The principal functions of the system are the following:

The software architecture of CHILI is based on the results of the European Project HELIOS, in which concepts for distributed, object-oriented software systems were developed [6]. The core of the HELIOS software engineering environment is the HELIOS Unification Bus (HUB), a middleware technology. Independent and distributed software components communicate by means of the HUB. The middleware of CHILI is based on the same concepts as HELIOS, but has been redesigned in several aspects to match the specific requirements of an image communication environment that has parts which are linked by low bandwidth networks and are not permanently reachable. Security enhancements have been added to fulfill the legal requirements for data security and privacy. The middleware supports the security standards SSL/TLS (Secure Sockets Layer/Transport Layer Security) for authentication and secure communication [7].

Figure 1 shows the main components of the architecture:

This overall design allows a very flexible configuration of the system which can easily be adapted to any specific environment. The different services and modules can be combined in different ways to provide custom solutions. Examples of typical configurations (in clinical use) are:

The most powerful communication protocol for data exchange (and the only protocol that allows teleconferencing) is the CHILI protocol, which includes a strong data security concept [8]. This includes all measures which are necessary to comply with German and European requirements and laws. As a result of the data security concept, it is possible to restrict the rights of the receiver on the image data. The following restrictions are possible:

All data packets to be transmitted are encrypted with the PGP (Pretty Good Privacy) public key encryption system [9]. The public key of the receiver is used for the encryption. A checksum of the data is calculated and signed with the digital signature of the sender. This protects data integrity, and allows for authentication of the sender and privacy. Furthermore, all local data are encrypted with a symmetric key encryption method. Transmitted and received images are logged in special log files which are protected with a checksum.

The communication between the CHILI components via the middleware can optionally be protected using the TLS standard [7]. This ensures authentication and privacy.

It cannot be expected that all communication partners have the same teleradiology system. Thus, CHILI offers additional communication methods (Figure 2):

The user of the CHILI system is not confronted with the various protocols directly. The definition of the protocol to be used is configured for each communication partner. It has to be defined if the partner is connected via LAN, Internet, or ISDN lines. The ISDN zone is defined as well to be able to estimate the transfer costs depending of the transmission time, duration, distance, and compression rates.

The broad set of transfer methods enables the CHILI user to send images to nearly anyone with access to a computer and a network. The limitations are that teleconferences are possible with the CHILI protocol only, and that the user has to take reasonable precautions for data privacy and security when he does not use the secure CHILI protocol.

Image data can be transmitted manually by the user or automatically by the auto-routing component which is activated by the import service. Certain data fields of the image header are used to determine if and to whom the incoming data should be transmitted automatically.

A typical problem in radiology is that nearly every application needs its own computer. Console computers are necessary to operate the imaging modalities. Viewing stations are used for reporting or further analysis of the image data. Radiology Information System (RIS) terminals are used to manage and organize the patient information flow. Additional Personal Computers (PCs) are used for word processing (e.g. reports, publications). New computers for teleradiology are now entering the radiology department.

Developers of new medical image analysis applications face the problem that the current computer applications (e.g. viewing stations) are not open enough. Every developer must consider the connection to the modality or viewing station, file formats, data storage of files on the disk or in a data base system, and the display of 12-bit images on 8-bit screens. For the radiologist, the development a new method means the invention of many wheels a second or third time before the real problems can be addressed. Most of the manufacturers of the existing viewing stations do not allow the installation of additional software. Usually the customers do not get the root password of the systems. An additional computer with different user interfaces and additional interfaces to the rest of the world has to be introduced, therefore, to get the new functionality to the users.

The result of the depicted problem is that it takes a lot of effort for developers to implement and introduce new methods; this is most of the time very costly for the end user. Therefore, we designed the CHILI PlugIn mechanism [10]. Users can extend the system by adding powerful image postprocessing functions [11] or interfaces to other information systems. PlugIns can be either existing applications or new modules with interfaces to the existing CHILI components (Figure 3).

The PlugIn developers' Application Programming Interface (API) includes interfaces to all software components and the message system [10]. The developer is free to choose programming languages and interface toolkits (e.g. C, C++, Tcl/Tk). The Java Virtual Machine [12] is a PlugIn itself; thus, Java programs can easily be integrated to communicate with CHILI.

Several existing applications have successfully been integrated into CHILI as useful additions to the main system. One example is a terminal window (xterm) which is connected to a RADOS-M RIS. This enables the user to see both the image data and RIS information on the same screen (Figure 4). Another example of such a simple PlugIn is a WWW browser, which allows the user to explore, for example, reference cases on the intra- or internet. The user manual of the system is integrated with the same mechanism.

Examples for ongoing PlugIn development projects are a general purpose 3D volume segmentation and visualization project [13] (Figure 5), a tool for the planning of liver surgeries [14], and a 3D ultrasound doppler project [15].

Over time, the system has become a PACS component which can be used as an interface to DICOM compliant archives, and also for DICOM printing and reporting. Images can be distributed in-house, and light-weight clients running on PCs can be used to view or process the images in teleconferences. Clients for MS Windows are available, in addition to web-based interfaces which can be accessed from any hardware or operating system with a web browser. The CHILI database can be configured to act as a cache for the PACS archive (Figure 6). A storage hierarchy can be configured to reduce the traffic on the archive and to optimize image transfers on the network. The user is released from the load of knowing where the data are actually stored.

The CHILI Viewer with the multi-head option and diagnostic monitors is suited for image reporting. Filmless radiology can be realized with this technology.

An overview about the state of teleradiology in the United States has been published by Grigsby in 1994 [19]. Regular telephone lines and modems were normally used at that time in the US. Most of the systems developed were based on PCs using video cameras with PC frame grabber cards or scanners to capture the images, e.g [19,20].. Oft-discussed topics in the field of teleradiology were spatial resolution and grayvalue depth of the video images or scanners. However, these problems do not exist when the original digital images are processed; the receiver gets the images in original quality. Such advanced teleradiology systems based on digital data and digital telephone lines (ISDN) have been developed and used in Europe since 1994, e.g [20-24]..

A prototype of a dedicated teleradiology system has been developed by Gomez and co-workers in a research project funded by the EU AIM Programme [21]. The system has been evaluated in two hospitals in Spain. One imaging modality was connected to one of two teleradiology workstations. Handels describes another German teleradiology system, called KAMEDIN, which has been used in a field test for several months but has never gone to clinical routine [22].

All major image modality vendors in the field of radiology have been talking about teleradiology since 1994 or 1995. The main difference of the industrial solutions is that they simply extend the local area network of the radiology department by an ISDN link to another location. Images are then sent with proprietary protocols (sometimes with DICOM) to a remote machine. Teleconferences with two users at different locations who work together on the same images are not possible. Security aspects are missing as well.

CHILI (and its predecessor MEDICUS) seems to be the only system which has a complete data security concept [8]. A great advantage of the system, compared to others, is the direct connection with the imaging modalities of different vendors. The DICOM protocol can be used to receive the images, or proprietary solutions can be enabled for older modalities without DICOM interfaces.

A major feature of CHILI is that it is not a dedicated teleradiology system, such as its prototype MEDICUS or the KAMEDIN system [22]. Instead, it is a general purpose multimodality workstation with strong teleradiology and security features. Cross-platform communication, vendor independence, and extensibility are major distinctive features in comparisons. The system is also used for advanced image analysis functions and image reporting.

New teleradiology systems are currently under development in the programming language Java [12]. This makes the systems independent of the operating system and hardware used. Examples are the CYPRIS system which is under development by Klaiber [25] or the Jive system by Kleber [26]. Both systems are not yet mature enough to be compared with the CHILI environment. They lack basic image analysis functions, cooperative teleconferences, strong security features, and the full spectrum of vendor and system independent communications. The performance drawback of Java is another aspect which has to be taken into account; but independence from the operating system and the hardware are fascinating features.

CHILI is not just a research project where the system has been used for some test cases. It is integrated in the clinical routine in many different locations and institutions. The accounting numbers and the feedback of the users are proving that the CHILI architecture provides a powerful and flexible environment for PACS and teleradiology. PlugIn modules will also extend the multimodality workstation for teleradiology by providing advanced image analysis and therapy planning functionality.

In the future, development will continue on the Java implementation of the complete CHILI architecture to offer the most flexible and hardware/software independent environment for PACS and teleradiology. The feedback and suggestions of the users has influenced the system architecture by a great extent. Our own experience is that the differences between PACS and teleradiology components will vanish. This observation is supported by a recent market investigation by Frost&Sullivan [27]. The "tele" aspect will become a natural feature of radiological workstations. Radiologists will profit from better performance; and as a consequence, the patients will receive better treatment.