Introduction
SZOTE-PACS is a joint software development for archiving and transferring medical studies of the József Attila University and the Albert Szent-Györgyi Medical University of Szeged. The system is able to collect different type of medical studies from different imaging modalities and store those studies on common standardised DICOM format in a central Oracle database. The end users are able to retrieve the proper images from this database. The archived medical images can be presented and processed at the viewing stations. There is a need for compressing the image series for the sake of image storage capacity and network traffic. Our system is able to store the images for 15 days now. According to our calculations this period can be doubled using data compression. This means that the transmission rate between the stations will be reduced as well. After retrieving the selected images, the users at the viewing stations will process the images. Hence the compressed image should be restored without any loss of information.
Material and Methods
Before starting the realization of any compression method in SZOTE-PACS, we made preliminary tests to select the most suitable technique for each type of imaging studies. The test data set consisted of image sets from various modalities such as CR (Computed Radiology), CT (Computed Tomography), MR (Magnetic such as CR Resonance), US (Ultrasound) and NM (Nuclear Medicine). We built a data set with respect to patient study types, image resolution and color depth, number of frames, within each modality. While evaluating the results, the partial results for each modality were carefully weighted, taking into account the characteristics of average traffic of our PACS system, i.e., the statistical distribution of the studies between the modalities.
|
|
|
|
|
|
|
|
4 | 4 | 34 | 1 | 12 | 2500x2000 |
|
52 | 2189 | 1038 | 188 | 12 | 340x340 , 512x512 |
|
68 | 615 | 106 | 41 | 12 , 16 | 256x256 , 512x512 |
|
56 | 1544 | 24 | 120 | 8 , 16 | 64x64 , 128x128, 256x256 , 126x1024 |
|
11 | 72 | 20 | 31 | 8 , 24 | 480x640 |
|
191 | 4424 | 1224 |
We used compression methods which - besides being efficient - seem to be on their way to standardization and medical imaging system vendors will use them in the future probably. PNG (Portable Network Graphics) is an emerging format, platform-independent, progressive, up to 48 bit color depth and using Huffman and LZ77 coding. JPEG-LS (Lossless JPEG) is an extension of the well-known JPEG standard, designed for lossless compression.
Results
The tests were run on several types of hardware, ranging from Intel 486 to Silicon
Graphics Challenge. Since the time factor is very important in our case we have
to make a trade-off between the time needed for compression/decompression and
the gain in storage place or network bandwidth required. Of course, there is
a hardware-limit under which the tested compression methods cannot be used efficiently.
This limit depends on the specific task which the computer in case has to perform.
Since the decompression times usually are significantly shorter than the compression
times, less powerful machines can be used for viewing stations (which only need
to decompress and then display images) than for input (admitting) stations which
are supposed to compress the images before sending them to the (central) archive.
A solution to this problem would be - if the additional network traffic can
be afforded - to send the uncompressed images from these input stations to a
dedicated machine which deals only with the compression. We plan to send DICOM
studies with their Pixel Data element in Encapsulated format which includes
data compression in our case. This is a standard way of storing pixel data in
DICOM studies in some encoding form. The system is currently being tested with
this new feature. We developed some new software components for lossless compression/decompression
of DICOM studies.
Conclusion
According to the results of our experiments sending compressed image data across
the network is useful. The software components making compression/decompression
of DICOM studies can be useful for any PACS system. We are going to test compression
methods for whole sequences of DICOM studies in order to reach an even higher
compression rate (c.f. MNG).
Corresponding Author: Attila Kuba Ph.D. Department of Applied Informatics József Attila University H-6701 Szeged, P.O. Box 652 , Hungary Fax: (36)-62-312-292 e-mail: kuba(at)inf.u-szeged.hu
EuroPACS '98
paper Martonossy Laszlo
Oral presentation at EuroPACS'98, Barcelona, Spain