CONFERENCE & SEMINAR
March 14-16, 2013
Bangalore, India
ABSTRACT
Title of presentation/poster / Smart Filtering in C MOVE SCPPrimary Author (to be listed if only one name is listed) -name, job title, employer/affiliation, and e-mail / I.Prabhu - Lead Engineer, Design and Development, Siemens Technology and Services Private Limited,
R.R.Meenatchi Aparna - Lead Engineer, Design and Development, Siemens Technology and Services Private Limited,
Other/Co- Author(s) name(s), job title(s), employer(s)/ affiliation(s)
Name of the author conducting correspondence on behalf of all authors / I.Prabhu
Postal address of corresponding author / No 84,Keonics Electronics city, Bangalore 100
Repeat e-mail address of corresponding author /
Telephone of corresponding author / +91 (80) 33132584
Preferred format: / Oral only
Abstract: (200-750 words, in English) / This paper focuses on the usage of particular medical imaging workflow C MOVE DICOM protocol in an optimal manner in a medical imaging application. This improvement in C MOVE handling is a value add to the product in terms of performance and bandwidth utilization and hence can be implemented in any of the medical imaging softwares available currently.
This paper addresses solution to the problems that occur when images are retrieved from a remote DICOM node or moved to a DESTINATION DICOM node using C MOVE. There are a lot of issues related to the uniqueness of the image retrieved, performance and optimal usage of bandwidth. Additionally there are other common problems faced with legacy retrieval methods such as
- When a DICOM object is retrieved, the existence of the same image in the system is not checked until the image is committed to the database or a DICOM archive store. (Here the DICOM Archive store refers to any storage device like local hard disk or remote hard disk or database.)
- When a patient is retrieved there are conflicts due to Patient ID
- The user imports and sends the instances present in the same system again and again which introduces redundancy and wastage of performance without any effective workflow.
Please refer the above figure for current behavior of C MOVE.
The medical imaging applications in the market currently have the following inherent issues because of the above workflow.
1. An unnecessary network send happens for the instances which are already present in the system. A sender system does not find the presence of images in the target before sending.
2. The presence is detected only while committing the instances to database in the SCP
3. This leads to performance degrade in normal send, because of redundant instances.
4. The receiver in case of passive send has no other way to receive the redundant images and finds the presence of same instance only during database commitment.
5. There is no mechanism to prevent the duplicate instance UID conflicts issue
We did further analysis on these issues with certain set of experiments to find a way to prevent these issues from occurring by improvising the request handling in C MOVE SCP. We took a sample data of 1000 images and analyzed the performance in two scenarios when the images are already present in the target machine and when the images are uniquely generated in the source machine. Our analysis led us to improve the way of handling C MOVE RQ, by adding a filtering mechanism in C MOVE SCP. Thus the tweaking of C MOVE RQ handling gave better results which are proved by the experiments.
For instance
consider a set of 1000 images of Patient A for moving out of which 500 images are already present in the destination machine and only 500 images are newly generated
time taken to send one image from target to destination = 68ms
so approximately time taken to send 1000 images = 68 secs
time taken for commitment for 1000 images = 3 secs
total time taken for one workflow = 71 secs
By implementing the suggested change
Time taken for initial commitment(filter request ) =3 secs
Time taken to send 500 images =32 secs
Time taken for final commitment =3 secs
Total time taken =38 secs
There is a saving of 33 secs which is almost equal to the percentage of images already present in the destination Node(ie 50%).
This clearly shows that there is a requirement to improve the handling of C MOVE RQ in C MOVE SCP. A filtering mechanism when implemented in C MOVE SCP can give a performance improvement and a better bandwidth utilization.
Suggested Improvement in C MOVE Handling
Please refer the above figure for the steps
1)Before initiating the C STORE RQ from TARGET NODE, A FILTER RQ can be initiated from TARGET NODE to DESTINATION NODE
2)This FILTER RQ would have the list of DICOM objects that has to be moved
3)DESTINATION NODE must process this FILTER RQ and send back FILTER ACKNOWLEDGE with list of DICOM objects present.
4)The TARGET Node must filter the DICOM objects from the MOVE list after processing the FILTER RQ Acknowledgement
5)For DICOM objects that have been filtered a C MOVE PENDING RESP has to be send to the SOURCE NODE
6)The remaining DICOM objects are moved from TARGET NODE to DESTINATION NODE
In the Normal workflow the TARGET NODE initiates the C STORE RQ for all the DICOM objects that has to be moved. This may be for a list of instances which are present in the DESTINATION NODE as well as list of instances which are not present in the DESTINATION NODE. There is high probability that a patient having old images and new set of additional images can be transferred from TARGET SYSTEM to the DESTINATION NODE. In this case there is no requirement to transfer the old set of images to the DESTINATION NODE and only the delta needs to be transferred.
The TARGET NODE which is the CMOVE SCP should impart intelligence by filtering out the data which is present in the destination by using a NACTION-NEVENT communication. The protocol gives a list of instances which are ultimately not present in the database. Based on the NEVENT report the target node filters the instances which are not present in the destination machine and initiates CSTORE to the destination node only for the filtered instances. These instances which ultimately are not a part of the DESTINATION NODE’s database are only transmitted and the other redundant instances are not transferred. By this method a lot of improvement in the performance can be achieved as well as the network is used optimally.
The suggested approach adds a lot of advantages such as
- Better transfer speed
- Better utilization of network bandwidth
- No duplicate SOP instance conflicts
- Optimal protocol use
2012 Rev. 5