Home > ICT-onderzoek NL > Uitgelicht > Lilit Axner
 
Personal
 

My name is Lilit Axner and I am a Ph.D. student at the University of Amsterdam, Institute of Informatics, Section Computational Science (SCS). I was born and lived most of my life in the capital of Armenia, Yerevan, where I also received my high school education. After high school I entered to the State Engineering University of Armenia (SEUA), Department of Computer Systems and Informatics. I graduated with honors and have been awarded the bachelor's degree in Engineering in the field of Software for Computing Techniques and Automated Systems. During my studies I worked as a senior laboratory assistant and later as an engineer-programmer in the same university, at the Faculty Development Center (FDC). Meantime, I completed a course in Wired education during the CEENET 2001 workshop on Network Technologies, in Budapest, Hungary. While working in FDC I also had a chance to study at the Cisco Network Academy Program and later worked as an instructor at Cisco Regional Academy at SEUA.

 

Although my interests in the world of computer science were widespread, I still felt the lack of opportunity to explore more in this area and was eager to get a higher education in Europe. In 2000/2001 Armenia was just recovering from war and earthquake and step by step assuring her situation as an independent republic after the collapse of the Soviet Union. The high technologies were furiously starting to grow and in the university we had the first "generous" 32MB per month limitation for Internet traffic access. I have used this great possibility to search the opportunities for study in Europe. Luckily I found a program organized by Swedish Institute organization that was granting scholarships to students from Eastern European countries to study for a master's degree in Sweden. After several exams I won this scholarship and was informed by the coordinator at the Royal Technical University (KTH) in Stockholm, department of Numerical Analysis and Computing Science (NADA) Lennart Edsberg that I was accepted for this program. After one and a half year study, my final thesis was about the Parallel Implementation of a Redundant Wavelet Transformation. This was my first step towards the research in the merged world of high performance computing with physics. I realized that this is the direction that attracts me most and where I can enlarge my knowledge in the exciting world of high performance computing and apply them to real life problems. During the writing of my master's thesis I already knew that I wanted to continue my research in this direction.

 
I eagerly started to look for the Ph.D. position and found the website of SCS (UvA), that announced an open position for Ph.D. research in the area of Massive Parallel Virtual Particle Models. More precisely, it is about high performance computing in computational hemodynamics. From my further study of the topic, I discovered the importance and usefulness of this types of models in the field of surgery, e.g. treatments of vascular disorders caused by atherosclerotic diseases. Here the computer based training and decision support for e.g. preoperative planning is rapidly gaining importance. I immediately send an email to the head of SCS, Prof. Peter M. A. Sloot, expressing my interest and wish to apply for this position. We had a discussion over the theme and my background and I received a mail informing that I have been accepted for this position. I was very happy to get the further chance of research in this area.

Research
 

The project is called DIME: Distributed Interactive Medical Exploratory and is founded by the Netherlands Organisation for Scientific Research (NWO). I am carrying out this research together with my promotor Peter M. A. Sloot and co-promotor Alfons Hoekstra.

 

In close collaboration with the Leiden University Medical Center (Prof. Hans Reiber, partner in this project) we have developed HemoSolve, a Problem Solving Environment (PSE) for image based hemodynamics. A PSE is a computer system that provides all the computational facilities necessary to solve a target class of problems [1, 2]. The target class of problems that we chose is associated with cardiovascular diseases, a predominant cause of death, in particular due to the vascular disorders caused by atherosclerosis. The goal of HemoSolve is to provide a fully integrated environment for simulation of blood flow in patient specific arteries.

Because of the complex structure of the human vascular system it is not always obvious how to solve the problem of bypass and/or stent placement on the deformed part of the artery. HemoSolve can serve as a pre-operational training for medical students to enlarge their medical skills and also as an environment for biomedical engineers that study e.g. new stent designs. Moreover HemoSolve is merged with Grid technology, thus offering a unified access to different and distant computational and instrumental resources.

 
HemoSolve consists of four parts:

1. Medical data segmentation to obtain arteries of interest

2. 3D editing and mesh generation, to prepare for the flow simulation

3. Flow simulation, computing of blood flow during systole

4. Analyzes of the flow, pressure, and stress fields


Fig.1

Abdominal aorta: main stages of HemoSolve. Segmented data (a) is first cropped (b) and inlet/outlet layers are added (c) and the mesh is generated (d). Simulation results of created mesh are presented (e).

The goal of the segmentation is to automatically find the lumen border between the blood and non-blood, i.e. the vessel wall, thrombus or calcified plague. The generated 3D surface model serves as an input for the 3D editing tool. Here students can execute their experimental visualization studies on realistic geometries by cropping parts of the artery, adding inlet/outlet layers and enhancing it with structures like bypasses or stents. The final stage is the mesh generation which used in next stage by flow simulators. Two different computational hemodynamic simulators can be used in HemoSolve:This new sparse-based lattice Boltzmann solver shows excellent performance on both vector machines and Intel Xeon clusters.  

1. Fully parallelized Lattice Boltzmann method (LBM)

2. Finite element method (FEM)

In both solvers the flow is time-harmonic and after simulation the pressure, velocity and shear stress fields during one harmonic period are produced and can be visualized in last stage. One of the visualization techniques used in HemoSolve is the simulated pathline visualization [3].

 

After multiple examinations and observations of HemoSolve we concluded that the improvement and speed-up of the environment, especially of the flow solver is needed. This became the next challenging step of our research. Even though the LBM is fully parallelized it is still hampered by extreme memory consumption and remain compute intensive. Moreover, frequently the partial refined meshes at the parts of interest of geometries are needed. With the current LBM solver the fluid flow simulation in partially refined meshes is not possible.

 

In collaboration within NEC research laboratory in Germany, Regional Rechenzentrum Erlangen (RRZE) in Germany, High Performance Computing Center Stuttgart (HLRS) in Germany and the Institute for Computer Application in Civil Engineering (CAB) in Germany we created a new sparse-data based lattice Boltzmann fluid flow simulator. This sparse-based structure allows us to avoid storing of unnecessary high solid fractions of data structure thus avoiding the extreme memory consumption, and exclude extra computational time over solid nodes. The indirect address-lookup algorithm allows an easy implementation of mesh refinement. And the efficient implementation of graph partitioning algorithm based on METIS [4, 5] library generates high quality partitions by preserving memory load balance and minimizing the total communication volume.


 

 

Fig. 2

(a) - Abdominal aorta partitioned into six partitions via METIS based partitioning software. Bounding box size is 353x115x63, 5% fluid cells.

(b) - Porous media partitioned into sixteen partitions via METIS based partitioning software. Bounding box size is 780x122x122, 43% fluid cells.

The last step is the replacement of current LBM by the new sparse-based lattice Boltzmann solver. This improvement of the flow solver will make HemoSolve fast, accurate and less memory consumable. It will also enlarge the opportunities of the user by allowing the refinement (zooming) of the part of interest (deformed parts) of arteries for better examinations.

 

Further information

More about our research can be found in e.g.:

 

A.M.M. Artoli; A.G. Hoekstra and P.M.A. Sloot: Mesoscopic simulations of systolic flow in the Human abdominal aorta, Journal of Biomechanics, vol. 39, nr 5 pp. 873-884. 2006. (DOI: 10.1016/j.jbiomech.2005.01.033)

 

E.V. Zudilova and P.M.A. Sloot: Bringing Combined Interaction to a Problem Solving Environment for Vascular Reconstruction, Int. J. Future Generation Computer Systems, vol. 21, nr 7 pp. 1167-1176. 2005.

 

L. Abrahamyan; J.A. Schaap; A.G. Hoekstra; D.P. Shamonin; F.M.A. Box; R.J. van der Geest; J.H.C. Reiber and P.M.A. Sloot: A Problem Solving Environment for Image-Based Computational Hemodynamics, Lecture Notes in Computer Science, vol. 3514, pp. 287-294. Springer, Berlin, Heidelberg, May 2005. ISBN 3-540-26032-3.

 
References

1. E. N. Houstis and J. R. Rice, Math. and Comp. in Sim. 54, 243 (2000);

2. Ch. A. Taylor and M. T. Draney, Annu. Rev. Fluid. Mech. 36, 197 (2004);

3. D. A. Steinman, Annals of Biom. Eng. 30, 483 (2002);

4. G. Karypis and V. Kumar, J. Parallel Distrib. Comput., 48:1, 96 (1998);

5. A. Abou-Rjeili and G. Karypis, IEEE Int. Par. & Distr. Proc. Symp. (IPDPS), (in press) (2006).

 


Uitgelicht

Evenementen ICT

ICT.OPEN 2012

Stelling

Roadmap ICT

I/O Magazine
                   

Uitgelicht

Evenementen ICT

ICT.OPEN 2012

Stelling

Roadmap ICT

I/O Magazine
01-01-2013
Russische universiteit zoekt Nederlandse partners in IT
17-12-2012
ICT Personality Award 2012
11-11-2012
Sentinels
» Meer nieuws

10-04-2013
Sense of Contact 15
» Activiteit toevoegen
» Volledige kalender

ICT OPEN 2012


» Lees verder

Wat is de grootste uitdaging?
Complexiteit
Maatschappelijke innovatie
Wetenschappelijke paradigma's
Voldoende gekwalificeerde wetenschappers
 

» Resultaten stelling

Roadmap ICT topsectoren

» Lees verder

I/O - december 2012

» Lees verder