1. A service for monitoring the quality of intraoperative cone beam CT images
Frank Heckel1,2, Jan Schlamelcher3, Hanna Rotermund4, Max Rockstroh2, Stephan Zidowitz1, Thomas Neumuth2
1 Fraunhofer Institute for Medical Image Computing MEVIS, Bremen, Germany
2 Innovation Center Computer Assisted Surgery, Leipzig, Germany
3 OFFIS e. V., Oldenburg, Germanyt
4 Ziehm Imaging GmbH, Nuremberg, Germany

2. Introduction Central role of medical images (CT, MRI, …)
Diagnosis, planning, guided interventions
Pre-operative vs. intra-operative imaging (US, CBCT)
Trend from independent devices to digital, integrated ORs
Device communication
Mutual control
Proprietary solutions vs. open protocols (e.g., OR.NET/OSCP)
Image data exchange via DICOM
Image data processing commonly done by dedicated workstations (e.g., navigation system or planning tools)
Proposal: Dynamic service component for image data processing in an integrated OR using OSCP and DICOM (use case: quality assessment)
OSCP = Open Surgical Communication Protocol
DICOM = Digital Imaging and Communications in Medicine

3. Automatic Image Quality Assessment Framework (AQUA)
Basis are fully automatic “Tests”
Compute specific (numerical) features given an input image (an other inputs)
Examples: SNR, sharpness (aka “quality index”), resolution, artifacts, …
Atomic vs. tests with preconditions (= complex tests)
On top: management and interpretation layers
Interpretation
Management
Atomic Tests
Test Preconditions
Complex Tests
Test Sets
Atomic: Only gets input image
Complex: Depends on Precondition (e.g., computation of a mask image)
Precondition: Only gets input image
Set: Collection of tests for a specific use case (includes preconditions and parameters)

4. Automatic Image Quality Assessment Framework
The liquid Python Library
Implemented on top of MeVisLab
Main purpose is usage in Python but for convenience there is the module “LiquidTestManager”
A MeVisLab module is turned into a liquid-Test by adding an XML-based description
Liquid is not restricted to run tests in MeVisLab
Test results can be stored in XML as well
* liquid = medical image quality judgment

5. AQUA Service Usa case: Intra-operative CBCT imaging
Utilized Tests: SNR, Quality Index (= Sharpness), Resolution, Metal artifacts
UI with traffic light metaphor

6. AQUA Service Voting System as Interpretation Layer
Options o: 0.0 = bad, 0.7 = acceptable, 1.0 = good
Each test t votes for each option o (1.0 = fully agree, 0.0 = fully disagree)
Note that the voting functions above are use-case-specific

7. AQUA Service
Voting System Example

8. AQUA Service Device Communication
Exchange of IP etc. via MDPWS discovery
AQUA provides two DICOM services:
Modality Performed Procedure Step (MPPS)  provides Series Instance UID
PACS image retrieval
DICOM communication is done transparently via the DICOM gateway
Results are send to other devices in terms of an OSCP provider
MDPWS = Medical Device Profile for Web Services (= „transportation layer “)
OSCP = Open Surgical Communication Protocol (= „message format“)

9. Results
In a demonstrator setting (Leipzig 12/2015, conhIT 04/2016) we had 3 exemplary images of a head phantom
Good / slightly noisy / with severe metal artifacts
AQUA service was able to distinguish between them and it returned to correct reason for bad quality

10. Results Overall time about 1 minute, which includes
Storing of C-Arm data in the PACS
Retrieval of data from PACS
Conversion to MeVisLab-specific image format
Quality assessment

11. Summary
Successful integration of a image processing service into an integrated OR
Completely based on open protocols
Fully automatic + nearly real-time access to newly acquired CBCT data
AQUA service could permanently review image data in background
AQUA was able to distinguish between the exemplary images
Image quality assessment is only one use-case
Surgery documentation support
Navigation system accuracy estimation
3D reconstruction for intra-operative navigation based on CBCT data …

12. Thank You

13. Appendix

14. Introduction OR.NET Funded by BMBF Duration: 09/2012 – 04/2016
Budget: 18 Mio. €
A LOT of partners: 49 regular + 39 associated
Goal: Development of open, save, certifiable, dynamic (plug & play), multi-vendor networking options for existing and future devices and software solutions in the medical environment

15. (Medical) Device Profile for Web Services ((M)DPWS)
OASIS Standard DPWS 1.1
Provides functionalities such as data exchange, dynamic discovery of devices, device description, events
MDPWS
Extension to DPWS:
Safe remote control via IP-based networks (error detection via checksums, safety context, …)
Streaming data compression
Restrictions of DPWS:
???

16. Medical Device Profile for Web Services (MDPWS)
Medical Device Information Base (MDIB)
Hierarchical Description of a medical device
MDS = Medical Device System
VMD = Virtual Medical Devices
Channels = physical or logical group of parameters (Metrics)

17. Medical Device Profile for Web Services (MDPWS)
Description of a medical device: Medical Device Information Base (MDIB)
State of a medical device
Value of each Metric
Alerts
Activation
Interaction between devices:
Get
Event report
Set (only for parameters that explicitly allow this)

18. Open Surgical Communication Protocol (OSCP)
OSCP defines the message format but not how they are transported
MDIB
Serialization (XML)
Transportation layer is only suggested (→ DPWS)
HL7 and DICOM are bypassed
For real time communication a specific bus is used (Ethernet Powerlink)
E.g. for controlling a mill via a foot switch
Communication with clinic IT systems (KIS, PACS, …) is done via gateways
Arbitrary proprietary interfaces via „connectors“
Provider vs. Consumer

19. Open Surgical Communication Protocol (OSCP)

20. Open Surgical Communication Protocol (OSCP)
Implementations
OpenSDC (Java)
OSCLib (C++)
Implemented by SurgiTAIX AG (Aachen)
Open source:
Required 3rd party libs: POCO, Xerces
License pending (might be some kind of dual license – free license could be GPL(!) …)

21. The liquid Python Library
Other modules that use liquid so far:
ORNET_DataAnnotationTool
ORNET_AnnotationAnalyzationTool
(ORNET_AQUAService)