatomy mismatched 5% or more of the contoured organ volumes by 15-25 mm. CONCLUSION: Pelvic soft-tissue structures undergo large deformations and displacements during the external-beam and multiple-ICT course of radiation therapy for locally advanced cervix cancer. These changes cannot be modeled by the conventional rigid landmark transformation method. In the current study, we found that the deformable anatomic template registration method, based on continuum-mechanics models of deformation, successfully described these large anatomic shape changes before and after ICT. These promising modeling results indicate that realistic registration of the cumulative dose distribution to the organs (or targets) of interest for radiation therapy of cervical cancers is achievable.

Evaluation of computer display systems using digital test patterns.

Two digital test patterns are used to evaluate computer video display systems. One pattern consists of linear and logarithmic gray-scale wedges, blocks of constant brightness, and horizontal and vertical bar patterns. The other is a cross-hatch bar pattern. These patterns are used to assess display systems' gray scale, resolution, and spatial linearity. Microdensitometry measurements were performed to determine brightness uniformity, spatial resolution, and to relate optical density to exposure. These test patterns, used qualitatively, are useful in the routine adjustment and quality control of digital systems. The precise quantitative characterization of a display system is important in research applications involving image processing.

Preoperative three-dimensional reconstruction of the heart and great vessels in patients with congenital heart disease. Technique and initial results.

Magnetic resonance images have been obtained preoperatively in six patients with congenital heart disease. Contiguous sequences of electrocardiogram-triggered spin-echo images have been reconstructed in three-dimensional form to define the size and anatomic relationships to the great vessels and internal cardiac structures. Findings of magnetic resonance imaging were corroborated by angiographic and sector-scan echocardiographic studies and at operation. Individual scan slices were manually edited to separate the heart and great vessels from the blood within them and from extracardiac structures. Surface reconstruction software originally developed for craniofacial and orthopedic surgical planning was adapted for processing of cardiac magnetic resonance image sequences. Preoperative three-dimensional magnetic resonance imaging reconstructions were obtained in patients with aortic coarctation with ventricular septal defect, hypoplastic left ventricle, pulmonary artery atresia with ventricular septal defect, atrial septal defect, partial atrioventricular canal defect with anomalous pulmonary venous drainage, and tetralogy of Fallot with peripheral pulmonary artery stenosis. The reconstructions showed anatomic findings consistent with two-dimensional magnetic resonance imaging, echocardiography, cineangiography, and intraoperative findings. The three-dimensional images have a format that is familiar and consistent with the gross intraoperative appearance of the heart and great vessels. These three-dimensional images can facilitate the interpretation of magnetic resonance scan findings for cardiac surgeons without the sacrifice of significant clinical information.

Binary nature and radiographic identifiability of craniosynostosis.

RATIONALE AND OBJECTIVES: Two independent gold standards and diagnoses from three-dimensional computed tomography (CT) images were used to examine the possibility that craniosynostosis is a binary abnormality that potentially may be diagnosed without error. METHODS: Surgical reports, histology of excised sutures, and three-dimensional CT images were compared for 25 children undergoing surgical management of craniosynostosis. Surgical reports identified sutures as normal or abnormal. Histology reported suture closure on a 5-point scale. Four radiologists used three-dimensional CT images to diagnose sutures on a 6-point rated response scale. RESULTS: Sutures with histology 0, 1, or 2 were normal on surgical reports, and those with histology 3 or 4 were abnormal. Most readers achieved nearly perfect sensitivity and specificity. Reader confidence was unrelated to degree of pathology. CONCLUSION: Craniosynostosis appears to be binary in our sample. Surgical reports, pathology results, and three-dimensional CT images read by experienced viewers achieved nearly perfect agreement.

Surface imaging from computerized tomographic scans.

A series of computer programs were created to display digital data obtained by the computerized tomographic (CT) scanner in a format readily comprehensible to the nonradiologist physician. The software requires no hardware beyond the CT scanner itself, is efficient in computational time and storage, and requires no operator intervention. The program manipulates standard CT axial data into images that simulate photographs of skin, muscle, or bone surfaces from a variety of perspectives. While the images themselves are planar, the surfaces they display are three dimensional. Portions of the images can be made transparent to permit viewing of previously concealed surfaces. Soft tissue and osseous surfaces may be superimposed to assist the localization of deep pathologic processes. Although the program was created to facilitate craniofacial reconstructive surgery, it has now been applied in over 150 patients for the evaluation of intracranial tumors, head and neck malignancies, cervical spine injuries, pelvic fractures, and deformities of the hands and feet. This ability to provide the surgeon with images that allow direct visualization of anatomic details has reduced operative time and improved surgical results in our management of congenital and acquired deformities.

Noninvasive three-dimensional reconstruction of the heart and great vessels by ECG-gated magnetic resonance imaging: a new diagnostic modality.

We have developed a new method of computer image processing that allows true three-dimensional (3-D) images of the heart and great vessels to be reconstructed from standard ECG-gated two-dimensional magnetic resonance (MR) images. Contiguous 5-mm thick MR images of the thorax from the level of the cardiac apex to the aortic arch were obtained in 4 normal volunteers and 3 patients with congenital heart disease: 1 with pseudotruncus arteriosus and 1 with a ventricular septal defect, each with Eisenmenger's complex, and 1 with aortic coarctation. Each image could be obtained at up to seven different intervals throughout the cardiac cycle with ECG gating. The scanning procedure is noninvasive and requires no contrast material. Using standard software, images from each interval in the cardiac cycle were edited to isolate pertinent cardiac and great vessel structures. High-resolution 3-D reconstructions were formed for each interval by stacking the edited images. Sequential projection of 3-D reconstructions from each interval yields four-dimensional (includes time) cine views. Both 3-D and cine views can be obtained from any axis or divided in any plane to allow accurate, noninvasive assessment of cardiac and great vessel anatomy, chamber volumes, and regional and global wall motion. Noninvasive 3-D reconstruction of the heart and great vessels provides accurate anatomical data not available from standard cardiac catheterization or other noninvasive diagnostic procedures, and aids in the preoperative planning of the procedure to correct complex congenital malformations.

Preliminary study of magnetic resonance imaging characteristics in 8- to 16-year-olds with mania.

OBJECTIVE: To examine magnetic resonance imaging (MRI) characteristics in children and adolescents with mania according to DSM-III-R criteria. METHOD: A convenience sample of consecutively referred 8- to 16-year-old manic (n = 10) and normal (n = 5) subjects were assessed using the Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present Episode Version, the Children's Global Assessment Scale, and the Family History-Research Diagnostic Criteria. MRI scans were obtained from unsedated subjects using a 1.5 T MR scanner to acquire T1-weighted coronal and sagittal images and T2-weighted axial images. Images were assessed by blind clinical interpretation, ratings of T2-weighted deep white matter hyperintensities and petalia, and computer-assisted volumetric analysis of ventricular and cerebral volumes. RESULTS: Eight of 10 manic subjects and all 5 controls completed the scans. Scans of 4 manic subjects and 1 control subject showed ventricular or white matter abnormalities by clinical interpretation. Significant findings were positive correlations between increasing age and both right and left ventricular volumes. Two of the 8 manic subjects and no controls had confluent subcortical hyperintensities. CONCLUSIONS: MRI brain scanning was feasible in 8- to 16-year-olds. Preliminary findings from clinical interpretations and structured ratings suggest structural differences between young manic and normal subjects. Investigations of larger samples are needed to better characterize the differences.

Regional myocardial stress distribution from magnetic resonance image-based mathematical models.

The instantaneous regional stress distribution within the myocardium, which cannot be directly measured, has been estimated using improved numerical methods and nonaxisymmetric biventricular geometry. To do this, we have employed computer-aided solid mathematical modeling to generate a three-dimensional representation for an ex vivo canine biventricular unit using magnetic resonance imaging. A two-dimensional transverse section was isolated from the solid mathematical model for regional stress analysis using p-version finite element analysis. Loading conditions and material property descriptions were taken from published reports. Analyses showed the maximum principal stresses to range from -1.76 X 10(5) to 8.52 X 10(5) dynes/cm2 during systolic loading, and from -3.85 X 10(4) to 1.13 X 10(5) dynes/cm2 during diastolic loading. This study demonstrates that magnetic resonance image-based solid mathematical biventricular models are suitable for regional stress analysis using p-version finite element analysis. p-Version finite element analysis using magnetic resonance image-based cardiac representations facilitates in vivo stress-strain analyses and may allow the clinical estimation of regional myocardial stress.

Three-dimensional display of intracranial soft-tissue structures.

Efficient three-dimensional surface reconstructions of intracranial bony, soft-tissue, and cerebrospinal fluid-filled structures have been created from serial narrowly collimated transaxial computed tomographic (CT) scans. These three-dimensional surface images are useful in visualizing the relations among intracranial details. Views may be produced in frontal, rear, 45 degrees oblique (anterior and posterior), top and bottom, and both lateral projections. This method has been applied to image the surface of intracranial structures by reformatting the images with elimination of the overlying bone. This method was realized as a computer program that operates on a conventional unmodified CT scanner without operator intervention.

Craniosynostosis: diagnostic imaging with three-dimensional CT presentation.

PURPOSE: To measure diagnostic performance and preference of two three-dimensional CT reconstruction modalities (voxel-gradient and surface-projection) displayed two ways (conventional and unwrapped) in craniosynostosis confirmed by surgical inspection and histologic analysis of resected sutures. METHODS: High-resolution 2-mm contiguous CT sections were obtained and three-dimensional reconstruction images generated for 25 infants and children with skull deformities before surgical treatment of craniosynostosis. Two pediatric radiologists and two neuroradiologists first ranked images by their own preferences for diagnostic use. Then they diagnosed craniosynostosis from images presented in random order and blinded. The standard of reference was inspection during surgery and histologic evaluation of excised sutures. Finally, reviewers repeated their subjective preference tests. RESULTS: The least experienced radiologist had 100% sensitivity for all imaging modalities and specificities ranging from 43% to 83%. The two most experienced radiologists performed nearly identically, with sensitivities of 96% and specificities of 100%. After performing diagnostic tasks using all image types, all radiologists preferred conventional surface projections. CONCLUSION: Experienced readers can achieve nearly perfect diagnostic performance using the latest three-dimensional CT reconstruction images, making it a contribution to the diagnostic process. Although performance is nearly identical for all modalities, readers strongly prefer conventionally presented surface-projection images.

Techniques of CT colonography (virtual colonoscopy).

Colorectal cancer is a leading cause of death in older adults, which usually involves a long-term progressive change of normal mucosa into adenomatous polyps and then cancer. The detection and treatment of this disease in an early stage can lead to a cure in most cases by simply removing the polyp. Computed tomographic colonography (CTC), also referred to as virtual colonoscopy (VC), is a recent advance that gives an intraluminal visualization of the colon that is similar to endoscopy. VC requires fast 3D display (at least 10 frames/sec) of the colon's mucosal surface on a computer screen. Spiral/helical computer tomography is used to gather 3D volume data prior to display. CTC has been demonstrated to be promising for colorectal cancer screening. Studies on unraveling of the colon are underway to map the convoluted tubular structure into a straightened and flattened image volume for global visualization. In this article, we review the current status of CTC with an emphasis on image processing and visualization algorithms. Clinical assessment results of existing techniques are summarized. Practical issues and future perspectives are also discussed.

Multispectral magnetic resonance image analysis.

Multiecho magnetic resonance (MR) scanning produces tomographic images with approximately equal morphologic information but varying gray scales at the same anatomic level. Multispectral image classification techniques, originally developed for satellite imaging, have recently been applied to MR tissue characterization. Statistical assessment of multispectral tissue classification techniques has been used to select the most promising of several alternative methods. MR examinations of the head and body, obtained with a 0.35, 0.5, or 1.5T imager, comprised data sets with at least two pulse sequences yielding three images at each anatomical level: (1) TR = 0.3 sec, TE = 30 msec, (2) TR = 1.5, TE = 30, (3) TR = 1.5, TE = 120. Normal and pathological images have been analyzed using multispectral analysis and image classification. MR image data are first subjected to radiometric and geometric corrections to reduce error resulting from (1) instrumental variations in data acquisition, (2) image noise, and (3) misregistration. Training regions of interest (ROI) are outlined in areas of normal (gray and white matter, CSF) and pathological tissue. Statistics are extracted from these ROIs and classification maps generated using table lookup, minimum distance to means, maximum likelihood, and cluster analysis. These synthetic maps are then compared pixel by pixel with manually prepared classification maps of the same MR images. Using these methods, the authors have found that: (1) both supervised and unsupervised classification techniques yielded theme maps (class maps) which demonstrated tissue characteristic signatures and (2) tissue classification errors found in computer-generated theme maps were due to subtle gray scale changes present in the original MR data sets arising from radiometric inhomogeneity and spatial nonuniformity.

Spatial variation of section sensitivity profile in spiral computed tomography.

The section sensitivity profile (SSP) is widely used to describe the longitudinal image resolution in computed tomography (CT). Due to asymmetry in spiral CT interpolation, the SSP depends on the transverse position of the associated longitudinal line. Spatial variation in the spiral CT SSP has not been investigated previously. To determine the SSP variation over the scan field, the SSP for the half-scan interpolation method in terms of its transverse position was formulated. Based on the SSP formula, mathematical analysis and numerical simulation were performed. It was shown that the SSP associated with a transverse slice is antisymmetric with respect to the abscissa axis (which connects the slice center and the source location in the slice). Moment analysis was conducted on the SSP up to order three. The area under the SSP is equal to that under the longitudinal detector response function. As a result, with a normalized SSP the total mass of an object approximately remains after reconstruction. The mean of the SSP is equal to zero, although the SSP is asymmetric. The standard deviation and skewness of the SSP were numerically simulated and plotted for typical imaging parameters. It was demonstrated that for a 50 degrees fan angle (which is extended by two extreme rays in a fan beam) the relative change in the slice thickness is no more than 10%, while the skewness of the SSP is within about 0.15. The smaller the fan angle, the less the SSP spatial variation. Therefore, in spiral CT the SSP along the longitudinal axis can be used as a representative of the SSP family.

Helical CT image noise--analytical results.

Helical CT is an important recent development in x-ray CT. In helical CT, planar projection sets are synthesized from raw projection data via interpolation. Among various interpolation schemes, linear interpolation is usually preferred due to its efficiency and performance. In this paper, image noise variance is derived for typical helical CT linear interpolation techniques, including the full scan (FS), under-scan (US), full scan with interpolation (FI), half-scan (HS), half-scan with interpolation (HI) and half-scan with extrapolation (HE) methods. Image noise deviation ratios of helical CT to conventional 360 degrees reconstruction (CR) are tabulated. These are consistent with previously reported simulation results. The theoretical results provide further understanding of helical CT noise performance. It is shown that helical CT image noise deviation is independent of transaxial position, proportional to the raw projection noise deviation, and not affected by the fan angle (approximately for the HE method). Also, helical CT image noise variance is proportional to the area under the square of the reconstruction filter.

Optimal pitch in spiral computed tomography.

The focus of this paper is to analytically optimize spiral/helical computed tomography (CT) protocols based on a simplified imaging model. Spiral CT was approximately modeled as follows: Using the half-scan raw data interpolation method, the variance of the spiral CT slice sensitivity profile is equal to the sum of squared detector collimation divided by 12 and squared table increment divided by 24. Image noise variance is inversely proportional to tube current and detector collimation. The maximum continuous scanning time is inversely proportional to tube current. Slice thickness, image noise, and signal-to-noise ratio were, respectively, optimized for a given scanning coverage, consistently resulting in pitch of square root of 2. To avoid longitudinal aliasing, at least 2-3 transverse slices should be reconstructed per collimation. When the simplified spiral CT model is valid and a scanning range specified, 1.4 pitch is required for optimal image quality. The method can be applied to more accurate spiral CT models.

The effect of pitch in multislice spiral/helical CT.

The purpose of this study is to understand the effect of pitch on raw data interpolation in multislice spiral/helical computed tomography (CT) and provide guidelines for scanner design and protocol optimization. Multislice spiral CT is mainly characterized by the three parameters: the number of detector arrays, the detector collimation, and the table increment per x-ray source rotation. The pitch in multislice spiral CT is defined as the ratio of the table increment over the detector collimation in this study. In parallel to the current framework for studying longitudinal image resolution, the central fan-beam rays of direct and opposite directions are considered, assuming a narrow cone-beam angle. Generally speaking, sampling in the Radon domain by the direct and opposite central rays is nonuniform along the longitudinal axis. Using a recently developed methodology for quantifying the sensitivity of signal reconstruction from nonuniformly sampled finite points, the effect of pitch on raw data interpolation is analyzed in multislice spiral CT. Unlike single-slice spiral CT, in which image quality decreases monotonically as the pitch increases, the sensitivity of raw data interpolation in multislice spiral CT increases in an alternating way as the pitch increases, suggesting that image quality does not decrease monotonically in this case. The most favorable pitch can be found from the sensitivity-pitch plot for any given set of multislice spiral CT parameters. An example for four-slice spiral CT is provided. The study on the effect of pitch using the sensitivity analysis approach reveals the fundamental characteristics of raw data interpolation in multislice spiral CT, and gives insights into interaction between pitch and image quality. These results may be valuable for design of multislice spiral CT scanners and imaging protocol optimization in clinical applications.