An approach to DICOM extension for medical image streaming.
The advance in technology enabled the use of handheld devices in medical environments. The mobile devices enable the physicians to access medical data wherever they are--in hospital buildings, outdoor or in their homes. Unfortunately, these devices are mainly used for receiving and displaying the medical record data without medical images. Even with the image compression, the size and resolution of medical images overcomes the capabilities of common mobile devices (Mirkovic et al., 2005). The mobile device limitations of processing power, memory capacity, display capabilities, and the mobile network band-with limitations increases the overall time needed to transmit, decompress and present the image. For example the calculated time that passes from requesting the JPEG2000 medical image until it is presented to the client on PDA is about one minute, which is unacceptable.
Picture Archiving and Communication System (PACS) is imaging system designed to steer medical image data and the corresponding information between various parts of healthcare system. The communication between various parts of PACS is usually achieved through Digital Image and Communication in Medicine (DICOM) protocol (Dragan & Ivetic, 2005). To enable the transmission of medical images to mobile devices, the image archive has to expand with the image copies suitable for different mobile devices. For example, to include mobile phone, PDA and Tablet PC into the PACS, the medical image archive has to contain images prepared for these devices, fig. 1. The PACS treats them as different medical images. Therefore, the DICOM client-side application has to know which image is adequate for its display and processing capabilities before requesting it. This is achieved only in querying the DICOM server prior to image request. There is no mechanism which will allow the client to request the image with the desired region of interest, resolution, and quality.
Medical image streaming is a more effective solution to enable the transmission of medical images to mobile devices (Skodras, 2006). Image streaming means that pixel data needed to represent a part of the image (or the whole image) in a certain resolution and quality are extracted from the one stored image codestream and transmitted to the client-side application, fig. 2. Therefore, the client only has to define the desired part, resolution, and quality of the image. The network communication is not overloaded with unnecessary data; the medical images are smaller in size; and the image resolution is suitable for display device resolution. The best results could be reached if image streaming is combined with a still image compression which achieves smaller spatial distortion.
[FIGURE 1 OMITTED]
Although it has been improved over the years, DICOM has not been modeled to support streaming of medical images. Therefore, DICOM restricts PACS to devices with higher power, storage, communication and display capacities. In this paper we describe the way in which DICOM can be extended to support image streaming. This will improve DICOM networks which will then support mobile client-side devices with limited processing and display capabilities. The proposed solution is transparent, simple to implement and easy to include into the existing DICOM based PACS.
The organization of the paper is as follows: the section 2 gives the overview of other's people researchers; the section 3 describes the proposed extension of DICOM; and the section 4 concludes the paper.
2. DICOM AND MEDICAL IMAGE STREAMING
DICOM defines the message format, the protocol for message interchange and the file structure for biomedical images and image-related information. Although it is not supported directly, the importance of medical image streaming is recognized by the DICOM standard. The standard defines the pixel data provider service which enables transmission of pixel data using a network protocol other than DICOM. Currently, the only service supported is the JPIP (JPEG2000 Interchange Protocol) pixel data provider which represents a mechanism for supporting the use of JPIP network protocol (Taubman & Prandolini, 2003). It enables interaction with JPEG2000 content and it is used for JPEG2000 image streaming. When this mechanism is used, pixel data are replaced with single DICOM attribute which contains URL string that represents the JPIP request, including the specific target information. The client-side application can expand the JPIP request with description of desired region of interest, resolution, and quality of the image. This service enables medical image streaming but not over DICOM network. Although the whole range of the JPEG2000 streaming capabilities is on disposal, this mechanism has several disadvantages. DICOM server has no control over the pixel data transmission. If, from some reason, image does not exists at JPIP server-side, or image can not be served, the transmission of the DICOM image has to be done from beginning, completely over DICOM protocol and the medical image has to be transmitted in its full size. Therefore, the JPEG2000 streaming can not be guaranteed. Also, the JPIP and the DICOM server archive have to be synchronized.
[FIGURE 2 OMITTED]
The streaming of DICOM images has been studied. The usual solution is based on two layer communication. One layer is used for communication with DICOM based PACS system and the second layer is used to stream images to the client-side devices. Tian proposed in his works (Tian et al., 2008) the two layer architecture in which the second part of the system is based on JPIP protocol. The medical image archive remained in original uncompressed DICOM format. The storage requirements are high, because the image streaming and compression are not combined. In this communication system, mobile devices are not internal part of the DICOM networks.
3. DICOM EXTENSION
It is possible to extend DICOM to support medical image streaming. Standard DICOM message should be extended with additional fields which will enable DICOM clients to request the desired region of interest, resolution, and quality of the medical image. Also, the extension should contain fields with the description of the DICOM server response. The additional fields should be transparently and easily incorporated inside standard DICOM message following DICOM rules for DICOM fields formatting and labeling.
The standard DICOM message can be extended in two ways: by a set of parameters or by a parameter string. The set of parameters would define a new DICOM field for each parameter of the requested medical image, fig. 3.A. For example, if DICOM client-side application requests part of a medical image of desired resolution and quality, DICOM request message should contain at least three attributes (one for each of the requested parameters). The parameter string would define only one new DICOM field, fig. 3.B. This DICOM field should contain the description of the requested image parameters, formatted according to some messaging format like HTTP, or JPIP. The choice for selecting one of the proposed extensions depends from the specific implementation.
The best result could be achieved when all the client requests are served from one basic image source. The basic image is processed according to the DICOM client request. The DICOM server would transmit only the data needed to represent the medical image in resolution and quality requested from the client, suitable for the client device. With this approach the DICOM medical image archive remains the same. It is even better to combine this mechanism with still image compression whit minimal spatial distortion. This combination leads to smaller medical image archive and high-quality resulting medical images.
The JPEG2000 compression supports image streaming from one JPEG2000 image with minimal spatial distortion in decompressed image (Taubman & Marcellin, 2001). The processed JPEG2000 images always yield the best quality for given resolution. Although DICOM does not support JPEG2000 streaming, it supports JPEG2000 image format and it is possible to replace native DICOM image format with JPEG2000 image format. It is possible to implement JPEG2000 streaming over DICOM network. The DICOM server should process JPEG2000 medical image according to client request, encapsulate it inside DICOM message and send it back to DICOM client. The JPEG2000 streaming has been tested in medical domain, but outside DICOM, and it achieved good results (Moshfeghi & Ta, 2004). Therefore, we propose the combination of the extended DICOM syntax with JPEG2000 image streaming.
We presented transparent extensions to standard DICOM syntax which enable streaming of medical images over DICOM networks. This enables fast browsing of medical images inside DICOM networks (even over low-band networks) and viewing of high-quality low-resolution medical images on limited size display devices. JPEG2000 compression combined with the DICOM extension represents the best solution for streaming of medical images. JPEG2000 reduces PACS storage demands more than 10 times and supports image streaming with minimal spatial distortion. Contrary to similar solutions, this extension offers transparent, efficient and simple way for achieving streaming of medical images over DICOM networks. Streaming mechanism is an integral part of DICOM network. Our future research is focused on an implementation of a hospital PACS based on the extended DICOM syntax and JPEG2000 compression, which will offer high quality medical images to physicians wherever they are, on devices they have.
Acknowledgements. This research was supported by IT Project No. 13013, financed by the government of Republic of Serbia.
Dragan, D. & Ivetic, D. (2005). DICOM overview, Proceedings of the VII International Symposium "Young People and Multidisciplinary Research"; Zaberca, V. (Ed.), pp. 105-115, ISBN 1453-7394, Resita, Romania, Sep 2005.
Mirkovic, J.; Ivetic, D. & Dragan, D. (2007). Presentation of Medical Images Extracted From DICOM Objects on Mobile Devices, Proceedings of the 9th International Symposium of Interdisciplinary Regional Research "ISIRR 2007" Hungary--Serbia--Romania, Novi Sad, Serbia, Jun 2007. In Press 2008.
Moshfeghi, M. & Ta, J. (2004). Efficient Image Browsing with JPEG2000 Internet Protocol. Medical Imaging 2004: PACS and Imaging Informatics, Proceedings of SPIE 2004, Vol. 5371 No. 08, pp 31-42, ISSN 0277-786X.
Skodras A. N. (2006). The JPEG2000 Image Compression Standard in Mobile Health. In: M-Health: Emerging Mobile Health Systems, Istepanian R., Laxminarayan S., Pattichis C. (Ed.), pp. 313-327, Springer US, ISBN 978-0-387-26558-2.
Taubman, D. & Marcellin, M. (2001). JPEG2000: Image compression fundamentals, standards and practice. ISBN 978-0-7923-7519-7, Boston, Kluwer Academic Publishers, 2001.
Taubman, D. & Prandolini, R. (2003). Architecture, Philosophy and Performance of JPIP: Internet Protocol Standard for JPEG2000. Available from: www.ee.unsw.edu.au/~taubman/publications_files/JPIP-architecture-vcip03.pdf Accessed: 2008-04-23
Tian, Y.; Cai, W.; Sun, J. & Zhang, J. (2008). Accessing medical image databases on the go. Available from: http://spie.or^x19505.xml?highlight=x2416 Accessed: 2008-02-26.
Fig. 3. DICOM extensions for medical image streaming; A--set of parameters; B--parameter string. A Requested Resolution 1024,1024 Requested Region 512,512 Requested Region Offset 256,256 B Request Parameter String fsiz = 1024,1024&rsiz = 512,512&roff=256,256