Voxel Similarity Measures for Automated Image Registration

Size: px
Start display at page:

Download "Voxel Similarity Measures for Automated Image Registration"

Transcription

1 Voxel Similarity Measures for Automated Image Registration Derek LG Hill, Cohn Studholme & David J Hawkes Radiological Sciences, UMDS, Guy's & St Thomas' Hospitals, London SE1 9RT, UK D.Hill@ umds.ac.uk, ABSTRACT We present the concept of the feature space sequence: 2D distributions of voxel features of two images generated at registration and a sequence of misregistrations. We provide an explanation of the structure seen in these images. Feature space sequences have been generated for a pair of MR image volumes identical apart from the addition of Gaussian noise to one, MR image volumes with and without Gadolinium enhancement, MR and PET-FDG image volumes and MR and CT image volumes, all of the head. The structure seen in the feature space sequences was used to devise two new measures of similarity which in turn were used to produce plots of cost versus misregistration for the 6 degrees of freedom of rigid body motion. One of these, the third order moment of the feature space histogram, was used to register the MR image volumes with and without Gadolinium enhancement. These techniques have the potential for registration accuracy to within a small fraction of a voxel or resolution element and therefore interpolation errors in image transformation can be the dominant source of error in subtracted images. We present a method for removing these errors using sinc interpolation and show how interpolation errors can be reduced by over two orders of magnitude. 1. INTRODUCTION Important clinical applications of multimodal imaging have emerged in functional neuro-imaging, surgical planning, tumour localisation and treatment assessment"2. Our objective is the development of clinically useable automatic 3D medical image registration algorithms. The approach to meet this objective which is described in this paper is motivated by the observation that medical images ofthe same region ofa patient acquired with different imaging modalities are usually recognisably similar, even to a non-expert. For example MR and CT images of many parts of the body contain broadly the same features, but they appear with different intensities and texture. Where such images are being combined, image registration can be accomplished by identifying a small number of equivalent features such as points or surfaces in the images16. An alternative approach is to perform registration using all, or at least a large number, of the voxels in the images rather than a small number of derived features. If achievable this might provide methods more robust to noise in the image data and less prone to truncation effects due to partially overlapping imaged volumes. The basis of this approach is the assumption that some arithmetic combination of voxel values in two images, when applied to each image voxel in turn, provides a similarity measure that has an optimum value when the images are aligned. The best known algorithm of this type is cross correlation. We present new work on a methodology for devising improved similarity measures Accurate registration using voxel intensity values The cross correlation of two functions is frequently used in signal processing and image processing as a measure of how well two functions A and B match when they are transformed with respect to each other. The transformation T that provides the best match between the functions has the highest correlation value. Roger Woods7'8 has proposed an algorithm that is related to cross correlation, hut incorporates an important modification. His algorithm is based on an idealised assumption that when two images are accurately aligned the value of any voxel in one image is related to the value of the corresponding voxel in the other image by a multiplicative factor R. In other words, for all voxels of intensity a and b, in images A and B respectively, = R. When A and B are acquired from the same patient using the same modality at different times, there will be a single value of R for all intensity values. When A and B are acquired from the same patient using different modalities, there might be a different value of R for each intensity value in either image. The Woods algorithm has been developed specifically for registration of multiple neurological PET images from the same patient, and for O /94/$6.OO SPIE Vol / 20S

2 registration of neurological PET images to MRI images of the same patient. The latter requires delineation of the outer surface of the brain in MR to exclude all but the intracranial structures. We have previously shown how this technique can be modified in order to automatically register MR and CT images of the head without the need for prior segmentation, provided there is sufficient axial sampling9. The success of these techniques for solving specific registration problems encouraged us to devise a methodology for studying this and other related voxel similarity measures Interpolation schemes for accurate image transformation Registration using voxel similarity measures has the potential for very high registration accuracy due to the noise averaging properties of the algorithms used. Registration accuracy better than image resolution or voxel dimensions is feasible. This is particularly true for intramodality image registration where very high accuracy is required in emerging clinical applications. These include subtraction of pre and post contrast spiral CT images to produce spiral CT angiograms, subtraction of pre and post Gadolinium MR images for improved delineation of enhancing lesions, monitoring disease progression and functional MR imaging, where resting state brain images are subtracted from images acquired whilst a cognitive task is being undertaken. For all these applications it cannot be assumed that the subject remains still between concurrent image acquisitions, nor that the patient can be repositioned accurately for subsequent studies. Small movements (considerably less than a voxel) can lead to edge artefacts in the subtracted images that might be misleading. Therefore accurate image registration is required. If this is achieved conventional image transformation using nearest neighbour or linear interpolation can introduce significant errors. A consideration of sampling theory can be useful in overcoming these interpolation artefacts. The k-space image representation of a correctly sampled band limited function can be recovered from the discrete Fourier transform by multiplication with a cuboidal window. The corresponding spatial domain operation is a convolution with the Fourier transform of the cuboidal window, which is a 3D sinc function. Intermediate sampling points in a band limited image can therefore be recovered in the spatial domain using a sinc function with zero crossing separation equal to the voxel dimensions. Medical images are invariably truncated in all three dimensions, with the consequence that they are not truly band limited. Consequently, sinc interpolation does not correctly transform the images and introduces artefacts such as shading. Extrapolation of the image prior to interpolation can enormously reduce these truncation artefacts. The extrapolation can be done by adding an average or "DC" intensity around the periphery of the image, or by data replication. The number of voxel neighbours used for the interpolation has a significant impact on the time taken to do the transformation. Therefore the sinc function has to be truncated after a certain number of zero crossings in order to transform the images in a reasonable length of time. The function used to truncate the sinc function affects the residual interpolation artefacts. 2. METHODS 2.1. Feature space sequences and their characteristics A possible way of examining the effect of misregistration useful for devising appropriate voxel similarity measures is to construct 2D distributions or feature spaces of image intensities (or other voxel features) of two images at registration and for a sequence of known misregistrations for each degree of freedom. We term these feature space sequences. A feature space of two images is a two dimensional histogram of the value of a numerical image feature from one image plotted against the corresponding value from the other image. The feature space may be constructed from all voxels where there is correspondence (i.e. overlap) between the two images. The image feature is usually voxel intensity, but could also be a derived value such as the output of a gradient operator or texture analysis. A feature space sequence can be calculated for each degree of freedom of the rigid body transformation. It provides a way of studying the change in the appearance of a feature space with misregistration. To understand the structure seen in feature space sequences it is informative to calculate the expected appearance using a simple model. Consider a tomographic slice consisting of two regions, a vertical band of uniform intensity surrounded by a uniform background intensity. The image intensities from image modality A, at,, are given by:- = a2 for u < i v else a1 = a1 and, similarly assuming identical voxel sizes for imaging modality B, image intensities b,1 are given by:- b,1 = b2 for u < i v else = where u and v are the bounds of the vertical band in the x direction and 0 i <Xm(,xand 0 <Ymax 206 / SPIE Vol. 2359

3 A feature space of this image will consist of two signals at coordinates,b1) and (a2,b2) ofintensities equal to the number of voxels within each region with zero elsewhere. With misregistration in the x direction by t pixels two other signals will appear at (a1,b2) and (a2,b1) of intensity ty,,. With the addition of noise in each modality the feature space image will be convolved with the noise distribution functions of image Aand B, A and in the a and b axes respectively. With image blurring, by the impulse response function 'Arof imaging modality A and 'Br imaging modality B, image intensities a, and are given, in the noise free case, by:- a. = IAr1(i_r)j and b11 = IBr1(i_r)j The feature space with image blurring will consist of the original two high signals at (a1,b1) and (a2,b2) connected by an arc with the parametric equation in p for the coordinates of arc (a1, b) given by:- a : ai+ia(r_p)(a2 al) and b = bl+ib(r_p)(b2bl). When the blurring functions are the same in the two modalities this parametric equation describes a straight line with:- (a a1)/(a2 a1) = (b b1)/(b2 b1) forb1 The intensity at each point, p, along this arc in feature space is proportional to the sum of products of the inverse of the impulse response functions ( ('A(.)'B())' + ('A(v-p)'B(v-p)Y1) and the inverse ofthe arc length. At misregistration by a distance t in the x direction the parametric equation of the arc becomes:- a = al+ia(r_p)(a2 al) andb = bl+ib(r_p_l)(b2bl) This equation generates an arc between points (a1,b1) and (a2,b2) similar in shape to the hysteresis curve of magnetisation of ferromagnetic material. The greater the misregistration the greater will be the area enclosed by the curve, until at the limit that IAand I - 0, when the arc will describe the rectangle (a1,b1 ) (a1,b2) (a2,b2) (a2,b1 ). The area between the two curves is a function of misregistration, contrast between the two regions and the impulse response functions of the two imaging modalities. In a real image the feature space will consist of a superimposition of a number of points connected by arcs with intensity proportional to image blurring. Image noise will blur the feature space and misregistration will generate hysteresis type curves from the connecting arcs. The misregistration of each boundary element along its surface normal will determine the area enclosed by the corresponding hysteresis curve in feature space. For convoluted structures surface orientations will be widely distributed and the feature space for a particular misregistration will be a superimposition of multiple hysteresis type curves, but the envelope of these curves will correspond to the maximum misregistration perpendicular to the surface. We have generated feature space sequences from various combinations of MR. CT and PET images in the head. The first feature space was generated from two MR images (Ti gradient echo, voxel size 0.859mm x 0.859mm x 0.9mm) that differ only by the addition of Gaussian noise with a standard deviation similar to that of the air signal. Feature spaces were also generated from images of patients scanned at different times or with different modalities that had been pre-registered using our interactive point landmark system3. These included:- pre and post Gadolinium MR images (Ti weighted, voxel sizes 0.898mm x 0.898mm x 2.0mm pre Gadolinium and 0.898mm x 0.898mm x 1.0mm post Gadolinium); MR images (Ti weighted, voxel size 0.918mm x 0.918mm x 2.5mm) and CT images (voxel size 0.488mm x 0.488mm x 3.0mm); and MR images (TI weighted, voxel size 0.859mm x 0.859mm x 2.5mm) and PET-FDG images (voxel sizes 2.0mm x 2.0mm x 3.75mm); all of the head. SPIE Vol / 207

4 Prior to the calculation of the feature spaces, each pair of images was blurred so that they both have the same resolution. For the work presented here all images were filtered with a Hanning window to give a full width half maximum of the impulse response function of 10mm. This will sharpen structure seen in the feature space sequences by reducing the blurring effect of image noise. In addition filtering to the same resolution in each modality will produce symmetric structure in the feature spaces. The feature spaces shown are produced using translation in units of a single pixel, but we are able to produce these using arbitrary transformations in all six degrees of freedom Similarity measures A number of measures of the feature space dispersion described above could be envisaged. We have examined 4 such meas- ures:- 1. Simple correlation; 2. The variance of intensity ratio of one modality to the second, computed for each intensity of the second and averaged (Woods' algorithm8); 3. As above but restricted to predefined bands of intensity, corresponding to rectangular regions of the feature space; 4. The third order moment of the intensity histogram of the feature space. The third measure will give a low signal at registration for regions of the feature space which include the arcs connecting two adjacent regions in the original images. The hysteresis effects outlined above will increase this variance. The histogram of the feature space contains information about the distribution of feature space intensities. It is weighted towards high values if a large number of feature space pixels have high intensity, and it is weighted towards low values if a large number of feature space pixels have low intensity. The higher order moments of the histogram measure this distribution. We used the third order moment of the feature space histogram, but the choice of this rather than any other moment of order 2 or above was arbitrary. Using registered reference images, the similarity measure is evaluated for the images at registration, and when misregistered by known transformations in each of the degrees of freedom of the desired registration transformation. For rigid body registration, this provides a series of six one dimensional curves, each of which is a plot of similarity measure value against misregistration for a single degree of freedom. We term the resulting graphs similarity measure plots. The similarity measures are converted to cost functions, so an ideal cost function has a minimum value at registration, and is a monotonically increasing function of misregistration. It must be emphasised that these similarity measure plots do not sample all of the parameter space, and there may well be many local minima of cost that are not visible in the similarity measure plots. These techniques have been applied to 2 copies (and therefore registered versions) of the same MR dataset in which one copy had Gaussian noise added (sigma equal to 6 standard deviations of the original MR image noise) and 3 sets of images preregistered using our interactive point based registration technique3 and assessed visually to ensure good registration. These 3 sets were a T1 weighted MR volume with and without Gadolinium enhancement, MR and PET-FDG image volumes and MR and CT image volumes. All images were of the head, but each pair of image volumes was from a different patient Image transformation using sinc interpolation The consequences of interpolation artefacts can easily be demonstrated by transforming an image and then transforming the transformed image back using the inverse transformation, and subtracting the original image from the twice transformed one. This was carried out using an arbitrary transformation through all 6 degrees of freedom using nearest neighbour and trilinear interpolation on an example MR image volume of the head. To demonstrate the reduction in artefact that can be achieved using sinc interpolation, the same double transformation was carried out using a 3D sinc function truncated after 12 zero crossings using a Welch window. The data was extrapolated using periodic replication of the image. In the slice plane, where the image object is surrounded by air, the image was simply replicated repeatedly, so that the pixel ordering becomes:, n i, n (1,2,3 n i, n 11,2,3 In the direction perpendicular to the slice plane, the object is not surrounded by air, so the truncation artefacts are more serious. In this direction, the extrapolated slice ordering was:, 3, 2, lii, 2, 3 n i, n I n, n i 208 / SPIE Vol. 2359

5 3.1. Feature 3. RESULTS Figure 1. A feature space sequence generated from two similar MR images (original intensity on horizontal axis and added noise on vertical axis) registered (left) and translated laterally by 2 pixels (centre), and 5 pixels (right). Figure 2. A feature space sequence generated from pre- (horizontal axis) and post- (vertical axis) Gadolinium MR images, registered (left) and translated laterally by 2 pixels (centre), and 5 pixels (right). The echo times, and hence contrast, were slightly different in the two MR images. A feature space sequence from the two copies of the same MR image volume is shown in Fig. 1.Feature space sequences for the MR image volumes with and without Gadolinium are shown on Fig. 2, for the MR and PET-FDG image volumes on Fig. 3 and for the MR and CT image volumes on Fig. 4. These feature space sequences are of very different overall appearance but they share common characteristics predicted in section 2.1:- 1. Diagonal features, corresponding to blurring between adjacent regions in the image volumes, disperse with misregistration. 2. This dispersion is accompanied by hysteresis type patterns which increase in area as misregistration increases until in the limit horizontal and vertical lines appear. 3. The hysteresis type patterns show clear bounds, presumably caused by the maximum misregistration along surface normals. 4. Except at the origin, the brightest feature space pixels decrease in intensity with misregistration. 5. The number of low intensity feature space pixels increases with misregistration. SPIE Vol / 209

6 Figure 3. A feature space sequence generated from MR (horizontal axis) and PET (vertical axis) images registered (left) and translated laterally by 2 pixels (centre), and 5pixels (right). Figure 4. A feature space sequence generated from CT (horizontal axis) and MR (vertical axis) images registered (left) and translated laterally by 2 pixels (centre), and 5pixels (right). Fig.5 shows the cost function with misregistration generated from the variance of intensity ratios for the PET-FDG and MR image volumes (PET-FDG intensity divided by MR intensity). The functions are reasonably well behaved with monotonically increasing cost with misregistration except for x and z rotations. Also for x and y translations the cost function starts to decrease once misregistration exceeds about 5mm which is probably related to the resolution distance of the image volumes. Fig. 6 shows that, by selecting a predefined band of intensities to perform the analysis, the cost function for x and z rotations are considerably improved. Fig.7 provides similarly well behaved cost functions for the MR and CT image volume, where the vanance of intensity ratios (MR intensity divided by CT intensity) was computed oven a range of intensities corresponding to fat and the skin surface. Fig. 8 shows that, as expected, conventional correlation provides a rather poor basis for computing cost for MR and CT registration. Similar results for correlation were obtained for the PET-FDG and MR image volumes. Fig. 9. shows the intensity regions for the MR PET-FDG and MR CT feature spaces that were used to compute the cost functions in Figs. 6 and 7, respectively, together with labels for different tissues on the MR PET-FDG feature space. Fig. 10 shows the cost functions for the pre and post Gadolinium MR image volumes. These images were acquired according to the normal routine clinical protocol. The patient was removed from the MR scanner for injection of contrast between acquisitions, and the voxel dimensions were different in the two images. These plots demonstrate that the cost increases almost monotonically with misregistration in each of the degrees of freedom. These image volumes have been successfully registered automatically using the third order moment of the feature space histogram as a cost function with optimisation using a genetic algorithm10 with a population size of 100 and 30 generations, at two scales. Fig. 11 shows the subtracted image generated from these registered images. 210/SPIEV0!. 2359

7 a) Ca > misregistration - translation (mm) or rotation (deg.) 10.0 Figure 5. Cost function calculated using the variance of intensity ratios for the MR and PET-FDG image volumes using the full range of intensities. a) > Cl) misregistration- translation (mm) or rotation (deg.) 10.0 Figure 6. Cost function calculated using the variance of intensity ratios for the MR and PET image volumes using a limited range of intensities corresponding to the outer surface of the brain and skull. SPIEVo!. 2359/211

8 U, > o 0 x translate 0 y translate o 0 z translate x rotate y rotate 1 Vzrotate C-) misregistration - translation (mm) or rotation (deg.) Figure7. Cost functions calculated using the variance of intensity ratios, for the MR and CTimage volumes using a limited range of low CT intensities corresponding to the skin surface and fat. a, CU > C 0 CU U, (3 o o x translate 0 C y translate 0 0 z translate x rotate y rotate V V z rotate misregistration - translation (mm) or rotation (deg.) 10.0 Figure 8. Correlation of MR and CT image volumes as a function of misregistration. Cost in this case will be the inverse of the correlation value. 212 ISPIE Vol. 2359

9 Skull Air Figure 9. Intensity ranges defined for computation of variance of intensity ratios for the MR PET-FDG feature space (left) and the MR CT Feature space (right). Different tissue regions are labelled on the MR PET-FDG feature space. a) > U) misregistration - translation (mm) or rotation (deg.) Figure 10. Cost function calculated using the 3rd order moment of the feature space histogram, for two T1 weighted MR images (pre and post Gadolinium) Image transformation using sine interpolation Fig. 12 shows a slice from the original image volume, and example slices from the difference images produced by subtracting the original image from the twice transformed one using nearest neighbour, linear and sinc interpolation. The intensity range in the original image is The subtracted images are displayed with an intensity range of -25 to +25. The mean square difference between the original image and the twice transformed image provides a quantitative indication of the interpolation errors clearly visible in this figuee. These values are given in table 1. SPIEVo!. 2359/213

10 Figure 11. Single slice from an image volume produced by subtracting the pre and post Gadolinium MR volume after registration and transformation. Table 1: Mean square difference values for interpolation artifacts. Interpolation algorithm Mean square difference Nearest Neighbour Trilinear (8 neighbour) 96.4 Sinc interpolation (12 zero crossing) DISCUSSION We have presented the concept of a feature space sequence, a series of 2D distributions of voxel features from two image volumes generated at registration and known misregistrations. We have analysed the type of structure visible in these feature spaces and described how this analysis might help to devise appropriate measures of misregistration, or similarity measures, which might be used in an optimisation process for automatic registration. Cost has been computed in 4 examples pairs of image volumes using 4 such measures: image correlation; the variance of intensity ratios of one modality to the other averaged over each intensity of the latter; the same restricted to certain bands of image intensity; and the third order moment of the feature space histogram. As expected correlation of intensity values is of little use in multi-modal registration while the last two measures show promise at providing an appropriate cost function for automated registration. The third order moment of the feature space histogram was successfully used for the automatic registration of MR images pre and post injection of Gadolinium. This is an important application of image registration, as subtraction of these images can provide important clinical informatio&1, and patients normally move between the acquisition of these sequences. It is an example of a large class of image registration problems, in which a time series of images need to be compared. Obvious examples include monitoring disease progression (e.g.: plaque volume in patients with multiple sclerosis), and correcting for movement artifacts in functional MR'2. In these applications it is essential to use accurate interpolation in the transformation process. We have shown how sinc interpolation can be used to compute accurate, if computationally expensive, image transformations once the correct transformation has been computed. Work is in progress to test the registration process on a large number of registered datasets from a wide range of modalities in different parts of the body. 214/SPIEV0!. 2359

11 Figure 12. Interpolation artefacts generated by transforming the image volume, transforming it back to its original position, and subtracting the original image from the twice transformed one. Top left: slice from original image. Bottom left: nearest neighbour interpolation. Top right: linear interpolation. Bottom right: interpolation using a sinc function truncated after 12 zero crossings using a Welch window. The diagonal features in the feature space images provide strong cues for registration. These cues only exist when registration of surface features is within the resolution distance of each image. For misregistrations beyond this distance there is no cue on the degree of misregistration and therefore, unlike methods based on the chamfer distance5'6, there is no indication of the appropriate direction of search. The solution might be to blur images to a greater extent when starting estimates are poor and progressively sharpening the images as the registration process proceeds. We are investigating such a multiscale approach. Alternatively, first and higher order gradient operators might also provide suitable features from which to construct feature spaces sequences and devise appropriate similarity measures. Such operators might be tuned to particular types of features SPIE Vol. 2359/215

12 such as ridges or troughs as proposed by Van den Elsen13 or the multiscale medial axis transform as proposed by Morse et al14 Truncation and only partially overlapping datasets cause difficulties for all automated registration algorithms. The feature space approach makes use of, in principle, all corresponding (overlapping) voxels in the two datasets and will therefore be less prone to truncation effects unless starting estimates are so poor that edge artefacts interfere with the generation of sufficiently coarse scale images. Our proposed techniques are generic in that they can be applied to many different combinations of multimodal 3D images in a wide range of applications. More work is necessary to test appropriate similarity measures and to devise algorithms which efficiently converge to the global minimum in the 6 dimensional search space of the rigid body transformation. This technique will also find applications in image guided intervention where interventional imaging devices such as ultrasound and MR could be used to update the pre-operative representation. We are exploring ways in which the matching process might be extended to provide multiple local matches in order to track deformations of soft tissue structures. 5. ACKNOWLEDGEMENTS This work was funded by UK EPSRC Grant GR/H We are grateful for the support and encouragement of our clinical colleagues in this work, in particular Mr. Michael Gleeson (ENT Surgeon), Mr. Anthony Strong (Neurosurgeon), Dr. Tim Cox (Neuroradiologist), Dr. Wai-Lup Wong (Radiologist)and Dr. Alan Colchester (Neurologist), and for the technical assistance of the Radiographic staff of Guy's, St. Thomas', The Maudsley and St. George's Hospitals in London. 6. REFERENCES 1. Hill DLG, Hawkes DJ, Gleeson MJ, Cox TCS, Strong AJ, Wong W-L, Ruff CF, Kitchen ND, Thomas DGT, Sofat A, Crossman JE, Studholme C, Gandhe A, Green SEM, Robinson GP. "Accurate frameless registration ofmr and CT images of the head: Applications in Planning Surgery and Radiotherapy". Radiology, 191 ; , Evans AC, Marrett S, Torrescorzo J, Ku 5, Collins L. "MRI-PET Correlation in Three Dimensions Using a Volume-of- Interest (VOI) Atlas". J Cereb Blood Flow Metab; 1 1 :A69-A Hill DLG, Hawkes DJ, Crossman JE, Gleeson MJ, Cox TCS, Bracey EECML, Strong AJ, Graves P. "Registration of MR and CT images for skull base surgery using point-like anatomical features." Br J Radiology. 64: Pelizzari CA, Chen GTY, Spelbring DR, Weichselbaum RR, Chen C-T. "Accurate three dimensional registration of CT, PET and/or MR images of the brain". J Comput Assist Tomogr 13: Jiang H, Robb RA, Holton KS. "New approach to 3-D registration of multimodality medical images by surface matching". In: Visualisation in Biomedical Computing, Proc Soc Photo-opt Instrum Eng 1808: Hill DLG, Hawkes DJ. "Medical image registration using knowledge of adjacency of anatomical structures". Image and Vision Computing 12: Woods RP, Cherry SR. Mazziotta JC. "A rapid automated algorithm for accurately aligning and reslicing PET images." J Comp Assis Tomogr 16: Woods RP, Mazziotta JC, Cherry SR. "MRI-PET registration with automated algorithm". J Comp Assis Tomogr 17: Hill DLG, Hawkes DJ, Harrison N, Ruff CF. "A strategy for automated multimodality registration incorporating anatomical knowledge and imager characteristics." In: Barrett HH, Gmitro AF, eds. Information Processing in Medical Imaging IPMI '93. Lecture Notes in Computer Science 687 Springer-Verlag, Berlin. ppl Goldberg DE. "Genetic algorithms in search optimisation and machine learning." Addison Wesley, Mass. USA Lloyd GAS, Barker PG, Phelps PD. "Subtraction gadolinium enhanced magnetic resonance for head and neck imaging". Br J Radiology 66: Hajnal JV, Myers R, Oatridge A, Schwieso JE, Young IR, Bydder GM. "Artifacts due to stimulus correlated motion in functional imaging of the brain". Mag. Reson. Med. 31: Van den Elsen PA. "Multimodality matching of brain images". Utrecht University Ph.D. Thesis Morse BS, Pizer SM, Liu A, "Multiscale medial analysis of medical images", In: Barrett HH, Gmitro AF, eds. Information Processing in Medical Imaging IPMI '93. Lecture Notes in Computer Science 687 Springer-Verlag, Berlin. ppll2-l3l ISPIE Vol. 2359

Medical Image Registration Using Knowledge of Adjacency of Anatomical Structures.

Medical Image Registration Using Knowledge of Adjacency of Anatomical Structures. Medical Image Registration Using Knowledge of Adjacency of Anatomical Structures. Derek LG Hill and David J Hawkes Image Processing Group, Radiological Sciences, UMDS, Guy's and St Thomas's Hospitals,

More information

Key words: automated registration, voxel similarity, multiresolution optimization, magnetic resonance, positron emission tomography

Key words: automated registration, voxel similarity, multiresolution optimization, magnetic resonance, positron emission tomography Automated three-dimensional registration of magnetic resonance and positron emission tomography brain images by multiresolution optimization of voxel similarity measures Colin Studholme, a) Derek L. G.

More information

Image Registration. Prof. Dr. Lucas Ferrari de Oliveira UFPR Informatics Department

Image Registration. Prof. Dr. Lucas Ferrari de Oliveira UFPR Informatics Department Image Registration Prof. Dr. Lucas Ferrari de Oliveira UFPR Informatics Department Introduction Visualize objects inside the human body Advances in CS methods to diagnosis, treatment planning and medical

More information

A Model-Independent, Multi-Image Approach to MR Inhomogeneity Correction

A Model-Independent, Multi-Image Approach to MR Inhomogeneity Correction Tina Memo No. 2007-003 Published in Proc. MIUA 2007 A Model-Independent, Multi-Image Approach to MR Inhomogeneity Correction P. A. Bromiley and N.A. Thacker Last updated 13 / 4 / 2007 Imaging Science and

More information

Annales UMCS Informatica AI 1 (2003) UMCS. Registration of CT and MRI brain images. Karol Kuczyński, Paweł Mikołajczak

Annales UMCS Informatica AI 1 (2003) UMCS. Registration of CT and MRI brain images. Karol Kuczyński, Paweł Mikołajczak Annales Informatica AI 1 (2003) 149-156 Registration of CT and MRI brain images Karol Kuczyński, Paweł Mikołajczak Annales Informatica Lublin-Polonia Sectio AI http://www.annales.umcs.lublin.pl/ Laboratory

More information

Intraoperative Prostate Tracking with Slice-to-Volume Registration in MR

Intraoperative Prostate Tracking with Slice-to-Volume Registration in MR Intraoperative Prostate Tracking with Slice-to-Volume Registration in MR Sean Gill a, Purang Abolmaesumi a,b, Siddharth Vikal a, Parvin Mousavi a and Gabor Fichtinger a,b,* (a) School of Computing, Queen

More information

Locating Motion Artifacts in Parametric fmri Analysis

Locating Motion Artifacts in Parametric fmri Analysis Tina Memo No. 200-002 Presented at MICCAI 999 Locating Motion Artifacts in Parametric fmri Analysis A.J.Lacey, N.A.Thacker, E. Burton, and A.Jackson Last updated 2 / 02 / 2002 Imaging Science and Biomedical

More information

Ensemble registration: Combining groupwise registration and segmentation

Ensemble registration: Combining groupwise registration and segmentation PURWANI, COOTES, TWINING: ENSEMBLE REGISTRATION 1 Ensemble registration: Combining groupwise registration and segmentation Sri Purwani 1,2 sri.purwani@postgrad.manchester.ac.uk Tim Cootes 1 t.cootes@manchester.ac.uk

More information

Image Segmentation and Registration

Image Segmentation and Registration Image Segmentation and Registration Dr. Christine Tanner (tanner@vision.ee.ethz.ch) Computer Vision Laboratory, ETH Zürich Dr. Verena Kaynig, Machine Learning Laboratory, ETH Zürich Outline Segmentation

More information

Functional MRI data preprocessing. Cyril Pernet, PhD

Functional MRI data preprocessing. Cyril Pernet, PhD Functional MRI data preprocessing Cyril Pernet, PhD Data have been acquired, what s s next? time No matter the design, multiple volumes (made from multiple slices) have been acquired in time. Before getting

More information

Abstract. 1. Introduction

Abstract. 1. Introduction A New Automated Method for Three- Dimensional Registration of Medical Images* P. Kotsas, M. Strintzis, D.W. Piraino Department of Electrical and Computer Engineering, Aristotelian University, 54006 Thessaloniki,

More information

Computational Neuroanatomy

Computational Neuroanatomy Computational Neuroanatomy John Ashburner john@fil.ion.ucl.ac.uk Smoothing Motion Correction Between Modality Co-registration Spatial Normalisation Segmentation Morphometry Overview fmri time-series kernel

More information

2 Michael E. Leventon and Sarah F. F. Gibson a b c d Fig. 1. (a, b) Two MR scans of a person's knee. Both images have high resolution in-plane, but ha

2 Michael E. Leventon and Sarah F. F. Gibson a b c d Fig. 1. (a, b) Two MR scans of a person's knee. Both images have high resolution in-plane, but ha Model Generation from Multiple Volumes using Constrained Elastic SurfaceNets Michael E. Leventon and Sarah F. F. Gibson 1 MIT Artificial Intelligence Laboratory, Cambridge, MA 02139, USA leventon@ai.mit.edu

More information

Medicale Image Analysis

Medicale Image Analysis Medicale Image Analysis Registration Validation Prof. Dr. Philippe Cattin MIAC, University of Basel Prof. Dr. Philippe Cattin: Registration Validation Contents 1 Validation 1.1 Validation of Registration

More information

Functional MRI in Clinical Research and Practice Preprocessing

Functional MRI in Clinical Research and Practice Preprocessing Functional MRI in Clinical Research and Practice Preprocessing fmri Preprocessing Slice timing correction Geometric distortion correction Head motion correction Temporal filtering Intensity normalization

More information

Image Registration I

Image Registration I Image Registration I Comp 254 Spring 2002 Guido Gerig Image Registration: Motivation Motivation for Image Registration Combine images from different modalities (multi-modality registration), e.g. CT&MRI,

More information

EE795: Computer Vision and Intelligent Systems

EE795: Computer Vision and Intelligent Systems EE795: Computer Vision and Intelligent Systems Spring 2012 TTh 17:30-18:45 WRI C225 Lecture 04 130131 http://www.ee.unlv.edu/~b1morris/ecg795/ 2 Outline Review Histogram Equalization Image Filtering Linear

More information

Super-resolution Reconstruction of Fetal Brain MRI

Super-resolution Reconstruction of Fetal Brain MRI Super-resolution Reconstruction of Fetal Brain MRI Ali Gholipour and Simon K. Warfield Computational Radiology Laboratory Children s Hospital Boston, Harvard Medical School Worshop on Image Analysis for

More information

Automatic Subthalamic Nucleus Targeting for Deep Brain Stimulation. A Validation Study

Automatic Subthalamic Nucleus Targeting for Deep Brain Stimulation. A Validation Study Automatic Subthalamic Nucleus Targeting for Deep Brain Stimulation. A Validation Study F. Javier Sánchez Castro a, Claudio Pollo a,b, Jean-Guy Villemure b, Jean-Philippe Thiran a a École Polytechnique

More information

SPM8 for Basic and Clinical Investigators. Preprocessing. fmri Preprocessing

SPM8 for Basic and Clinical Investigators. Preprocessing. fmri Preprocessing SPM8 for Basic and Clinical Investigators Preprocessing fmri Preprocessing Slice timing correction Geometric distortion correction Head motion correction Temporal filtering Intensity normalization Spatial

More information

MEDICAL IMAGE ANALYSIS

MEDICAL IMAGE ANALYSIS SECOND EDITION MEDICAL IMAGE ANALYSIS ATAM P. DHAWAN g, A B IEEE Engineering in Medicine and Biology Society, Sponsor IEEE Press Series in Biomedical Engineering Metin Akay, Series Editor +IEEE IEEE PRESS

More information

New Technology Allows Multiple Image Contrasts in a Single Scan

New Technology Allows Multiple Image Contrasts in a Single Scan These images were acquired with an investigational device. PD T2 T2 FLAIR T1 MAP T1 FLAIR PSIR T1 New Technology Allows Multiple Image Contrasts in a Single Scan MR exams can be time consuming. A typical

More information

EPI Data Are Acquired Serially. EPI Data Are Acquired Serially 10/23/2011. Functional Connectivity Preprocessing. fmri Preprocessing

EPI Data Are Acquired Serially. EPI Data Are Acquired Serially 10/23/2011. Functional Connectivity Preprocessing. fmri Preprocessing Functional Connectivity Preprocessing Geometric distortion Head motion Geometric distortion Head motion EPI Data Are Acquired Serially EPI Data Are Acquired Serially descending 1 EPI Data Are Acquired

More information

Methodological progress in image registration for ventilation estimation, segmentation propagation and multi-modal fusion

Methodological progress in image registration for ventilation estimation, segmentation propagation and multi-modal fusion Methodological progress in image registration for ventilation estimation, segmentation propagation and multi-modal fusion Mattias P. Heinrich Julia A. Schnabel, Mark Jenkinson, Sir Michael Brady 2 Clinical

More information

Introduction to fmri. Pre-processing

Introduction to fmri. Pre-processing Introduction to fmri Pre-processing Tibor Auer Department of Psychology Research Fellow in MRI Data Types Anatomical data: T 1 -weighted, 3D, 1/subject or session - (ME)MPRAGE/FLASH sequence, undistorted

More information

Basic fmri Design and Analysis. Preprocessing

Basic fmri Design and Analysis. Preprocessing Basic fmri Design and Analysis Preprocessing fmri Preprocessing Slice timing correction Geometric distortion correction Head motion correction Temporal filtering Intensity normalization Spatial filtering

More information

Whole Body MRI Intensity Standardization

Whole Body MRI Intensity Standardization Whole Body MRI Intensity Standardization Florian Jäger 1, László Nyúl 1, Bernd Frericks 2, Frank Wacker 2 and Joachim Hornegger 1 1 Institute of Pattern Recognition, University of Erlangen, {jaeger,nyul,hornegger}@informatik.uni-erlangen.de

More information

Image Acquisition Systems

Image Acquisition Systems Image Acquisition Systems Goals and Terminology Conventional Radiography Axial Tomography Computer Axial Tomography (CAT) Magnetic Resonance Imaging (MRI) PET, SPECT Ultrasound Microscopy Imaging ITCS

More information

2D Rigid Registration of MR Scans using the 1d Binary Projections

2D Rigid Registration of MR Scans using the 1d Binary Projections 2D Rigid Registration of MR Scans using the 1d Binary Projections Panos D. Kotsas Abstract This paper presents the application of a signal intensity independent registration criterion for 2D rigid body

More information

Basic principles of MR image analysis. Basic principles of MR image analysis. Basic principles of MR image analysis

Basic principles of MR image analysis. Basic principles of MR image analysis. Basic principles of MR image analysis Basic principles of MR image analysis Basic principles of MR image analysis Julien Milles Leiden University Medical Center Terminology of fmri Brain extraction Registration Linear registration Non-linear

More information

ANALYSIS OF PULMONARY FIBROSIS IN MRI, USING AN ELASTIC REGISTRATION TECHNIQUE IN A MODEL OF FIBROSIS: Scleroderma

ANALYSIS OF PULMONARY FIBROSIS IN MRI, USING AN ELASTIC REGISTRATION TECHNIQUE IN A MODEL OF FIBROSIS: Scleroderma ANALYSIS OF PULMONARY FIBROSIS IN MRI, USING AN ELASTIC REGISTRATION TECHNIQUE IN A MODEL OF FIBROSIS: Scleroderma ORAL DEFENSE 8 th of September 2017 Charlotte MARTIN Supervisor: Pr. MP REVEL M2 Bio Medical

More information

UvA-DARE (Digital Academic Repository) Motion compensation for 4D PET/CT Kruis, M.F. Link to publication

UvA-DARE (Digital Academic Repository) Motion compensation for 4D PET/CT Kruis, M.F. Link to publication UvA-DARE (Digital Academic Repository) Motion compensation for 4D PET/CT Kruis, M.F. Link to publication Citation for published version (APA): Kruis, M. F. (2014). Motion compensation for 4D PET/CT General

More information

Chapter 3 Set Redundancy in Magnetic Resonance Brain Images

Chapter 3 Set Redundancy in Magnetic Resonance Brain Images 16 Chapter 3 Set Redundancy in Magnetic Resonance Brain Images 3.1 MRI (magnetic resonance imaging) MRI is a technique of measuring physical structure within the human anatomy. Our proposed research focuses

More information

Performance Evaluation of the TINA Medical Image Segmentation Algorithm on Brainweb Simulated Images

Performance Evaluation of the TINA Medical Image Segmentation Algorithm on Brainweb Simulated Images Tina Memo No. 2008-003 Internal Memo Performance Evaluation of the TINA Medical Image Segmentation Algorithm on Brainweb Simulated Images P. A. Bromiley Last updated 20 / 12 / 2007 Imaging Science and

More information

Non-rigid Image Registration

Non-rigid Image Registration Overview Non-rigid Image Registration Introduction to image registration - he goal of image registration - Motivation for medical image registration - Classification of image registration - Nonrigid registration

More information

Automatic Vascular Tree Formation Using the Mahalanobis Distance

Automatic Vascular Tree Formation Using the Mahalanobis Distance Automatic Vascular Tree Formation Using the Mahalanobis Distance Julien Jomier, Vincent LeDigarcher, and Stephen R. Aylward Computer-Aided Diagnosis and Display Lab, Department of Radiology The University

More information

Spatio-Temporal Registration of Biomedical Images by Computational Methods

Spatio-Temporal Registration of Biomedical Images by Computational Methods Spatio-Temporal Registration of Biomedical Images by Computational Methods Francisco P. M. Oliveira, João Manuel R. S. Tavares tavares@fe.up.pt, www.fe.up.pt/~tavares Outline 1. Introduction 2. Spatial

More information

SPM8 for Basic and Clinical Investigators. Preprocessing

SPM8 for Basic and Clinical Investigators. Preprocessing SPM8 for Basic and Clinical Investigators Preprocessing fmri Preprocessing Slice timing correction Geometric distortion correction Head motion correction Temporal filtering Intensity normalization Spatial

More information

Methods for data preprocessing

Methods for data preprocessing Methods for data preprocessing John Ashburner Wellcome Trust Centre for Neuroimaging, 12 Queen Square, London, UK. Overview Voxel-Based Morphometry Morphometry in general Volumetrics VBM preprocessing

More information

Image Sampling & Quantisation

Image Sampling & Quantisation Image Sampling & Quantisation Biomedical Image Analysis Prof. Dr. Philippe Cattin MIAC, University of Basel Contents 1 Motivation 2 Sampling Introduction and Motivation Sampling Example Quantisation Example

More information

1 Introduction Motivation and Aims Functional Imaging Computational Neuroanatomy... 12

1 Introduction Motivation and Aims Functional Imaging Computational Neuroanatomy... 12 Contents 1 Introduction 10 1.1 Motivation and Aims....... 10 1.1.1 Functional Imaging.... 10 1.1.2 Computational Neuroanatomy... 12 1.2 Overview of Chapters... 14 2 Rigid Body Registration 18 2.1 Introduction.....

More information

Supplementary methods

Supplementary methods Supplementary methods This section provides additional technical details on the sample, the applied imaging and analysis steps and methods. Structural imaging Trained radiographers placed all participants

More information

An Automated Image-based Method for Multi-Leaf Collimator Positioning Verification in Intensity Modulated Radiation Therapy

An Automated Image-based Method for Multi-Leaf Collimator Positioning Verification in Intensity Modulated Radiation Therapy An Automated Image-based Method for Multi-Leaf Collimator Positioning Verification in Intensity Modulated Radiation Therapy Chenyang Xu 1, Siemens Corporate Research, Inc., Princeton, NJ, USA Xiaolei Huang,

More information

Medical Image Analysis

Medical Image Analysis Computer assisted Image Analysis VT04 29 april 2004 Medical Image Analysis Lecture 10 (part 1) Xavier Tizon Medical Image Processing Medical imaging modalities XRay,, CT Ultrasound MRI PET, SPECT Generic

More information

HST.582J / 6.555J / J Biomedical Signal and Image Processing Spring 2007

HST.582J / 6.555J / J Biomedical Signal and Image Processing Spring 2007 MIT OpenCourseWare http://ocw.mit.edu HST.582J / 6.555J / 16.456J Biomedical Signal and Image Processing Spring 2007 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.

More information

3D Voxel-Based Volumetric Image Registration with Volume-View Guidance

3D Voxel-Based Volumetric Image Registration with Volume-View Guidance 3D Voxel-Based Volumetric Image Registration with Volume-View Guidance Guang Li*, Huchen Xie, Holly Ning, Deborah Citrin, Jacek Copala, Barbara Arora, Norman Coleman, Kevin Camphausen, and Robert Miller

More information

Medical Image Registration

Medical Image Registration Medical Image Registration Submitted by NAREN BALRAJ SINGH SB ID# 105299299 Introduction Medical images are increasingly being used within healthcare for diagnosis, planning treatment, guiding treatment

More information

Image Sampling and Quantisation

Image Sampling and Quantisation Image Sampling and Quantisation Introduction to Signal and Image Processing Prof. Dr. Philippe Cattin MIAC, University of Basel 1 of 46 22.02.2016 09:17 Contents Contents 1 Motivation 2 Sampling Introduction

More information

2D-3D Registration using Gradient-based MI for Image Guided Surgery Systems

2D-3D Registration using Gradient-based MI for Image Guided Surgery Systems 2D-3D Registration using Gradient-based MI for Image Guided Surgery Systems Yeny Yim 1*, Xuanyi Chen 1, Mike Wakid 1, Steve Bielamowicz 2, James Hahn 1 1 Department of Computer Science, The George Washington

More information

Multi-modal Image Registration Using the Generalized Survival Exponential Entropy

Multi-modal Image Registration Using the Generalized Survival Exponential Entropy Multi-modal Image Registration Using the Generalized Survival Exponential Entropy Shu Liao and Albert C.S. Chung Lo Kwee-Seong Medical Image Analysis Laboratory, Department of Computer Science and Engineering,

More information

Classification of Subject Motion for Improved Reconstruction of Dynamic Magnetic Resonance Imaging

Classification of Subject Motion for Improved Reconstruction of Dynamic Magnetic Resonance Imaging 1 CS 9 Final Project Classification of Subject Motion for Improved Reconstruction of Dynamic Magnetic Resonance Imaging Feiyu Chen Department of Electrical Engineering ABSTRACT Subject motion is a significant

More information

FMRI Pre-Processing and Model- Based Statistics

FMRI Pre-Processing and Model- Based Statistics FMRI Pre-Processing and Model- Based Statistics Brief intro to FMRI experiments and analysis FMRI pre-stats image processing Simple Single-Subject Statistics Multi-Level FMRI Analysis Advanced FMRI Analysis

More information

NIH Public Access Author Manuscript Proc Soc Photo Opt Instrum Eng. Author manuscript; available in PMC 2014 October 07.

NIH Public Access Author Manuscript Proc Soc Photo Opt Instrum Eng. Author manuscript; available in PMC 2014 October 07. NIH Public Access Author Manuscript Published in final edited form as: Proc Soc Photo Opt Instrum Eng. 2014 March 21; 9034: 903442. doi:10.1117/12.2042915. MRI Brain Tumor Segmentation and Necrosis Detection

More information

Deformable Registration Using Scale Space Keypoints

Deformable Registration Using Scale Space Keypoints Deformable Registration Using Scale Space Keypoints Mehdi Moradi a, Purang Abolmaesoumi a,b and Parvin Mousavi a a School of Computing, Queen s University, Kingston, Ontario, Canada K7L 3N6; b Department

More information

Fmri Spatial Processing

Fmri Spatial Processing Educational Course: Fmri Spatial Processing Ray Razlighi Jun. 8, 2014 Spatial Processing Spatial Re-alignment Geometric distortion correction Spatial Normalization Smoothing Why, When, How, Which Why is

More information

Robust Realignment of fmri Time Series Data

Robust Realignment of fmri Time Series Data Robust Realignment of fmri Time Series Data Ben Dodson bjdodson@stanford.edu Olafur Gudmundsson olafurg@stanford.edu December 12, 2008 Abstract FMRI data has become an increasingly popular source for exploring

More information

Hybrid Spline-based Multimodal Registration using a Local Measure for Mutual Information

Hybrid Spline-based Multimodal Registration using a Local Measure for Mutual Information Hybrid Spline-based Multimodal Registration using a Local Measure for Mutual Information Andreas Biesdorf 1, Stefan Wörz 1, Hans-Jürgen Kaiser 2, Karl Rohr 1 1 University of Heidelberg, BIOQUANT, IPMB,

More information

Computational Medical Imaging Analysis Chapter 4: Image Visualization

Computational Medical Imaging Analysis Chapter 4: Image Visualization Computational Medical Imaging Analysis Chapter 4: Image Visualization Jun Zhang Laboratory for Computational Medical Imaging & Data Analysis Department of Computer Science University of Kentucky Lexington,

More information

Digital Volume Correlation for Materials Characterization

Digital Volume Correlation for Materials Characterization 19 th World Conference on Non-Destructive Testing 2016 Digital Volume Correlation for Materials Characterization Enrico QUINTANA, Phillip REU, Edward JIMENEZ, Kyle THOMPSON, Sharlotte KRAMER Sandia National

More information

An Iterative Approach for Reconstruction of Arbitrary Sparsely Sampled Magnetic Resonance Images

An Iterative Approach for Reconstruction of Arbitrary Sparsely Sampled Magnetic Resonance Images An Iterative Approach for Reconstruction of Arbitrary Sparsely Sampled Magnetic Resonance Images Hamed Pirsiavash¹, Mohammad Soleymani², Gholam-Ali Hossein-Zadeh³ ¹Department of electrical engineering,

More information

Image Registration + Other Stuff

Image Registration + Other Stuff Image Registration + Other Stuff John Ashburner Pre-processing Overview fmri time-series Motion Correct Anatomical MRI Coregister m11 m 21 m 31 m12 m13 m14 m 22 m 23 m 24 m 32 m 33 m 34 1 Template Estimate

More information

Slide 1. Technical Aspects of Quality Control in Magnetic Resonance Imaging. Slide 2. Annual Compliance Testing. of MRI Systems.

Slide 1. Technical Aspects of Quality Control in Magnetic Resonance Imaging. Slide 2. Annual Compliance Testing. of MRI Systems. Slide 1 Technical Aspects of Quality Control in Magnetic Resonance Imaging Slide 2 Compliance Testing of MRI Systems, Ph.D. Department of Radiology Henry Ford Hospital, Detroit, MI Slide 3 Compliance Testing

More information

doi: /

doi: / Yiting Xie ; Anthony P. Reeves; Single 3D cell segmentation from optical CT microscope images. Proc. SPIE 934, Medical Imaging 214: Image Processing, 9343B (March 21, 214); doi:1.1117/12.243852. (214)

More information

MR IMAGE SEGMENTATION

MR IMAGE SEGMENTATION MR IMAGE SEGMENTATION Prepared by : Monil Shah What is Segmentation? Partitioning a region or regions of interest in images such that each region corresponds to one or more anatomic structures Classification

More information

Lucy Phantom MR Grid Evaluation

Lucy Phantom MR Grid Evaluation Lucy Phantom MR Grid Evaluation Anil Sethi, PhD Loyola University Medical Center, Maywood, IL 60153 November 2015 I. Introduction: The MR distortion grid, used as an insert with Lucy 3D QA phantom, is

More information

TUMOR DETECTION IN MRI IMAGES

TUMOR DETECTION IN MRI IMAGES TUMOR DETECTION IN MRI IMAGES Prof. Pravin P. Adivarekar, 2 Priyanka P. Khatate, 3 Punam N. Pawar Prof. Pravin P. Adivarekar, 2 Priyanka P. Khatate, 3 Punam N. Pawar Asst. Professor, 2,3 BE Student,,2,3

More information

Deviceless respiratory motion correction in PET imaging exploring the potential of novel data driven strategies

Deviceless respiratory motion correction in PET imaging exploring the potential of novel data driven strategies g Deviceless respiratory motion correction in PET imaging exploring the potential of novel data driven strategies Presented by Adam Kesner, Ph.D., DABR Assistant Professor, Division of Radiological Sciences,

More information

Atlas Based Segmentation of the prostate in MR images

Atlas Based Segmentation of the prostate in MR images Atlas Based Segmentation of the prostate in MR images Albert Gubern-Merida and Robert Marti Universitat de Girona, Computer Vision and Robotics Group, Girona, Spain {agubern,marly}@eia.udg.edu Abstract.

More information

Biomedical Image Analysis based on Computational Registration Methods. João Manuel R. S. Tavares

Biomedical Image Analysis based on Computational Registration Methods. João Manuel R. S. Tavares Biomedical Image Analysis based on Computational Registration Methods João Manuel R. S. Tavares tavares@fe.up.pt, www.fe.up.pt/~tavares Outline 1. Introduction 2. Methods a) Spatial Registration of (2D

More information

Obtaining Feature Correspondences

Obtaining Feature Correspondences Obtaining Feature Correspondences Neill Campbell May 9, 2008 A state-of-the-art system for finding objects in images has recently been developed by David Lowe. The algorithm is termed the Scale-Invariant

More information

Clinical Importance. Aortic Stenosis. Aortic Regurgitation. Ultrasound vs. MRI. Carotid Artery Stenosis

Clinical Importance. Aortic Stenosis. Aortic Regurgitation. Ultrasound vs. MRI. Carotid Artery Stenosis Clinical Importance Rapid cardiovascular flow quantitation using sliceselective Fourier velocity encoding with spiral readouts Valve disease affects 10% of patients with heart disease in the U.S. Most

More information

3D Registration based on Normalized Mutual Information

3D Registration based on Normalized Mutual Information 3D Registration based on Normalized Mutual Information Performance of CPU vs. GPU Implementation Florian Jung, Stefan Wesarg Interactive Graphics Systems Group (GRIS), TU Darmstadt, Germany stefan.wesarg@gris.tu-darmstadt.de

More information

3-D Compounding of B-Scan Ultrasound Images

3-D Compounding of B-Scan Ultrasound Images 3-D Compounding of B-Scan Ultrasound Images Jochen F. Krücker, Charles R. Meyer, Theresa A. Tuthill, Gerald L. LeCarpentier, J. Brian Fowlkes, Paul L. Carson University of Michigan, Dept. of Radiology,

More information

Segmentation and Modeling of the Spinal Cord for Reality-based Surgical Simulator

Segmentation and Modeling of the Spinal Cord for Reality-based Surgical Simulator Segmentation and Modeling of the Spinal Cord for Reality-based Surgical Simulator Li X.C.,, Chui C. K.,, and Ong S. H.,* Dept. of Electrical and Computer Engineering Dept. of Mechanical Engineering, National

More information

Métodos de fusão e co-registo de imagem

Métodos de fusão e co-registo de imagem LICENCIATURA EM ENGENHARIA BIOMÉDICA Fundamentos de Imagem Diagnóstica e Terapeutica Métodos de fusão e co-registo de imagem Jorge Isidoro Sumário Introdução à imagem médica Registo de imagens Metodologia

More information

Biomedical Imaging Registration Trends and Applications. Francisco P. M. Oliveira, João Manuel R. S. Tavares

Biomedical Imaging Registration Trends and Applications. Francisco P. M. Oliveira, João Manuel R. S. Tavares Biomedical Imaging Registration Trends and Applications Francisco P. M. Oliveira, João Manuel R. S. Tavares tavares@fe.up.pt, www.fe.up.pt/~tavares Outline 1. Introduction 2. Spatial Registration of (2D

More information

Good Morning! Thank you for joining us

Good Morning! Thank you for joining us Good Morning! Thank you for joining us Deformable Registration, Contour Propagation and Dose Mapping: 101 and 201 Marc Kessler, PhD, FAAPM The University of Michigan Conflict of Interest I receive direct

More information

Computer-Aided Diagnosis in Abdominal and Cardiac Radiology Using Neural Networks

Computer-Aided Diagnosis in Abdominal and Cardiac Radiology Using Neural Networks Computer-Aided Diagnosis in Abdominal and Cardiac Radiology Using Neural Networks Du-Yih Tsai, Masaru Sekiya and Yongbum Lee Department of Radiological Technology, School of Health Sciences, Faculty of

More information

White Pixel Artifact. Caused by a noise spike during acquisition Spike in K-space <--> sinusoid in image space

White Pixel Artifact. Caused by a noise spike during acquisition Spike in K-space <--> sinusoid in image space White Pixel Artifact Caused by a noise spike during acquisition Spike in K-space sinusoid in image space Susceptibility Artifacts Off-resonance artifacts caused by adjacent regions with different

More information

An Adaptive Eigenshape Model

An Adaptive Eigenshape Model An Adaptive Eigenshape Model Adam Baumberg and David Hogg School of Computer Studies University of Leeds, Leeds LS2 9JT, U.K. amb@scs.leeds.ac.uk Abstract There has been a great deal of recent interest

More information

Sources of Distortion in Functional MRI Data

Sources of Distortion in Functional MRI Data Human Brain Mapping 8:80 85(1999) Sources of Distortion in Functional MRI Data Peter Jezzard* and Stuart Clare FMRIB Centre, Department of Clinical Neurology, University of Oxford, Oxford, UK Abstract:

More information

3/27/2012 WHY SPECT / CT? SPECT / CT Basic Principles. Advantages of SPECT. Advantages of CT. Dr John C. Dickson, Principal Physicist UCLH

3/27/2012 WHY SPECT / CT? SPECT / CT Basic Principles. Advantages of SPECT. Advantages of CT. Dr John C. Dickson, Principal Physicist UCLH 3/27/212 Advantages of SPECT SPECT / CT Basic Principles Dr John C. Dickson, Principal Physicist UCLH Institute of Nuclear Medicine, University College London Hospitals and University College London john.dickson@uclh.nhs.uk

More information

Automatic Parameter Optimization for De-noising MR Data

Automatic Parameter Optimization for De-noising MR Data Automatic Parameter Optimization for De-noising MR Data Joaquín Castellanos 1, Karl Rohr 2, Thomas Tolxdorff 3, and Gudrun Wagenknecht 1 1 Central Institute for Electronics, Research Center Jülich, Germany

More information

Learning-based Neuroimage Registration

Learning-based Neuroimage Registration Learning-based Neuroimage Registration Leonid Teverovskiy and Yanxi Liu 1 October 2004 CMU-CALD-04-108, CMU-RI-TR-04-59 School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 Abstract

More information

Probabilistic Tracking and Model-based Segmentation of 3D Tubular Structures

Probabilistic Tracking and Model-based Segmentation of 3D Tubular Structures Probabilistic Tracking and Model-based Segmentation of 3D Tubular Structures Stefan Wörz, William J. Godinez, Karl Rohr University of Heidelberg, BIOQUANT, IPMB, and DKFZ Heidelberg, Dept. Bioinformatics

More information

RADIOMICS: potential role in the clinics and challenges

RADIOMICS: potential role in the clinics and challenges 27 giugno 2018 Dipartimento di Fisica Università degli Studi di Milano RADIOMICS: potential role in the clinics and challenges Dr. Francesca Botta Medical Physicist Istituto Europeo di Oncologia (Milano)

More information

Mutual Information Based Methods to Localize Image Registration

Mutual Information Based Methods to Localize Image Registration Mutual Information Based Methods to Localize Image Registration by Kathleen P. Wilkie A thesis presented to the University of Waterloo in fulfilment of the thesis requirement for the degree of Master of

More information

Correction of Partial Volume Effects in Arterial Spin Labeling MRI

Correction of Partial Volume Effects in Arterial Spin Labeling MRI Correction of Partial Volume Effects in Arterial Spin Labeling MRI By: Tracy Ssali Supervisors: Dr. Keith St. Lawrence and Udunna Anazodo Medical Biophysics 3970Z Six Week Project April 13 th 2012 Introduction

More information

Machine Learning for Medical Image Analysis. A. Criminisi

Machine Learning for Medical Image Analysis. A. Criminisi Machine Learning for Medical Image Analysis A. Criminisi Overview Introduction to machine learning Decision forests Applications in medical image analysis Anatomy localization in CT Scans Spine Detection

More information

UNIVERSITY OF SOUTHAMPTON

UNIVERSITY OF SOUTHAMPTON UNIVERSITY OF SOUTHAMPTON PHYS2007W1 SEMESTER 2 EXAMINATION 2014-2015 MEDICAL PHYSICS Duration: 120 MINS (2 hours) This paper contains 10 questions. Answer all questions in Section A and only two questions

More information

Interactive Differential Segmentation of the Prostate using Graph-Cuts with a Feature Detector-based Boundary Term

Interactive Differential Segmentation of the Prostate using Graph-Cuts with a Feature Detector-based Boundary Term MOSCHIDIS, GRAHAM: GRAPH-CUTS WITH FEATURE DETECTORS 1 Interactive Differential Segmentation of the Prostate using Graph-Cuts with a Feature Detector-based Boundary Term Emmanouil Moschidis emmanouil.moschidis@postgrad.manchester.ac.uk

More information

SUMMARY: DISTINCTIVE IMAGE FEATURES FROM SCALE- INVARIANT KEYPOINTS

SUMMARY: DISTINCTIVE IMAGE FEATURES FROM SCALE- INVARIANT KEYPOINTS SUMMARY: DISTINCTIVE IMAGE FEATURES FROM SCALE- INVARIANT KEYPOINTS Cognitive Robotics Original: David G. Lowe, 004 Summary: Coen van Leeuwen, s1460919 Abstract: This article presents a method to extract

More information

THE geometric alignment or registration of multimodality

THE geometric alignment or registration of multimodality IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 16, NO. 2, APRIL 1997 187 Multimodality Image Registration by Maximization of Mutual Information Frederik Maes,* André Collignon, Dirk Vandermeulen, Guy Marchal,

More information

Object Identification in Ultrasound Scans

Object Identification in Ultrasound Scans Object Identification in Ultrasound Scans Wits University Dec 05, 2012 Roadmap Introduction to the problem Motivation Related Work Our approach Expected Results Introduction Nowadays, imaging devices like

More information

Computer Vision I. Announcements. Fourier Tansform. Efficient Implementation. Edge and Corner Detection. CSE252A Lecture 13.

Computer Vision I. Announcements. Fourier Tansform. Efficient Implementation. Edge and Corner Detection. CSE252A Lecture 13. Announcements Edge and Corner Detection HW3 assigned CSE252A Lecture 13 Efficient Implementation Both, the Box filter and the Gaussian filter are separable: First convolve each row of input image I with

More information

Automatic Generation of Training Data for Brain Tissue Classification from MRI

Automatic Generation of Training Data for Brain Tissue Classification from MRI Automatic Generation of Training Data for Brain Tissue Classification from MRI Chris A. COCOSCO, Alex P. ZIJDENBOS, and Alan C. EVANS http://www.bic.mni.mcgill.ca/users/crisco/ McConnell Brain Imaging

More information

Convolution-Based Truncation Correction for C-Arm CT using Scattered Radiation

Convolution-Based Truncation Correction for C-Arm CT using Scattered Radiation Convolution-Based Truncation Correction for C-Arm CT using Scattered Radiation Bastian Bier 1, Chris Schwemmer 1,2, Andreas Maier 1,3, Hannes G. Hofmann 1, Yan Xia 1, Joachim Hornegger 1,2, Tobias Struffert

More information

Multimodal Image Fusion Of The Human Brain

Multimodal Image Fusion Of The Human Brain Multimodal Image Fusion Of The Human Brain Isis Lázaro(1), Jorge Marquez(1), Juan Ortiz(2), Fernando Barrios(2) isislazaro@gmail.com Centro de Ciencias Aplicadas y Desarrollo Tecnológico, UNAM Circuito

More information

REAL-TIME ADAPTIVITY IN HEAD-AND-NECK AND LUNG CANCER RADIOTHERAPY IN A GPU ENVIRONMENT

REAL-TIME ADAPTIVITY IN HEAD-AND-NECK AND LUNG CANCER RADIOTHERAPY IN A GPU ENVIRONMENT REAL-TIME ADAPTIVITY IN HEAD-AND-NECK AND LUNG CANCER RADIOTHERAPY IN A GPU ENVIRONMENT Anand P Santhanam Assistant Professor, Department of Radiation Oncology OUTLINE Adaptive radiotherapy for head and

More information

Deformable Segmentation using Sparse Shape Representation. Shaoting Zhang

Deformable Segmentation using Sparse Shape Representation. Shaoting Zhang Deformable Segmentation using Sparse Shape Representation Shaoting Zhang Introduction Outline Our methods Segmentation framework Sparse shape representation Applications 2D lung localization in X-ray 3D

More information