Image registration is the task of aligning two or more images into the same reference frame to compare or distinguish the images. The majority of registration methods deal with registering only two images at a time. Recently, a clustering method that concurrently registers more than two multi-sensor images was proposed, dubbed ensemble clustering. In this thesis, we apply the ensemble clustering method to deformable registration scenario for the first time. Non-rigid deformation is implemented by a FFD model based on B-splines. A regularization term is added to the cost function of the method to limit the topology and degree of the allowable deformations. However, the increased degrees of freedom in the transformations caused the Newton-type optimization process to become ill-conditioned. This made the registration process unstable. We solved this problem by using the matrix approximation afforded by the singular value decomposition (SVD). Experiments showed that the method is successfully applied to non-rigid multi-sensor ensembles and overall yields better registration results than methods that register only 2 images at a time. In addition, we parallelized the ensemble clustering method to accelerate the performance of the method. The parallelization was implemented on GPUs using CUDA (Compute Unified Device Architecture) programming model. The GPU implementation greatly reduced the running time of the method.

This paper presents a set of Image Registration, Image Integration, interpolation and image restoration and other technology, the number of low-resolution images synthesized high-resolution image. Relative to the existing image fusion technology, the method provided in this paper has more advantages, such as: (1) high-precision value; (2)low computation cost; (3)a compact system; (4) applicable to noise images; (5) robotic and automatic performance.

This paper presents a novel integration technique for segmented 3D profiles measured by projected fringe profilometry. Fringe patterns are projected to the inspected surface. The projected patterns fix their positions relative to the tested object during two segmented measurements. Thus, finding two matched surface points becomes a problem of searching for two identical phases in the fused data sets. This novel integration technique can match images successfully and achieve pixel-to-pixel registration easily even in the presence of geometric deformation, illumination changes, and severe occlusions. It is superior to the other methods because of its: (1) High matching accuracy; (2) Improved robustness; (3) Reduced computational time; (4) Capability of compensating distortions of the optical system at every pixel location; (5) Suitable for images rotating or scaling; and (6) Suitable for any other projected fringe measurement method. We also propose a method to design and fabricate a 2-D fringe pattern which can be applied to the integration technique for segmented 3D profiles. Campered with using 1-D fringe patterns for image registration, using a 2-D fringe pattern saves the measurement time and further proveds more tolerence to hand the shadow and noise problems. Tests of the system performance have been carried out that the accuracy of the registration scheme is 5.96% of image pixel size. Therefore, this technique can be extensively used in modern high technology industry. Especially when it requires higher resolution close-up images or overcomes the issue of not every inspected object can be fully expressed just by a single full-field measurement, it is necessary to use this integration technique.

Thesis (Ph.D)--Computing, Georgia Institute of Technology, 2009. / Committee Chair: Vuduc, Richard; Committee Member: Biros, George; Committee Member: Davatzikos, Christos; Committee Member: Tannenbaum, Allen; Committee Member: Zhou, Hao Min. Part of the SMARTech Electronic Thesis and Dissertation Collection.

Το πρόβλημα της αντιστοίχισης εικόνων είναι ένα από τα σημαντικότερα στο πεδίο της υπολογιστικής όρασης, αφού η ευθυγράμμιση δύο ή περισσότερων εικόνων χρησιμοποιείται τουλάχιστον σαν στάδιο προεπεξεργασίας σε ένα μεγάλο αριθμό εφαρμογών. Στην εργασία αυτή μας απασχόλησε το πρόβλημα της στοίχισης εικόνων στις οποίες οι φωτομετρικές παραμορφώσεις είναι τοπικές και δεν μπορούν να μοντελοποιηθούν με το γενικό σφαιρικό μοντέλο της αντίθεσης και της φωτεινότητας, ή/και τμήματα των προς στοίχιση εικόνων είναι αποκλεισμένα από τη μια από αυτές. Για την αντιμετώπιση των παραπάνω προβλημάτων, η αντιστοίχηση των εικόνων προσεγγίστηκε μέσω της σταθμισμένης ελαχιστοποίησης μετρικών σφάλματος που βασίζονται στο τετραγωνικό σφάλμα. Συγκεκριμένα, εκμεταλλευόμαστε την αμεταβλητότητα της κανονικοποιημένης κλίσης μιας εικόνας σε τοπικές φωτομετρικές παραμορφώσεις και τη δυνατότητα στοίχισης κάθε ζεύγους αντίστοιχων εικονοστοιχείων των υπό στοίχιση εικόνων με την μεγιστοποίηση της μεταξύ τους συσχέτισης. Έτσι πετυχαίνουμε την αποσύνδεση του αρχικού προβλήματος σε δύο υποπροβλήματα η λύση των οποίων καταλήγει σε δύο υπερκαθορισμένα συστήματα γραμμικών εξισώσεων, καθένα εκ των οποίων έχει ως αγνώστους τις ανά κατεύθυνση παράμετρες του μετασχηματισμού που αναζητούμε για την εξάλειψη της γεωμετρικής παραμόρφωσης και ως δεξιό μέλος τις τιμές των φωτομετρικών παραμορφώσεων. Τελικά, με την επιλογή δύο κατάλληλων υποσυνόλων των προαναφερθέντων γραμμικών εξισώσεων, που εξασφαλίζουν την εφικτότητα των επιμέρους λύσεων οδηγούμαστε στον προσδιορισμό των βέλτιστων παραμέτρων. Η προτεινόμενη τεχνική δοκιμάστηκε στη βάση προσώπων Yale Β που έχει χρησιμοποιηθεί από άλλες τεχνικές αντιστοίχισης που είναι ειδικά προσαρμοσμένες για την αντιστοίχιση προσώπων. Η απόδοση της προτεινόμενης τεχνικής είναι πολύ καλή και υπερτερεί και στα ποσοστά σύγκλισης αλλά και στην ακρίβεια των λύσεων από την απόδοση των άλλων τεχνικών τόσο στη στοίχιση εικόνων που έχουν υποστεί γεωμετρικές παραμορφώσεις (από πολύ μικρές μέχρι και πολύ έντονες) όσο και σε εικόνες με διαφορετικές έντονες φωτομετρικές παραμορφώσεις. Επίσης, η προτεινόμενη τεχνική δοκιμάστηκε στις βάσεις του Affine Covariance Regions του University of Oxford στις οποίες το περιεχόμενο των εικόνων είναι γενικό και οι ειδικού σκοπού τεχνικές αποτυγχάνουν, με εξίσου πολύ καλή απόδοση. / The image registration problem is one of the most important problems in the field of computer vision, since the process of aligning two or more images is used, at least as a preprocessing step, in many applications. In this work, we employed the problem of image alignment in which the photometric deformations are local and can not be modeled with the general spherical model of contrast and brightness, and / or portions of images to align are occluded. To address these problems, the image registration was approached by minimizing the weighted error metric based on squared error. In particular, we exploit the invariance of the normalized image gradient in local photometric deformations so we can align each pair of corresponding pixels in the images by maximizing the correlation between them. Thus, we achieve to dissolve the original problem into two subproblems the solution of which leads to two over-determined systems of linear equations, each of which has the direction parameters of the transformation we seek to estimate as unknowns and as right member the values of photometric deformations. Ultimately, the choice of two suitable subsets of the above linear equations, ensuring the feasibility of individual solutions we are lead to the identification of best parameters. The proposed technique was tested in Yale B face database which has been used by other mapping techniques adapted to matching persons. The performance of the proposed technique is very good and superior at the convergence rates and the accuracy of the solutions to the performance of other techniques concerning both images that have undergone geometrical deformation (from very small to very intense) and images in different intense photometric deformations. Also, the proposed technique was tested on database of Affine Covariance Regions of the University of Oxford in which the content of the images is general and special-purpose techniques fail, with equally good performance.

No description available.

Anatomical changes can have significant clinical impact during head and neck radiotherapy. Adaptive radiotherapy (ART) may be applied to account for such changes. Implementation of ART to alter dose delivery requires deformable image registration (DIR) to assess 3D deformations. This study evaluates the performance and accuracy of a commercial DIR system for clinical applications. The investigations in this project were carried out using images of induced changes in two standard radiotherapy phantoms (RANDO® and CIRS®) and one in-house built phantom. CT image data before and after deformation of the phantoms were processed using Eclipse / SmartAdapt® v.10 system employing a Demons-based algorithm. A DIR protocol was designed, and algorithm performance was assessed quantitatively, using volume analysis and the Dice Similarity Index (DSI), and also evaluated qualitatively. In addition, algorithm performance was assessed for 5 head and neck cancer patients using clinical CT images. Each original planning CT image containing contours of 10 volumes of interest including treatment target volumes and organs at risk was deformed to match a second CT image acquired during the course of the treatment. The original structures were deformed, copied onto the target image and compared to reference contours drawn by 3 radiation oncologists. Phantom investigations gave varied results with average DSI scores ranging from 0.69 to 0.93, with an overall average of 0.86 ± 0.08. These quantitative results were reflected qualitatively, with generally accurate matching between reference and DIR-generated structures. Although air gaps in the phantoms compromised algorithm performance and gave rise to physically aberrant results. Clinical results were generally better with a DSI range of 0.75-0.99 and an overall average of 0.89 ± 0.05, suggesting high DIR accuracy. Qualitatively, some minor contour deformations were noted, as well as artefacts in the axial direction that were due to the CT slice resolution (3 mm) that was used to scan the patients. In addition, contour propagation between images using DIR reduced the time required by physicians to contour the images of head and neck cancer patients by ~47%. This study demonstrated that deformable image registration using a Modified Demons algorithm yields clinically acceptable results and time-saving benefits in contouring that improve clinical workflow. The study also showed that it is feasible to incorporate deformable image registration as part of an adaptive radiotherapy strategy for head and neck cancer, provided further studies are designed to carry out accurate and verifiable dose deformation.

The modern day radiotherapy treatments are almost exclusively based on computed tomography (CT) images. The CT images are acquired using x-rays, and therefore reflect the radiation interaction properties of the material. This information is used to perform accurate dose calculation by the treatment planning system, and the data is also well suited for creating digitally reconstructed radiographs for comparing patient set up at the treatment machine where x-ray images are routinely acquired for this purpose. The magnetic resonance (MR) scanner has many attractive features for radiotherapy purposes. The soft tissue contrast as compared to CT is far superior, and it is possible to vary the sequences in order to visualize different anatomical and physiological properties of an organ. Both of these properties may contribute to an increase in accuracy of radiotherapy treatment. Using the MR images by themselves for treatment planning is, however, problematic. MR data reflects the magnetic properties of protons, and thus have no connection to the radiointeraction properties of the material. MRI also has inherent difficulty in imaging bone, which will appear in images as areas of no signal similar to air. This makes both dose calculation and patient positioning at the treatment machine troublesome. There are several clinics that use MR images together with CT images to perform treatment planning. The images are registered to a common coordinate system, a process often described as image fusion. In these cases, the MR images are primarily used for target definition and the CT images are used for dose calculations. This method is now not ideal, however, since the image fusion may introduce systematic uncertainties into the treatment due to the fact that the tumor is often able to move relatively freely with respect to the patients’ bony anatomy and outer contour, especially when the image registration algorithms take the entire patient anatomy in the volume of interest into account. The work presented in the thesis “Integration of MRI into the radiotherapy workflow” aim towards investigating the possibilities of workflows based entirely on MRI without using image registration, as well as workflows using image registration methods that are better suited for targets that can move with respect to surrounding bony anatomy, such as the prostate. / Modern strålterapi av cancer baseras nästan helt på datortomografiska (CT) bilder. CT bilder tas med hjälp av röntgenfotoner, och återger därför hur det avbildade materialet växelverkar med strålning. Denna information används för att utföra noggranna dosberäkningar i ett dosplaneringssystem, och data från CT bilder lämpar sig också väl för att skapa digitalt rekonstruerade röntgenbilder vilka kan användas för att verifiera patientens position vid behandling. Bildgivande magnetresonanstomografi (MRI) har många egenskaper som är intressanta för radioterapi. Mjukdelskontrasten i MR bilder är överlägsen CT, och det är möjligt att i stor utstäckning variera sekvensparametrar för att synliggöra olika anatomiska och funktionella attribut hos ett organ. Dessa bägge egenskaper kan bidra till ökad noggrannhet i strålbehandling av cancer. Att använda enbart MR bilder som planeringsunderlag för radioterapi är dock problematiskt. MR data reflekterar magnetiska attribut hos protoner, och har därför ingen koppling till materialets egenskaper då det gäller strålningsväxelverkan. Dessutom är det komplicerat att avbilda ben med MR; ben uppträder som områden av signalförlust i bilderna, på samma sätt som luft gör. Detta gör det svårt att utföra noggranna dosberäkningar och positionera patienten vid behandling. Många moderna kliniker använder redan idag MR tillsammans med CT under dosplanering. Bilderna registreras till ett gemensamt koordinatsystem i en process som kallas bildfusion. I dessa fall används MR bilderna primärt som underlag för utlinjering av tumör, eller target, och CT bilderna används som grund för dosberäkningar. Denna metod är dock inte ideal, då bildregistreringen kan införa systematiska geometriska fel i behandlingen. Detta på grund av att tumörer ofta är fria att röra sig relativt patientens skelett och yttre kontur, och många bildregistreringsalgoritmer tar hänsyn till hela bildvolymen. Arbetet som presenteras i denna avhandling syftar till att undersöka möjligheterna med arbetsflöden som baseras helt på MR data utan bildregistrering, samt arbetsflöden som använder bildregistrerings-algoritmer som är bättre anpassade för tumörer som kan röra sig i förhållande till patientens övriga anatomi, som till exempel prostatacancer.

Inaccurate dose calculation in radiotherapy can lead to errors in treatment delivery and evaluation of treatment efficacy. Respiration can cause of intra-fractional motions, leading to uncertainties in tumor targeting. These motions should therefore be included in dose calculation. The finite element method-based deformable registration platform MORFEUS is able to accurately quantify organ deformations. The dose accumulation algorithm included in MORFEUS takes organ deformation and tumor movement into account. This study has experimentally validated this dose accumulation algorithm by combining 3D gel dosimetry, respiratory motion-mimicking actuation mechanism, and finite element analysis. Results have shown that within the intrinsic measurement uncertainties of gel dosimetry, under normal conformal dose distribution conditions, more than 90% of the voxels in MORFEUS generated dose grids have met the criterion analogous to the gamma test. The average (SD) distance between selected pairs of isodose surfaces on the gel and MORFEUS dose distributions is 0.12 (0.08) cm.

Effective early detection and monitoring of patient response to cancer therapy is important for improved patient outcomes, avoiding unnecessary procedures and their associated toxicities, as well as the development of new therapies. Dynamic contrast-enhanced magnetic resonance imaging shows promise as a way to evaluate tumour vasculature and assess the efficacy of new anti-angiogenic drugs. However, unavoidable patient motion can decrease the accuracy of subsequent analyses rendering the data unusable. Motion correction algorithms are challenging to develop for contrast-enhanced data since intensity changes due to contrast-enhancement and patient motion must somehow be differentiated from one another. A novel method is presented that employs a motion-informed intensity correction in order to facilitate the registration of contrast enhanced data. The intensity correction simulates the presence or absence of contrast agent in the image volumes to be registered in an attempt to emulate the level of contrast-enhancement present in a single reference image volume.