swift pre rendering volumetric visualization of magnetic ...424 nikhilkumar p. patel et al. /...

Procedia Technology 14 ( 2014 ) 422 – 429

2212-0173 © 2014 Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).Peer-review under responsibility of the Organizing Committee of ICIAME 2014.doi: 10.1016/j.protcy.2014.08.054

Available online at www.sciencedirect.com

ScienceDirect

2nd International Conference on Innovations in Automation and Mechatronics Engineering, ICIAME 2014

Swift Pre Rendering Volumetric Visualization Of Magnetic Resonance Cardiac Images Based On Isosurface Technique

Nikhilkumar P. Patela* ,Shankar K. Parmarb† ,Kavindra R. Jainc‡

a,b,c G.H.Patel College of Engineering & Technology,V.V.Nagar-388120,India

Abstract

Magnetic Resonance imaging (MRI) is a medical imaging procedure which uses strong magnetic fields and radio waves to produce cross sectional images of organs and internal structures of the body. Three dimensional (3D) models of CT is available and it has been practiced by almost all radiologists for pre-diagnosis. But in MRI still there is a scope for researcher to improvise a 3D model. Two dimensional images are taken from different viewpoints to reconstruct them in 3D, which is known as rendering process. In this paper, we have proposed a rendering concept for Medical (cardiac MRI) images based on iso values and number of marching cubes. Designercan place colors and textures over the 3D model to make it look realistic. This makes it easier for people to observe and visualize a substance in a better sense. The algorithm basically works on triangulation methods with various isovalue and different combination of marching cube pairs. As a result of an application of marching cube concept, volumetric data (voxels) is generated. Voxels are then arranged and projected to visualize a 3D scene. Approximate processing time for various iso values are also compared in this paper.

© 2014 The Authors. Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Organizing Committee of ICIAME 2014.

Keywords:Marching Cubes;Rendering;Magnetic Resonance Imaging-MRI;Three dimension(3D);Voxels;isosurface.

* Student in the masters of communication Engineering,GCET, V.V.Nagar.(email: [email protected])Tel.: +91-8866328620.† Assistant Professor in Electronics & Communication Engineering department, GCET, V.V.Nagar. (email :[email protected])‡ Assistant Professor in Electronics & Communication Engineering department, GCET, V.V.Nagar .(email: [email protected])

© 2014 Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).Peer-review under responsibility of the Organizing Committee of ICIAME 2014.

423 Nikhilkumar P. Patel et al. / Procedia Technology 14 ( 2014 ) 422 – 429

1. Introduction

3D imaging represents an advanced technology in the area of medical imaging. We all are interested to watch high quality videos, movies or three dimensional (3D) movies for just enjoyment but if we are talking about medical images and its ability to provide better diagnosis for radiologist, it has its own importance. The technology has been witnessing increasing popularity and adoption due to benefits offered such as faster and accurate diagnosis.A revolutionary image processing technique will sort out various problem related with poor quality, low resolution (or contrast) medical (x-ray, CT, MRI) images for doctors [3].

Traditionally, a description of the 3D scene being rendered is provided by a detailed and complex model of the scene. To avoid the expense of modelling a complicated scene, it is sometimes more convenient to capture a scene from different viewpoints. To create images for novel viewpoints that were not captured, an interpolation scheme may be applied. This whole process of modelling a 3D image is called image-based rendering [4].Same concept may be applicable for medical images for betterment of pre-diagnosis for radiologist. 3D CT (3-dimensional computed tomography rendering of virtual images has been used in a surgery curriculum, as well as in surgical planning for human medicine[6].

An application of 3D rendering can be exemplified with the help of artificial bone fitting. Earlier for artificial bone fitting, orthopedic/radiologist will observe x-ray images taken from all sides and by visualizing those x-ray image he/she will come to know about perfect dimension (especially depth & width) of bone. This process was time consuming and less accurate. The problem being faced can be resolved using 3D image model (rendering) from raw image database [1,6].

1.1 Magnetic Resonance Imaging.

The MRI technology makes use of the hydrogen atoms present in form of molecules inside the human body as shown in fig.1(a). When the magnetic field applied on the protons present in the hydrogen atoms, they are aligned in one direction as shown in fig.1(b).After reaching this stage Radio waves are imparted on human body. When the body lies in a magnet, it becomes temporarily magnetized. This state is achieved when the hydrogen nuclei in the body align with the magnetic field. When magnetized, the body responds to exposure to radio waves at a particular frequency by sending back a radio wave signal at resonance and that will generate image(MRI).[1]

a b

Fig. 1. Photon Movement in human body [(a) normal (b) After Magnetic Field]

1.2 Research Case

Medical Imaging is constant growing technology for betterment of diagnosis for radiologist (Doctors).The post processing of medical image include various processes like image correction ,edge detection, image negative, some morphological operations and rendering(3D visualization). Visualization of medical images in three dimension (3D) would enable a better perception of surgeon for pre-diagnosis(for surgery, Precautions).Earlier for High Quality Magnetic resonance imaging(MRI) doctors used special serum or they may go for higher field MRI(3Tesla,7Tesla) MRI scanner(very costly) .Limitation of such processes lead to the use of rendered MRI images. In this paper we


have developed and tabulated 3D render view of cardiac/chest MRI with various iso values along with respective processing time.

2. Materials and Methods

2.1 Image Rendering

Rendering is a computer graphics process for visualizing 2D images with 3D effects without changing hardware. Rendering with graphical processing unit will boost up speed of process. There are two major types of rendering.

Real-Time Rendering (GPU): Real-Time Rendering is used most prominently in gaming and interactive graphics, where images must be computed from 3D information at interactivity speed.Offline or Pre-Rendering (CPU): Offline rendering is used in situations where speed is less of an issue, with calculations typically performed using multi-core CPUs rather than dedicated graphics hardware.

The major difference in this technique is of speed at which these images are computed, finalized and displayed. In case of Offline (CPU) rendering processing, dedicated graphics memory is not required. The rendering process is primarily based on ‘Marching Cubes’ Algorithm along with selection of proper iso values. The rendered image quality depends on value of ‘iso’(i.e 10,30,80,100 taken for simulation) as shown in fig.2(a). Effective shapes would be generated from voxels by use of cubes as shown in fig. 2(b).

Fig. 2. (a) Impact of ISO selection (b) Vertices (intersection) Calculation.

In Fig. 2(a), the lowest (Vmin) and highest (Vmax) vertex value are represented by a point in span space. Given an isovalue (Viso), only cells that satisfy both (Vmin iso and Vmax iso) and contain the isosurface are further processed.Midpoints of cube (eij) are calculated using equation 1.

(1)

Where, ,i jv v = Vertex values ; ije = Intermediate edge points of cube and ie = Edge midpoints.

(2)

Where, (3)

.(1 )j jij

i j

v j v iev v

0 0 0( ) ( , , ) ( )

n n n

n ijk ijkk j i

V x V i j k L x

w

0 0 0

( )( )( )( )( )( )( )

n n nu v

ijki u j v k wu v w

u i v j w k

x x y y z zL xx x y y z z

a b


Where, i,j,k are coordinates of voxel cube; n: Selected vertices point and x,y,z are 3D mesh index value.

a b c

d e

Fig. 3. Pre-Processing For Marching Cube.(A)Image Object (B) Segmented Image (C) Index Point Finding(inside Image).(D)Interpolated Points(Final Indexing).(E) Join All indexed points.

As shown in Fig. 3 step by step rendering process is formulated. The given image is first of all divided in to small parts (Fig.3(b)) after that the intersection between image and 3D mesh is being done(Fig.3(c)).The points outside the boundary of the image region are calculated using Lagrange’s Interpolation(Eq. 2) as shown in Fig. 3(d). As a next step all points are joined to make a 3D images from vertex values using Table 2

2.2 Marching Cube

The standard marching cube constructs a facetized isosurface by processing the data set in a sequential, cube-by-cube (scanline) manner. Thus, the approach sses the m cubes of the set in sequential order: C1,C2,C3,..,Cm(Fig. 4).During this processing, each cube vertex Vi that has a value equal to or above the iso value is marked; all other vertices are left unmarked. [13]

Fig. 4 The 15 basic Marching cube (C1, C2,.., C15) pair.


The isosurface intersects each cube edge Ei terminated by one marked vertex Vi, and one unmarked vertex Vi any cube that contains an intersected edge is active. The computations that ive cubes can be viewed as the active cube determination component of marching cube. This component can be implemented as a stand-alone early processing step or integrated with other processing, but in either case it involves data set traversal in sequential, forward-marching order. Since each of the eight vertices of a cube can be either marked or unmarked, there are 256 (28) possible marking scenarios for a cube. Each cube marking scenario encodes cube isosurface intersection pattern (i.e., con

a b

c

d

Fig. 5.Render Image view from voxels (a) Points for Triangulation (b) Join them by 15 cube pair (c)Triangulation from Boundary line(isosurfaces) (d) Render View.

We are mainly focus to apply rendering algorithm to chest/cardiac Magnetic Resonance Images (MRI).

3. Proposed work

Block diagram of proposed approach is shown in Fig. 6

Fig. 6. Proposed block diagram for MRI rendering.

It is known that MRI images are taken at various point of time to cover different cross sectional view of same part of a body .As a first step of simulation to render MRI images is to take MRI raw data and arrange them in a vector form from MRI slices having time as a reference. Generated vector is in such a form that time dimension is now squeezed. It is required that we need to fix iso value at this juncture before moving to next step. This vector is further processed in next step to generate 3D mesh and to find intersected points with MRI vector data based on predefined iso value. Algorithm steps to achieve a result up to this point are described in table 1. Due to time complexity to achieve the simulation result we have divided whole process in two steps which are part 1 and 2 of proposed algorithm.


Table 1. Algorithm for rendering process.

Algorithm Part 1

1. Take Raw MRI(Cardiac) data.2. 3D mesh generation for image partition.3. Find Intersected points of given Image for given iso value.4. If all points covers image & are within the limit of iso values go to step 5 else use interpolation to find mid-point.5. Triangulation method for boundary finding..6. Move to Algorithm 2(Marching Cubes).

Part 2 of a rendering process is to create and project voxels as shown in Table 2. Intersection point obtained at the end of part1 are applied for shape generation using triangulation method as described in section 2.2. Duplication of vertices formed in triangulation method which can form duplication of isosuface , are removed by arranging vertices in vectorized array form. Hence the speed of rendering is boosted up. Finally, they will be visualized by projecting this array on the screen.

Table 2. 3D faces from Voxels (volumetric Pixel) Generation.

Algorithm Part 2

1. From calculated index find edges (lookup table) for triangulation process.

2. By using scalar value in each vertex of edge, find surface and edge intersection by interpolation (bilinear).

3. Put all surfaces in vector space for plotting.

4. Remove duplicate vertices by differentiation after sorting of vertex vector process.

5. Visualization of rendered image.

To achieve quality rendered image at least 15 unique pair of marching cubes must be used. If we further increase pair of marching cubes obviously it leads towards increased quality of rendered image at the cost of processing time required compared to time required with 15 pairs of marching cubes.

4. Simulation and Results

The achieved 3D MRI image rendering (CPU) based are tabulated in Table 3 with different iso values.The system used for this process is comprising of 2GB RAM,Intel Core i5 processor along with MATLAB 7 software.

Table 3. Simulated results of cardiac MRI database.

SliceTime

ISO=10 ISO=30 ISO=80 ISO=100

T1


T2

T3

T4

First column of table 3 represents the MRI(cardiac) image slices at different time instants (T1=10-15s,T2=20-25s,T3=30-35s,T4=40-45s).The simulated results are shown for iso values of 10,30,80,100.The time required for processing is tabulated in table 4 for different iso value cases. It can be inferred from Table 4 that the iso value of 30 is giving better results for this database [16].

Table 4 Simulation processing time for particular iso values.

MRI slice Ti T1 T2 T3 T4

ISO value 10 30 80 100 10 30 80 100 10 30 80 100 10 30 80 100

Approximate Processing Timewithout OpenGl

(minutes)

3 4 2 3 3.5 3 3 3 3 5 2.5 2.5 3 4 3 3.5

Approximate Processing Time

with OpenGl

(seconds)

15 23 10 15 17 15 15 15 15 30 13 13 15 23 15 17


5. Conclusion and Future work

3D imaging represents an advanced technology in the area of medical imaging. The technology has been witnessing increasing popularity and adoption due to benefits offered such as faster and accurate diagnosis. Thoughthe same thing without changing hardware is more challenging. The utmost solution to the above problem is successfully achieved and testified by 3D rendering algorithm based on marching cube for cardiac MRI data base along with different iso values and its processing time. Improvement in the processing speed is achieved by integrating OpenGl with MATLAB software on Magnetic Resonance Cardiac Images. In future, 3D reconstruction of magnetic resonance neural (brain) images can be carried out with some advancement as brain images are having more complexity compared to cardiac MRI.

Acknowledgements

We are sincerely thankful to Dr. Deepak V. Mehta (Radiologist, Shree Krishna Medical Hospital, Karamsad) and Dr. Hiren P. Patel (First Year Resident Doctor (M.D.), SSG Hospital,Vadodara) for their massive support in rendering immense knowledge of MRI physics and 3D techniques for this work. We also acknowledge cancerimagingarchive.net and Department of Computer science, YORK University for providing data base of cardiac MRI.

References

[1] Ulrich Bonse,Frank Busch,Olaf Günnewig,Felix Beckmann,Reinhard Pahl,Günter Delling,Michael Hahn,Walter Graeff . 3D computed X-ray tomography of human cancellous bon tial and 10 Bone and Mineral-Volume 25, Issue 1, ELSEVIER; April 1994. p. 25-38.

[2] Olivier Cartiaux,Laurent Paul,Bernard G. Francq,Xavier Banse,Pierre-Louis Docquier. Improved Accuracy with 3D Planning and Patient-Specific Instruments During Simulated Pelvic Bone Tumor Surgery. Annals of Biomedical Engineering,Volume 42, Issue 1(Volume 18),Springer link;2014. p. 205-213. ISSN 0938-7994

[3] Laura Faneacorresponding author and Andrew J. Fagan. Review:Magnetic resonance imaging techniques in ophthalmology. Mol Vis.; 2012.p. 2538–2560.PMCID: PMC3482169.

[4] Vishal Verma,Ekta Walia. 3D Rendering - Techniques and Challenges. International Journal of Engineering and Technology Vol.2(2); 2010. ISSN:0975-4024

[5] William R. Mark Leonard McMillan Gary Bishop. Post-Rendering 3D Warping. In Proceedings of 1997 Symposium on Interactive 3D Graphics, Providence, RI; April 27-30, 1997, p. 7-16.

[6] H. Lee,J. Kim,Y. Cho,M. Kim,N. Kim,K. Lee. Three-dimensional computed tomographic volume rendering imaging as a teaching tool in veterinary radiology instruction. Veterinarni Medicina, 55; 2010. p. 603–609.

[7] Chyi-Cheng Lin, Yu-Tai Ching. An efficient volume-rendering algorithm with an analytic Approach. Springer-Verlag(The Visual Computer ,Volume 12; 2006. p.515–526.

[8] Marcos Vinicius Mussel Cirne & Hélio Pedrini. Marching cubes technique for volumetric visualization accelerated with graphics processing units. Springer London(J B raz Comput Soc),Volume 19; 2013. p. 223–233. ISBN:1678-4804.

[9] Boguslaw Cyganek,J. Paul Siebert. Space Reconstruction and Multiview Integration. John Wiley & Sons, Ltd(An Introduction to 3D Computer Vision Techniques and Algorithms(Chapter 7); 2009. ISBN: 9780470699720

[10]Alexandre Beaudoin. A Thesis On Parallel Algorithms For Rendering Large 3D Models On A Graphics Cluster. The Department of Computer Science And Computer Engineering. Concordia University,montr´eal, Qu´ebec, Canada; 2012.

[11]D. Martin. A book on Clinical Radiology. ELSEVIER Journal Imprint: SAUNDERS, ISSN: 00099260.[12]Arie Kaufman and Klaus Mueller. Overview of Volume Rendering Chapter for the Visualization Handbook, eds. C. Johnson and C.

Hansen. Academic Press; 2005.[13]Timothy S. Newman,Hong Yi. A survey of the marching cubes algorithm. ELSEVIER, Science Direct (Computer & Graphics (30)); 2006.

p. 854-879.[14]NCBI MRI database from http://www.ncbi.nlm.nih.gov/ (Last visited on 3rd January, 2014).[15]2D image database taken from https://public.cancerimagingarchive.net/ (Last visited on 3rd January, 2014).[16]Cardiac MRI(.mat) data base from http://www.cse.yorku.ca/~mridataset/(Last visited on 3rd January,2014).

swift pre rendering volumetric visualization of magnetic ...424 nikhilkumar p. patel et al. /...

Documents