New Project

Distributed Interactive Medical Exploratorium for 3D Medical Images

A VR-based pre-surgical planning and teaching environment

DiME

Partners

 

Prof Dr. P. Sloot (UvA)

Dr. P.A.M. Kommers (UT)

Dr. B. Geelkerken  (MST)

Prof  Dr. JHC Reiber (RUL)

 

The focus of this research is explicitly on the medical care sector. There are 3 information components in the research: Data Integration: Coherently and reliable integration of distributed patient information specifically 3D images.

Information Retrieval: Automatic analyses of images, like extraction of 3D anatomical structures.

Data Presentation: Providing a VR-based cognitive interpretation/learning environment to extract medical knowledge from the information present.

Medical scanners produce three-dimensional pictures of the human body. These 3D data sets contain a large amount of information. In today’s clinical practice, much of the information is not utilised because the 3D data is examined by the medical experts (e.g. radiologists and surgeons) using two-dimensional greyscale images. Modern post-processing technology allows the extraction and visualisation of 3D anatomical structures from these large data sets. These results will be included in the Electronic Patient Records (EPRs) of hospitals for subsequent dissemination to other and referring physicians. This is the data-delivery part of the system, and it is completely distributed, and based on ideas stemming from recent grid-computing research.

Technological developments will lead to pre-surgical planning and teaching applications, which will most likely result in better post-surgical results, lower health care costs and increased efficiency in the training of fellow-surgeons.

DiME brings the use of 3D images in medicine a few steps further along these lines. This is realised through combination of recent results in grid-computing, Visualisation/VR, and Image analysis, into a distributed data delivery and data enrichment test-bed for 3D medical images. In addition a training support system for learning and exploration in virtual reality will be developed. This is the data enrichment part of the project. The system will be applicable in any medical fields using 3D imaging, but we will tailor it towards the field of vascular surgery using Magnetic Resonance Angiography (MRA) as a prototypical example. The multidisciplinary character of DiME requires a close collaboration between 3 disciplines: Medicine, Informatics and Cognition. This is not trivial. Fortunately the 3 groups involved have long standing expertise in such multidisciplinary collaborations and have been working together previously in various projects.

 

We humans are highly visually oriented. We regularly transfer data into pictures, be it mentally, physically on a piece of paper, or virtually on a computer screen. Next we interpret these pictures by visual inspection or by more quantitative measurements. In any case we turn the data into information and the information into knowledge, and based on that knowledge we take appropriate actions.

 

In medicine, pictures play a very prominent role in diagnosis and planning of treatment, but also in training of medical professionals. A large number of techniques are currently available to medical professionals to obtain pictures from (the inside of) the human body. Many of these imaging modalities, such as Computed Tomography or Magnetic Resonance Imaging provide a three-dimensional (3D) data set of parts of a patient’s body. In typical use this 3D data is visualised as stacks or sequences of two-dimensional (2D) gray-scale pictures, such that the 3D information must be reconstructed mentally by the observer, which is far from ideal. Such series of 2D pictures are then visually inspected, after which a diagnosis is made or treatment is planned. It is generally acknowledged that more accurate knowledge can be drawn from the data, by displaying it as true 3D images (applying dedicated image segmentation software and using high-end graphical workstations or immersive virtual reality). When quantitative data analysis tools are applied, even more accurate knowledge can be obtained. An overlay of different imaging techniques, or the accumulation of the same image in a time series (e.g. to monitor progress of treatment), calls for even more advanced data analysis, visualisation, and exploration techniques. Finally, the availability of realistic 3D medical images and tools to explore these images provide new and exiting opportunities for training of medical specialists and pre-treatment planning procedures.

 

There are two main reasons that these technological opportunities have not yet reached the stage of regular use in medicine. First, the envisioned data analysis, visualisation and exploration requires much computational power, typically not available in hospitals. Secondly, until today a correct and accurate interpretation of 3D images presented in virtual reality is a tedious job and requires special and dedicated training and support.

 

It is our ambition to remove these two burdens, and make this technology widely available in clinical practice. First, by developing a distributed data delivery test bed in which the 3D data sets are read from a patient database, then transformed into medically meaningful 3D images, using remote compute servers, and finally sending the images to a low end (desktop) virtual reality system, installed in a hospital, for further analysis and interpretation. In a later stage, drawing from the results from another project that we are currently initiating, we also aim to allow roaming access to the images, using web-enabled mobile telephones or Personal Digital Assistants (PDA’s). Secondly, we aim to develop strategies for exploration and learning in virtual reality systems, thus enriching the data. In other words, turning hidden information into true medical knowledge. The exploration and learning environment will be focussed on supporting the education of surgeons in training.

 

Although the proposed test bed and exploration and learning procedures are generic and can be put into practice in any medical specialisation where such 3D imaging techniques are of use, in this project we specifically focus on vascular surgery. The project is a collaboration of a ‘computer science’ group from the University of Amsterdam (UvA), a ‘medically oriented computer science’ group from the Leiden University Medical Center (LUMC), a ‘cognition’ group from the University of Twente (UT) and medical specialists from LUMC (Dr. M. Wasser) and The Medical Spectrum Twente.

 

The LUMC group currently runs another project (financed a.o. by Senter, and in collaboration with MEDIS medical imaging systems bv, 2Cure bv, and a hospital in Haarlem) in which results obtained from medical images are to be included into the hospital patient information system in Haarlem, thus enriching the EPD with this information.

 

Transforming Data into Information

Recent achievements in the three fields of Computer Science research, being Distributed Computing and Networking, Visualisation and Virtual Reality, and Imaging and Image analysis, open up a new and exiting avenue in medical diagnosis, pre-treatment planning, and training of physicians. It now becomes possible to extract much more information from 3D data sets taken from patients in medical scanners, and to present this data to medical professionals in a time- and place independent way, thus allowing them to obtain the desired knowledge in a much more flexible and efficient way.

One main research goal of this project is to integrate the knowledge of these three fields into a test bed to demonstrate the viability and usability of a multimedia, distributed, interactive information system for 3D medical images. Building upon results of previous research, we specifically focus on the field of vascular surgery. Nevertheless, the system that we envision has general applicability in any medical specialisation requiring access to and analysis of 3D images.

 

Figure 1 presents a schematic overview of the functional components of the proposed test bed. Medical images are acquired in a scanner, e.g. by Magnetic Resonance Angiography, which is a non-invasive and therefore patient-friendly technique specifically suited for the generation of high resolution images of vascular structures. The raw data is stored in a database for further analysis. Next, using advanced image segmentation techniques, relevant 3D structures such as arteries, are extracted from the raw data. The 3D structures may then be stored in a database for later visualisation, or may be sent directly to a visualisation and exploration environment. Depending on the specific application, the visualisation and exploration environment may be anything ranging from a high-end virtual reality environment such as the Cave to a desktop PC with stereoscopic viewing, or even the screen of a hand held device such as a PDA or a 3G WAP enabled mobile telephone.

              

Figure 1 : A scheme showing the main components of the test bed. The picture on the right shows an example of a Cave session, visualising the lower abdominal arteries, together with some exploration tools in action.

 

A key feature of this infrastructure is that all components are distributed. The MRI scanner and database for raw data may be situated in e.g. Leiden, the image segmentation software may be running on a compute server located in Amsterdam, and the visualisation and exploration clients may be in any location, depending on the end user. In other words, we assume time- and place-independent access to data and compute servers. Seamless integration therefore requires a dedicated middleware layer, which we base on grid technology (Globus) and a specifically designed client-server system.

 

We envision many user scenarios for our test bed. In a diagnostic setting a patient may go to a hospital with MRI facilities close to his home, and after the scan has been acquired this data is stored in a database. The medical specialist who requested the scan subsequently will be informed that the data is ready for analysis (through e.g. email, SMS or WAP inline instruction). Next, the data may be segmented automatically (after a request from the specialist, or as part of standard procedure). This segmentation can be done in batch mode or in real time in an interactive session. Next, the medical specialist can visualise the data and draw his conclusions. Depending on the situation, the user may decide on ‘traditional’ 2D gray scale plots or on more advanced 3D visualisation using VR techniques. In this more advanced mode the user may also want to exploit powerful exploration tools. In the example of arteries, he may wish to measure e.g. the cross-sectional areas of the normal and enlarged vessel segments, calculate the vessel wall thickness, or calculate the extend and severity of any narrowing in the vessel. All this will also depend on the visualisation hardware that the user has available at the moment that he wants to consult the data, e.g. a hand held device at the patient’s bed or while travelling, a PC in the office, or an advanced virtual reality theatre for more complex analysis. Moreover, two or more people, not necessarily at the same location, may access and visualise the data and together analyse it.

 

LUMC and UvA currently initiate another project in which they attempt to calculate blood flow in the segmented native arterial structure, and subsequently assess the changes in the blood flow distribution as a result of an intervention, thereby mimicking a surgical intervention, such as a bypass operation, or a radiological intervention such as the placement of a stent. The structural information may now be supplemented with blood flow information and the overall system then becomes suitable for pre-treatment or pre-surgical planning. This however still requires much more work, but the results of this current proposal will be instrumental for the further evolution of the test bed into a pre-treatment planning system.

 

Another user scenario may be training of medical specialists. Here, a large number of images related to a range of relevant clinical cases, preferably in combination with calculated blood flows, may be stored in the database. The visualisation client now is embedded in an educational setting, and again in a time – and place independent way trainees (such a surgeons in training) may access the data and go trough a specific training cycle. It will be clear that this will improve significantly the training facilities for these medical experts. In addition, it will very likely be cost-effective and also decrease the burden on the senior surgeons who otherwise would have to train these surgeons in training. Even more important, the post-surgical results on the patients will improve as optimised surgical procedures can be selected from various choices (see also Section 6.3).

 

In this project we focus particularly on this final scenario. Furthermore, we assume that in this user mode, visualisation is always performed in a virtual reality setting, and that the visualisation of the data is supported by instructional strategies for learning in virtual reality (see next sections).

 

Relevance of Medical Training in Virtual Reality

The expectations of patients concerning the outcome of medical interventions are rapidly increasing. In the beginning of the last century the only demand about the outcome of medical interventions was preservation of one’s life. The diseases which were subject to intervention were life threatening and it was generally accepted that interventions were mutilating. Nowadays the treated diseases are more complex. The patients are older and consequently the tolerance of collateral damage is declined. Contrary, the modern patient does not accept any mutilation attributed to the disease nor to the intervention. Besides the preservation of patient life, the quality of life after the intervention becomes very important as well. To fulfil these demands of the “market”, surgical techniques become more complex (e.g. endoscopic surgery) and parallel to these more complex processes the learning curve to master these new techniques becomes steeper and longer. A nice example of this is the laparoscopic cholecystectomy. Willem van Erp in Eindhoven introduced this technique only 11 years ago in the Netherlands. Nowadays the laparascopic cholecystectomy is the preferred technique in many hospitals. However, the majority of the surgeons performing laparoscopic cholecystectomies are autodidactic. They heard about the technique on congresses. They visited clinical demonstrations in centres of excellence. Thereafter they planned the first procedures in their own hospital. It is not surprising that the results are not as good as reported in the literature in the early periods. The steep learning curve was moved on patients. A sufficient training and formal assessment of the surgical team before introducing the new technique into the hospital is not available. Moreover, more complex intervention techniques are rapidly developed and introduced in the daily practice. An example of this is the endovascular exclusion of infernal aortic aneurysms with an endograft. Cuijper recently reported in his thesis that only after an endovascular experience of 30 electively treated triple A the complication ratio is sloping down to acceptable levels. In the Netherlands only a few hospitals have such an experience. Also the first ruptured aortic aneurysms are treated endovascularly with very good outcome. However, this emergency procedure demands a steep endovascular experienced team available during day and night. Gaining enough experience with this procedure is not possible in most of the hospitals in the Netherlands. Unfortunately, it is not possible to transport patient with a ruptured triple A to centers of excellence because of hemodynamical instability. In other words, the patients do not survive delay of treatment due to transportation. The next generation, more complex endografts with the possibility of perirenal sealing is underway. The results of the first clinical experiments came from “down under”. It is clear that a perfect preoperative visualisation and planning and a dummy operation of the whole procedure is essential.

 

The conventional and still actual surgical teaching method, introduced more than a century ago, is a close daily working relation between the practised teacher (a consultant surgeon) and the unskilled pupil (the surgical resident). Working-weeks of 70 to 90 hours were accepted and after 6 years of gaining theoretical and especially practical skills under direct supervision of the consultant the resident becomes a surgeon. Nowadays our society does not accept such long periods of formal learning and the working week is shortened to a maximum of 48 hours. This results in a 40% decline of directly supervised practical experience of residents in their first years of surgical training. Moreover, the government asked the surgical society to offer the basic surgical training in only 5 years! On the other side the same government makes laws as the BIG and the WGBO. The surgeons are obliged to be qualified (formal licensed) and to be skilled and properly trained to offer and execute an intervention to or on a patient. Due to the “Schengen convention” there is a right of free establishing of the citizens of the European community in the participating countries. The formal training of resident and consultants in the European countries showed large diversities. Objective and proper methods for assessment and comparison of the outcome of the surgical training in the European countries are not available.

 

It is clear that the conventional surgical teaching methods do not fulfil the demands of patients, the society, the government and the surgical profession. New training methods have to be developed. The development of virtual reality surgical tools for theoretical and practical training and assessment of the resident is urgent. Moreover, virtual surgical tools should be available for life-long medical education and assessment of the surgical consultant maintaining a high level of expertise and skills in the profession. This project intends to contribute to the steep increasing need for practical training and objective assessment for the becoming surgeon.

Data Enrichment: Transforming Information into Knowledge

The tendency to allow trainees to explore critical situations and, in case of medical students, to let them experiment with an underlying model of the phenomena and processes in the human body, is not new; (Psotka, 1995, McLellan, 1993). There is a considerable amount of evidence that training by computerized models facilitates the learning process (Moshell and Hughes, 1994). Also we see a growing interest for offering fully 3D environments that allow a continuous fading between the full realistic reality like looking through glasses to the real patient and at the other side seeing the superimposed vectorized model of the same patient but now with only the structure of the blood vessels (and if available, the computed blood flow and the animated pressure on critical spots). In the pre-service training novice surgeons may primarily explore the models at the vector side of the spectrum. As more experience builds up the captured videos of the same patient may become involved. Finally students may undertake real operations, supervised by a specialist. As this stage of the training is costly, it is quite desirable that an intense VR stage of the learning has taken place. In this project we will establish the most effective instructional sequence in the VR-based training simulation.

Constructionism for experiential learning

The traditional approach to optimise learning through ‘just-in-time’ information and feedback can be summarised as instruction. As the trainee has a larger repertoire of prior knowledge and skills, this ‘cybernetic’ approach has the disadvantage of not enough stimulating the meta-cognition and the potential to learn-to-learn by the trainee; (Kommers & Zhiming, 1998; Kommers, 2000). The complement and to a certain extend the alternative for instruction is learning by construction. Constructionism is the awareness that learners undergo a highly personal process, due to their cognitive style, various ways of mental imagination and differences in prior knowledge as well. Under this paradigm, trainees in surgery need the opportunity to acquire a model like the rules underlying the trade-off in artery intervention, in an unthreatening situation with a larger bandwidth for experimentation and reflection, before the actual practice with real patients take place.

The connotation of constructivism is that the learner actually builds his/her conceptual knowledge upon prior analogue knowledge. In case of artery surgery, models from hydrodynamics, the principles in flow theory and many more, play a crucial role. Besides that there is a set of specific facts that play a role around blood vessels. During learning the trainee attempts to reconcile earlier experiences with factual rules as formulated by experts in the field.

-          If the factual data dominate the trainees’ intuition and imagination, we may expect that his/her performance later will be brittle and not flexible enough for the large variety of complex situations.

-          If the trainee relies too heavily on similar domain knowledge and intuition, there may arise a discrepancy with the ‘golden rules’ and the statistical lines based upon experience among the colleagues. Also the more or less ‘standard’ operations may not become automated and the communication with colleagues about taken decisions may become troubled.

In the underlying project proposal, the first stage of ‘Medical Consultation’ aims at identifying the more ‘objective’ training elements that need to be conveyed before the constructionistic learning starts. This is the reason that the first stage arranges teams with the various expertise that plays a role for the definition of the anchoring points in the learning of the future surgeon.

Prior and Parallel Projects on VR for Medical Training

Medical training is one of the most prominent application areas for virtual reality. Some of them show a large overview of the larger and more successful projects. Most of the projects have invested in the actual building of the models and have no didactic interface yet. The underlying project proposal sees this need and aims at defining a generic instructional method that intermediates between a VR medical model and a novice who needs to understand and optimise its functioning. Existing medical VR projects can be found under:

 http://www.hitl.washington.edu/projects/knowledge_base/medapps.html

 

This identifies a large overview of the larger and successful projects. Most of the projects have invested in the actual building of the models and have no didactic interface.  One of the example projects is the VRASP Project:

… Virtual Reality Assisted Surgery Planning (VRASP) is being developed for implementation into the hospital operating room. It will give the surgeon flexible computational support intra-operatively, permitting modification and control of large, patient specific datasets in real time. It will render and transmit imagery in response to the surgeon's commands without interfering with normal surgical activities…

Here the support is for the experienced surgeon to give him/her an additional scope of analysis and subsequent alternative plans for treatments. This approach is typical for the many VR projects in the field of medicine. The underlying project plan goes beyond this goal; it is meant to superimpose a procedure for human learning just before the real surgery exercises start.

 

The project “Virtual Environments and Real-time Deformations for Surgery Simulation” is concerned with the simulation of the perceived environment that a surgeon encounters when he or she is performing an endoscopic surgery procedure. The goal is to explore physically based deformations of several organs in a scene while emphasising real-time interaction. For our prototyping we are focusing specifically on abdominal procedures that target the removal of the gall bladder. We have already developed a simple system that allows us to explore the feasibility of computing the surgeon/organ interactions in real-time. Using the results of the first prototype, the next step in this project is to develop a realistic simulator that will provide detailed visual and tactile feedback to the user.

In approach it is congruent to our project plan, as it also aims at real-time interaction preserving the key actions/decisions to be taken by the surgeon. Seen this attempts it comes close to our goal to bring the trainee to the key skills before entering his/her first operation.

 

Work Program

The research that we propose in this project falls into four categories:

1.        Further develop building blocks (segmentation, virtual reality environments)

2.        Develop middleware for integration of distributed data delivery system;

3.        Develop instructional strategies for learning through virtual reality

4.        Utilisation

The project will have a duration of 4 years. For tasks (1) and (2) we need an OIO (the UvA currently has a very good candidate available) and a PostDoc (the LUMC currently has a very good candidate available). Task 3 requires an OIO (the UT has a very good candidate available). This means that as soon as funding is available we can guarantee an immediate start of the project. The first three tasks can be worked on more or less in parallel, of course with enough interaction between the participating groups. The utilisation part will be scheduled at the end of the project.

 

Building blocks

Each of these blocks in Fig. 1 are by itself now well-established, although they require further scientific inquiry and technological improvement, which is challenging in itself, and will also be addressed in this project. Specifically, the image segmentation needs to be enhanced to the demands of this project. The currently available detection-algorithm for the lumen is too simple, as it is shown to be too sensitive for noise and nearby anatomical structures. Therefore, we will include models of the shape of different parts of the vessel. With such knowledge about the shape it will be easier to detect abnormalities in the segmentation, which can then be corrected. One example where shape models can be useful is in the detection of bifurcations, where the vessel subdivides into two separate smaller vessels, a Y shape. An important characteristic of a bifurcation is that the sum of the cross-sectional areas of the two smaller vessels is about the same as the cross-sectional area of the main vessel, and that there are no sharp angles between the vessels. On the other hand, in a side-branch structure, one smaller vessel branches off from the main vessel in a T-shape fashion. This structure has different characteristics, i.e. the cross-sectional area of the main vessel remains about the same before and after the side-branch, and the angle between the side-branch and the main vessel is sharp, up to about perpendicular. The basic idea is to define a small number of vessel building blocks with special characteristics, which can be used in an assembly to form a model of the entire vessel. With this generic approach it is possible to make a specific model for any vessel in the body. We will also look into quantifying the thickness of the vessel wall and the volume of plaque areas.

This segmentation process is run semi-automatically and requires minimal user interaction. All steps are continuously visualised to give the physician insight in the data. This visualisation will be place and platform independent, an will even allow physicians in different locations to view the same data at the same time while consulting one another.

Instructional strategies

Building upon previous and current research, this project aims at the further evolution into a training support system that orient novice surgeons to explore and learn the critical dimensions in the anticipation of blood vessel transplantation and the placement of a stent or a by-pass. The project covers the data delivery, medical, VR-interactive and instructional procedures. The transformation into a training support system needs two stages.

 

  1. Medical consultation: The collaborative use of VA by medical experts who discuss and negotiate about the interpretation of 3D visualisations of a patient’s artery system. The calculation of the pulsatile blood flow and the deduction into shear stresses on the vessel wall is an important but not satisfactory argument to decide upon medical intervention. This stage of the research aims at a multidisciplinary collaboration between actors from the various medical disciplines, VR specialists, imagery experts and the training designers.
  2. Instructional design: The design and implementation of a learning system that allows novices to prepare themselves for the supervised in vivo operations. The system should both allow a tutorial situation, where the experienced surgeon explains and demonstrates the key functionality and simulation mechanism. But also the trainee should be allowed to make an in-depth tour through specimens of the cases. After this phase a wrapping-up conversation with the expert and the trainee should take place, eventually leading to the stage of supervised interventions in real patients.

 

Stage 1 will start right in the beginning of the envisaged project and will continue until the end of the third year. Stage 2 will start in the middle of year 2 and will end at the end of year 4. The overlap between the medical consultation and the instructional design phase is needed as we expect additional evidences to arise during this process.

 

A large part of the test bed is dedicated to the learning / teaching goal, as it is an articulation of the much wider potential such a system has for research in general. A relevant side effect of VR-supplied training situations is the fact that it allows a smooth transition between ‘being locally present in the training institute’ versus ‘participating via telematic facilities at remote sites’. This aspect of ubiquity is of importance both for the trainee and the specialist/trainer as (s)he does not necessarily need to monitor the student’s prior learning in his physical presence. Looking back to the logged students’ actions may enable the trainer to decide upon one’s qualifications before entering a real operating room.

 

Utilisation

The building blocks that need to be developed will, in the end, be integrated into a PC-based workstation with optional connections to large compute servers. Different partners will provide different building blocks as described above. The LUMC will carry out the research needed for the segmentation, quantitation and visualisation of the medical image data sets. These software packages can be integrated into the CMS-workstation by MEDIS medical imaging systems. The UvA will provide the virtual reality software needed for the immersive VR visualisation of the segmented data sets. Simple visualisation will be available as well on a PC-based workstation. Work between MEDIS and UvA is needed to come to an optimal integration. The results of all the analyses need to be stored in the Electronic Patient Record through a collaboration with 2Cure.

Expected Use of Instrumentation

MRI and CT scanners from Leiden and Twente. Computing infrastructure in A’dam and Leiden. VR infrastructure in A’dam CAVE desktop VR system from the SCS research group.

Furthermore the research builds on existing (commodity) soft- and hardware, with the exception of WAP-based PDA’s that are needed to implement and validate roaming access to the decision support system. We need 2 PDA 3G devices for development and 2 more for in-field testing. Expected cost for hardware and support software cost 12 kfl. The access to Dbase servers, (super)computers, Surfnet etc. will be provided by the various groups.

 

Literature: Key Publications of Research Team

 

P.M.A. Sloot: Simulation and Visualisation in Medical Diagnosis: Perspectives and Computational Requirements, in A. Marsh; L. Grandinetti and T. Kauranne, editors, Advanced Infrastructures for Future Healthcare, IEEE-EMBS Press, 1999

 

Schaap JA, Koning PJH de, Geest RJ van der, Reiber JHC. 3D Quantification and visualization of MRA. In: Computer Assisted Radiology and Surgery - CARS 2001. HU Lemke, MW Vannier, K Inamura, AG Farman, K Doi (eds.). Elsevier, Amsterdam, 2001: 928-33.

 

Lelieveldt BPF, Geest RJ van der, Ramze Rezaee M, Bosch JG, Reiber JHC. Anatomical model matching with fuzzy implicit surfaces for segmentation of thoracic volume scans. IEEE Transactions on medical imaging 1999; 18, No. 3, 218-230.

 

Reiber JHC, Serruys PW, Kooijman CJ, Wijns W, Slager CJ, Gerbrands JJ, Schuurbiers JCH, Boer A den, Hugenholtz PG. Assessment of short-, medium-, and long-term variations in arterial dimensions from computer-assisted quantitation of coronary cineangio-grams. Circulation 71, 1985: 280-288.

 

R.G. Belleman and P.M.A. Sloot: The Design of Dynamic Exploration Environments for Computational Steering Simulations, in M. Bubak; J. Mošcinski and M. Noga, editors, Proceedings of the SGI Users' Conference 2000, pp. 57-74. Academic Computer Centre CYFRONET AGH, Krakow, Poland, October 2000. ISBN 83-902363-9-7.

 

A.G. Hoekstra and P.M.A. Sloot: Distributed Particle Simulation of Flow in Complex Geometries, in M. Ingber; H. Powers and C.A. Brebbia, editors, High Performance Computing in Engineering VI, pp. 371-380. WIT Press, Southampton, 2000.

 

P.M.A. Sloot and A.G. Hoekstra: Cellular Automata as a Mesoscopic Approach to Model and simulate Complex Systems, in V.N. Alexandrov; J.J. Dongarra; B.A. Juliano; R.S. Renner and C.J.K. Tan, editors, Proceedings of the 2001 International Conference on Computational Science ICCS 2001, in series Lecture Notes in Computer Science, pp. 518-527. May 2001.

 

Kommers, P.A.M.; (2000) "Imagination through Virtuality for in-depth Learning" Key note speech for the International Workshop on Advanced Learning Technologies (IWALT 2000) Palmerston North, New Zealand.

 

Kommers, P.A.M & Zhiming, Z.; (1998) Conceptual Support with Virtual Reality in Web-based Learning. (Co-author Zhao Zhiming). In: International Journal of Continuing Engineering Education and Life-Long Learning. ISSN 0957-4344. Volume 8, No 1/2. pp 184-204.

 

Z. Zhao; R.G. Belleman; G.D. van Albada and P.M.A. Sloot: System integration for interactive simulation systems using intelligent agents, in R.L. Lagendijk; J.W.J. Heijnsdijk; A.D. Pimentel and M.H.F. Wilkinson, editors, Proceedings of the 7th annual conference of the Advanced School for Computing and Imaging, pp. 399-406. ASCI, May 2001. ISBN 90-803086-6-8.

 

References