Attila Tanács

System for robotically assisted prostate biopsy and therapy with intraoperative CT guidance

RATIONALE AND OBJECTIVES The purpose of this study was to assess the work-in-progress prototype of an image-guided, robotic system for accurate and consistent placement of transperineal needles into the prostate with intraoperative image guidance inside the gantry of a computed tomographic (CT) scanner. MATERIALS AND METHODS The coach-mounted system consists of a seven-degrees-of-freedom, passive mounting arm: a remote-center-of-motion robot; and a motorized, radiolucent needle-insertion device to deliver 17-18-gauge implant and biopsy needles into the prostate with the transperineal route. The robot is registered to the image space with a stereotactic adapter. The surgeon plans and controls the intervention in the CT scanner room with a desktop computer that receives DICOM images from the CT scanner. The complete system fits in a carry-on suitcase, does not need calibration, and does not utilize vendor-specific features of the CT scanner. RESULTS In open air, the average accuracy was better than 1 mm at a 5-8-cm depth. In various phantoms, the average orientation error was 1.3 degrees, and the average distance between the needle tip and the target was 2 mm. CONCLUSION Results of preliminary experiments indicate that this robotic system may be suitable for transperineal needle placement into the prostate and shows potential in a variety of other percutaneous clinical applications.

Transrectal Prostate Biopsy Inside Closed MRI Scanner with Remote Actuation, under Real-Time Image Guidance

We present the proof-of-concept prototype of a prostate biopsy robot to be used inside a conventional high-field MRI scanner. A three degree-of-freedom (DOF) mechanical device translates and rotates inside the rectum and enters a needle into the body, and steers the needle to a target point preselected by the user. The device is guided by real-time images from the scanner. Networked computers process the medical images and enable the clinician to control the motion of the mechanical device that is operated remotely from outside the imager. The system is also applicable to localized prostate therapy and also demonstrates potential in other intra-cavitary procedures.

Estimation of linear deformations of 2D and 3D fuzzy objects

Registration is a fundamental task in image processing, it is used to determine geometric correspondences between images taken at different times and/or from different viewpoints. Here we propose a general framework in n-dimensions to solve binary shape/object matching problems without the need of establishing additional point or other type of correspondences. The approach is based on generating and solving polynomial systems of equations. We also propose an extension which, provided that a suitable segmentation method can produce a fuzzy border representation, further increases the registration precision. Via numerous synthetic and real test we examine the different solution techniques of the polynomial systems of equations. We take into account a direct analytical, an iterative least-squares, and a combined method. Iterative and combined approaches produce the most precise results. Comparison is made against competing methods for rigid-body problems. Our method is orders of magnitude faster and is able to recover alignment regardless of the magnitude of the deformation compared to the narrow capture range of others. The applicability of the proposed methods is demonstrated on real X-ray images of hip replacement implants and 3D CT volumes of the pelvic area. Since the images must be parsed through only once, our approach is especially suitable for solving registration problems of large images. Highlightsn-Dimensional image registration method by shape/object matching.The approach is based on generating and solving polynomial systems of equations.Detailed discussion of fast and efficient algorithmic solution techniques.Efficiency confirmed by a thorough evaluation performed on synthetic and real data.Especially suitable, due to time efficiency, for registration of large objects.

Car recognition from frontal images in mobile environment

Recognition of car make and model from frontal images is a common problem in computer vision. We refined existing approaches based on ROIs defined relative to the number plate. Square-Mapped-Gradient features are extracted from the ROI and recognition is accomplished by classification utilizing a learning set. The classifier is evaluated using ground truth data provided manually. Via numerical simulations we evaluated the detection tolerance of the method and proposed semi-automatic and fully automatic methods. The SMG-based classification is able to give nearly perfect results when there is no outlier class, which decreases to 92% and 87% in case of the semi-automatic and fully automatic methods, respectively. Separation between outliers and known types can be balanced by a threshold. Since the size of the learning set can be kept low and the size of the SMG features are small, this approach can be successfully used to solve mobile client-server scenarios.

Estimation of linear deformations of 3D objects

We propose a registration method to find affine transformations between 3D objects by constructing and solving an overdetermined system of polynomial equations. We utilize voxel coverage information for more precise object boundary description. An iterative solution enables us to easily adjust the method to recover e.g. rigid-body and similarity transformations. Synthetic tests show the advantage of the voxel coverage representation, and reveal the robustness properties of our method against different types of segmentation errors. The method is tested on a real medical CT volume.

Recovering Affine Deformations of Fuzzy Shapes

Fuzzy sets and fuzzy techniques are attracting increasing attention nowadays in the field of image processing and analysis. It has been shown that the information preserved by using fuzzy representation based on area coverage may be successfully utilized to improve precision and accuracy of several shape descriptors; geometric moments of a shape are among them. We propose to extend an existing binary shape matching method to take advantage of fuzzy object representation. The result of a synthetic test show that fuzzy representation yields smaller registration errors in average. A segmentation method is also presented to generate fuzzy segmentations of real images. The applicability of the proposed methods is demonstrated on real X-ray images of hip replacement implants.

Fast linear registration of 3D objects segmented from medical images

In this paper a linear registration framework is used for medical image registration using segmented binary objects. The method is best suited for problems where the segmentation is available, but we also propose a general bone segmentation approach for CT images. We focus on the case when the objects to be registered differ considerably because of segmentation errors. We check the applicability of the method to bone segmentation of pelvic and thoracic CT images. Comparison is also made against a classical mutual information-based registration method.

Establishing Correspondences between Planar Image Patches

Finding correspondences between image pairs is a fundamental task in computer vision. Herein, we focus on establishing matches between images of urban scenes which are typically composed of planar surface patches with highly repetitive structures. The latter property makes traditional point-based methods unreliable. The basic idea of our approach is to formulate the correspondence problem in terms of homography estimation between planar image regions: given a planar region in one image, we are simultaneously looking for its corresponding segmentation in the other image and the planar homography acting between the two regions. We will show, that due to the overlapping views the general 8 degree of freedom (DOF) of the homography mapping can be geometrically constrained to 3 DOF and the resulting segmentation/registration problem can be efficiently solved by finding the regions occurrence in the second image using pyramid representation and normalized mutual information as the intensity similarity measure. The method has been validated on a large database of building images taken by different mobile cameras and quantitative evaluation confirms robustness against intensity variations, occlusions or the presence of non-planar parts. We also show examples of 3D planar surface reconstruction as well as 2D mosaicking.

Application of automatic image registration in a segmentation framework of pelvic CT images

In radiation treatment (RT) planning, clinicians must trace the outline of a few critical structures on a large number of images. Using automated image segmentation could save tremendous time and effort. Segmentation of the organs near the pubic bone (prostate and bladder) is an important and challenging task: Some of the neighboring organs have similar density values in the CT images and the border between the different organs is hardly visible. In a segmentation framework, transforming a CT study to a common reference frame is used in two tasks: For statistical atlas (model) generation, and in the clinical application, establishing the voxel-to-voxel correspondence between the study and the model. In these cases precise alignment of all anatomical structures is not crucial, the focus is on proper alignment of the pubic bone area and fast execution. Our proposed method solves this by a new, two step process based on a voxel similarity-based registration algorithm.

Evaluation of point matching methods for wide-baseline stereo correspondence on mobile platforms

Wide-baseline stereo matching is a common problem of computer vision. By the explosion of smartphones equipped with camera modules, many classical computer vision solutions have been adapted to such platforms. Considering the widespread use of various networking options for mobile phones, one can consider a set of smart phones as an ad-hoc camera network, where each camera is equipped with a more and more powerful computing engine in addition to a limited bandwidth communication with other devices. Therefore the performance of classical vision algorithms in a collaborative mobile environment is of particular interest. In such a scenario we expect that the images are taken almost simultaneously but from different viewpoints, implying that the camera poses are significantly different but lighting conditions are the same. In this work, we provide quantitative comparison of the most important keypoint detectors and descriptors in the context of wide baseline stereo matching. We found that for resolution of 2 megapixels images the current mobile hardware is capable of providing results efficiently.