Parallel Medical Imaging Transmission
 
Authors:
Rouzbeh Maani, MSc, University of Manitoba, TRLabs; Sergio Camorlinga, PhD; Neil Arnason, PhD
 
Background:
One of the main challenges of the current medical imaging systems is dealing with large amounts of data acquired by the modern modalities. Indeed, new modalities can generate large amounts of high-quality images in a short time. If we count the number of studies performed per patient and the number of patients examined per day, this advancement in the modern modalities is leading to an information explosion. As a result the transmission, access, display and navigation through these large amounts of data can turn into a real challenge to achieve[1].

In fact, one of the most important demands toward dealing with large amount of data is to provide a way for fast data transmission between medical imaging applications. In other words, as long as we deal with large amounts of data, we need to transfer these large amounts to the other applications, where the data needs to be processed or stored, within relatively small time delays.

The most common approach toward enhancing the speed of data transmission is using compression techniques. In fact, there have been many attempts to improve the speed of data transmission between DICOM application entities, which are fundamentally based on compression idea[2-11]. However, on reasonably fast networks the overhead of compression/decompression slows down the total transmission operation rather than speeding it up[12]. In other words, this approach is only effective when the speed of the connecting media is not much compared to the compression/decompression computing time, so adding the overhead of compression/decompression times is worth. Hence, as a matter of fact, compression techniques are ineffectual in high-speed networks.

In this paper, using parallel transmission between two medical imaging systems is proposed and assessed. The proposed approach is based on Digital Imaging and Communications in Medicine (DICOM) protocol[13]. DICOM is the standard storage format as well as the transmission protocol, for medical images. It has several advantages such as interoperability, integrity and consistency, which have made it the world wide practical standard for interconnecting medical imaging systems[2,14].

The proposed method uses parallel connections to carry out the image transformation between two DICOM application entities. These parallel connections are used in the Storage Services in the DICOM protocol. In fact, the method is a way to improve the speed of data transmission in high-speed networks where data compression techniques are not effective. To implement the parallelism, a multi-threading technique has been used. The implementation of the method is in Java, and is based on the dcm4che2 tool[15] and the libraries provided by this tool.
 
Evaluation:
We first examine the impact of data compression in a high-speed network. For this we used several compression techniques on a defined dataset consisting of 303 images with aproximately 150 MB. The dataset includes a range of different modalities and SOPs (Service Object Pair). The complete properties of the utilized dataset are shown in Table 1.
 

Table 1
Table 1: The properties of the dataset used in the experiments
 

The evaluation is performed on a 100Mbps LAN. The Sender computer was a Pentium 4 Dual-Core with 1 GB RAM running Windows XP and the receiver was a Pentium 2 Quad-Core with 2 GB RAM running Linux CentOS.

We used five lossless compression techniques for the experiments. In DICOM terminology, the data encoding is represented by the Transfer Syntax term. The chosen Transfer Syntaxes are lossless because we want the final images to be the same as the original images and, therefore, the impact of compression overheads can be compared with the default method of communication (i.e., using the default Transfer Syntax without parallelism). The properties of the chosen Transfer Syntaxes are shown in Table 2.
 
Table 2
Table 2: The properties of the chosen Transfer Syntaxes
 
Figure 1 shows the ratio of data compression for each method. The Transfer Syntaxes are presented with numbers 1-5 represented in Table 2.
 
Figure 1
Figure 1. Comparison of compression ratios between the Transfer Syntaxes
 
The total transmission time (including compression overheads) with these compression techniques and the default Transfer Syntax, which is non-compressed, is depicted in Figure 2.
 
Figure 2
Figure 2. Total transmission time for different Transfer Syntaxes
 

As you can see the overheads of compression techniques make them inferior to the default uncompressed transfer method.

Now, as the second experiment, we used parallelism to connect two applications. To examine the impact of parallelism, different number of connections was examined. The total time of data transmission by different number of parallel connections is presented in Figure 3.
 
Figure 3
Figure 3. Comparison of total transmission time for different number of parallel connections
 
Discussion:
Using compression techniques is ineffectual in reasonably high-speed networks. The main reason is rooted in the fact that, although these techniques decrease the total volume of data to half or less, the imposed overhead times are pretty huge and comparable to the saved time reached by sending the smaller amount of data. This fact was examined for the intended dataset. We observed that the total transmission time in the uncompressed default Transfer Syntax rose about 42% more for Transfer Syntaxes 1, 2, and 3 and increased drastically about nearly 3 times for Transfer Syntaxes 4 and 5.

On the other side, the proposed method sped up the transmission time in high-speed networks and, as you can see in Figure 3, the best transmission time was reached by four parallel threads (with about 19% speed-up improvements).
 
Conclusion:
The staggering number of images acquired by modern modality requires new approaches in medical imaging systems for transmission, organization, and storage. As an important part of medical imaging workflow, data communication plays a critical role. There have been many attempts made for speeding up the data communication between medical imaging systems. Those attempts were mostly based on the compression idea. Although the compression techniques are so valuable in low-speed networks, they are ineffectual in high-speed networks. The proposed method is an effective approach for high-speed networks, and by using parallel connection for Storage Services in DICOM protocol, we may improve the speed of transmission in those networks. Experimental evidence shows that few parallel transmission links (3-4) are needed to achieve this goal.
 
References:
[1] Andriole KP, Barish MA, Khorasani R. “Advanced Image Processing in the Clinical Arena: Issues to Consider.” Journal of the American College of Radiology. 2006;3(4):296-298.

[2] Ramakrishnana B, Sriraam N. “Internet transmission of DICOM images with effective low bandwidth utilization.” Digital Signal Processing. 2006;16(6):825-831.

[3] Clunie DA. “Lossless compression of grayscale medical images: effectiveness of traditional and state-of-the-art approaches.” Proc SPIE Med Imaging. 2000;3980:74-84.

[4] Foos DH, Muka E, Slone RM, et al. “JPEG 2000 compression of medical imagery.” Proc SPIE Med Imaging. 2000;3980:85-96.

[5] Kalyanpur A, Neklesa VP, Taylor CR, Daftary AR, Brink JA. “Evaluation of JPEG and wavelet compression of body CT images for direct digital teleradiologic transmission.” Radiology. 2000;217:772–779.

[6] Sung MM, Kim HJ, Yoo SK, et al. “Clinical evaluation of compression ratios using JPEG2000 on computed radiography chest images.” Journal of Digit Imaging. 2002;15(2):78–83.

[7] Sung MM, Kim HJ, Kim EK, Kwak JY, Yoo JK, Yoo HS. “Clinical evaluation of JPEG2000 compression for digital mammography.” IEEE Transactions on Nuclear Science. 2002;49(3)827-832.

[8] Erickson BJ. “Irreversible compression of medical images.” Journal of Digital Imaging. 2002;15(1):5-14.

[9] Amor NB, Essoukri N, Amara B. “DICOM image compression by wavelet transform.” IEEE International Conference on Systems. 2002;2:553- 557.

[10] Seeram E. “Irreversible compression in digital radiology. A literature review.” Radiography. 2006;12(1):45-59.

[11] Logeswaran R. “Compression of medical images for Teleradiology.” Teleradiology. 2008;21-31.

[12] Pianykh OS. “Digital Imaging and Communications in Medicine (DICOM): A practical introduction and survival guide.” Springer-Verlag, Berlin Heidelberg. 2008.

[13] Mildenberger P, Eichelberg M, Martin E. “Introduction to the DICOM standard.” European radiology. 2002;12(4):920-927.

[14] Vossberg M, Tolxdorff T, Krefting D. “DICOM image communication in Globus-based medical grids.” IEEE Transactions on Information Technology in Biomedicine. 2008;12(2):145-153.

[15] http://www.dcm4che.org.