Parallelization of Mutual Information Registration

Size: px
Start display at page:

Download "Parallelization of Mutual Information Registration"

Transcription

1 Parallelization of Mutual Information Registration Krishna Sanka Simon Warfield William Wells Ron Kikinis May 5, 2000 Abstract Mutual information registration is an effective procedure that provides an accurate spatial relationship between two volumetric medical images of differing modalities. This had aided surgeons in precisely locating pathological tissues and avoiding critical structures. This paper focuses on accelerating the registration procedure, while preserving accuracy, by implementing parallelism with POSIX threads. When compared to the currently used serially executing registration process, the time needed to align two sets of volumetric imagery was sharply reduced. Although thread overhead and maintenance presented some limitations, the registration speedup increased - within a range of available CPUs - linearly with respect to the available CPUs with a slope close to one, and with accuracy preserved. Rapid alignment through parallel registration allows for shorter interventions and improved surgical navigation. This leads to a superior execution of clinical procedures, such as tumor resections, which leads to a better surgical outcome and faster recovery for the patient. Keywords: Mutual information, POSIX threads, symmetric multi-processor, medical image registration 1 Introduction Registration is the process of determining an accurate spatial relationship between sets of medical image data. Applications of registration range from image guided surgery to the analysis of functional images, and include the characterization of normal and abnormal anatomical variability (brain mapping), the detection of change in disease state over time, the visualization of multimodality data sets, and modeling anatomy in the process of segmentation. This paper presents registration in the context of providing an accurate spatial relationship between intra-operative volumetric image data and pre-operative volumetric image data. There are several advantages to obtaining intra-operative images. For example, using a high field interventional MR imager allows the quick acquisition of cross-sections of a patient s anatomy and can portray the current state of the anatomy in terms of characteristics such as deformation. Intra-operative MR images acquired during surgery serve 1

2 the purpose of guiding the surgeon s actions. It provides additional valuable information by allowing the surgeon to see beneath the surface of structures. However, this quick procurement of intra-operative images comes with a price. Compared to a conventional clinical scanner, a lower magnet field strength is used in obtaining intra-operative images, which leads to a poorer signal-to-noise ratio. In addition, there is a smaller field of view, which results in only the region of interest being scanned. Finally, while acquiring images with a larger voxel size accelerates the process with an improved signal-to-noise ratio, it also leads to larger sampling artifacts than are usually seen in conventional MR scanners [1]. However, the value of intra-operative imaging is greatly increased by registration. By registering the intra-operative data with the pre-operative data, one could compare and determine whether changes in tissue structure have occurred. For instance, this would be extremely useful in evaluating the extent of a tumor. In general, utilizing such a registration would allow a surgeon to more precisely identify and avoid critical structures and more accurately locate pathological tissues during a procedure. There are several algorithms that one can use in order to perform a registration. Here, the registration process involves the utilization of the mutual information that is shared between the sets of data. Mutual information (MI) is an information-theoretic measure of the relatedness of two sets of data that is defined in terms of various entropies of the data [2, 3]. When registration is conducted by maximizing mutual information, the spatial relationship of two data sets is adjusted to maximize their mutual information. It has been shown to be an effective mechanism for multi-modality registration of volumetric medical image data in surgical cases [4]. As shown in intra-operative imaging and image guided surgery, high performance computing is becoming increasingly important in imaging applications because radiological image analysis is shifting from a visual analysis of collections of planar images to a computerized quantitative image analysis of volumetric images [5]. This method of analysis is increasingly limited by memory and speed constraints of conventional workstations. Therefore, in order for registration (utilizing any given algorithm) to become a widely used clinical tool, techniques need to be available that can execute in a clinically compatible time frame on widely available hardware. It has been shown that the current state of performing registrations between intraoperative and pre-operative image data has room for improvement. In particular, it would be desirable to employ parallelism, through the use of threads, in order to accelerate the MI-based registration process. In essence, the registration process maximizes mutual information through many iterations. Although there can be several ways in which to implement parallelism in the registration process, it seems that the most efficient implementation would require a small change in the execution of the MI algorithm so that these iterations can be performed in parallel. This paper focuses on expediting the registration process through this method while preserving the accuracy of the coordinate transformation that relates the coordinate systems of the pre-operative and intra-operative image data. 2

3 for (i = 0; i < 20000; i++) { // sample randomly selected coordinates // perform calculations to get di/dt // T <- T + (lambda * di/dt) } Figure 1: The implementation of the current system runs iterations in order to calculate an appropriate pose. di/dt is the derivative of the mutual information. lambda is the learning rate. T is the coordinate transform (pose). Our implementation of registration by maximization of mutual information makes use of the EMMA approach to estimating derivatives of MI with respect to the coordinate transformation parameters [3, 2, 6]. EMMA is an example of the stochastic gradient descent approach to optimization, which has been described by Widrow and Hoff [7]. This method has been commonly used in the neural network community where it is referred to as the LMS rule. In the present work we will describe a method for parallelizing the stochastic gradient descent approach. In the remainder of the paper we will endeavor to answer the following questions. First, will a parallel MI-based registration provide the same accuracy as the current system? Second, will there be an appropriate speedup as the number of CPUs in the system increase? Finally, will the overhead associated with implementing threads be tolerable enough in order to see an appropriate speedup compared to the current system? Through the execution and comparison of many trials of the parallel registration and the current system, this paper will show that the above questions can be answered, within a certain scope, in the affirmative. 2 Method The current system for conducting registration consists of many iterations that maximize the mutual information between two sets of volumetric medical image data. In every iteration, a pre-determined number of coordinates are randomly sampled from one of the sets of image data. Unless said otherwise, 50 coordinates were sampled. Several calculations that maximize mutual information are then performed. This leads to obtaining an estimate of the derivative of the mutual information. This derivative is then multiplied by a learning rate parameter and used to update the coordinate transform (a pose ) to a more accurate value. After all iterations have been completed the pose has been modified sufficiently to provide a precise spatial relationship between the two sets of volumetric data. This can be seen in Figure 1. As shown in Figure 1, the new value of the pose depends on its value in the previous iteration. In order to use POSIX threads to implement parallelism in an effective manner, it was decided that a number of iterations (equal to the number of threads) would be 3

4 for (i = 0; i < 20000; i += (number of threads)) { for (j = 0; j < (number of threads); j++) { // sample randomly selected coordinates // perform calculations to get di/dt // MI[j] = di/dt} for (k = 0; k < (number of threads); k++) { // T += (lambda * MI[k]) } Figure 2: The algorithm has been changed slightly in order to implement parallelism. This is the serial implementation of the algorithm. As expected, it has the same performance characteristics when compared to the original system. In the parallel version, the for loop that is indexed by j is calculated by all the threads in parallel which allows the overall computation to proceed at a faster rate. Test Machine Architecture CPUs Ecache Operating System 1 Enterprise 450 4x300MHz 2MB SunOS Ultra 80 4x450MHz 4MB SunOS Ultra 60 2x360MHz 4MB SunOS Wildfire Architecture 28x250MHz 4MB SunOS 5.6 Table 1: The characteristics of the machines on which the registration trials were performed. Features listed are the machine architecture, number and type of CPUs, amount of Ecache, and operating system. All machines were manufactured by Sun Microsystems, Inc. and use Ultrasparc II processors. performed in parallel. The derivative of mutual information obtained from each of those iterations would then be accumulated before the pose was updated. Figure 2 shows how this changed algorithm would be executed serially. Although this amounts to a slight change in the algorithm, it will be shown in the results section to be just as accurate as the current one. In introducing parallelism into mutual-information based registration, a standard workpile decomposition [8] is implemented with POSIX threads to parallelize the calculations of the derivative of the mutual information in each iteration. One process manages the workpile by controlling access to the units of work and collating partial results, while other processes request work units (in this case, an iteration of calculations) and carry out the work independently. The expected speedup of the system is linear, with a slope close to one, with the number of threads (and therefore CPUs) available for computation. The parallel version of the MI-based registration was tested and compared to its serially executing counterpart as well as with the original serial version. These tests were performed on several machines. The characteristics of these machines are depicted in Table 1. 4

5 It should be noted that the architecture of machine #4 in Table 1 is experimental 1 [9]. On each machine, the number of threads were varied while executing the registration code. For each variation, 100 trials were performed with each version of the code. Before each trial began, one of the volumetric image data sets was randomly offset by a pose so that registration could be performed to have the two data sets converge. This random offset was determined by two factors: a rotational perturbation about a randomly selected axis by a random angle uniformly selected from a pre-determined interval, and a uniformly random translational offset selected from a pre-determined interval that was added to each translational axis (X, Y, and Z). The translational interval was 20 mm and the rotational interval was 30 degrees. At the end of each trial, the standard deviation of each translational axis for the final pose was calculated. An average rotation angle from an average rotation was also computed. The final pose was also evaluated for accuracy. It was deemed successful if all the translational components were within 3 mm of the translational components of the original pose. In addition, the duration of each trial was measured in seconds in order to evaluate the effectiveness of parallelism in MI-based registration. 3 Results Figure 3 displays the results of a successful trial. The results obtained from all trials are displayed in graphical or tabular format. Unless specified otherwise, all data is taken from trials that randomly sampled 50 coordinates during each iteration of the registration process. 4 Analysis The results show that in each case the parallel registration converged to a successful registration result. The variance of the difference between the ideal and obtained translations was small, as shown in Table 2, and the maximum error over all trials was less than 3mm which is comparable to the voxel size. It must also be kept in mind that the equivalent of one round of registration was tested. In a clinical setting, several more rounds of refined registration would be performed that would even further reduce the maximum error. The modified parallel algorithm has similar accuracy to the original serial algorithm, and provides a significant speedup on small symmetric multi-processor (SMP) machines. It can be seen in Table 8 that the time taken to perform parallel registration with one CPU available was larger than the time elapsed for the serial registration procedure. One would expect these times to be the same, because both types of registration are operating under 1 WildFire is Sun s internal codename for an advanced Server architecture that is under development. WildFire prototype systems have not been tested, optimized, or qualified for sale; they have been built for evaluative purposes only. Elements of the WildFire prototype may or may not be used in future Sun products. 5

6 Figure 3: Display of a successful trial. The 3 images on the left are obtained from an interventional MRI scan that represents intra-operative data and the 3 images on the right are conventional MRI images that represent pre-operative data Speedup Factor Number of Available CPUs Figure 4: Speedup Factors for MACHINE 1 6

7 MACHINE CODE CPUs T θ FINAL TRIALS SUCCESS XYZ σ X σ Y σ Z θ ±mm mm % 4 P P P P P P P P P P SP S P P SP S P P P P SP S P P P P SP S Table 2: MRT MRI volumetric registration results table. The MACHINE column refers to the test machines in Table 1. In the CODE column, P refers to the parallel registration program, SP refers to the serial implementation of the parallel registration program, and S refers to the currently used serially executing system. The CPUs column refer to the number of CPUs available for computation. From a known position and orientation, a random offset uniformly selected from the interval T was added to each translational axis ( T) after the reference volume had been rotated about a randomly selected axis by a random angle uniformly selected from the interval θ. The distributions of the final poses can be evaluated by comparing the standard deviations of the location of the center, computed separately in X, Y and Z. Furthermore, the average rotation angle from an average rotation is computed ( θ ). Finally, the number of trials that succeeded in converging to near the correct solution (by visual inspection) is reported. 7

8 Speedup Factor Number of Available CPUs Figure 5: Speedup Factors for MACHINE 2 8

9 2 Speedup Factor Number of Available CPUs Figure 6: Speedup Factors for MACHINE 3 9

10 30 TEST MACHINE 4-50 Samples TEST MACHINE Samples TEST MACHINE Samples Speedup Factor Number of Available CPUs Figure 7: Speedup Factors for MACHINE 4 shown for three sets of trials at different coordinate sample sizes 10

11 MACHINE CODE CPUs TRIALS AVERAGE EXECUTION TIME 3 P P SP S P P P P SP S P P P P SP S Table 3: Average execution time in seconds over 100 trials for all the different registration programs. Column headings are as in the previous tables. conditions that allow serial computation only. This difference in time, which is probably due to increased costs of thread synchronization, is described further in Section 4.2. Machine 1, 2, and 3 all showed a significant amount of speedup as the number of CPUs used was increased. Figures 4, 5, and 6 show that the speedup in registration was linear, with a slope that was close to one. When compared to the currently used serial registration system, the slope of the speedup was not as close to one as desired. Nonetheless, there was still a significant speedup. The reduction in speedup is probably due to the overhead of the POSIX threads, which is described further in Section 4.2. The performance of the serial implementation of the parallel registration system was very similar to the currently used system. This was expected because neither system utilized threads to implement parallelism. 4.1 Performance Analysis on Wildfire Architecture Machine 4, which operates on the experimental Wildfire architecture, provided some interesting results. Three sets of 100 trials were performed on Machine 4 with the parallel registration system. The only difference in the trials is that one set of trials randomly sampled 50 coordinates every time the derivative of the mutual information was found. Another set of trials sampled 100 coordinates for every instance in which the derivative of mutual information was obtained, and the last set of trials sampled 200 coordinates. As more coordinates were sampled, the trials would take longer to compute the final pose because there would be more work being done in each iteration of the registration process. 11

12 MACHINE CODE CPUs TRIALS AVERAGE EXECUTION TIME 4 P P P P P P P P P P P P P P P P P P P P P P P P P P SP S Table 4: Average execution time in seconds over 100 trials for TEST MACHINE 4. In each trial, 50 coordinates are randomly sampled. Column headings are as in the previous tables. 12

13 MACHINE CODE CPUs TRIALS AVERAGE EXECUTION TIME 4 P P P P P P P P P P P P P P P P P P P P P P P P P P Table 5: Average execution time in seconds over 100 trials for TEST MACHINE 4. In each trial, 100 coordinates are randomly sampled. Column headings are as in the previous tables. 13

14 MACHINE CODE CPUs TRIALS AVERAGE EXECUTION TIME 4 P P P P P P P P P P P P P P P P P P P P P P P P P P Table 6: Average execution time in seconds over 100 trials for TEST MACHINE 4. In each trial, 200 coordinates are randomly sampled. Column headings are as in the previous tables. SAMPLES TIME WITH 26 CPUs TIME WITH 1 CPU SPEEDUP Table 7: Speedup associated with 26 CPUs available on MACHINE 4 for sets of trials with differing sample sizes. Average times are reported in seconds. 14

15 MACHINE PARALLEL TIME SERIAL TIME ADDITIONAL TIME CPU CLOCK FOR 1 CPU FOR 1 CPU PER ITERATION (MHz) Table 8: Additional time spent per iteration by parallel code for registration. The time to execute the parallel code with 1 CPU available is reported in seconds in Column 2. The time to execute the serial code with 1 CPU available is reported in seconds in Column 3. The additional time spent per iteration by the parallel code is reported in milliseconds in Column 4. In each implementation, the registration consisted of iterations. The speed of each machine s CPUs are reported in Column 5. In all cases, the slope of speedup of the parallel computation was close to 1 up to about 6 CPUs. However, as more CPUs were being used, the nature of the speedup changed for the three sets of trials. The speedup in the set of trials that sampled 50 coordinates no longer varied and seemed to reach a plateau with very minor fluctuations. The trials that sampled 100 coordinates still had a positive speedup after 6 CPUs, but the rise was much smaller with a slope below 1. Eventually, the speedup also plateaued with minor fluctuations. The exception to this was seen when a drop in speedup occurred with available CPUs. In the trials that sampled 200 coordinates, the speedup s slope fell below 1, but kept rising without plateauing for 20 CPUs. There was a slight drop in speedup when 21 CPUs became available, but continued to rise afterwards. This can be seen in Figure 7. In the set of trials that sampled 200 coordinates, there was a greater speedup overall compared to the set of trials with 100 samples. Not surprisingly, the set of trials that sampled 100 coordinates had a greater speedup overall compared to the set of trials with 50 samples. As the data shows, using 26 CPUs with 50 samples showed a speedup of approximately When 26 CPUs were utilized with 100 coordinates being sampled, the computation time was larger, but the speedup was When 200 coordinates were sampled, there was an even greater increase in computation time. However, a greater speedup of was associated with it. This seems to imply that the threads were not being given enough work for computation, and that speedup can be dependent on the amount of work that each individual thread must accomplish. On occasion, it was also observed that some threads spent more time on computation than others, which would lead to a degradation in performance. As described below, each thread executes the same code on the same amount of data and each is expected to complete in the same amount of time, so we attribute this difference to different times spent in contention for the workpile mutual exclusion lock, condition variable signalling and LWP scheduling. In any case, the threads did not provide the desired performance when more CPUs were available for computation. 15

16 4.2 Possible Limitations on Threads Why might there not be more of a speedup when more CPUs are utilized? Ideally speaking, if a number of threads equal to the number of CPUs available are to be utilized in performing registration, the rate at which the registration was performed would rise with a slope of 1 whenever an additional CPU was made available. One reason why this ideal might not be reached is that each thread does not have enough work to compute. The work that is needed to create, manage, and destroy the job structure for the thread may be greater than the work required to perform the desired computation. If a thread were forced to perform more work, then the requirements for computation would probably take more time to accomplish than the requirements for thread maintenance. As the time required for each thread job increases the relative importance of the fixed cost of building the job structure decreases. The cost of thread synchronization also becomes relatively less significant. It must be kept in mind that the time required for each iteration is on the order of milliseconds. A good approximation of this time can be obtained by dividing the total execution time of a trial by the total number of iterations, which was in this case. As was shown with Machine 4, when each thread was given more work to do, there was an increase in speedup. The speedup is also probably limited because of increased thread synchronization costs that do not scale with the available CPUs. In particular, the ratio of time to compute one iteration to the time required to synchronize threads between iterations might not be as high as desired when the number of CPUs increase. Our measurements indicate that the increased synchronization costs can be accounted for by the fact that as the number of CPUs increase, the threads don t execute at the same time because they are spending as much time waiting for work as they are when they execute an iteration. Another reason for not achieving the ideal could also lie in Amdahl s Law, which takes into account the fraction of original execution time that is amenable to improvement. The overall speedup of the registration process may be affected due to this limitation. It was observed that the threads did not start performing computations at the same time. Ideally, if N threads are to be used for a computation that take T seconds to complete, all N threads should begin at the same time so that the total time elapsed will be T seconds. The worst case scenario is where the threads executed serially so that NxT seconds elapse. However, even in the ideal situation, threads do not begin computation at exactly the same time due to several factors, such as mutex competition. In any case, there were cases of some threads executing after other threads had finished. Although each thread might have taken the same amount of time to perform computation, this situation led to the total elapsed time being doubled. If this happened often enough, there would definitely be a drop in speedup. Again, this can be solved if each thread has more work to do because that would result in a reduced probability that a thread would begin computation after another thread has finished computation. Although this would lead to an increase in speedup, the time needed to finish all computation would increase. If one were to have each thread perform more computations in parallel registration, special 16

17 care would have to be taken to preserve the accuracy of the process so that a credible final pose could be obtained. Finally, we observed that the time taken to perform the parallel registration process using one CPU was larger than the time taken to perform the serial registration process. Theoretically, these times should be identical since the parallel registration is operating under the conditions of a serial system. However, as Table 8 shows, the time difference amounts to a fraction of a millisecond more being spent on each iteration of the parallel registration process. This fraction of a millisecond could easily be attributed to the synchronization overhead that limited the speedup. This synchronization is controlled with the POSIX condition variables. As Table 8 also shows, the magnitude of the extra time per iteration could be influenced by the speed of the processors being used by the machine. For instance, Machine 4 is made up of 250 MHz processors and has the largest amount of extra time per iteration at 0.64 ms. However, Machine 2 is comprised of 450 MHz processors and has the smallest amount of extra time per iteration at 0.31 ms. As a whole, the data shows that extra time per iteration is inversely proportional to the speed of a CPU. 5 Future Work This paper addressed the possibility of accelerating the registration process while preserving its accuracy by implementing parallel computing through the use of POSIX threads. In the future, one could attempt to further reduce the synchronization overhead for even better performance. This could be achieved through algorithmic changes and different implementations of parallelism. For instance, utilizing the master-slaves paradigm may reduce the synchronization overhead for better performance. In this scenario, each slave would compute an iteration of the registration process after the work has been prepared. Through changes of this sort, the registration process could evolve to become even more efficient and effective. If a similar success ratio can be maintained within similar translational and rotational perturbations, then it would be very beneficial to implement this change. 6 Conclusion In an effort to accelerate the process of volumetric mutual information registration, parallelism was implemented using POSIX threads. By using threads to execute a standard workpile decomposition that parallelizes the computation of the derivative of mutual information from different iterations of the registration process, the time needed for two sets of volumetric medical image data to be aligned was sharply reduced compared with the current system already in use. Trials performed on several machines showed that the speedup of the registration process was linear with the number of CPUs available for computation. The slope of this speedup was close to one for a small number of processors, but 17

18 was reduced as the number increased. In addition, the accuracy of the registration process was preserved as these changes were implemented. The successful results of these trials show that the implementation of parallelism can greatly speed up the registration of intraoperative data with pre-operative data in order to reduce the length of an intervention and increase the overall comfort of a patient. In addition, the rapid alignment achieved with parallelism would allow a better use of two-dimensional, orthogonal two-dimensional, or low resolution three-dimensional images and hence better surgical navigation. As a result, it would become more practical to acquire more images during surgery, which would lead to a better surgical outcome for the patient. References [1] Simon K. Warfield, Ferenc A. Jolesz, and Ron Kikinis. Real-Time Image Segmentation for Image-Guided Surgery. In Supercomputing 1998, page 1028, November [2] P. Viola and W. Wells. Alignment by maximization of mutual information. In Proceedings of the 5th International Conference on Computer Vision, [3] W. Wells, P. Viola, S. Nakajima, H. Atsumi, and R. Kikinis. Multi-modal volume registration by maximization of mutual information. Medical Image Analysis, 1(1):35 53, [4] J. West, J.M. Fitzpatrick, et al. Comparison and Evaluation of Retrospective Intermodality Brain Image Regsistraion Techniques. Journal of Computer Assisted Tomography, July/Aug [5] G. D. Rubin, S. Napel, and A. N. Leung. Volumetric analysis of volumetric data: Achieving a paradigm shift. Radiology, 200: , [6] P. Viola. Alignment by Maximization of Mutual Information. PhD thesis, Massachusetts Institute of Technology, [7] B. Widrow and M. Hoff. Adaptive switching circuits. IRE WESCON Convection Record, [8] S. Kleiman, D. Shah, and B. Smaalders. Programming with Threads. Prentice Hall, [9] L. Noordergraaf and R van der Pas. Performance Experiences on Sun s Wildfire Prototype. In Supercomputing 1999,

CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION

CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION 6.1 INTRODUCTION Fuzzy logic based computational techniques are becoming increasingly important in the medical image analysis arena. The significant

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

Multi-Modal Volume Registration Using Joint Intensity Distributions

Multi-Modal Volume Registration Using Joint Intensity Distributions Multi-Modal Volume Registration Using Joint Intensity Distributions Michael E. Leventon and W. Eric L. Grimson Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA leventon@ai.mit.edu

More information

Efficient population registration of 3D data

Efficient population registration of 3D data Efficient population registration of 3D data Lilla Zöllei 1, Erik Learned-Miller 2, Eric Grimson 1, William Wells 1,3 1 Computer Science and Artificial Intelligence Lab, MIT; 2 Dept. of Computer Science,

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

Medical Image Registration by Maximization of Mutual Information

Medical Image Registration by Maximization of Mutual Information Medical Image Registration by Maximization of Mutual Information EE 591 Introduction to Information Theory Instructor Dr. Donald Adjeroh Submitted by Senthil.P.Ramamurthy Damodaraswamy, Umamaheswari Introduction

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

Assessing Accuracy Factors in Deformable 2D/3D Medical Image Registration Using a Statistical Pelvis Model

Assessing Accuracy Factors in Deformable 2D/3D Medical Image Registration Using a Statistical Pelvis Model Assessing Accuracy Factors in Deformable 2D/3D Medical Image Registration Using a Statistical Pelvis Model Jianhua Yao National Institute of Health Bethesda, MD USA jyao@cc.nih.gov Russell Taylor The Johns

More information

William Yang Group 14 Mentor: Dr. Rogerio Richa Visual Tracking of Surgical Tools in Retinal Surgery using Particle Filtering

William Yang Group 14 Mentor: Dr. Rogerio Richa Visual Tracking of Surgical Tools in Retinal Surgery using Particle Filtering Mutual Information Computation and Maximization Using GPU Yuping Lin and Gérard Medioni Computer Vision and Pattern Recognition Workshops (CVPR) Anchorage, AK, pp. 1-6, June 2008 Project Summary and Paper

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

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

Video Registration Virtual Reality for Non-linkage Stereotactic Surgery

Video Registration Virtual Reality for Non-linkage Stereotactic Surgery Video Registration Virtual Reality for Non-linkage Stereotactic Surgery P.L. Gleason, Ron Kikinis, David Altobelli, William Wells, Eben Alexander III, Peter McL. Black, Ferenc Jolesz Surgical Planning

More information

A Study of Medical Image Analysis System

A Study of Medical Image Analysis System Indian Journal of Science and Technology, Vol 8(25), DOI: 10.17485/ijst/2015/v8i25/80492, October 2015 ISSN (Print) : 0974-6846 ISSN (Online) : 0974-5645 A Study of Medical Image Analysis System Kim Tae-Eun

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

Magnetic Resonance Elastography (MRE) of Liver Disease

Magnetic Resonance Elastography (MRE) of Liver Disease Magnetic Resonance Elastography (MRE) of Liver Disease Authored by: Jennifer Dolan Fox, PhD VirtualScopics Inc. jennifer_fox@virtualscopics.com 1-585-249-6231 1. Overview of MRE Imaging MRE is a magnetic

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

Registration concepts for the just-in-time artefact correction by means of virtual computed tomography

Registration concepts for the just-in-time artefact correction by means of virtual computed tomography DIR 2007 - International Symposium on Digital industrial Radiology and Computed Tomography, June 25-27, 2007, Lyon, France Registration concepts for the just-in-time artefact correction by means of virtual

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

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

Non linear Registration of Pre and Intraoperative Volume Data Based On Piecewise Linear Transformations

Non linear Registration of Pre and Intraoperative Volume Data Based On Piecewise Linear Transformations Non linear Registration of Pre and Intraoperative Volume Data Based On Piecewise Linear Transformations C. Rezk Salama, P. Hastreiter, G. Greiner, T. Ertl University of Erlangen, Computer Graphics Group

More information

Automatic Optimization of Segmentation Algorithms Through Simultaneous Truth and Performance Level Estimation (STAPLE)

Automatic Optimization of Segmentation Algorithms Through Simultaneous Truth and Performance Level Estimation (STAPLE) Automatic Optimization of Segmentation Algorithms Through Simultaneous Truth and Performance Level Estimation (STAPLE) Mahnaz Maddah, Kelly H. Zou, William M. Wells, Ron Kikinis, and Simon K. Warfield

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 Integrated Visualization System for Surgical Planning and Guidance using Image Fusion and Interventional Imaging

An Integrated Visualization System for Surgical Planning and Guidance using Image Fusion and Interventional Imaging An Integrated Visualization System for Surgical Planning and Guidance using Image Fusion and Interventional Imaging David T. Gering 1,AryaNabavi 2, Ron Kikinis 2,W.EricL.Grimson 1, Noby Hata 2, Peter Everett

More information

Fiber Selection from Diffusion Tensor Data based on Boolean Operators

Fiber Selection from Diffusion Tensor Data based on Boolean Operators Fiber Selection from Diffusion Tensor Data based on Boolean Operators D. Merhof 1, G. Greiner 2, M. Buchfelder 3, C. Nimsky 4 1 Visual Computing, University of Konstanz, Konstanz, Germany 2 Computer Graphics

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

Advanced Visual Medicine: Techniques for Visual Exploration & Analysis

Advanced Visual Medicine: Techniques for Visual Exploration & Analysis Advanced Visual Medicine: Techniques for Visual Exploration & Analysis Interactive Visualization of Multimodal Volume Data for Neurosurgical Planning Felix Ritter, MeVis Research Bremen Multimodal Neurosurgical

More information

RIGID IMAGE REGISTRATION

RIGID IMAGE REGISTRATION RIGID IMAGE REGISTRATION Duygu Tosun-Turgut, Ph.D. Center for Imaging of Neurodegenerative Diseases Department of Radiology and Biomedical Imaging duygu.tosun@ucsf.edu What is registration? Image registration

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

CHAPTER 2. Morphometry on rodent brains. A.E.H. Scheenstra J. Dijkstra L. van der Weerd

CHAPTER 2. Morphometry on rodent brains. A.E.H. Scheenstra J. Dijkstra L. van der Weerd CHAPTER 2 Morphometry on rodent brains A.E.H. Scheenstra J. Dijkstra L. van der Weerd This chapter was adapted from: Volumetry and other quantitative measurements to assess the rodent brain, In vivo NMR

More information

Rigid and Deformable Vasculature-to-Image Registration : a Hierarchical Approach

Rigid and Deformable Vasculature-to-Image Registration : a Hierarchical Approach Rigid and Deformable Vasculature-to-Image Registration : a Hierarchical Approach Julien Jomier and Stephen R. Aylward Computer-Aided Diagnosis and Display Lab The University of North Carolina at Chapel

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

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

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

Automatic MS Lesion Segmentation by Outlier Detection and Information Theoretic Region Partitioning Release 0.00

Automatic MS Lesion Segmentation by Outlier Detection and Information Theoretic Region Partitioning Release 0.00 Automatic MS Lesion Segmentation by Outlier Detection and Information Theoretic Region Partitioning Release 0.00 Marcel Prastawa 1 and Guido Gerig 1 Abstract July 17, 2008 1 Scientific Computing and Imaging

More information

A Non-Linear Image Registration Scheme for Real-Time Liver Ultrasound Tracking using Normalized Gradient Fields

A Non-Linear Image Registration Scheme for Real-Time Liver Ultrasound Tracking using Normalized Gradient Fields A Non-Linear Image Registration Scheme for Real-Time Liver Ultrasound Tracking using Normalized Gradient Fields Lars König, Till Kipshagen and Jan Rühaak Fraunhofer MEVIS Project Group Image Registration,

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

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

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

Fast Fuzzy Clustering of Infrared Images. 2. brfcm

Fast Fuzzy Clustering of Infrared Images. 2. brfcm Fast Fuzzy Clustering of Infrared Images Steven Eschrich, Jingwei Ke, Lawrence O. Hall and Dmitry B. Goldgof Department of Computer Science and Engineering, ENB 118 University of South Florida 4202 E.

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

Experiments with Edge Detection using One-dimensional Surface Fitting

Experiments with Edge Detection using One-dimensional Surface Fitting Experiments with Edge Detection using One-dimensional Surface Fitting Gabor Terei, Jorge Luis Nunes e Silva Brito The Ohio State University, Department of Geodetic Science and Surveying 1958 Neil Avenue,

More information

Knowledge-Based Deformable Matching for Pathology Detection

Knowledge-Based Deformable Matching for Pathology Detection Knowledge-Based Deformable Matching for Pathology Detection Thesis Proposal Mei Chen CMU-RI-TR-97-20 The Robotics Institute Carnegie Mellon University Pittsburgh, Pennsylvania 15213 May 1997 c 1997 Carnegie

More information

XI Signal-to-Noise (SNR)

XI Signal-to-Noise (SNR) XI Signal-to-Noise (SNR) Lecture notes by Assaf Tal n(t) t. Noise. Characterizing Noise Noise is a random signal that gets added to all of our measurements. In D it looks like this: while in D

More information

Automatic Cerebral Aneurysm Detection in Multimodal Angiographic Images

Automatic Cerebral Aneurysm Detection in Multimodal Angiographic Images Automatic Cerebral Aneurysm Detection in Multimodal Angiographic Images Clemens M. Hentschke, Oliver Beuing, Rosa Nickl and Klaus D. Tönnies Abstract We propose a system to automatically detect cerebral

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

Parallel Non-Rigid Registration on a Cluster of Workstations

Parallel Non-Rigid Registration on a Cluster of Workstations Parallel Non-Rigid Registration on a Cluster of Workstations Radu Stefanescu, Xavier Pennec, Nicholas Ayache INRIA Sophia, Epidaure, 004 Rte des Lucioles, F-0690 Sophia-Antipolis Cedex {Radu.Stefanescu,

More information

Compressed Sensing Algorithm for Real-Time Doppler Ultrasound Image Reconstruction

Compressed Sensing Algorithm for Real-Time Doppler Ultrasound Image Reconstruction Mathematical Modelling and Applications 2017; 2(6): 75-80 http://www.sciencepublishinggroup.com/j/mma doi: 10.11648/j.mma.20170206.14 ISSN: 2575-1786 (Print); ISSN: 2575-1794 (Online) Compressed Sensing

More information

Tetrahedral Mesh Generation for Medical Imaging

Tetrahedral Mesh Generation for Medical Imaging Tetrahedral Mesh Generation for Medical Imaging Andriy Fedorov 1,2,3, Nikos Chrisochoides 1,2,3, Ron Kikinis 2, and Simon Warfield 2,3 1 Department of Computer Science, College of William and Mary, Williamsburg,

More information

Pathology Hinting as the Combination of Automatic Segmentation with a Statistical Shape Model

Pathology Hinting as the Combination of Automatic Segmentation with a Statistical Shape Model Pathology Hinting as the Combination of Automatic Segmentation with a Statistical Shape Model Pascal A. Dufour 12,HannanAbdillahi 3, Lala Ceklic 3,Ute Wolf-Schnurrbusch 23,JensKowal 12 1 ARTORG Center

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

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

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

Cortical Surface Registration Using Texture Mapped Point Clouds and Mutual Information

Cortical Surface Registration Using Texture Mapped Point Clouds and Mutual Information Cortical Surface Registration Using Texture Mapped Point Clouds and Mutual Information Tuhin K. Sinha, David M. Cash, Robert J. Weil, Robert L. Galloway, and Michael I. Miga Vanderbilt University, Nashville

More information

coding of various parts showing different features, the possibility of rotation or of hiding covering parts of the object's surface to gain an insight

coding of various parts showing different features, the possibility of rotation or of hiding covering parts of the object's surface to gain an insight Three-Dimensional Object Reconstruction from Layered Spatial Data Michael Dangl and Robert Sablatnig Vienna University of Technology, Institute of Computer Aided Automation, Pattern Recognition and Image

More information

Pathology Hinting as the Combination of Automatic Segmentation with a Statistical Shape Model

Pathology Hinting as the Combination of Automatic Segmentation with a Statistical Shape Model Pathology Hinting as the Combination of Automatic Segmentation with a Statistical Shape Model Pascal A. Dufour 1,2, Hannan Abdillahi 3, Lala Ceklic 3, Ute Wolf-Schnurrbusch 2,3, and Jens Kowal 1,2 1 ARTORG

More information

Distributed Scheduling for the Sombrero Single Address Space Distributed Operating System

Distributed Scheduling for the Sombrero Single Address Space Distributed Operating System Distributed Scheduling for the Sombrero Single Address Space Distributed Operating System Donald S. Miller Department of Computer Science and Engineering Arizona State University Tempe, AZ, USA Alan C.

More information

Grid-Enabled Software Environment for Enhanced Dynamic Data-Driven Visualization and Navigation during Image-Guided Neurosurgery

Grid-Enabled Software Environment for Enhanced Dynamic Data-Driven Visualization and Navigation during Image-Guided Neurosurgery Grid-Enabled Software Environment for Enhanced Dynamic Data-Driven Visualization and Navigation during Image-Guided Neurosurgery Nikos Chrisochoides 1, Andriy Fedorov 1, Andriy Kot 1, Neculai Archip 2,

More information

Automatic Generation of Training Data for Brain Tissue Classification from MRI

Automatic Generation of Training Data for Brain Tissue Classification from MRI MICCAI-2002 1 Automatic Generation of Training Data for Brain Tissue Classification from MRI Chris A. Cocosco, Alex P. Zijdenbos, and Alan C. Evans McConnell Brain Imaging Centre, Montreal Neurological

More information

Towards Breast Anatomy Simulation Using GPUs

Towards Breast Anatomy Simulation Using GPUs Towards Breast Anatomy Simulation Using GPUs Joseph H. Chui 1, David D. Pokrajac 2, Andrew D.A. Maidment 3, and Predrag R. Bakic 4 1 Department of Radiology, University of Pennsylvania, Philadelphia PA

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

Segmentation and Tracking of Partial Planar Templates

Segmentation and Tracking of Partial Planar Templates Segmentation and Tracking of Partial Planar Templates Abdelsalam Masoud William Hoff Colorado School of Mines Colorado School of Mines Golden, CO 800 Golden, CO 800 amasoud@mines.edu whoff@mines.edu Abstract

More information

ADVANCED IMAGE PROCESSING METHODS FOR ULTRASONIC NDE RESEARCH C. H. Chen, University of Massachusetts Dartmouth, N.

ADVANCED IMAGE PROCESSING METHODS FOR ULTRASONIC NDE RESEARCH C. H. Chen, University of Massachusetts Dartmouth, N. ADVANCED IMAGE PROCESSING METHODS FOR ULTRASONIC NDE RESEARCH C. H. Chen, University of Massachusetts Dartmouth, N. Dartmouth, MA USA Abstract: The significant progress in ultrasonic NDE systems has now

More information

Landmark-based 3D Elastic Registration of Pre- and Postoperative Liver CT Data

Landmark-based 3D Elastic Registration of Pre- and Postoperative Liver CT Data Landmark-based 3D Elastic Registration of Pre- and Postoperative Liver CT Data An Experimental Comparison Thomas Lange 1, Stefan Wörz 2, Karl Rohr 2, Peter M. Schlag 3 1 Experimental and Clinical Research

More information

Data mining for neuroimaging data. John Ashburner

Data mining for neuroimaging data. John Ashburner Data mining for neuroimaging data John Ashburner MODELLING The Scientific Process MacKay, David JC. Bayesian interpolation. Neural computation 4, no. 3 (1992): 415-447. Model Selection Search for the best

More information

Using Probability Maps for Multi organ Automatic Segmentation

Using Probability Maps for Multi organ Automatic Segmentation Using Probability Maps for Multi organ Automatic Segmentation Ranveer Joyseeree 1,2, Óscar Jiménez del Toro1, and Henning Müller 1,3 1 University of Applied Sciences Western Switzerland (HES SO), Sierre,

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

Comparing Gang Scheduling with Dynamic Space Sharing on Symmetric Multiprocessors Using Automatic Self-Allocating Threads (ASAT)

Comparing Gang Scheduling with Dynamic Space Sharing on Symmetric Multiprocessors Using Automatic Self-Allocating Threads (ASAT) Comparing Scheduling with Dynamic Space Sharing on Symmetric Multiprocessors Using Automatic Self-Allocating Threads (ASAT) Abstract Charles Severance Michigan State University East Lansing, Michigan,

More information

High-Performance Image Registration Algorithms for Multi-Core Processors. A Thesis. Submitted to the Faculty. Drexel University

High-Performance Image Registration Algorithms for Multi-Core Processors. A Thesis. Submitted to the Faculty. Drexel University High-Performance Image Registration Algorithms for Multi-Core Processors A Thesis Submitted to the Faculty of Drexel University by James Anthony Shackleford in partial fulfillment of the requirements for

More information

Hybrid Approach for MRI Human Head Scans Classification using HTT based SFTA Texture Feature Extraction Technique

Hybrid Approach for MRI Human Head Scans Classification using HTT based SFTA Texture Feature Extraction Technique Volume 118 No. 17 2018, 691-701 ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version) url: http://www.ijpam.eu ijpam.eu Hybrid Approach for MRI Human Head Scans Classification using HTT

More information

CT NOISE POWER SPECTRUM FOR FILTERED BACKPROJECTION AND ITERATIVE RECONSTRUCTION

CT NOISE POWER SPECTRUM FOR FILTERED BACKPROJECTION AND ITERATIVE RECONSTRUCTION CT NOISE POWER SPECTRUM FOR FILTERED BACKPROJECTION AND ITERATIVE RECONSTRUCTION Frank Dong, PhD, DABR Diagnostic Physicist, Imaging Institute Cleveland Clinic Foundation and Associate Professor of Radiology

More information

Lecture 13 Theory of Registration. ch. 10 of Insight into Images edited by Terry Yoo, et al. Spring (CMU RI) : BioE 2630 (Pitt)

Lecture 13 Theory of Registration. ch. 10 of Insight into Images edited by Terry Yoo, et al. Spring (CMU RI) : BioE 2630 (Pitt) Lecture 13 Theory of Registration ch. 10 of Insight into Images edited by Terry Yoo, et al. Spring 2018 16-725 (CMU RI) : BioE 2630 (Pitt) Dr. John Galeotti The content of these slides by John Galeotti,

More information

Performance impact of dynamic parallelism on different clustering algorithms

Performance impact of dynamic parallelism on different clustering algorithms Performance impact of dynamic parallelism on different clustering algorithms Jeffrey DiMarco and Michela Taufer Computer and Information Sciences, University of Delaware E-mail: jdimarco@udel.edu, taufer@udel.edu

More information

3D Surface Reconstruction of the Brain based on Level Set Method

3D Surface Reconstruction of the Brain based on Level Set Method 3D Surface Reconstruction of the Brain based on Level Set Method Shijun Tang, Bill P. Buckles, and Kamesh Namuduri Department of Computer Science & Engineering Department of Electrical Engineering University

More information

Intelligent 3DHD Visualization for Microsurgery

Intelligent 3DHD Visualization for Microsurgery Intelligent 3DHD Visualization for Microsurgery Neurosurgery TrueVision has been used as both the primary and secondary visualization for neurosurgical procedures that include aneurysms, brain and pituitary

More information

Image registration for motion estimation in cardiac CT

Image registration for motion estimation in cardiac CT Image registration for motion estimation in cardiac CT Bibo Shi a, Gene Katsevich b, Be-Shan Chiang c, Alexander Katsevich d, and Alexander Zamyatin c a School of Elec. Engi. and Comp. Sci, Ohio University,

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

Analysis of Functional MRI Timeseries Data Using Signal Processing Techniques

Analysis of Functional MRI Timeseries Data Using Signal Processing Techniques Analysis of Functional MRI Timeseries Data Using Signal Processing Techniques Sea Chen Department of Biomedical Engineering Advisors: Dr. Charles A. Bouman and Dr. Mark J. Lowe S. Chen Final Exam October

More information

Prostate Detection Using Principal Component Analysis

Prostate Detection Using Principal Component Analysis Prostate Detection Using Principal Component Analysis Aamir Virani (avirani@stanford.edu) CS 229 Machine Learning Stanford University 16 December 2005 Introduction During the past two decades, computed

More information

Volumetric Deformable Models for Simulation of Laparoscopic Surgery

Volumetric Deformable Models for Simulation of Laparoscopic Surgery Volumetric Deformable Models for Simulation of Laparoscopic Surgery S. Cotin y, H. Delingette y, J.M. Clément z V. Tassetti z, J. Marescaux z, N. Ayache y y INRIA, Epidaure Project 2004, route des Lucioles,

More information

Using K-means Clustering and MI for Non-rigid Registration of MRI and CT

Using K-means Clustering and MI for Non-rigid Registration of MRI and CT Using K-means Clustering and MI for Non-rigid Registration of MRI and CT Yixun Liu 1,2 and Nikos Chrisochoides 2 1 Department of Computer Science, College of William and Mary, enjoywm@cs.wm.edu 2 Department

More information

Statistical Analysis of MRI Data

Statistical Analysis of MRI Data Statistical Analysis of MRI Data Shelby Cummings August 1, 2012 Abstract Every day, numerous people around the country go under medical testing with the use of MRI technology. Developed in the late twentieth

More information

arxiv: v1 [cs.cv] 6 Jun 2017

arxiv: v1 [cs.cv] 6 Jun 2017 Volume Calculation of CT lung Lesions based on Halton Low-discrepancy Sequences Liansheng Wang a, Shusheng Li a, and Shuo Li b a Department of Computer Science, Xiamen University, Xiamen, China b Dept.

More information

Advanced Operating Systems (CS 202) Scheduling (2)

Advanced Operating Systems (CS 202) Scheduling (2) Advanced Operating Systems (CS 202) Scheduling (2) Lottery Scheduling 2 2 2 Problems with Traditional schedulers Priority systems are ad hoc: highest priority always wins Try to support fair share by adjusting

More information

CHAPTER 3 TUMOR DETECTION BASED ON NEURO-FUZZY TECHNIQUE

CHAPTER 3 TUMOR DETECTION BASED ON NEURO-FUZZY TECHNIQUE 32 CHAPTER 3 TUMOR DETECTION BASED ON NEURO-FUZZY TECHNIQUE 3.1 INTRODUCTION In this chapter we present the real time implementation of an artificial neural network based on fuzzy segmentation process

More information

Silhouette-based Multiple-View Camera Calibration

Silhouette-based Multiple-View Camera Calibration Silhouette-based Multiple-View Camera Calibration Prashant Ramanathan, Eckehard Steinbach, and Bernd Girod Information Systems Laboratory, Electrical Engineering Department, Stanford University Stanford,

More information

Introduction to Parallel Computing

Introduction to Parallel Computing Introduction to Parallel Computing Introduction to Parallel Computing with MPI and OpenMP P. Ramieri Segrate, November 2016 Course agenda Tuesday, 22 November 2016 9.30-11.00 01 - Introduction to parallel

More information

Use of MRI in Radiotherapy: Technical Consideration

Use of MRI in Radiotherapy: Technical Consideration Use of MRI in Radiotherapy: Technical Consideration Yanle Hu, PhD Department of Radiation Oncology, Mayo Clinic Arizona 04/07/2018 2015 MFMER slide-1 Conflict of Interest: None 2015 MFMER slide-2 Objectives

More information

Image Compression: An Artificial Neural Network Approach

Image Compression: An Artificial Neural Network Approach Image Compression: An Artificial Neural Network Approach Anjana B 1, Mrs Shreeja R 2 1 Department of Computer Science and Engineering, Calicut University, Kuttippuram 2 Department of Computer Science and

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

Distance Transforms in Multi Channel MR Image Registration

Distance Transforms in Multi Channel MR Image Registration Distance Transforms in Multi Channel MR Image Registration Min Chen 1, Aaron Carass 1, John Bogovic 1, Pierre-Louis Bazin 2 and Jerry L. Prince 1 1 Image Analysis and Communications Laboratory, 2 The Laboratory

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

Computational Fluid Dynamics as an advanced module of ESP-r Part 1: The numerical grid - defining resources and accuracy. Jordan A.

Computational Fluid Dynamics as an advanced module of ESP-r Part 1: The numerical grid - defining resources and accuracy. Jordan A. Computational Fluid Dynamics as an advanced module of ESP-r Part 1: The numerical grid - defining resources and accuracy Jordan A. Denev Abstract: The present paper is a first one from a series of papers

More information

Hybrid Formulation of the Model-Based Non-rigid Registration Problem to Improve Accuracy and Robustness

Hybrid Formulation of the Model-Based Non-rigid Registration Problem to Improve Accuracy and Robustness Hybrid Formulation of the Model-Based Non-rigid Registration Problem to Improve Accuracy and Robustness Olivier Clatz 1,2,Hervé Delingette 1, Ion-Florin Talos 2, Alexandra J. Golby 2,RonKikinis 2, Ferenc

More information

Tetrahedral Mesh Generation for Medical Imaging

Tetrahedral Mesh Generation for Medical Imaging Tetrahedral Mesh Generation for Medical Imaging Andriy Fedorov 1,2,3, Nikos Chrisochoides 1,2,3, Ron Kikinis 2, and Simon Warfield 2,3 1 Department of Computer Science, College of William and Mary, Williamsburg,

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

Automatic segmentation of cortical vessels in pre- and post- tumor resection laser range scan images

Automatic segmentation of cortical vessels in pre- and post- tumor resection laser range scan images Automatic segmentation of cortical vessels in pre- and post- tumor resection laser range scan images Siyi Ding* a, Michael I. Miga b, Reid C. Thompson c, Ishita Garg b, Benoit M. Dawant a a Dept. of Electrical

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

GE Healthcare CLINICAL GALLERY. Discovery * MR750w 3.0T. This brochure is intended for European healthcare professionals.

GE Healthcare CLINICAL GALLERY. Discovery * MR750w 3.0T. This brochure is intended for European healthcare professionals. GE Healthcare CLINICAL GALLERY Discovery * MR750w 3.0T This brochure is intended for European healthcare professionals. NEURO PROPELLER delivers high resolution, motion insensitive imaging in all planes.

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

Automated Medical Image Registration Using the Simulated Annealing Algorithm

Automated Medical Image Registration Using the Simulated Annealing Algorithm Automated Medical Image Registration Using the Simulated Annealing Algorithm Ilias Maglogiannis 1 and Elias Zafiropoulos 2 1 University of the Aegean, Dept. of Information and Communication Systems Engineering

More information