Goal - TUM
1
Computer Aided Medical Procedures Technische Universität München A Lightweight and Portable Communication Framework for Multimodal Image-Guided Therapy A. Schoch, B. Fuerst, F. Achilles, S. Demirci and N. Navab Computer Aided Medical Procedures (CAMP), TU Munich, Germany Bladder Rectum Prostate US Transducer PET Head PET Detector Plate PET Field of View US Field of View Devices and labels © 2013 TUM-CAMP, DESY and EndoTOFPET-US Collaboration, see [1] Anatomy © 2005 Terese Winslow, U.S. Govt. has certain rights NDI Polaris NDI Aurora EMT Targets OT Targets Introduction Motivation • Mul:modal and IntraOpera:ve Imaging (EndoTOFPETUS) [1] • Collabora:on of researchers from areas : • computer science (GUI, tracking, reconstruc:on) • physics (crystals, op:cs, aNenua:on and scaNer correc:on) • medicine (tracers, preop imaging) • electronics (photo electronics, readout, DAQ cards) • Full monitoring and centralized control of system required Goal • Simple to use communica:on framework (“one header only”) • PlugIns for visualiza:on (CAMPvis) • Opensource: hNps://campar.in.tum.de/Main/CAMPCom • Control & Data Flow: communicate data and system messages between various components • Efficiency: lightweight & efficient communica:on • Pla5orm Independency: serializa:on and transmission • Abstrac=on & Extensibility: support extension, general interface • Priori=za=on: asynchronous func:on to support efficiency Methods Query Handling & Prioritization • Ac:vity diagram for client library • Simplified (without dispatcher thread pool) Inter-Device Communication • Sample setup for EndoTOFPETUS Project • Server handles connec:ons based on subscrip:on model Evaluation Client/Server Communication Stability and Reliability • 24h random data test • No data loss, connec:on issues, memory leaks, or CPU loads • Server memory approx. 1.4MB, CPU < 3% Transfer rates • Average transfer rate 90% of SMB; Compared to [2] • 6 Experiments: different 100MB data packages Loading data Serialization Sending Receiving Deserialization Total OpenIGTLink Duration [sec] 16.464 0.728 8.837 11.258 0.052 37.311 Transfer rate [MB/sec] 6.074 137.269 11.316 8.883 4035.152 2.680 CAMPCom (Compression and Prioritization off) Duration [sec] 14.026 0.666 9.272 12.247 0.882 37.094 Transfer rate [MB/sec] 7.129 150.039 10.785 8.165 113.399 2.696 Prioritization and Compression • zlib based compression of MNI BITE dataset [3, 4] • 67.60% compression overall transfer rate 3.35MB/sec • 19.52% overall improvement compared to no compression Conclusion • Stable, fast, reliable framework for IGT • Easy integra:on, complex communica:on process hidden • Similar performance compared to state of art • Added features: priori :za:on, compression, subscrip:on, thread pool, memory efficiency, simple integra:on [1] Erika Garutti. EndoTOFPET-US: a Novel Multimodal Tool for Endoscopy and Positron Emission Tomography. arXiv preprint arXiv:1303.4503, 2013. [2] Junichi Tokuda, Gregory S Fischer, Xenophon Papademetris, Ziv Yaniv, et. al. OpenIGTLink: an open network protocol for image-guided therapy environment. INT J MED ROBOT COMP, 5(4):423–434, 2009. [3] http://zlib.net/feldspar.html [4] Laurence Mercier, Rolando F Del Maestro, Kevin Petrecca, David Araujo, Claire Haegelen, and D Louis Collins. Online database of clinical MR and ultrasound images of brain tumors. Med. Phys, 39:3253, 2012. EndoTOFPETUS has received funding from the European Union 7 th Framework Program (FP7/ 2007-2013) under Grant Agreement No. 256984. PicoSEC MC-Net Project is supported by a Marie Curie Early Initial Training Network Fellowship of the European Community’s Seventh Framework Programme under contract number (PITN-GA-2011-289355-PicoSEC-MCNet).
Transcript of Goal - TUM
Compu
ter A
ided
Med
ical Procedu
res
Techn
ische
Universitä
t Mün
chen
A Lightweight and Portable Communication
Framework for Multimodal Image-Guided Therapy
A. Schoch, B. Fuerst, F. Achilles, S. Demirci and N. Navab Computer Aided Medical Procedures (CAMP), TU Munich, Germany
Bladder
Rectum
Prostate
US Transducer PET Head
PET Detector Plate
PET Field of View
US Field of View
Devices and labels © 2013 TUM-CAMP, DESY and EndoTOFPET-US Collaboration, see [1] Anatomy © 2005 Terese Winslow, U.S. Govt. has certain rights
NDI Polaris
NDI Aurora
EMT Targets OT Targets
Intro
duct
ion
Motivation • Mul:modal and Intra-‐Opera:ve Imaging (EndoTOFPET-‐US) [1] • Collabora:on of researchers from areas:
• computer science (GUI, tracking, reconstruc:on) • physics (crystals, op:cs, aNenua:on and scaNer correc:on) • medicine (tracers, pre-‐op imaging) • electronics (photo electronics, read-‐out, DAQ cards)
• Full monitoring and centralized control of system required
Goal • Simple to use communica:on framework (“one header only”) • Plug-‐Ins for visualiza:on (CAMPvis) • Open-‐source: hNps://campar.in.tum.de/Main/CAMPCom
• Control & Data Flow: communicate data and system messages between various components
• Efficiency: lightweight & efficient communica:on • Pla5orm Independency: serializa:on and transmission • Abstrac=on & Extensibility: support extension, general interface • Priori=za=on: asynchronous func:on to support efficiency
Met
hods
Query Handling & Prioritization • Ac:vity diagram for client library • Simplified (without dispatcher thread pool)
Inter-Device Communication • Sample setup for EndoTOFPET-‐US Project • Server handles connec:ons based on subscrip:on model
Eval
uatio
n Client/Server Communication Stability and Reliability • 24h random data test • No data loss, connec:on issues, memory leaks, or CPU loads • Server memory approx. 1.4MB, CPU < 3%
Transfer rates • Average transfer rate 90% of SMB; Compared to [2] • 6 Experiments: different 100MB data packages
7
Loading data Serialization Sending Receiving Deserialization TotalOpenIGTLink
Duration [sec] 16.464 0.728 8.837 11.258 0.052 37.311Transfer rate [MB/sec] 6.074 137.269 11.316 8.883 4035.152 2.680
CAMPCom (Compression and Prioritization off)Duration [sec] 14.026 0.666 9.272 12.247 0.882 37.094Transfer rate [MB/sec] 7.129 150.039 10.785 8.165 113.399 2.696
Table 1: By averaging over six experiments, the transfer rates of CAMPCom and OpenIGTLink whensending a 100MB raw data package on a round trip (from client to server and back) are compared. Bothframeworks are built in the simplest implementation.
itization and compression are absolutely necessary for progressing the state of art in IGT systems and aretherefore the most crucial contributions of our proposed framework.
4 Conclusion and Outlook
We have presented a stable, reliable, and fast framework for research purposes in the field of image guidedinterventions. The CAMPCom framework provides an efficient way to connect various devices in an intra-operative environment. Its lightweight nature and extensibility allows easy integration. At the same time itis equipped with various features and hides most of the complexity of the communication process from theuser.
We have shown that the framework reaches performance similar to the state of the art software solution,while offering additional features, such as prioritization in communication, and compression of data toreduce the required bandwidth. We have provided our evaluation results and most of the implementationdetails in this paper, and will make the entire framework openly available13 through our webpage14. Webelieve that the framework will progress the state of the art in IGT systems and therefor provides a valuablecontribution to the community. Furthermore, we encourage fellow researchers to adapt the framework andcontribute new features to CAMPCom.
We are planing to add new features like XML parser in combination with a settings manager to ease duringrun-time, as well as a code generator to easily create new compile-time data exchange types and automati-cally add them to the system.
Acknowledgements: This research was performed in the course of the EndoTOFPET-US project fundedwithin the EU FP7 framework (FP7/2007-2013, Grant Agreement No. 256984). The authors would like tothank Christian Schulte zu Berge for providing and maintaining CAMPVis.
References
[1] Andinet Enquobahrie, Patrick Cheng, Kevin Gary, Luis Ibanez, David Gobbi, Frank Lindseth, ZivYaniv, Stephen Aylward, Julien Jomier, and Kevin Cleary. The image-guided surgery toolkit IGSTK:an open source C++ software toolkit. Journal of Digital Imaging, 20(1):21–33, 2007. 1.2
13Please find a summary of CC BY-NC 3.0 DE on http://creativecommons.org/licenses/by-nc/3.0/de/
14http://campar.in.tum.de/view/Main/CAMPCom
Latest version available at the Insight Journal [ http://hdl.handle.net/10380/3408]Distributed under Creative Commons Attribution License
Prioritization and Compression • zlib based compression of MNI BITE dataset [3, 4] • 67.60% compression à overall transfer rate 3.35MB/sec • 19.52% overall improvement compared to no compression Conclusion
• Stable, fast, reliable framework for IGT • Easy integra:on, complex communica:on process hidden • Similar performance compared to state of art • Added features: priori:za:on, compression, subscrip:on,
thread pool, memory efficiency, simple integra:on
[1] Erika Garutti. EndoTOFPET-US: a Novel Multimodal Tool for Endoscopy and Positron Emission Tomography. arXiv preprint arXiv:1303.4503, 2013. [2] Junichi Tokuda, Gregory S Fischer, Xenophon Papademetris, Ziv Yaniv, et. al. OpenIGTLink: an open network protocol for image-guided therapy environment. INT J MED ROBOT COMP, 5(4):423–434, 2009. [3] http://zlib.net/feldspar.html [4] Laurence Mercier, Rolando F Del Maestro, Kevin Petrecca, David Araujo, Claire Haegelen, and D Louis Collins. Online database of clinical MR and ultrasound images of brain tumors. Med. Phys, 39:3253, 2012. EndoTOFPETUS has received funding from the European Union 7th Framework Program (FP7/ 2007-2013) under Grant Agreement No. 256984. PicoSEC MC-Net Project is supported by a Marie Curie Early Initial Training Network Fellowship of the European Community’s Seventh Framework Programme under contract number (PITN-GA-2011-289355-PicoSEC-MCNet).
