In craniofacial surgery simulations, the realistic prediction of a patient's postoperative appearance is of utmost importance. Due to the complexity of biomechanical behavior of soft tissue simplified models such as widespread linear elastic models are usually applied. However, large deformations often occurring in craniofacial surgery can't be accurately modeled via linear elastic approach, since it is generally limited by the assumption of small deformations. In this paper, we present an adaptive nonlinear elastic approach based on the finite element discretization of 3D patient's model, which is free of linearization limitations and yields more realistic results in simulations of large facial tissue deformation.

Bone graft surgery is often necessary for reconstruction of craniofacial defects after trauma, tumor, infection or congenital malformation. In this operative technique the removed or missing bone segment is filled with a bone graft. The mainstay of the craniofacial reconstruction rests with the replacement of the defected bone by autogeneous bone grafts. To achieve sufficient incorporation of the autograft into the host bone, precise planning and simulation of the surgical intervention is required. The major problem is to determine as accurately as possible the donor site where the graft should be dissected from and to define the shape of the desired transplant. A computer-aided method for semi-automatic selection of optimal donor sites for autografts in craniofacial reconstructive surgery has been developed. The non-automatic step of graft design and constraint setting is followed by a fully automatic procedure to find the best fitting position. In extension to preceding work, a new optimization approach based on the Levenberg-Marquardt method has been implemented and embedded into our computer-based surgical planning system. This new technique enables, once the pre-processing step has been performed, selection of the optimal donor site in time less than one minute. The method has been applied during surgery planning step in more than 20 cases. The postoperative observations have shown that functional results, such as speech and chewing ability as well as restoration of bony continuity were clearly better compared to conventionally planned operations. Moreover, in most cases the duration of the surgical interventions has been distinctly reduced.

The interpretation of radiological image is routine but it remains a rather difficult task for physicians. It requires complex mental processes, that permit translation from 2D slices into 3D localization and volume determination of visible diseases. An easier and more extensive visualization and exploitation of medical images can be reached through the use of computer-based systems that provide real help from patient admission to post-operative followup. In this way, we have developed a 3D visualization interface linked to a PACS database that allows manipulation and interaction on virtual organs delineated from CT-scan or MRI. This software provides the 3D real-time surface rendering of anatomical structures, an accurate evaluation of volumes and distances and the improvement of radiological image analysis and exam annotation through a negatoscope tool. It also provides a tool for surgical planning allowing the positioning of an interactive laparoscopic instrument and the organ resection. The software system could revolutionize the field of computerized imaging technology. Indeed, it provides a handy and portable tool for pre-operative and intra-operative analysis of anatomy and pathology in various medical fields. This constitutes the first step of the future development of augmented reality and surgical simulation systems.

We have designed and constructed a liver respiratory motion simulator as a first step in demonstrating the feasibility of using a new magnetic tracking system to follow the movement of internal organs. The simulator consists of a dummy torso, a synthetic liver, a linear motion platform, a graphical user interface for image overlay, and a magnetic tracking system along with magnetically tracked instruments. While optical tracking systems are commonly used in commercial image-guided surgery systems for the brain and spine, they are limited to procedures in which a line of sight can be maintained between the tracking system and the instruments which are being tracked. Magnetic tracking systems have been proposed for image-guided surgery applications, but most currently available magnetically tracked sensors are too small to be embedded in the body. The magnetic tracking system employed here, the AURORA from Northern Digital, can use sensors as small as 0.9 mm in diameter by 8 mm in length. This makes it possible to embed these sensors in catheters and thin needles. The catheters can then be wedged in a vein in an internal organ of interest so that tracking the position of the catheter gives a good estimate of the position of the internal organ. Alternatively, a needle with an embedded sensor could be placed near the area of interest.

Image guided neurosurgery systems rely on rigid registration of the brain to preoperative images, not taking into account the displacement of brain tissue during surgery. Co-registered ultrasound appears to be a promising means of detecting tissue shift in the operating room. Although the use of ultrasound images alone may be insufficient for adequately describing intraoperative brain deformation, they could be used in conjunction with a computational model to predict full volume deformation. We rigorously test the assumption that co-registered ultrasound is an accurate source of sparse displacement data. Our co-registered ultrasound system is studied in both clinical applications as well as in a series of porcine experiments. Qualitative analysis of patient data indicates that ultrasound correctly depicts displaced tissue. The porcine studies demonstrate that features from co-registered ultrasound and CT or MR images are properly aligned to within approximately 1.7 mm. Tissue tracking in pigs suggests that the magnitude of displaced tissue may be more accurately predicted than the actual location of features. We conclude that co-registered ultrasound is capable of detecting brain tissue shift, and that incorporating displacement data into computational model appears feasible.

We created a method for three-dimensional registration of medical scanner image volumes to images of physical tissue sections or other volumes, and evaluated its accuracy. The method is applicable for many animal experiments, and we are applying it to evaluate interventional MRI imaging of thermal ablation and to quantify in vivo drug release from a new device for localized, controlled release. The method computes an optimum set of rigid body registration parameters by iterative minimization of the Euclidean distances between automatically generated correspondence points, along manually selected fiducial needle paths, and optional point landmarks. For numerically simulated registrations, using two needle paths over a range of needle orientations, median voxel displacement errors depended only on needle localization error when the angle between needles was at least 15 degrees. For parameters typical of our in vivo experiments, the median error was <EQ0.18 mm. In addition, we determined that the distance objective function was a useful diagnostic for predicting registration quality. To evaluate the registration quality of physical specimens, we computed the misregistration for a needle not considered during the optimization procedure. We registered an ex vivo sheep brain MR volume with another MR volume and tissue section photographs, using various combinations of needle and point landmarks. Registration error was always <EQ0.65 mm for MR-to-MR registrations and <EQ0.9 mm for MR to tissue section registrations. We conclude that our method provides sufficient spatial correspondence to facilitate comparison of 3D image data with data from gross pathology tissue sections and histology.

Microscope-based image-guided neurosurgery can be divided into three steps: calibration of the microscope optics; registration of the pre-operative images to the operating space; and tracking of the patient and microscope over time. Critical to this overall system is the temporal retention of accurate camera calibration. Classic calibration algorithms are routinely employed to find both intrinsic and extrinsic camera parameters. The accuracy of this calibration, however, is quickly compromised due to the complexity of the operating room, the long duration of a surgical procedure, and the inaccuracies in the tracking system. To compensate for the changing conditions, we have developed an adaptive procedure which responds to accruing registration error. The approach utilizes miniature fiducial markers implanted on the bony rim of the craniotomy site, which remain in the field of view of the operating microscope. A simple error function that enforces the registration of the known fiducial markers is used to update the extrinsic camera parameters. The error function is minimized using a gradient descent. This correction procedure reduces RMS registration errors for cortical features on the surface of the brain by an average of 72%, or 1.5 mm. These errors were reduced to less than 0.6 mm after each correction during the entire surgical procedure.

This study proposes a new method for matching vascular imaging modalities without the use of external frame or external landmarks. We first perform a 3D reconstruction of a piece of the cerebral vascular tree using Magnetic Resonance Angiography (MRA). Then, this structure is projected on the Digital Subtracted Angiography (DSA) images until its best position and orientation are found. As the 3D structure is known in the MRA referential, this method enables us to match information from DSA and MRA. The complete matching of all the DSA images in many incidences and the MRA set have been obtained. For the DSA images, the epipolar constraint has been verified between all the incidences. This new approach in medical imaging brings a very original method, making easier and more efficient visualization and quantification of vascular information.

Piecewise removal of myomata (fibroids, benign tumors) from the uterine wall is performed using a hysteroresectoscope inserted through the vagina. Only radical removal of the myoma prevents regrowth, whereas penetration of the uterine wall should be avoided. Preoperative MR (Magnetic Resonance) images show the full outline and the location of the myoma in the uterine wall. Preoperative 3D US (ultrasound) reveals the part of the myoma that protrudes into the uterine cavity, but the part located in the uterine wall is almost invisible. The images and extracted preoperative information must be registered to the preoperative situation. A pragmatic solution in the form of three point incremental matching is applied. Three patients were scanned and the matching was performed on the data. The procedure provided a good match on two patients.

In this paper, we propose to use 2D image projections to automatically segment a needle in a 3D ultrasound image. This approach is motivated by the twin observations that the needle is more conspicuous in a projected image, and its projected area is a minimum when the rays are cast parallel to the needle direction. To avoid the computational burden of an exhaustive 2D search for the needle direction, a faster 1D search procedure is proposed. First, a plane which contains the needle direction is determined by the initial projection direction and the (estimated) direction of the needle in the corresponding projection image. Subsequently, an adaptive 1D search technique is used to adjust the projection direction iteratively until the projected needle area is minimized. In order to remove noise and complex background structure from the projection images, a priori information about the needle position and orientation is used to crop the 3D volume, and the cropped volume is rendered with Gaussian transfer functions. We have evaluated this approach experimentally using agar and turkey breast phantoms. The results show that it can find the 3D needle orientation within 1 degree, in about 1 to 3 seconds on a 500 MHz computer.

In this contribution a postprocessing method is introduced that enables dynamic accuracy examinations of position and angle measurements of two not interfering localizing systems describing the same subspace of Euclidian R³. Furthermore, the method can be used for realization of a hybrid localizing system given a common temporal synchronization of the measurements. Therewith, this article provides a flexible method for examining the influence of the operating room on magnetic tracking by dynamic comparison with reference measurements of an optical localizing system.

With the invention of miniaturized electromagnetic digitizers comes a variety of potential clinical applications for computer aided interventions using flexible instruments; it has become possible to track endoscopes or catheters within the body. To evaluate the reliability of a new commercial tracking system, we measured the systematic distortions induced by various materials such as closed metallic loops, wire guides, catheters and ultrasound scan heads. The system under evaluation was the electromagnetic tracking system Aurora (Mednetix/CH, NDI/Can); data were acquired using the serial port of a PC running SuSE Linux 7.1 (SuSE, Gmbh, Nuernberg). The objects suspected to cause distortions were brought into the digitizer volume. Beside this, we evaluated the influence of a C-arm fluoroscopy unit. To quantify the reliability of the system, the miniaturized sensor was mounted on a nonmetallic measurement rack while the transmitter was fixed at three different distances within the digitizer range. The tracker is more sensitive to distortions caused by materials close to the emitter (average value 13.6 mm +/- 16.6mm) for wire loops positioned at a distance between 100 mm and 200 mm from the emitter). Distortions caused by materials near the sensor (distances smaller than 100 mm) are small (typical error: 2.2 mm +/- 1.9 mm) in comparison to the errors of other electromagnetic systems published in an earlier study of our group where we found an average error of 3.4 mm. Considerable distortions are caused by the C-arm fluoroscopy unit and limits the reliability of the tracker (error: 18.6 mm +/- 24.9 mm). The US scan head was found to cause significant distortions only at a distance between the emitter and the scan head less than 100 mm from the emitter in contrast to the average error of 3.8 mm +/- 6.3 mm at distances greater than 100 mm. Taking into account that significant distortions only occur in the presence of metallic objects close to the emitter, these results indicate the opportunities which are now available in surgical applications where flexible instruments may need to be monitored within the patient.

Many image guided surgical (IGS) procedures use surface representations of anatomical features to provide a registration between physical space and image space. While anatomical fiducial registration techniques can often be compromised by soft tissue deformation, laser-scanning systems offer a non-contact method for accurately resolving geometric surfaces. This digitization approach is particularly well suited for liver resection surgery, where registration is accomplished by aligning the segmented preoperative liver image volume with the intraoperative presentation of the organ. Further, this three-dimensional technology affords a method of rapid surface data acquisition that can be used for the measurement and compensation of tissue deformation. A phantom was constructed to assess the laser scanner's capability of acquiring accurate shape information and robust surface registration. The phantom contained a wealth of point and surface geometry that could be identified in both a CT image volume and a range scan image. Point based registrations were calculated and served as a reference transformation for comparison of the results from surface registration using the iterative closest point (ICP) algorithm. The laser range scanner used the optical principle of triangulation to obtain a dense point set of surface data with a grid spacing of 0.6 mm. The scanner was able to localize and track fiducials on the order of 0.1mm. Phantom experiments demonstrated the ability to perform point-based registrations to CT with a Fiducial Registration Error of 0.5 mm. The scanner accurately resolves geometric surface information and achieved surface-based registrations within a mean distance residual of 0.4 mm. The principal sources of error were systemic misrepresentations of the position by the range scanner, which can be accounted for through calibration or other procedures. The laser scanner system was successfully used to register images to sub-millimetric accuracy. Future work involves acquisition of liver surfaces during surgical procedures for use in registration and deformation measurements.

Image guidance during endovascular intervention is predominantly provided by two-dimensional (2D) digital radiographic systems used for vessel visualization and localization of clips and coils. This paper describes the propagation of 2D noise in the determination of three-dimensional (3D) object position from a single perspective view. In our system, a view is obtained by a digital fluoroscopic x-ray system, corrected for XRII distortions (+/- 0.035mm) and mechanical C-arm shifts (+/- 0.080mm). The tracked object contains high-contrast markers with known relative spacing, allowing for identification and centroid calculation. A least-square projection-Procrustes analysis of the 2D perspective projection is used to determine the 3D position of the object. The effect of uncertainty in 2D marker position on the precision of the 3D object localization using simulations and phantoms was investigated and a nearly linear relationship was found; however, the slope of this relationship is not unity. The slope found indicates a significant amplification of error due to the least-square solution, which is not equally distributed among the 3 major axes. In order to obtain a 3D localization error of less than +/- 1mm, the 2D localization precision must be better than +/- 0.2mm for each marker.

Combining endoscopic video with overlaid reconstructed 3D images allows surgeons to see past opaque surfaces and provides 2D and 3D planning and navigation information. Ensuring surgical quality necessitates accurate calibration and accuracy testing of such systems, as errors from sources such as uncompensated lens distortion and the tracking system cause image misregistration. A method for the calibration of an image-enhanced endoscopy system is presented. Constant-radius linear compensation is proposed as an alternative to computationally expensive nonlinear radial lens-distortion compensation. The results of a system accuracy test using these methods are discussed and the linear lens-distortion compensation method is shown to be an effective compromise between computational speed and accuracy.

The aim of conformal radiation therapy and of radiosurgery (Gamma Knife and Multi-beam radiosurgery) is to irradiate the pathological target volume with ionizing radiation while avoiding as well as possible the surrounding normal tissues. Recently, new micro multileaf collimator ((mu) MLC) devices are available for conformal therapy. A (mu) MLC is formed by narrow sliding leafs in such a manner that the irradiation field can be adjusted to the shape of the target. It is interesting to compare the different techniques to evaluate their effectiveness and their accuracy. This comparison involves 8 clinical cases. For each treatment modality, we compare indexes defined in the international literature by the Radiation Therapy Oncology Group (RTOG). This theoretical study shows (i) the interest of the use of intensity modulation in the case of conformal radiation therapy and (ii) the improvement of RTOG indexes with using the conformal radiotherapy although the volumes of irradiated normal tissue remains lower with the radiosurgery than those with the (mu) MLC. However the comparison between these three techniques for the brain tumors shows that in complex cases it is more effective to use the fractionated conformal therapy with intensity modulation instead of radiosurgery. It is already sure that the micro multileaf collimator holds an important place in conformal therapy.

Patient positioning accuracy remains an issue for external beam radiotherapy. Currently, kilovoltage verification images are used as reference by clinicians to compare the actual patient treatment position with the planned position. These images are qualitatively different from treatment-time megavoltage portal images. This study will investigate the feasibility of using PEREGRINE, a 3D Monte Carlo calculation engine, to create reference images for portal image comparisons. Portal images were acquired using an amorphous- silicon flat-panel EPID for 1) the head and pelvic sections of an anthropomorphic phantom with 7-8 mm displacements applied, and 2) a prostate patient on five treatment days. Planning CT scans were used to generate simulated reference images with PEREGRINE. A correlation algorithm quantified the setup deviations between simulated and portal images. Monte Carlo simulated images exhibit similar qualities to portal images, the phantom slabs appear clearly. Initial positioning differences and applied displacements were detected and quantified. We find that images simulated with Monte Carlo methods can be used as reference images to detect and quantify set-up errors during treatment.

In most planning procedures of intravascular brachytherapy the lesion path is assumed to be a straight line and the dose distribution is essentially specified with respect to the seed. We propose a new approach for the planning of peripheral intravascular brachytherapy based on the characterization of the vessel by virtual exploratory navigation. In order to make the procedure more secure and reliable, the dosimetry planning takes actually account of the geometrical characteristics of the vessel (length, diameter, curvature). Virtual exploratory navigation constitutes the basis of optimal path planning (centering, positioning, trajectory definition) and dosimetry planning for high dose rate brachytherapy. In order to estimate the post-angioplasty vessel centerline and inner surface, a simplified simulation of the transluminal angioplasty procedure is performed using a single pre-angioplasty CT data set. Thanks to the dose rate computation performed in the whole CT volume, the irradiation of the surrounding anatomical structures can be anticipated. A virtual angioscopy like process is used to check the homogeneity of the dose distribution in the vessel wall. Preliminary results obtained from a phantom model and from an animal model of stenosis are reported.

This paper presents the results from applying a computer- based methodology for making precise measurements of longitudinal changes in a patient's digital retinal images presenting with age-related macular degeneration. The digital retinal image analysis system applies recognized principles in automatic image segmentation and integrates the automation with a graphical user interface. Drusen, retinal lesions associated with age-related macular degeneration (ARMD), were segmented using a region-growing algorithm. The algorithm calculates the 76 percentile intensity in a region to provide seed points for the neighborhood-growing algorithm. Twenty-one cases were analyzed. Agreement statistics (kappa) were determined by comparing the automated results with those provided from manually derived measurements. Agreement statistics ranged from 0.49 to 0.71 for different regions of the retina. The manual analysis ground truth was performed by trained graders from the University of Wisconsin Reading Center using guidelines found in the Wisconsin Age-Related Maculopathy Degeneration Grading Scheme (WARMGS). Because of the time required, the ophthalmic graders can only grade (size, area, type) the most prominent drusen in specific regions, resulting in a small sampling of drusen lesions in the retina. The computer-based approach allows one to efficiently and comprehensively grade all of the lesions for larger numbers of images. The additional advantage, however, is in the precision and total area that can be graded with the computer-aided technology. Computer-registered longitudinal images produced a precise determination of the temporal changes in the individual lesions. This study has demonstrated a robust segmentation and registration methodology for automatic and semiautomatic detection and measurement of abnormal regions in longitudinal retinal images.

The Visible Korean Human dataset is currently being made. The performance period for the experiment is March 2000- August 2005. The complete MR and CT images of the Korean cadaver are acquired. The cadaver is serially sectioned (interval: 0.2 mm) and inputted into the personal computer to make anatomical images (pixel size: 0.2 mm) without missing images. Anatomical structures in the anatomical images are segmented. The Visible Korean Human dataset is expected to be more beneficial than the Visible Human Project dataset in many ways. First, the Korean images will be more helpful in diagnosing and treating patients belonging to the Oriental race. Second, the complete MR and CT images will be more helpful in studying MR and CT images. Third, the anatomical images without missing images will be more beneficial by making complete 3D images. Fourth, the anatomical images with thin interval (0.2 mm) and small pixel size (0.2 mm) will show small anatomical structures. Fifth, the additional segmented images will make 3D images and virtual dissection software easily.

Stereotactic neurosurgery for movement disorders involves the accurate localization of functionally distinct subcortical anatomy that appears homogeneous on magnetic resonance or computed tomographic images. To aid localization of these surgical targets on patient images, we have developed a visualization oriented searchable and expandable database of functional organization representing bilaterally the sensorimotor thalamus, pallidum, internal capsule, and subthalamic nucleus. Data were obtained through microelectrode recording and stimulation mapping routinely performed during 123 functional stereotactic procedures. Electrophysiologic data were standardized using a multi-parameter coding system and annotated to their respective MRIs at the appropriate position in patient stereotactic space. To accommodate for normal anatomical variability, we have developed an intensity-based nonlinear registration algorithm that rapidly warps a patient's volumetric MRI to a high-resolution MRI average brain. The annotated functional data are subsequently transformed into the average brain coordinate system using the displacement grids generated by the algorithm. When the database is searched, clustering of like inter-patient physiologic responses within target anatomy and adjacent structures is revealed. These data may in turn be registered to a preoperative MRI using a desktop computer enabling prior to surgery interactive delineation of surgical targets. The database is expandable, fully searchable, and provides a visual 3D representation of subcortical functional organization.

A substantial component of an image-guided surgery system (IGSS) is the kind of three-dimensional (3D) presentation to the surgeon because the visual depth perception of the complex anatomy is of significant relevance for orientation. Therefore, we examined for this contribution four different visualization techniques, which were evaluated by eight surgeons. The IGSS developed by our group supports the intraoperative orientation of the surgeon by presenting a visualization of the spatially tracked surgical instruments with respect to vitally important intrahepatic vessels, the tumor, and preoperatively calculated resection planes. In the preliminary trial presented here, we examined the human ability to perceive an intraoperative virtual scene and to solve given navigation tasks. The focus of the experiments was to measure the ability of eight surgeons to orientate themselves intrahepatically and to transfer the perceived virtual spatial relations to movements in real space. With auto-stereoscopic visualization making use of a prism-based display the navigation can be performed faster and more accurate than with the other visualization techniques.

The success of a cochlear implant is dependent on a functioning auditory nerve. An accurate noninvasive method for screening cochlear implant patients to help determine viability of the auditory nerve would allow physicians to better predict the success of the operation. In this study we measured the size of the auditory nerve relative to the size of the juxtaposed facial nerve and correlated these measurements with audiologic test results. The study involved 15 patients, and three normal volunteers. Noninvasive high-resolution bilateral MRI images were acquired from both 1.5T and 3T scanners. The images were reformatted to obtain an anatomically referenced oblique plane perpendicular to the auditory nerve. The cross- sectional areas of the auditory and facial nerves were determined in this plane. Assessment of the data is encouraging. The ratios of auditory to facial nerve size in the control subjects are close to the expected value of 1.0. Patient data ratios range from 0.73 to 1.3, with numbers significantly less than 1.0 suggesting auditory nerve atrophy. The acoustic nerve area correlated to audiologic test findings, particularly (R²equals0.68) to the count of words understood from a list of 100 words. These preliminary analyses suggest that a threshold of size may be determined to differentiate functional from nonfunctional auditory nerves.

A robust near real-time MRI based surgical guidance scheme has been developed and used in neurosurgical procedure performed in our combined 1.5 Tesla MR operating room. Because of the increased susceptibility difference in the area of surgical site during surgery, the preferred real- time imaging technique is a single shot imaging sequence based on the concept of the half acquisition with turbo spin echoes (HASTE). In order to maintain sufficient spatial resolution for visualizing the surgical devices, such as a biopsy needle and catheter, we used focused field of view (FOV) in the phase-encoding (PE) direction coupled with an out-volume signal suppression (OVS) technique. The key concept of the method is to minimize the total number of the required phase encoding steps and the effective echo time (TE) as well as the longest TE for the high spatial encoding step. The concept has been first demonstrated with a phantom experiment, which showed when the water was doped with Gd- DTPA to match the relaxation rates of the brain tissue there was a significant spatial blurring primarily along the phase encoding direction if the conventional HASTE technique, and the new scheme indeed minimized the spatial blur in the resulting image and improved the needle visualization as anticipated. Using the new scheme in a typical MR-guided neurobiopsy procedure, the brain biopsy needle was easily seen against the tissue background with minimal blurring due the inevitable T2 signal decay even when the PE direction was set parallel to the needle axis. This MR based guidance technique has practically allowed neurosurgeons to visualize the biopsy needle and to monitor its insertion with a better certainty at near real-time pace.

Contrast enhanced breast MRI is currently used as an adjuvant modality to x-ray mammography because of its ability to resolve ambiguities and determine the extent of malignancy. This study described techniques to create and visualize 3D geometric models of abnormal breast tissue. MRIs were performed on a General Electric 1.5 Tesla scanner using dual phased array breast coils. Image processing tasks included: 1) correction of image inhomogeneity caused by the coils, 2) segmentation of normal and abnormal tissue, and 3) modeling and visualization of the segmented tissue. The models were visualized using object-based surface rendering which revealed characteristics critical to differentiating benign from malignant tissue. Surface rendering illustrated the enhancement distribution and enhancement patterns. The modeling process condensed the multi-slice MRI data information and standardized its interpretation. Visualizing the 3D models should improve the radiologist's and/or surgeon's impression of the 3D shape, extent, and accessibility of the malignancy compared to viewing breast MRI data slice by slice.

Advanced Diagnostics, Inc. (ADI) has developed a technique using Diffractive Energy Imaging (DEI) to provide an accurate, non-ionizing method for continuously tracking biopsy needles. The method utilizes ultrasound to generate large field-of-view, high resolution, multiplanar images. Accurate, real-time biopsy guidance using DEI is possible by identifying the physical relationship between a lesion and a focal plane or image slice. The lesion is located by progressively focusing through the breast volume until the lesion is visualized with enough clarity to distinguish it from surrounding tissue. The selected focal plane identifies the lesion location along a medial-lateral axis through the breast. This plane is identified for the physician using a laser point to paint a line on the skin surface. Cartesian coordinates applied to the image slice define a location for surgical incision and needle entry along this focal plane. Maintaining the needle within the defined focal plane and tracking the needle in real-time on a high-resolution monitor ensures a direct line approach through the tissue to the targeted lesion. Data indicate that physicians experienced in biopsy but without prior DEI operational experience are successful in acquiring simulated targets embedded in biopsy phantoms.

Local motion on the beating heart is investigated in the context of minimally invasive robotic surgery. The focus lies on the motion remaining in the mechanically stabilised field of surgery of the heart. Motion is detected by tracking natural landmarks on the heart surface in 2D video images. An appropriate motion model is presented with a discussion of its degrees of freedom and a trajectory analysis of its parameters.

Besides the static soft tissue prediction, the estimation of basic facial emotion expressions is another important criterion for the evaluation of craniofacial surgery planning. For a realistic simulation of facial mimics, an adequate biomechanical model of soft tissue including the mimic musculature is needed. In this work, we present an approach for the modeling of arbitrarily shaped muscles and the estimation of basic individual facial mimics, which is based on the geometrical model derived from the individual tomographic data and the general finite element modeling of soft tissue biomechanics.

Current handheld Personal Digital Assistants (PDAs) can be used to view radiology teaching files. We have developed a toolkit that allows rapid creation of radiology teaching files in handheld formats from existing repositories. Our toolkit incorporated a desktop converter, a web conversion server and an application programming interface (API). Our API was integrated with an existing pilot teaching file database. We evaluated our system by obtaining test DICOM and JPEG images from our PACS system, our pilot database and from personal collections and converting them on a Windows workstation (Microsoft, Redmond, CA) and on other platforms using the web server. Our toolkit anonymized, annotated and categorized images using DICOM header information and data entered by the authors. Image processing was automatically customized for the target handheld device. We used freeware handheld image viewers as well as our custom applications that allowed window/level manipulation and viewing of additional textual information. Our toolkit provides desktop and web access to image conversion tools to produce organized handheld teaching file packages for most handheld devices and our API allows existing teaching file databases to incorporate handheld compatibility. The distribution of radiology teaching files on PDAs can increase the accessibility to radiology teaching.

To meet the computational requirements of mid-range and high-end programmable ultrasound systems, multiple processors are currently required. Algorithms optimized specifically for a single processor-based system may not perform well in a multiprocessor environment. They need to be efficiently remapped on multiple processors to take advantage of the increased computing power while minimizing the interprocessor data transfer and the latency between data acquisition and display. In this paper, we describe a multiprocessor-based implementation of scan conversion, a key processing task in an ultrasound system that geometrically transforms the acquired polar ultrasound data to Cartesian coordinates for display. The single processor-based scan conversion algorithm that was reported previously uses inverse mapping for geometric transformation, where the pixel values in the Cartesian display are determined from data in the polar domain. Inverse mapping requires access to a full frame of pre-scan-converted ultrasound data, which in a multiprocessor system can be located across multiple processors, thus requiring a significant amount of interprocessor data communications. Our modified scan conversion algorithm reduces the data movement by performing inverse-mapped scan conversion locally on the polar-domain data present in each processor's memory. Each processor handles a smaller amount of data, thus reducing the latency. The raster pixels generated by each processor are combined later. Interprocessor synchronization is used to ensure that each processor displays data belonging to the same frame. Data overlapping between processors avoids boundary artifacts between regions that are processed on different processors. Using four Hitachi/Equator Technologies' 300-MHz MAP-CA processors, scan conversion requires 5.6 ms for a 600x420 RGB frame, as compared to 14.6 ms using a single processor, and the latency is reduced by 33.3%. We believe that this type of parallel algorithms will facilitate the development and deployment of flexible multiprocessor-based ultrasound and other medical imaging systems.

The main subjects of the paper are the methodology of characterizing liquid crystal displays (LCDs) and the properties of a three-million-pixel monochrome display system. The system is characterized by display function and dynamic range as a function of viewing angle, spatial luminance uniformity, flicker, peak-to-peak temporal modulation transfer, spatial modulation transfer function (MTF), spatial noise power spectra, and single-pixel signal-to-noise ratios. The evaluated LCD has image quality that, in most respects, is superior to CRT monitors of comparable addressable pixel matrix. In particular, the LCD has perfect spatial modulation transfer. For the evaluated monochrome display system, the general limitation of liquid crystal display (LCD) drivers to 8-bits of grayscale precision is overcome by spatial as well as temporal modulation techniques. The architecture of the control electronics of the system is presented, and as part of it, the implementation of the modulation techniques. The spatial or a combination of spatial and temporal modulation techniques increases the precision with which luminance levels can be defined to 9.58 or 11.58 bits, respectively. The conformance of the calibrated display with the DICOM Standard Display Function is demonstrated without and with application of the modulation techniques. Excellent conformance is achieved for the combination of spatial and temporal modulation.

In this paper, we present results concerning the development of advanced characterization methods and their application to the evaluation of a high-end medical imaging monitor. The display is a 20.8-inch diagonal high-resolution (2048 x 1536 pixels, 123 dpi) monochrome active-matrix liquid-crystal display (AM-LCD) based on a-Si:H TFT active-matrix, dual- domain and in-plane switching technologies. We report the luminance characteristics of the AM-LCD, its grayscale performance, and the panel deviation from the DICOM standard grayscale function. The display luminance for different gray levels under both normal and off-axis viewing directions is described, together with the resulting contrast ratio. The viewing angle dependence of the luminance and contrast ratio is also studied in terms of display compliance to DICOM's grayscale function at off- normal viewing angles. Small-spot contrast ratio performances are investigated in relation to the display technology and especially parasitic phenomena such as electronic crosstalk. We also present the effect of target size and luminance on small-spot contrast ratio results, and we report the orientation dependence of this phenomenon. In addition, we present results on reflections, spatial resolution and display noise and texture and the AM-LCD performances are compared to typical parameters obtained for medical imaging CRTs.

The measurement of image quality in diagnostic monochrome display devices has been the focus of recent efforts. However, there has not been a complete understanding of the capabilities of such methods to appropriately characterize display resolution and noise. In addition to being influenced by the luminance level of the scene being displayed, both properties are affected by the intrinsic details of the pixel emissive structures. In this paper, we report on the resolution and noise properties of cathode-ray tubes (CRTs) and active-matrix liquid crystal displays (AMLCDs). We present results on the resolution and noise characteristics of 5-megapixel monochrome CRTs and a 3-megapixel monochrome AMLCD with dual-domain in-plane switching pixel design. The CCD camera was calibrated using a uniform light source constructed specifically for excellent spatial uniformity. The methods employed for resolution are based on either the measurement of line patterns or the broadband transfer method. Noise is analyzed in terms of 2-dimensional noise power and signal-to-noise ratios. We found that all 4 methods considered are affected by residual deterministic structure in the recorded images. Especially in AMLCDs, the MTFs computed by the line method and broadband response method do not agree, due to contributions of the non-stochastic component that remain in the analyzed images after subtraction of a representative background.

We have developed three hydrogenated amorphous silicon (a-Si:H) thin-film transistors (TFTs)-based pixel electrode circuits for active-matrix organic polymer light-emitting displays (AM-OP-LEDs). Our 3- a-Si:H TFTs voltage driven pixel electrode circuits can supply a continuous output current for AM-OP-LEDs. Each pixel circuit has compensation circuits that can adjust for the OP-LED and a-Si:H TFTs electrical characteristics shift. This pixel electrode circuit fabricated in our laboratory had the driving output current level of 1.13 (mu) A that is equivalent to a pixel current density of over 20 mA/cm². The display brightness of ~4500 cd/m² for green light (540 nm) can be achieved with this type of pixel electrode circuits for the organic polymer light-emitting devices having an external quantum efficiency of 5%.

Cardiac MRI is a technique that provides information about morphology and function of the cardiovascular system in the form of four-dimensional (4D) scalar data sets. Visualization and extraction of clinically relevant parameters from these data sets may help to diagnose cardiac diseases and malfunctions. Some of these parameters are left (right) ventricle volume, ejection fraction, flow measurements, and wall motion and thickening. Although cardiac MRI is a rapidly growing technique, it must overcome several problems (such as poor spatial resolution, flow and motion artifacts, and low signal-to-noise ratio) in order to produce images with sufficient quality to be used in clinical applications. Existing approaches to visualize cardiac MRI data sets in 4D are based on rendering a geometrical model extracted from the data. In most cases, these models are polygon meshes describing the epicardial and endocardial surfaces of the heart. A wide range of different techniques can be found in the literature to achieve this geometrical model extraction. Our approach consists of applying an iso-surface volume rendering technique in order to visualize the data sets. This visualization includes shape visualization and functional mapping. With this technique, the medical data itself is rendered instead of rendering an extracted geometrical model. This technique has been successfully applied to 3D MRI and CT data sets. Even though the extension of this technique to 4D data sets is not straightforward, the preliminary results are very promising.

Recently, 3D virtual endoscopy has been used as an alternative noninvasive procedure for visualization of a hollow organ. In this paper, we propose a fast volume rendering scheme based on perspective ray casting for virtual endoscopy. As a pre-processing step, the algorithm divides a volume into hierarchical blocks and classifies them into opaque or transparent blocks. Then, the rendering procedure is as follows. In the first step, we perform ray casting only for sub-sampled pixels on the image plane, and determine their pixel values and depth information. In the second step, by reducing the sub-sampling factor by half, we repeat ray casting for newly added pixels, and their pixel values and depth information are determined. Here, the previously obtained depth information is utilized to reduce the processing time. This step is performed recursively until the full-size rendering image is acquired. Experiments conducted on a PC shows that the proposed algorithm can reduce the rendering time by 70-80% for the bronchus and colon endoscopy, compared with the brute-force ray casting scheme. Thereby, interactive rendering becomes more realizable in a PC environment without any specific hardware.

3D Digital mammography (3DDM) is a new technology that provides high resolution X-ray breast tomographic data. Like any other tomographic medical imaging modalities, viewing a stack of tomographic images may require time especially if the images are of large matrix size. In addition, it may cause difficulty to conceptually construct 3D breast structures. Therefore, there is a need to readily visualize the data in 3D. However, one of the issues that hinder the usage of volume rendering (VR) is finding an automatic way to generate transfer functions that efficiently map the important diagnostic information in the data. We have developed a method that randomly samples the volume. Based on the mean and the standard deviation of these samples, the technique determines the lower limit and upper limit of a piecewise linear ramp transfer function. We have volume rendered several 3DDM data using this technique and compared visually the outcome with the result from a conventional automatic technique. The transfer function generated through the proposed technique provided superior VR images over the conventional technique. Furthermore, the improvement in the reproducibility of the transfer function correlated with the number of samples taken from the volume at the expense of the processing time.

Direct volume rendering (DVR) is a very useful visualisation technique. However, the difficulty in specifying a suitable transfer function often discourages its application to visualisation problems. This work presents a technique for providing meaningful and fast visual feedback that greatly facilitates the transfer function specification process. The feedback is based on a slice-based preview of the actual volume rendering that can be calculated in real-time and can be superimposed on a slice-based view of the data that is being rendered.

Light field algorithm is one of the most famous image-based rendering techniques. In this paper, an improved light field algorithm - sphere light field algorithm - is proposed. This new algorithm replaces the parallel planes used in traditional light field algorithms by a triangularly parameterized sphere surface. Comparing to the traditional light field algorithms, this new algorithm achieves a more uniform distribution of light slabs in the three dimensional space. This improves the quality of the rendered images. Using the triangular parameterization and subdivision of the whole light field, less calculation and less memory are needed, resulting in improved real-time rendering.

In Medical Imaging, shell rendering and shear-warp rendering are two of the most efficient and effective voxel-based techniques for volume visualization. This work presents a comparative analysis of shell rendering and shear-warp rendering in terms of storage, speed, and image quality. We have chosen 10 different objects of various sizes, shapes and topologies and one 1-GHz Pentium-III PC with 512 MB RAM for our experiments. Hard and fuzzy boundaries of up to 2,833 K voxels in size have been created to test both methods in surface and volume rendering, respectively. Hard surface shell rendering and surface shear-warp rendering required less than 0.5 second. In the worst case, volume shell rendering required 1.45 second, while volume shear-warp rendering spent 0.65 second for the same task. Shear-warp rendering uses on average from 3 to 6 times more memory space than shell rendering, but it can be up to 2.79 times faster than shell rendering. On average, shear-warp rendering is as fast as shell rendering for hard boundaries and 1.7 times faster than shell rendering for fuzzy boundaries. We have also observed that both can produce similar high-quality images.

Repair of a defect heart valve is of great advantage for the patient in comparison to replacement with a prosthesis. The applicability and the success of heart valve repair can be improved by an exact diagnosis of the valve's pathological modification. The best way for imaging heart valve insufficiencies is echocardiography, since it is fast, relatively cheap, can be used intraoperatively and provides information about morphology as well as blood flow. Three-dimensional echocardiography has been proven to be superior to conventional echocardiographic techniques. Although the overall structures are much better displayed by three-dimensional visualization methods, it is sometimes difficult to comprehend the orientation of the scene, since anatomical landmarks like the aortic outflow tract may be hidden by other structures. Also, such anatomical landmarks often are only partly contained in the acquired data set so that they are clearly visible in a few slices only, making them difficult to find in a three-dimensional visualization. The knowledge of the absolute orientation is of essential value for the surgeon to mentally transfer the preoperatively acquired data to the intraoperative situs. Therefore, it is desirable to have additional hints for orientation in the three-dimensional scene. We present methods that enable better and easier orientation and therefore improve the usability of three-dimensional echocardiography.

In this paper we present a method for interactive analysis of non-segmented medical volume data. We discuss both different rendering methods for visualization and different possibilities for interaction in relation to segmentation results. Furthermore, the adaptive region growing approach is applied to both segmentation of a structure of interest, as well as generation of transfer function for volume rendering of the same structure. The adaptive region growing method is based on the statistical evaluation of 3D-neighbourhood. The method is used for determination of a homogeneity criterion for the structure of interest. Subsequently this criterion is used for segmenting of data and for generating of an initial transfer function for volume rendering. We utilize this for displaying a hybrid 3D-visualization of the segmented structure and the specific gray-value interval of original data. Based on this rendering we discuss possibilities for user-guided validation of segmentation results, based on the variation of several rendering parameters.

This paper describes two new presentation methods that are intended to improve the ability of radiologists to visualize abnormalities in mammograms by enhancing the appearance of the breast parenchyma pattern relative to the fatty-tissue surroundings. The first method, referred to as mountain- view, is obtained via multiscale edge decomposition through filter banks. The image is displayed in a multiscale edge domain that causes the image to have a topographic-like appearance. The second method displays the image in the intensity domain and is referred to as contrast-enhancement presentation. The input image is first passed through a decomposition filter bank to produce a filtered output (I_d). The image at the lowest resolution is processed using a LUT (look-up table) to produce a tone scaled image (I'). The LUT is designed to optimally map the code value range corresponding to the parenchyma pattern in the mammographic image into the dynamic range of the output medium. The algorithm uses a contrast weight control mechanism to produce the desired weight factors to enhance the edge information corresponding to the parenchyma pattern. The output image is formed using a reconstruction filter bank through I' and enhanced I_d.

The gold standard for NASCET-type stenosis measurements is currently 2D digital subtraction angiography (DSA). In this paper, we evaluate the efficacy of 3D cone-beam, Volumetric Subtraction Angiography (VSA) for assessing internal carotid artery stenosis, by comparison with conventional DSA. VSA perspective maximum intensity projections (MIPs) and DSAs were assessed separately for NASCET-type, minimum stenosis measurements. Although virtually any viewing angle of the VSA was possible, the minimum stenosis grades were not significantly higher than that of the DSAs. Our study of 38 arteries yielded a sensitivity and specificity of 100% (using a clinically relevant 60% stenosis threshold). Measurements from three neuroradiologists provided an average stenosis grade of 75 +/- 6% and 76 +/- 7% for the DSA and VSA respectively. A paired student t-test indicated a 98% confidence of no statistical difference in the means. Thus, VSA provides gold standard 3D information about carotid lumen geometry. While not intended to supplant noninvasive techniques during routine clinical diagnosis, it does provide a 3D reference standard for research investigations. Additionally, cone-beam CT can provide quantification of calcification around the carotid bifurcation.

Cone-beam reconstruction (CBR) is growing in importance, but current computer systems are slower than desirable for clinical use. We have built a high-speed system for high-quality, 3D imaging. We partitioned the problem into input, filtering, backprojection, postprocessing, and output components. We mapped most of the components to standard RACE++ processing nodes. The backprojection component is very compute-intensive; we mapped it to a field-programmable gate array (FPGA)-based adjunct processor. We built a prototype FPGA card, optimized for flexibility, and implemented the backprojection in that FPGA. This strategy allows for redesigning the backprojection function when necessary, and keeps the other details of the CBR algorithm in easily programmable processors. We present a system that performs Feldkamp CBR of 300 projections into a 512³ cubical image in 38.7 seconds. The system is designed to be scalable, so that Feldkamp CBR of 21.4 seconds can be performed with two adjunct processors, and Feldkamp CBR of other regions of interest or dimensions could be performed in proportionately shorter times. Further optimization and faster-processing parts will also contribute to continual speed improvements. This system is flexible and can be extended to perform other imaging functions, such as real-time planar angiography, with the same hardware.

Based on the Varioscope, a commercially available head mounted operating binocular, we have developed a head mounted display for augmented reality visualization that seamlessly fits into the infrastructure of a surgical navigation system. This head mounted display, called the Varioscope AR, is equipped with two miniature computer monitors that merge computer graphics with the view of the operating field as seen by the surgeon. Since the position of the Varioscope AR is being tracked by the navigation system's optical tracker, planning data such as the location of a lesion identified on preoperative volume images can be displayed in the correct position transparently overlaying the optical field of view. In order to assess the system's accuracy and the depth perception of a user aiming at a given target, we have designed a phantom for skull base surgery. 16 steel spheres were fixed at the base of a bony skull, and several typical craniotomies were applied. After having taken CT scans, the skull was filled with opaque gelatine in order to simulate brain tissue. The positions of the spheres were registered using VISIT, a system for computer aided surgical navigation. Then attempts were made to locate the steel spheres with a bayonet probe through the craniotomies using VISIT and the Varioscope AR as a stereoscopic display device. Localization using stereoscopic vision with this novel device had a success rate (defined as a first trial hit rate) of 81,5%. Using monoscopic vision, the success rate was found to be 50%. We conclude that the Varioscope AR is now mature for further cadaver tests and clinical studies.

Augmented reality systems (ARS) allow the transparent projection of preoperative CT images onto the physicians view. A significant problem in this context is the registration between the patient and the tomographic images, especially in the case of soft tissue deformation. The basis of our ARS is a volume rendering component on standard PC platform, which allows interactive volumetric deformation as a supplement to the 3D-texture based approaches. The volume is adaptively subdivided into a hierarchy of sub-cubes, each of which is deformed linearly. In order to approximate the Phong illumination model, our system allows pre-calculated gradients to be deformed efficiently. The registration is realized by the introduction of a two-stage procedure. Firstly, we compute a rigid pre-registration by the use of fiducial markers in combination with an electromagnetic navigation system. The second step accounts for the nonlinear deformation. For this purpose, several views of an object are captured and compared with its corresponding synthetic renderings in an optimization method using mutual information as metric. Throughout the experiments with our approach, several tests of the rigid registration has been carried out in a real laparoscopic intervention setup as a supplement to the actual clinical routine. In order to evaluate the nonlinear part of the registration, up until now several dummy objects (synthetically deformed datasets) have been successfully examined.

We developed an augmented reality system targeting image guidance for surgical procedures. The surgeon wears a video- see-through head mounted display that provides him with a stereo video view of the patient. The live video images are augmented with graphical representations of anatomical structures that are segmented from medical image data. The surgeon can see, e.g., a tumor in its actual location inside the patient. This in-situ visualization, where the computer maps the image information onto the patient, promises the most direct, intuitive guidance for surgical procedures. In this paper, we describe technical details of the system and its installation in UCLA's iMRI operating room. We added instrument tracking to the capabilities of our system to prepare it for minimally invasive procedures. We discuss several pre-clinical phantom experiments that support the potential clinical usefulness of augmented reality guidance.

Cataplexy is a sudden loss of voluntary muscle control experienced by narcolepsy patients. It is usually triggered by strong, spontaneous emotions and is more common in times of stress. The Sleep Disorders Unit and the Biomedical Imaging Resource at Mayo Clinic are developing interactive display technology for reliably inducing cataplexy during clinical monitoring. The project is referred to as the Cataplexy/Narcolepsy Activation Program, or CatNAP. We have developed an automobile driving simulation that introduces humorous, surprising, and stress-inducing events and objects as the patient attempts to navigate a vehicle through a virtual town. The patient wears a head-mounted display and controls the vehicle via a driving simulator steering wheel and pedal cluster. As the patient attempts to drive through the town, various objects, sounds or conditions occur that distract, startle, frustrate or amuse. These responses may trigger a cataplectic episode, which can then be clinically evaluated. We believe CatNAP is a novel and innovative example of the effective application of virtual reality technology to study an important clinical problem that has resisted previous approaches. An evaluation phase with volunteer patients previously diagnosed with cataplexy has been completed. The prototype system is being prepared for a full clinical study.

In this paper we present a novel approach for the calibration of video cameras in an augmented reality image-guided surgery system. Whereas most calibration algorithms rely on the extraction of features such as points or lines, our proposed calibration algorithm determines the intrinsic and extrinsic camera calibration parameters by maximising the similarity between the real view of a calibration object and the synthetic view of the same calibration object. Our new method offers a number of advantages over existing calibration techniques: First, our calibration algorithm does not require the identification of fiducials such as points or lines in the video images. As a result of this the algorithm does require any feature extraction such as a corner or edge detection. Instead the algorithm uses the image intensities directly. Second, the calibration algorithm is model-based, which means that different camera or lens distortion models can be easily integrated into the algorithm. We have applied the calibration algorithm for the calibration of a head-mounted augmented reality system for image-guided neurosurgery. Our results show that the proposed calibration algorithm can lead to improved accuracy compared to conventional feature-based calibration techniques.

In the field of image-guided surgery, registration of pre- and intra-operative images is an important, yet sometimes cumbersome procedure. Visual matching, consisting of a manual alignment of such images can improve the efficiency of registration. In this paper we investigate the accuracy of a visual matching system. A number of tests, including mono and stereo visualization of real and virtual objects, was set up. Seven test subjects were asked to register pre- and intra-operative images while accuracy and registration time were measured. The error analysis led to the conclusion that one can accurately match a virtual phantom with a real one in directions perpendicular to the viewing direction. The depth accuracy though is insufficient for registration purposes, even when using stereo perception. We conclude that visual matching can be complementary to more automatic registration methods or even a valuable alternative in applications where the viewing directions are nearly perpendicular as in bi-plane fluoroscopy.

We have investigated a registration method between pre-operative 3D CT angiography (3D-CTA) and intra-operative fluoroscopic image sequence (with/without contrast injection) during intervention. Most registration methods are developed for assisting neurosurgery or orthopedic surgery, but our method is developed for interventional procedure such as endovascular stent grafting. In our method, DRR (Digitally Reconstructed Radiograph) are generated by voxel projection of 3D-CTA after extracting an aorta region. By increasing/decreasing CT value in the aorta region of CTA, DRR with/without contrast media injection are obtained. Subsequently we calculate distance (or similarity) measures between DRR and fluoroscopic image iteratively by changing imaging parameters. The most similar DRR to fluoroscopy is selected. We validated our algorithm by using simulated/clinical fluoroscopic images and DRR (with/without contrast media injection) of thorax and abdomen. Several distance (or similarity) measures were investigated in this experiment. Validation results show that M-estimator of residual is good as matching measure, and registration is well performed for almost all cases. However, accuracy is not enough for non-contrasted thoracic images, and calculation time should be reduced for all cases.

In cranial neurosurgery, a common approach to matching image space to physical space in the operating room is to attach markers to the skin of the patient. By localizing these markers in image space, and subsequently in physical space before the surgical procedure, a transformation between the two spaces may be determined. However, the accuracy of this transformation is to a large extent determined by the accuracy with which the markers may be identified in each space. In this paper, we investigate the effect of zoom and slice direction on the localization error of the markers in image space. We use a dataset of three-dimensional head images taken of patients wearing skin markers. We ask a set of qualified observers to localize the centroids of the markers as accurately as possible in each image, both with and without an enhancement designed to facilitate the localization process. This enhancement is an image reformatting step to ensure that the 2D image slices lie in planes aligned with the principal axes of the marker being localized.

Despite the growing popularity of frameless image-guided surgery systems, stereotactic head frame systems are widely accepted by neurosurgeons and are still commonly used to perform stereotactic biopsy, functional procedures, and stereotactic radiosurgery. In this study, we investigate the accuracy of the Cosman-Roberts-Wells (CRW) stereotactic frame system when the mechanical load on the frame changes between pre-operative imaging and the intervention due to different patient position - supine during imaging, prone during intervention. We analyze CT images acquired from 12 patients who underwent stereotactic biopsy or stereotactic radiosurgery. Two CT images were acquired for each patient, one with the patient in the supine position and one in the prone position. The prone images were registered to the respective supine images using an intensity-based registration algorithm, once using only the frame and once using only the head. The difference between the transformations produced by these two registrations describes the movement of the patient's head with respect to the frame due to mechanical distortion of the latter. The maximum frame-based registration error between supine and prone positions was 2.8 mm, greater than 2 mm in two patients, and greater than 1.5 mm in five patients. Anterior-posterior translation is the dominant component of the difference transformation for most of these patients. In general, the magnitude of the movement increased with brain volume, which is an index of head weight. We conclude that in order to minimize frame-based registration error due to a change in the mechanical load on the frame, frame-based stereotactic procedures should be performed with the patient in the identical position during imaging and intervention.

We are developing a neurosurgical planning system, and this paper presents a method of automatic path searching from skin surface of the head to tumor inside the brain. The path obtained by our system is an optimal one in minimally invasive neurosurgery and it is calculated automatically based on a 3D atlas. The 3D atlas is a voxel based 3D matrix, and consists of two parts, structural part and knowledge part. The structural part includes the geometrical information with respect to each tissue. The knowledge part is generated based on experiences of doctors, and numeric value is provided as importance value for each tissue. The path searching consists of two steps, path finding and path smoothing. Path finding is to select an optimal path from surface of the head to the affected part based on minimization of importance value integrating along with the path. We applied the algorithm into a head MRI data set. The path obtained by path finding is reformed smoothly to approach the natural path of surgeon.

In robotics, the term workspace denotes that volume of space which the end-effector of a robot can reach. Similarly we denote the workspace of a flexible endoscope as the volume of space that can be occupied by the entire endoscope. From the endoscope's perspective, the workspace is determined by its internal (mechanical) constraints, as well by its external constraints, like organ geometry and gravity. In this paper we show how to calculate the workspace of a given endoscope under insertion into a patient specific organ model. We propose a model of an endoscope that takes the internal constraints into account. We also propose an algorithm which simulates the insertion of an endoscope into an organ. This algorithm calculates the workspace by recursively building a spatial tree whose growth is constrained by a series of filter functions. As Kukuk et al. showed, knowing a real endoscope's workspace can be used to significantly increase the success rate of a blind biopsy. We have built an experimental setup (lung phantom) that allows us to assess the accuracy of our model by comparing it to a real endoscope.

A new three-dimensional optical digitizer working with white light and high accuracy has been developed for applications in orthodontic diagnosis, to store and analyze data taken from plaster dental casts. The scanner consists basically of a triangulation device made of a CCD camera and an active light projector using Gray code and the phase stepping method to sample the surfaces under tests. Dense clouds of 3D points of the object's surface, related to the device's reference system, are generated, from which very exact measurements can be taken. This technique allows us to greatly improve the accuracy of results that previously were obtained by using mainly manual methods. Our system has been tested successfully on many different samples.

The objective of this study is firstly to define system configuration for an interactive telesurgery with open MRI, and secondly to elucidate optimum bandwidths of B-ISDN in the standpoint view of functionality to cost ratio. The test bed of B-ISDN network was established between Osaka University Hospital (OUH) and Kawasaki Hospital (KH). The interactive telesurgery has divided into three phases. Phase 1 is to transfer DICOM data of a patient from KH to OUH where image files are archived for data processing. Phase 2 is to make interactive videoconferencing with simultaneous sharing of the DICOM images. Phase 3 is to assist surgical navigation from OUH to KW with open MRI images and three dimensional images overlay. In order to satisfy these functions of the three phases and optimize running cost and functionality, the survey were made and the results were summarized as functional requirements. By using scalable bandwidths control, the optimum bandwidths in each phase were decided. The data obtained indicate functional requirements for telesurgical operations and will provide the basis for future challenging tele-robotic surgery which is under developing in our project between OUH and KH.

Brachytherapy of prostate cancer involves permanent implantation of radioactive seeds in the prostate gland under transrectal ultrasound (TRUS) and fluoroscopic guidance with the goal to treat the entire gland to a prescribed tumoricidal dose. The seed coordinates are calculated prior to the procedure. However, due to uncertainties associated with the seed insertion process and changes in the anatomy, it is difficult to faithfully reproduce the planned dose distribution. The procedure can be improved substantially by identifying the underdosed regions via real-time dosimetry before the completion of implantation. The physician can then insert additional seeds to achieve clinically needed dose distribution. One possible method of performing such dosimetry is to fuse TRUS (which images the prostate gland very well but not the metallic seeds) and fluoroscopy (which images metallic seeds well but not soft tissue) data. We have developed a technique to register the 3D seed positions reconstructed from multiple fluoroscopic images with the prostate TRUS images. The technique utilizes a set of fiducials that are detectable by both imaging modalities. The method was tested using a prostate mimicking phantom. The results demonstrate that the method is sufficiently accurate and practical for clinical use.

Unsharp masking is a widely used image enhancement method in medical imaging, e.g., in computed radiography, digital radiography, and digital mammography. It mainly consists of 3 steps: (1) convolving an input image with a lowpass filter, (2) obtaining a highpass-filtered image by subtracting the lowpass-filtered image from the original image, and (3) adding the weighted highpass-filtered image to the original image. It is computationally expensive, e.g., convolving a 2k x 2k image with a 21 x 21 kernel alone requires about 3.7 billion arithmetic operations. To support this high computational demand for unsharp masking, hardware-based solutions using ASIC, FPGA and FPLD have been developed and used. While they have reasonably met the computing requirement, they suffer from limited flexibility. On the other hand, software solutions using programmable processors are more flexible and can easily change algorithmic parameters, such as filter kernel size, and incorporate new features, but they have not been able to meet the fast computing requirement. Modern programmable mediaprocessors, such as MAP-CA and Texas Instruments TMS320C64x, can meet both fast computing and flexibility requirements due to their high computing power and full programmability. In this paper, we present an efficient implementation of adaptive unsharp masking on a MAP-CA mediaprocessor. For a 2k x 2k 16-bit image, our adaptive unsharp masking operation with a 149 x 149 boxcar kernel takes only 300 ms. This fast unsharp masking not only reduces the overall processing time in imaging modalities, but also allows the operator to adjust the selected parameters interactively for optimal image quality. Our implementation on the MAP-CA can be easily extended to other high-performance mediaprocessors, such as TMS320C64x and Pentium 4.

Adjunct processors have traditionally been used for certain tasks in medical imaging systems. Often based on application-specific integrated circuits (ASICs), these processors formed X-ray image-processing pipelines or constituted the backprojectors in computed tomography (CT) systems. We examine appropriate functions to perform with adjunct processing and draw some conclusions about system design trade-offs. These trade-offs have traditionally focused on the required performance and flexibility of individual system components, with increasing emphasis on time-to-market impact. Typically, front-end processing close to the sensor has the most intensive processing requirements. However, the performance capabilities of each level are dynamic and the system architect must keep abreast of the current capabilities of all options to remain competitive. Designers are searching for the most efficient implementation of their particular system requirements. We cite algorithm characteristics that point to effective solutions by adjunct processors. We have developed a field- programmable gate array (FPGA) adjunct-processor solution for a Cone-Beam Reconstruction (CBR) algorithm that offers significant performance improvements over a general-purpose processor implementation. The same hardware could efficiently perform other image processing functions such as two-dimensional (2D) convolution. The potential performance, price, operating power, and flexibility advantages of an FPGA adjunct processor over an ASIC, DSP or general-purpose processing solutions are compelling.

Telepathology is a means of practicing pathology at a distance, viewing images on a computer display rather than directly through a microscope. Without compression, images take too long to transmit to a remote location and are very expensive to store for future examination. However, to date the use of compressed images in pathology remains controversial. This is because commercial image compression algorithms such as JPEG achieve data compression without knowledge of the diagnostic content. Often images are lossily compressed at the expense of corrupting informative content. None of the currently available lossy compression techniques are concerned with what information has been preserved and what data has been discarded. Their sole objective is to compress and transmit the images as fast as possible. By contrast, this paper presents a novel image compression technique, which exploits knowledge of the slide diagnostic content. This 'content based' approach combines visually lossless and lossy compression techniques, judiciously applying each in the appropriate context across an image so as to maintain 'diagnostic' information while still maximising the possible compression. Standard compression algorithms, e.g. wavelets, can still be used, but their use in a context sensitive manner can offer high compression ratios and preservation of diagnostically important information. When compared with lossless compression the novel content-based approach can potentially provide the same degree of information with a smaller amount of data. When compared with lossy compression it can provide more information for a given amount of compression. The precise gain in the compression performance depends on the application (e.g. database archive or second opinion consultation) and the diagnostic content of the images.

A complete Web based hospital information system, which can allow medical doctors to access and modify patient information and records anywhere in the world via the Internet, was developed. More specifically, this Web information system can be linked seamlessly to our fully computerized MR image-guided neurosurgery suite. This information system, which utilizes the unprecedented Internet infrastructure and adopts the most updated software technologies, addresses the urgent need for handling today's hospital information flow and management. With this new information system in our surgery suite, images and records that have been transferred directly from a diagnostic system such as MR, CT, etc. to the DICOM archive are accessible via a secured Internet connection. When data is accessed via the Web, it can be retrieved in several formats, including raw DICOM and binary, which are extremely useful for various research and development purposes, as well as new applications that require access to the original image data. The Internet-based Web Hospital Information System (WHIS) can easily match the existing standards for this type of information system in a hospital and can accommodate any anticipated requirements for image-guided minimally invasive surgery in the future. A practical and potentially low cost Web Hospital information system, which is functionality- driven, will be presented in this paper. It provides an extremely intuitive interactive environment, as well as a very user-friendly interface for use by both medical doctors and patients.

New high performance programmable processors, called mediaprocessors, have been emerging since the early 1990s for various digital media applications, such as digital TV, set-top boxes, desktop video conferencing, and digital camcorders. Modern mediaprocessors, e.g., TI's TMS320C64x and Hitachi/Equator Technologies MAP-CA, can offer high performance utilizing both instruction-level and data-level parallelism. During this decade, with continued performance improvement and cost reduction, we believe that the mediaprocessors will become a preferred choice in designing imaging and video systems due to their flexibility in incorporating new algorithms and applications via programming and faster-time-to-market. In this paper, we will evaluate the suitability of these mediaprocessors in medical imaging. We will review the core routines of several medical imaging modalities, such as ultrasound and DR, and present how these routines can be mapped to mediaprocessors and their resultant performance. We will analyze the architecture of several leading mediaprocessors. By carefully mapping key imaging routines, such as 2D convolution, unsharp masking, and 2D FFT, to the mediaprocessor, we have been able to achieve comparable (if not better) performance to that of traditional hardwired approaches. Thus, we believe that future medical imaging systems will benefit greatly from these advanced mediaprocessors, offering significantly increased flexibility and adaptability, reducing the time-to-market, and improving the cost/performance ratio compared to the existing systems while meeting the high computing requirements.

MR diagnostics with obtaining images cleaned of strong signals of water and fat for discovery small tissue contrast variations related with a pathology is discussed. With help of water suppression, one can reveal the pathological changes of brain in small zones near the liquor spaces, particularly disseminated sclerosis cortical plaques, cerebral lesions, subarachnoid hemorrhage. Under the fat suppression T₂-weighted images, inflammatory changes in orbits, cranial base, and spine are surely determined. Among several methods of water or fat suppression, one can indicate methods based on frequency selective saturation of fat signals and methods based on inversion-recovery (IR) pulse sequence. Simultaneous suppression of fat and water opens up additional possibilities in medical diagnostics. We realized simultaneous suppression of fat and water signals at the MR scanner Tomikon S50 (Bruker). We used a pulse sequence based on the IR technique, supplemented with special procedure of frequency-selective fat saturation. Sometimes only simultaneous suppression of fat and water signal made possible successful MRI diagnostics.

A processing method is described which allows to present images with equalized detail contrast, i.e., contrast that is the same in all parts of the image, independent of the chosen look-up table (LUT) or the local signal level. Basically, a multiresolution algorithm is used which splits the image in a number of bandpass images. Only the lowest band (low-pass image) is transmitted through the intensity LUT, while all higher frequency subimages are nonlinearly enhanced and added to the LUT-transformed low-pass image. Nonlinear enhancement is used in order to improve the visibility of weakly contrasting details while minimizing artifacts at high contrast edges. Unfavorable noise enhancement can be avoided by limiting the enhancement in the low-dose areas of the highest frequency subimages. The resulting images show good detail visibility in all parts. Detail rendition in images with equalized contrast is independent of image latitude and of slight variations in overall image brightness or density. Preliminary experience with clinical images show that the display does not need individual parameter tuning for different images, yet allows producing artifact-free, naturally looking images with improved detail visibility for a wide variety of input images. Improvements compared to standard (intensity) equalization processing are evident especially for the mediastinum and subdiaphragmal regions of chest images and in lateral spine examinations that have an inherently large dynamic range.

One of the advantages of digital mammography is to display mammograms on softcopy (electronic displays). Softcopy display of mammography is challenging because of the spatial and contrast resolution demands present in mammograms. We have designed and developed a softcopy mammography display application, Mammoview, which is capable of allowing radiologists to read mammograms as quickly and as accurately as they can on film alternators. We review the studies using Mammoview to elucidate the requirements of a successful softcopy display state. The design and development of the Mammoview softcopy display station are described. Results of several studies using Mammoview are reviewed, including subjective feedback from RSNA demonstrations, and clinical studies measuring performance in terms of speed and accuracy. Additional analysis of user interactions and user feedback is used to study the successes and shortcomings of mammography display stations like Mammoview.

Since the introduction of digital flat panel detectors into clinical routine the discussion on monitor specifications for primary soft copy reading has gained new impetus. Major concerns exist for viewing of tiny opacities such as pulmonary nodules. In this study CDRAD phantom images were acquired on a caesium iodid/amorphous silicon detector at varying exposure levels. Images were read three times by three observers on a clinical 1K and 2K monitor workstation. All typical workstation functions such as magnification and window/level setting were applied during image reading. Correct detection ratios were calculated according to the CDRAD evaluation manual. Observer ratings were highest for high dose exposure and 2K monitor reading. No significant difference was detected in the correct detection ratio of observers. However, the difference between the two types of workstations (1K versus 2K monitors) despite less than 3% was significant at a 95% confidence level. This is in good accordance with recently published clinical studies. However, further clinical work will be needed to strengthen this laboratory based impression. Given these subtle differences in low contrast detail detection on 1K and 2K clinical PACS workstation we should probably rethink the recommendations of various national boards for the use of 2K monitors.

Our project was to design and implement a ceiling-mounted multi monitor display unit for use in a high-field MRI surgical suite. The system is designed to simultaneously display images/data from four different digital and/or analog sources with: minimal interference from the adjacent high magnetic field, minimal signal-to-noise/artifact contribution to the MRI images and compliance with codes and regulations for the sterile neuro-surgical environment. Provisions were also made to accommodate the importing and exporting of video information via PACS and remote processing/display for clinical and education uses. Commercial fiber optic receivers/transmitters were implemented along with supporting video processing and distribution equipment to solve the video communication problem. A new generation of high-resolution color flat panel displays was selected for the project. A custom-made monitor mount and in-suite electronics enclosure was designed and constructed at UCLA. Difficulties with implementing an isolated AC power system are discussed and a work-around solution presented.

The aim of this work is to visualize 3D objects in volume data with minimum numbers of user-defined model or parameters. In this report, we propose a novel method that utilizes the distances along the optimum paths between a seed voxel in a target object and other voxels. The distance is defined using gradient between adjacent voxels or using difference between the seed voxel and other voxels along the optimum path. The optimization of a path is carried out by selecting the path where the largest value of absolute gradient or difference along a path is minimum, and the distance of each voxel is the largest value along the optimum path. The visualization is performed by rendering the volume where the initial voxel values are replaced with the distances. By the proposed method, the volume visualization can be accomplished only by setting a seed voxel in the target object. From experiments for visualizing human embryos obtained with MR microscopy, we confirmed that the proposed method successfully visualized the objects.

In tomographic images, partial volume effects (PVE) cause several artifacts in volume renditions. In x-ray CT, for example, soft-tissue-like pseudo structures appear in bone-to-air and bone-to-fat interfaces. Further, skin, which is identical to soft tissue in terms of CT number, obscures the rendition of the latter. The purpose of this paper is to demonstrate these phenomena and to provide effective solutions that yield significantly improved renditions. Here, we introduce two methods that detect and classify voxels with PVE in x-ray CT. A method is described to automatically peel skin so that PVE-resolved renditions of bone and soft tissue reveal considerably more details. In the first method, the fraction of each tissue material in each voxel v is estimated by taking into account the intensities of the voxels neighboring v. The second method is based on the following postulate (IEEE PAMI, vol. 23 pp. 689- 706, 2001): In any acquired image, voxels with the highest uncertainty occur in the vicinity of object boundaries. The removal of skin is achieved by means of mathematical morphology. Volume renditions have been created before and after applying the methods for several patient CT datasets. A mathematical phantom experiment involving different levels of PVE has been conducted by adding different degrees of noise and blurring. A quantitative evaluation is done utilizing the mathematical phantom and clinical CT data wherein an operator carefully masked out voxels with PVE in the segmented images. All results have demonstrated the enhanced quality of display of bone and soft tissue after applying the proposed methods. The quantitative evaluations indicate that more than 98% of the voxels with PVE are removed by the two methods and the second method performs slightly better than the first. Further, skin peeling vividly reveals fine details in the soft tissue structures.

The Visible Human Project planned and promoted by National Library of Medicine (NLM) provides cryosection images of the normal male and female human bodies. The anatomy of human vocal organ is difficult to understand and to imagine due to its complexity. The purpose of this study is to develop the three-dimensionally computerized atlas of the human vocal organ using Visible Human male dataset. A self-developed program with C language and a recent personal computer can show specific organs and structures separately or together, rotate them at three axes, cross-section them transparently at any angles, and zoom them in and out. As a result, our own PC-based program will be a more interactive, more detailed, and more realistic three-dimensional computerized atlas of a human vocal organ including larygopharynx.

We demonstrate that the volume enclosed by triangulated surfaces can be computed efficiently in the same elegant way the volume enclosed by digital surfaces is computed by digital surface integration. Although digital surfaces are good for visualization and volume measurement, their drawback is that surface area measurements are inaccurate. On the other hand, triangulated surfaces give more accurate surface area measurements, but volume measurements and visualization are less efficient. The T-shell data structure previously proposed retains advantages and overcomes difficulties of both the digital and the triangulated approaches. We create a lookup table with area and volume contributions for each of the 256 Marching Cubes configurations. When scanning the shell (e.g., while creating it), the surface area and volume are incrementally computed by using the lookup table and the current x co-ordinate, where the sign of the x component of the triangle normal indicates the sign of the volume contribution. We have computed surface area and volume for digital and triangulated surfaces for digitized mathematical phantoms, physical phantoms, and real objects. The computations show that triangulated surface area is more accurate, triangulated volume follows digital volume closely, and that the values get closer to the true value with decreasing voxel size.

In this paper, a new feature based rendering algorithm for partial volume is presented. This algorithm utilizes both surface and volume information for the rendering of the partial volume segmentation data with associated features. First principal directions are extracted from the partial volume segmentation dataset to construct a feature vector dataset. Using the improved LIC (line integral convolution) algorithm, our rendering algorithm can integrate the feature dataset into the traditional volume rendering process. Combining the features and the traditional volume rendering, the partial volume could be visualized clearly. This can result in an improved diagnosis in clinic, because the partial volume quantities carry very rich diagnostic information.

Three-dimensional (3D) computed rotational angiography (CRA) using an X-ray image intensifier based C-arm imaging system is a relatively new modality useful in cerebral angiography and neuro-interventional surgical procedures. In this paper, we present and study several algorithms for estimating the perspective maximum intensity projection (MIP) -- PMIP -- of CRA (neuro-angiographic) data. The nature of the data, intended applications and the associated approximations differ from those in spiral CT and MR angiography (in which parallel MIP is used extensively). Of relevance is perspective geometry, the intrinsic parameters associated with which are those of the C-arm imaging system. Its relatively small X-ray cone-angle makes the voxel driven method viable for certain applications. Consequently, the voxel-driven method is considered in greater detail. Sparseness of neuro-angiographic data allows thresholding, and hence acceleration. This feature is useful for a rapid preview and as an aid to 2D/3D registration. Post-processing for improving the quality of voxel-driven PMIP is discussed, and an adaptive algorithm is proposed in the context. Results of PMIP of data acquired by scanning patients with a prototype C-arm system are presented.

Magnetic resonance angiography (MRA) is an imaging technique to show the blood vessels but suppress signals from all the other tissues. There are two approaches to acquire an image of MRA. One is done in two-dimension (2D) by projecting the three-dimensional (3D) vessels onto 2D plane. The other is to directly obtain the complete 3D information. The advantage of 3D MRA is that one can view the data from arbitrary direction. However, the scan time is usually very long for 3D MRA. When the scan time is limited, we must use 2D MRA and the depth information is sacrificed. The depth information can be partially recovered using stereoscopic angiography, i.e., acquiring two 2D images from different viewing angles. The modality used more frequently for stereoscopic angiography is digital subtraction X-ray angiography (DSA). For X-ray DSA, the pixel intensity is directly related to the integration of the attenuation value in the path of the X-ray. We can derive the vessel shape by solving the integrals from the two views. On the other hands, there are many imaging parameters in MRA (such as T₁, T₂, and proton density). Therefore, it is difficult to obtain the relation between the vessel shape and the pixel intensity. For this reason, we attempt to reconstruct the shape of the vessels simply from the edge information of the two views. We assume that the shape of the vessel on every cross-section is an ellipse. Then, we develop an algorithm to estimate the parameters of the ellipse from the boundaries of the projective images. A 3D MRA data set was used to test the capability of our algorithm. From these data, we make two projective images. The two projections are 30 degree(s) apart. We employed our algorithm to estimate all the ellipses and reconstruct the 3D model of the vessels. Comparing the boundaries of the original projective images with the boundaries of the reconstructed 3D model, the average error is 0.471 pixels.

Phase contrast angiography (PCA) delineates the vasculature morphology based on the fact that the blood is flowing inside the vessels. And the source images of PCA have significantly less background signals than that of the time of flight (TOF) MRA. Although the PCA scan requires a slightly longer acquisition time than that of the TOF, it contains a lot more information regarding to the flow direction and velocity of the blood. Unfortunately, all these are discarded except the signal enhancement due to flow when the resulting angiogram is reconstructed and displayed on a clinical MR machine. To reveal more information contained in the PCA a presentation scheme was developed, in which each pixel after the angiographic projection has assigned a color in addition to the intensity. Employing the concept similar to the three elemental color mixing, the directional information of flow can be conveyed in an easy to comprehend format. Using the maximum intensity projection (MIP) algorithm along the viewing trace, the pixel intensity value was obtained and signed to the corresponding pixel on the PP. The final pixel color is determined by the flow direction. To demonstrate the effectiveness of the method, we obtained a PC angiography data set (TR/TE/Flip equals 20ms/3.9ms/20 degree) of the circle of Willis in transverse orientation from both normal volunteers and patient heads at 1.5T using Venc of 60 cm/sec. Results showed additional information can be gained from the PCA data, which may be useful for selected patients and sometimes crucial in the routine clinical diagnosis.

Virtual dissection refers to a display technique for polyp detection, where the colon is digitally straightened and then flattened using multirow detector Computed Tomograph (CT) images. As compared to virtual colonoscopy where polyps may be hidden from view behind the folds, the unravelled colon is more suitable for polyp detection, because the entire inner surface of the colon is displayed in a single view. The method was tested both on artificial and cadaveric phantoms. All polyps could be recognized on both phantoms. This technique for virtual dissection requires only a minimum of operator interaction.

Two main concepts of Head Mounted Displays (HMD) for augmented reality (AR) visualization exist, the optical and video-see through type. Several research groups have pursued both approaches for utilizing HMDs for computer aided surgery. While the hardware requirements for a video see through HMD to achieve acceptable time delay and frame rate seem to be enormous the clinical acceptance of such a device is doubtful from a practical point of view. Starting from previous work in displaying additional computer-generated graphics in operating microscopes, we have adapted a miniature head mounted operating microscope for AR by integrating two very small computer displays. To calibrate the projection parameters of this so called Varioscope AR we have used Tsai's Algorithm for camera calibration. Connection to a surgical navigation system was performed by defining an open interface to the control unit of the Varioscope AR. The control unit consists of a standard PC with a dual head graphics adapter to render and display the desired augmentation of the scene. We connected this control unit to a computer aided surgery (CAS) system by the TCP/IP interface. In this paper we present the control unit for the HMD and its software design. We tested two different optical tracking systems, the Flashpoint (Image Guided Technologies, Boulder, CO), which provided about 10 frames per second, and the Polaris (Northern Digital, Ontario, Canada) which provided at least 30 frames per second, both with a time delay of one frame.

Embryology is one of the basic subjects in medical education, to learn the process of human development especially from fertilization to birth. The shape deformation in the development of human embryo is one of the most important points to be comprehended, but it is difficult to illustrate the deformation by texts, 2D drawings, photographs and so on, because it is extremely complicated. The purpose of our research is to construct a 3D model sequence to illustrate the deformation of human embryo, and to make the model sequence into the teaching materials for medical education. Firstly, 3D images of the specimens of human embryo were acquired using MR microscopy. Next, an initial 3D model sequence was manually modified by comparing with the features of the acquired images under the supervision of medical doctors, because the images were influenced not only by the noise or limitation of resolution in MR image acquisition, but also by the variation of shape depending on the difference of subject. Using the constructed 3D model sequence, CG animations and an interactive VRML system were composed as the teaching materials for embryology. These materials were quite helpful to understand the shape deformation compared with the conventional materials.

In this note, we employ a conformal mapping technique for flattening branching tubular structures for visualization of MRA and CT volumetric vessel imagery. This may be used for the study of possible vessel pathology (cerebral embolism, cerebral thrombosis, or lung nodules near vessels). Our method is based on a discretization of the Laplace-Beltrami operator to flatten a surface onto a planar polygonal region in an angle-preserving manner.

This is the first report where desktop x-ray microtomography was implemented for quantitative analysis of the level of calcification in different organic material from early stage embryos to bones. A technique for the definition of local calcium density was developed. Differences in local calcium density from 0.7*10⁹ to 11.7*10⁹ (atoms per micrometers ³) were detected.