Showing papers presented at "Computer Assisted Radiology and Surgery in 2001"

Total knee arthroplasty: coordinate system definition and planning based on 3-D ultrasound image volumes

Abstract: The main problem for the long-term quality of total knee arthroplasties (TKA) is the correct placement, especially for the femoral prosthesis. Most pre-operative planning procedures and intra-operative realizations actually rely on lower limb X-rays and extra- or intramedullary devices. Due to imprecise alignment of images and devices and fuzzy referencing procedures, this may lead to implant displacements. We defined precise coordinate systems referencing on superficially located direct visible bone landmarks. By use of 3-D ultrasound (US) and automatic image segmentation, we generated virtual models of the distal femur for 10 knee joints. The defined landmark systems could be detected on these models, which then were used for interactive planning. The investigation demonstrates that 3-D US can provide data of position and size of bony structures. Bone models created by analysis of 3-D US image volumes are useable for the simulation of TKA implantations.

Practical model “3DX” of limited cone-beam X-ray CT for dental use

Abstract: Objective: The authors have developed a clinical model of limited cone-beam X-ray CT for dental use and started to use the model in clinical practice. It is called “3DX multi image micro CT” (3DX, J. Morita, Kyoto, Japan). Presented here is a report about the result. Method: We made a design of limited cone-beam X-ray CT so that it could be most effective for dentistry use. This machine is a kind of cone-beam type X-ray CT equipped with a 4-in. imaging intensifier (II) as a sensor. The size of the X-ray at its rotational center is 29 mm high and 39 mm wide. While a subject is seated for shooting, a focal spot and the X-ray rotational center are put into agreement with a light beam of X, Y and Z directions. Images are taken under the exposure condition of 80 kV (X-ray tube voltage) and 2 mA (X-ray tube electric current). Exposure time is 17 s, during which 512 frames of two-dimensional images are recorded. These images are reconstructed with a personal computer into digital images in a columnar field of 38 mm in diameter and 29 mm in height. It takes about 10 min to carry out arithmetic calculation for the image reconstruction. Clinical use of this machine started in January 2001 at the Department of Radiology, Nihon University of Dental Hospital. Result: The 3DX produced highly clear, three-dimensional tomographic images of impacted teeth. Conclusion: Since the 3DX is capable of producing clear images with a small radiation dose, we consider it to be effective as a tool to support three-dimensional imaging diagnosis for dental implants, impacted teeth, etc.

Robot-assisted insertion of craniofacial implants—clinical experience

Abstract: A medically approved robot system was clinically used to insert craniofacial implants into the skull for anchoring a silicone ear prosthesis. Additionally, a new process was developed for the preoperative manufacturing of the prosthesis using the computed tomography image data and a rapid prototyping technique. The navigated robot showed the surgeon intraoperatively the planned implant positions and guided the insertion procedure. The robot worked not automatically but interactively with the surgeon. In 13 patients, 30 implants were inserted with no intraoperative injuries. An absolute implant position accuracy of about ±1 mm and a relative accuracy between the implants of about ±0.2 mm were reached. This accuracy made it possible to apply the preoperatively manufactured ear prosthesis directly after surgery. The rehabilitation time for the patient was shortened. These clinical results were reached only by careful optimisation of each step of the intervention, the image acquisition, patient fixation and the intraoperative execution. The experience is good cause to use the robot system and the new manufacturing concept for anaplastology in other areas of the head as well.

Influence of bone milling parameters on the temperature rise, milling forces and surface flatness in view of robot-assisted total knee arthroplasty

Abstract: The authors have developed a robot-assisted procedure to prepare the tibial surface in total knee arthroplasty (TKA). Using hybrid force/velocity control, the surgeon machines the bone by means of an end mill attached to the robot-arm. The milling forces can give information on the local bone quality intra-operatively. This paper describes experiments on human tibial bone to analyse the influence of the feed rate and the milling speed on the temperature rise, the milling forces and the surface flatness.

Stereoscopic augmented reality system for computer-assisted surgery

Abstract: A first architecture for an augmented reality system in computer-assisted surgery is presented in this paper. Like in “X-ray vision” systems, a stereoscopic overlay is visually superimposed on the patient. The main purpose of our approach is user-friendliness for the surgeon: no additive wearing equipment is required. Registration, rigid body location and 3D volume computation are proven to respect real-time processing, thanks to an optical navigation system and our integrated software framework. Studies are undertaken to replace our actual monitor display by an upcoming holographic screen.

Simulated vascular reconstruction in a virtual operating theatre

Abstract: We present an experimentation environment that combines interactive visualisation of patient-specific vascular medical data with a flow simulation environment into an interactive exploration environment that provides a virtual operating theatre in which vascular reconstruction procedures can be simulated.

Image fusion of ultrasound and MRI as an aid for assessing anatomical shifts and improving overview and interpretation in ultrasound-guided neurosurgery

Abstract: We have developed an image fusion module for combined 3D ultrasound and MRI/CT visualization in image-guided neuronavigation. The module demonstrates alternative image visualization possibilities, which might be integrated in future ultrasound-based neuronavigation systems. The image fusion visualization module reads 3D ultrasound, MRI/CT image data sets and patient registration information which are saved during the operation using the SonoWand ultrasound-based neuronavigation system earlier presented by our group [1] . The images are visualized simultaneously in the same scene by overlay, compositing or splitting visualization techniques which makes perception of information easier as compared to visualizing images in different scenes. Quantitation of brain shift between similar structures recognized in both preoperative MRI and intraoperative ultrasound images was done using both a manual and an automatic image processing algorithm. The results show that alternative image visualizations may be useful in order to get immediate feedback regarding brain shift in the early beginning of the operation. The present paper summarizes the experiences gained using this visualization module on patient image data acquired in our clinic during the last year using the SonoWand ultrasound-based neuronavigation system.

Optimal thresholding for 3D segmentation of pulmonary nodules in high resolution CT

Abstract: For the segmentation of a lung nodule out of a given volume of interest for subsequent rendering and measurement, it is crucial to find an appropriate Hounsfield threshold. A threshold finding method has been suggested by Zhao et al. [Med. Phys. 26 (6) (1999) 889]. By application of the divergence theorem to the computation of the mean gradient of the iso-surfaces for all possible thresholds, we were able to speed up this method by approximately a factor 100. Our novel computation scheme completes the task in under a second of computation time (on comparable hardware). It can thus be used in real-time and as an interactive tool in a medical image processing workstation. The algorithm presented here requires only a single pass through the data volume of interest and is guaranteed to determine the optimal threshold of all possible thresholds within the entire gray-value range (Hounsfield units). In contrast to the method suggested by Zhao et al. that works slice-wise in 2D, the algorithm presented here is fully 3D oriented, and particularly suited for use on high resolution CT volume data.

Comparison of task performance of robotic camera holders EndoAssist and Aesop

Abstract: Background: Robot and operator should be considered as a system. Evaluation of this system, not of the robot itself, will assess the quality of the robot. EndoAssist and Aesop, laparoscopic camera-holder surgical robots, were compared. Methods: Diagonal, vertical, sideways, zoom and complex movement tasks were timed. The whole procedure was videotaped in a simulated environment and the time of each movement was taken from the tape. Results: EndoAssist performed simple downward, sideways and diagonal tasks as well as complex movement faster than Aesop. There was no difference between the robots for zoom movements. Aesop was quicker only in the preprogrammed mode for most complex tasks. Conclusion: EndoAssist performed better overall due to its greater accuracy and reduced number of erratic movements.

Realism in surgical simulators with free-form geometric modeling

Abstract: Robotic systems for cardiac surgery have been introduced in clinical trials to facilitate minimally invasive techniques. Widespread use of surgical robotics necessitates new training methods to improve skills and continued practicing as the robotic systems are frequently being upgraded. Today, robotic training is performed on expensive animal models. Integration of a digital trainer with the two present robotic systems applied in coronary artery bypass procedures on beating heart requires real time simulation of tissue mechanics, sutures, instruments and bleeding, but no extra haptic device, since the robotic master is the haptic apparatus itself. We have demonstrated the possibility of obtaining surgical simulation on a standard PC Linux system by integrating free-form geometry data structures integrated with spring-mass tissue models. This technology also makes benefits when simulation is to be used over a network with limited bandwidth, especially when it comes to handling of soft tissue dynamics.

Stereoscopic digital mammography: improving detection and diagnosis of breast cancer

Abstract: Detecting and interpreting subtle breast lesions with standard film mammography is often difficult because of underlying and overlying normal tissue superimposed on the lesion in standard 2D views. Stereoscopic digital mammography may solve this and other problems of ambiguity by permitting the radiologist to view the lesion in the volume, separated from the superimposed tissue in depth. We have acquired both film and stereo mammograms on women prior to biopsy of a suspicious lesion. In a reading study with experienced mammographers, stereo mammography, as an adjunct to film, was shown to significantly improve diagnostic accuracy. But even more importantly, readers detected many new lesions in the stereo mammograms that were not visible in the films.

An automated classification scheme for mammograms based on amount and distribution of fibroglandular breast tissue density

Abstract: It is very important to assess the fibroglandular breast tissue density to define the degree of the risk that the lesions are obscured by normal breast tissue We developed an automated classification method, in which the mammograms were divided into three regions by both the variance histogram analysis and discriminant analysis and were classified into four categories based on the ratios of each of three regions The classification results by the normal images showed the high agreement rate between physicians and computer As a result of malignant images' classification, the influence of the existence of mass regions to classification results is dependent not only on mass sizes but also on these positions Because the rate of different classifications of right and left images in malignant database is larger than that in normal one, it may be possible to apply this scheme for potential indication of the detection of mass lesions

Computer-aided planning of trocar placement and robot settings in robot-assisted surgery

Abstract: The Zeus robotic system consists of three separate arms. Before surgery, each arm has to be attached to the operating table. Preoperatively, a total of 21 variables have to be set manually. This includes 15 variables for all three arms in addition to three trocars placed with two degrees of freedom. There are significant constrains in changing these variables during surgery, especially because maintained sterility is demanded in the surgical field. Thus, the initial positioning of the arms and trocars is important. Algorithms for calculating robot movements and distance between arms are developed and meant to be an integrated part of a system for computer-aided planning of trocar placement and robot settings in robot-assisted surgery. The mathematical model of the robotic arms and robotic set-up is made in Matlab. The movements of all three robotic arms are recorded during a robotic procedure by reading the position of three LED pointers attached to each robotic arm. By using the recorded movements, the effect of changing external parameters, like port placement, patient position and elbow direction can be simulated through the model. It is shown that even small changes in initial robot settings may result in decreased distance, or collision, between the robotic arms.

Intravenous coronary angiography with synchrotron radiation

Abstract: Efforts have been made to image coronary arteries by non-invasive or minimally invasive methods. One method is dichromography after intravenous injection of a contrast agent. The method and the NIKOS system, which was developed for investigations of patients using this method, are described. Up to now, 379 patients were investigated. The results of a validation study are given.

Automatic border detection in IntraVascular UltraSound images for quantitative measurements of the vessel, lumen and stent parameters

Abstract: This paper describes a knowledge- and model-guided system for the automatic contour detection of the vessel, lumen and stent borders in IntraVascular UltraSound (IVUS images). The different steps of a complete analysis of an IVUS pullback sequence are carried out such that the different components assist each other. The stent detection assists the vessel detection and both detections assist the lumen detection. Finally, the applications of the quantification software in the catheterization laboratory (on-line) as well as for research studies (off-line) are discussed.

A CAD system for lung cancer based on CT image

Abstract: This paper presents a prospective evaluation of lung cancer screening based on low-dose single helical 3D CT images. This CT screening uses a computer-aided diagnosis (CAD) system to efficiently detect lung nodules from 3D images. The performance of this screening is evaluated by over 5,000 cases for 5 years. The effectiveness of the screening using the CAD system is certified from a large data test.

The PREPaRe system—patient-oriented access to the “personal electronic medical record”

Abstract: The PREPaRe system is a patient-oriented internet-based information system that is able to store, combine, process and visualize all types of medical data that are part of a “personal electronic medical record” (PEMR). In this paper, the concept and architecture of the system as well as functional components, such as image processing, visualization, storage, linking and personalization mechanisms, will be presented.

Virtual endoscopy for cardio vascular exploration

Abstract: With the recent introduction of multi-slice CT, which acquire four image slices at the same time (which will increase to 16 image slices in the near future), the possible size of the generated image stack has increased tremendously. Traditional cine mode explorations of the image stack are rendered impractical by the sheer quantity of possible 1000 image slices of the resolution of 512×512 pixels. In this paper, we discuss the use of VIVENDI, a virtual endoscopy system applied for the three-dimensional exploration of cardio vascular structures. In particular, stenoses of the coronary arteries can be visualized, if depicted by the scan.

A fully automated method for segmentation and thickness determination of hip joint cartilage from 3D MR data

Abstract: A fully automated method for segmentation of hip joint cartilage, including both femoral head and acetabular cartilage, and determination of its thicknesses from in vivo 3D magnetic resonance (MR) data is described. The method utilizes the constraint of the hip joint shape. That is, it assumes that both the femoral and acetabular cartilage is distributed on a spherical surface whose center corresponds to the rotational center of hip joint motion. The method consists of the following four steps. (1) Determination of the center of a sphere that approximates the femoral head using Hough transform. (2) Enhancement of cartilage regions and their inner edges (bone-attached) using directional derivatives along radial directions originating from the sphere center. (3) Segmentation of femoral head and acetabular cartilage regions using radial directional derivative images. (4) Subvoxel localization of cartilage boundaries for thickness determination. The proposed method was successfully applied to 13 sets of in vivo MR data of normal volunteers and actual patients.

Computer-aided detection of polyps in CT colonography: evaluation of volumetric features in differentiating polyps from false positives

Abstract: We evaluated the ROC performance of six volumetric features in differentiating between the true-positive and false-positive polyp findings of our CAD system for CT colonography. Among the single feature statistics, the mean shape index had the best ROC performance, followed by a new D-iris feature. The ROC results of the round-robin discriminant analysis indicate that the performance of our CAD system can potentially be improved by use of a combination of the shape index, CT value, and the D-iris feature statistics.

Towards automatic quantitative analysis of cardiac MR perfusion images

Abstract: Magnetic Resonance Imaging (MRI) is a powerful technique for imaging cardiovascular diseases. The introduction of cardiovascular MRI into clinical practice is however hampered by the lack of efficient and reliable automatic image analysis methods. This paper focuses on the automatic evaluation of the perfusion of blood in the myocardium (the heart muscle) from cardiac MR perfusion image series, acquired using contrast-enhanced ECG-triggered MRI. We have developed a semi-automatic quantitative analysis method with which the perfusion image series can be analysed in only a few minutes. The method is described in this paper and preliminary validation results are presented.

Intraoperative detection of malignant gliomas using 5-Aminolevulinic acid-induced protoporphyrin fluorescence, openMRI and real-time navigation system

Abstract: It is generally accepted by neurosurgeons that the complete tumor resection is one of the factors to improve the prognosis of malignant glioma patient. However, the extensive resection remains controversial because the tumor margin is difficult to be distinguished from surrounding normal and/or edematous brain. It has been established that photosensitizer protoporphyrin IX (PpIX) induced by 5-Aminolevulinic acid (5-ALA) is used as a fluorescence detection marker for photodiagnosis of malignant gliomas. We tried to study whether or not the location of fluorescence detection using 5-ALA is compatible with image-guided tumor location and whether or not the fluoresce-positive tissue contained tumor cell by histological staining. Patients with malignant gliomas received oral doses of 5-ALA before anesthesia. During surgery, an MR image was scanned and registered for navigation system. Tumor fluorescence was visualized under Ultra-violet (UV) light illumination. The locations of fluorescence detection were confirmed by the real-time navigation. The marked fluorescing tissues were removed and used for following histological analysis. The locations of fluorescence detection were confirmed to be in agreement with Gd-enhanced region indicated by real-time navigation, whereas the area was macroscopically impossible to distinguish from the regions in which there was no radiological enhancement. Histological study showed that the tissues from fluorescence positive regions contained tumor cells.

Fast computation of isosurface contour spectra for volume visualization

Abstract: Surface rendering and the construction of transfer functions for direct volume rendering are often based on thresholding and the resulting isosurfaces. Several approaches have been proposed to simplify the process of finding meaningful isosurface thresholds. However, optimal rendering parameters are determined in many approaches by analysis of the rendered images and not of the original volume data, thus requiring a cost-intensive thresholding/rendering optimization cycle. In this paper we present a time- and memory-efficient method for automated detection of meaningful intensity transitions between different tissue types to support visualization of CT volume data. In a single pass through the data volume a contour spectrum is determined which in particular indicates the total gradient integral of the isosurface for each possible threshold.

Mobile teleradiology: all images everywhere

Abstract: This paper describes the ongoing work on mobile teleradiology systems in the EC-funded project MTM. The current development is based on the CHILI software architecture which provides a PACS and tel ...