 |
 |
|
| MEG project team's final goal is to figure out the function of human brain. As genrally known, human brain is complex compared with other human organs such as liver and lung. In brain science, scientific technologies show remarkable development. However, a brain consists many parts, each of which biomechanism are not yet known. |
|
|
|
 |
|
Our engineering goal is to testify new internet technologies not only theoretically but also in a practical manner. Recently, development of internet technologies is remarkable, but these technologies tend to be developed independently. This tendency may just enhance complexity of Internet Technologies and so maybe it may prevent development of the whole science.
From this perspective, we are working on MEG data analysis problem, one of the most important medical problems to be solved. (MEG is highly sophisticated brain imaging technologies as described below) Specifically, we are working to build MEG data analysis environment by making full use of emerging Internet Technologies such as Grid and IPv6. Moreover, in order to analyze and verify complex signals generated from brain activity, we plan to utilize leading-edge signal processing technologies on a new environment proposed and developed by us.
| |
|
|
 |
|
Developments of recent measurement technologies have made possible to increase the amount of data collected from measurement instruments, for example, medical data such as MR, CT images and MEG. Although recent processor technologies ensure performing fast analyzing computation, the increase of data amount increase the analysis time as well. This phenomenon is observed not only in medical scene but also other scientific fields.
Network technologies are improving day by day and parallel computing is possible, but medical data analyses are often performed on single-processor basis. This situation prevents medical doctors from detecting early symptom of diseases. It means that we are now in much danger of diseases even if we are able to treat these diseases.
|
|
|
|
 |
|
As consciousness towards interdisciplinary researches grows, virtual research environment is getting required for the interdisciplinary researches in many distributed organizations. This reason can be explained from three facts.
Distribution of knowledge and technologies
In any field of current medical scene, high degree of knowledge is required. Given an example from our assumption, MEG data analysis requires sophisticated knowledge and much experience. This may be easy to be understood if you take the complexity of signal processing techniques into consideration. In addition, medical knowledge is required for precious and adequate analysis and high degree of computer knowledge and skills to integrate vast number of computers into our system. These specialists hardly belong to a single organization. They mostly belong to different organizations or they locally gather as one research team. For example, while medical doctors work in a hospital or medical research center, computer scientists who have detail knowledge on parallel processing maybe work in an engineering research centers, and specialists with signal processing techniques may be in university laboratory. Given such a circumstance, efficient collaboration among brain scientists are expected.
Distribution of resources
Development of measurement technologies improves accuracy and temporal and spatial resolution in sensing device. On the other hand, the development drives up prices of measurement instruments such as telescope and accelerators. MEG is not an exception, either. Medical instrumentation systems are generally expensive, but MEG is more expensive than we expect. Price and maintenance costs are one of the factors that there are few MEG all over the world, despite its very promising feature in brain science. At the same time, high degree of computational resources is also expensive. High-performance graphical computers and supercomputers in some organizations are located at remote locations. It is hard to find an organization like our Cybermedia Center, where such computational resources are locally gathered.
Development of network technologies
Recent development of network technologies enables us to transfer the data with the order of gigabit in a few seconds. Within a few years, it is expected that one terabyte data will be transferred in a second. In addition, priority control of network traffic is being explored as QoS (Quality of Service). Development of these technologies is expected to be guragually improved. However, these technologies has not been shifted to next generation's application stage. In other words, they are still in experiment stage even though they are ready to be applicable in real world. These network technologies are now in hope that they can solve the above geographical distribution.
Such changes can be seen in medical field. From this point of view, parallel computing is necessary for near future medical data analysis.
|
|
|
|
 |
|
The above figure illustrates a simple conception of our MEG system. Our MEG data analysis system is composed of the following three parts (modules) on the wide-area network or the Internet. In fact, these three parts are assumed to be distributed geographically or organizationally.
Data Acquisition Part (Module)
MEG data reaches 0.9GB in an hour measurement if a 64-sensor MEG with sampling frequency 1kHz is used. So far, these data have been stored and managed in magnetic device such as Optical disk. Presently, the data are passed by hand among researchers. This situation may lead the data inconsistency, the loss of data itself and the leak of patient's privacy. In addition, according to the increase of data amount, it is difficult to store the data in a single optical disk.
To solve the problem, we have assumed to directly store the data into medical database and simultaneously transfer the data to computation part from the data acquisition part via the Internet. This would realize the seamless processing of the data. As a result, medical doctors would obtain the analysis results as quick as possible. Real-time analysis may be realized.
Furthermore, taking the use of network into consideration, a worldwide medical database is possible to be built. Such database is very useful and epoch-making for statistical investigation because it can offer large amount of medical data that has never been seen until now. In addition, it may play an important role to support the medicine in aged society. Medicine must be reached every home because it is easy to predict increase in the number of the elderly people who live alone. For the purpose, medical doctors need to retrieve the medical information from everywhere.
However, medical data are sensitive to security problems because they have much privacy. We have to protect the medical data from the external threats in transferring them on the Internet. In order to do so, we have to utilize advanced cryptography and network technologies. In our project, we have planned to realize secure data transfer environment by adopting PKI (Public Key Infrastructure), IPSec (IP Security) and SSH (Secure Shell) as central technologies.
Computation Part (Module)
MEG data analysis is very computation-intensive and so time-consuming as described below in detail. We have to obtain the computational power enough to adequately analyze the data as quickly as possible. In order to achieve this, we are focusing on enormous computation power on the Internet. Recently, there are many computers connected in the Internet. A wide range of computer from an ordinary personal computer to a supercomputer is connected there. However, these computers are not working in full time. In short, the workload is not always high. Our idea is to make full use of such computers on the Internet and realize quick and adequate data analysis. Hence, computation part in our system can have the characteristics of dynamics in the number and composition according to around surroundings such as network traffic and processor workload.
I/O Part (Module)
The most important role in I/O partis to convey user's will to our system and offer analysis results in intuitive-understandable visualization form. For the purpose, the easy-to-use GUI (Graphical User Interface) and intuitive visualization methods are the key of success.
Our approach to visualization is superimposing analysis results on human anatomical images such as MR and CT. For the purpose, we are planning to reconstruct 3D images from that. We also need to discuss how to express and superimpose analysis results.
We have adopted Globus grid toolkit as a building block of our system's. In our system, the Globus plays an important role in integrating the above three modules. Globus grid toolkit is being developed in hope that enables the universal access to diverse resources from a personal computer to a medical device connecting to the Internet. The Globus's goal is just like to create an environment that we want to achieve. The Globus facilitates us to easily integrate diverse distributed resources, even if they belong to different organizations.
|
|
|
|
|
|
|
|
