Presentation is loading. Please wait.

Presentation is loading. Please wait.

Combined DICOM and HL7 viewer in support of a bridge from content-based image retrieval to computer-aided diagnosis - ITiB - June 7, 2010 Welter P1,

Similar presentations

Presentation on theme: "Combined DICOM and HL7 viewer in support of a bridge from content-based image retrieval to computer-aided diagnosis - ITiB - June 7, 2010 Welter P1,"— Presentation transcript:

1 Combined DICOM and HL7 viewer in support of a bridge from content-based image retrieval to computer-aided diagnosis - ITiB - June 7, 2010 Welter P1, Topal F1, Jansen S1, Deserno TM1, Riesmeier J2, Grouls C3, Günther RW3 1Dept. of Medical Informatics, RWTH Aachen University 2ICSMED AG, Oldenburg 3Dept. of Diagnostic Radiology, RWTH Aachen University

2 Content-Based Image Retrieval (CBIR)
Search in large image repositories using features describing the image content extraction of features texture shape gray value … DB comparison of features CBIR system query by example result images CBIR process typically…: User/physican queries similar images to a given query image (Query by Example)

3 CBIR for Computer-Aided Diagnosis
fat embolism Goodpasture’s syndrome silicosis IRMA Physicians usually refer in their decision making of a diagnose to already closed cases with verified diagnostic results. In the context of CADx, the typical application of CBIR supports differential diagnosis, i.e. the distinguishing between two or more diseases by systematical comparison. The aim of CBIR is to provide the radiologist with a diagnostic aid by automatically identifying relevant past cases along with their diagnosis and other suitable information. This provides the radiologist with a second opinion, thereby enhancing the diagnostic assessment. CBIR has a high potential of improving the quality and efficiency of clinical care processes and entails proved high benefits for CADx. There are promising CBIR based CADx systems, the majority concentrating on a particular application area, e.g., lung cancer, mammography, computed tomographic images of the chest. However, CBIR is not established yet in clinical routine. CAD schemes using CBIR approaches face a number of challenges and this technology needs more research work. One essential factor is the seamless and robust integration into Picture Archiving and Communication Systems (PACS) and the radiologist’s workflow. 2nd opinion IRMA: Image Retrieval in Medical Applications

4 Challenge: Handling of CBIR Results
Aim use of open standards integration with a hospital-wide communication structure DICOM Structured Report CBIR-System This includes a suitable handling of the CBIR results. Because the appropriate and contextualized preparation of necessary data required for differential diagnosis delivers an indispensable brick in the bridge from CBIR to real CADx. Digital Imaging and Communications in Medicine (DICOM) Structured Reporting (SR) is a mean for encoding structured information and is an approved standard format for exchanging CADx results in clinical environments, e.g., for mammography, which can be stored to and retrieved from PACS. Because of the many advantages, the Integrating the Healthcare Enterprise (IHE) employs DICOM SR in several integration profiles. Our group defined a SR template for the application of CBIR to CADx. Hospital Information System (HIS) Picture archiving and communication system (PACS) Radiology Information System (RIS) Welter P, Deserno TM, Gülpers R et al. (2010) Exemplary Design of a DICOM Structured Report Template for CBIR Integration into Radiological Routine. Procs SPIE Medical Imaging 7628.

5 Challenge: Handling of CBIR Results
Aim use of open standards integration with a hospital-wide communication structure DICOM Structured Report comprehensive and complete representation Query examination image Identified similar examination images Diagnostic findings Patient information

6 State of the Art Representation
There are viewers capable of interpreting SR documents and producing a generic layout. As most of the SR viewers DICOMscope produces HTML output. The layout is kept generally. A conversion into XML format offers the adapted representation by using XSLT and CSS. OFFIS, DICOMscope DICOM viewer,

7 DICOMscope (2) Viewers like DICOMscope are not adjusted to the special needs of CBIR for CADx. Diagnostic findings from different clinical departments, e.g. pathology, laboratory or endoscopy, are usually archived centrally by the Hospital Information System (HIS), utilizing the Health Level Seven (HL7) protocol. Clinical viewers for diagnostic findings are usually not capable of integrating CBIR results or even CADx results in general. We close this gap and present the novel IRMACON viewer, a combined viewer for CBIR SR documents together with referenced diagnostic findings from HIS.

8 diagnostic workstation
CBIR based CADx scheme PACS DB Central storage of CBIR results Retrieval of CBIR results CBIR system Retrieval of similar examinations HIS DB diagnostic workstation For our approach, we designed the following CADx scheme. The physician examines medical images to make a diagnosis. Either, the CADx request is done automatically, e.g., by configuring an internal mechanism inside of the PACS, or the physician manually demands the CADx request. Data from the current patient’s context is then transferred to the CBIR system. The CBIR system processes the query input and stores the CBIR results in PACS. The DICOM archive holds all SR documents and delivers requested CBIR results to the physician at his diagnostic workstation, who is awaiting the CBIR support for his diagnosis. The CBIR SR document lists identified images with a certain minimum similarity to the current examination image. Normally, the physician then requests diagnostic details of selected similar images from HIS manually to collect all relevant information for his decision making. We suggest to automatically integrate the corresponding findings in the presentation of the CBIR results. Transfer of patient context Retrieval of diagnostic findings Flow of data from source to destination

9 diagnostic workstation
System design PACS DB SR IRMACON Viewer Findings diagnostic workstation HIS DB In our system design, the IRMACON viewer is responsible for the retrieval of the CBIR results and the corresponding diagnostic findings, and the suitable representation to the physician. HL7 DICOM PHP/HTML

10 diagnostic workstation
Implementation PACS DB C-MOVE XML PNG PHP, Smarty Templates Mapping DICOM - HL7 IRMACON Viewer QRY^R02 ORF^R04 diagnostic workstation HIS DB The IRMACON viewer is a web-based application implemented in Hypertext Preprocessor (PHP) that dynamically generates output in Hypertext Markup Language (HTML) format. OFFIS DCMTK [24] is used for realizing DICOM commands from within IRMACON. HL7 messages are implemented using HAPI (HL7 Application Programming Interface) [25]. SR documents are converted to Extensible Markup Language (XML) by DCMTK and then parsed by the XML extension of PHP. The retrieved DICOM images are converted to Portable Network Graphics (PNG) in order to have an image format suitable for the representation in web browser. We created a testbed for our viewer limiting PACS and HIS to their relevant functionalities needed in our scenario. It includes a DICOM database accomplished by DCMTK configured with an DICOM Application Entity Title (AET) for accessing DICOM images and SR documents. Further, the HIS is simulated by a simple HL7 application. The IRMACON viewer queries diagnostic findings by HL7 protocol that relate to patients and examinations whose medical images have been retrieved by the CBIR system. This requires a mapping of identifiers between the DICOM and the HL7 domain. HAPI, DCMTK,

11 Mapping DICOM - HL7 … Patienten-ID (0010,0020)
Accession-Number (0008,0050) QRD-9 Query for results CBIR SR-Document … QRD| |R|I|Q4412|||10^RD| |RES| … Relevant keys are patient and order which are used in the HL7 query message to query particular diagnostic findings. There is no official standard that defines a mapping. But based on generally applied principles and best practice, we decided to use the following mapping: HL7 QRY^R02 QRD-8 Patienten-ID QRD-10 Order-No

12 Exemplary HL7 messages Query for results of observation QRY^R02
MSH|^~\&|IRMACON||EKG||||QRY^R02|CDB22222|P|2.5 QRD| |R|I|Q4412|||10^RD| |RES| QRF|EKG|| Response ORF^R04 MSH|^~\&|EKG||IRMACON||||ORF^R04|X981672|P OBX|1|ST| ^VENTRICULAR RATE(EKG)||91|/MIN|60-100 OBX|10|FT|93000&ADT^EKG COMMENT||\.br\ 1. Ein Vergleich mit einem EKG vom zeigt, dass die ventrikuläre Frequenz um 30 bpm gestiegen ist.\.br\2. Die Kriterien für einen Seitenwandinfarkt sind nicht länger gegeben. The actual diagnostic information has to be extracted from the message and to be displayed to the radiologist. A HL7 query message is composed by a message header (“MSH”), followed by a query definition (“QRD”) [Patient-ID, and a query filter (“QRF”). Hier ist das erste Feld definiert als das erste auf „MSH“ folgende Zeichen welches den Feldtrenner (1) definiert; MSH-Segmentes (2) weitere Definitionen für Trennzeichen (Komponententrenner („^“)); sendende (3, 4) und empfangende Anwendung (5) und Einrichtung (6); optional und dienen zu Routing-Zwecken ; Datum und Zeit des Nachrichtenversands (7), ein optionales Sicherheitsfeld (8) und der Nachrichtentyp (9). MSH-9 Nachrichtentyps ;Nachrichten-Kontrollnummer (10), eine Prozess-ID (11) und die Versionsnummer des verwendeten Standards (12). Die Daten sollen strukturiert im Record-Format („R“) übermittelt und die Abfrage sofort („I“) beantwortet werden. Fortlaufende eindeutige Nummer der Abfrage ist „Q4412“. Sie wird bei jeder neuen Abfrage um eins hoch gezählt und dient der Identifizierung der Abfrage. Die Antwort darf maximal 10 Datensätze enthalten („10^RD“). QRD-9: „What subject fi lter“) enthält RES (Anfrage nach Ergebnissen). Die Filter schränken die Ergebnisse auf das EKG-System („EKG“) und den spätesten Zeitpunkt der Erstellung oder Änderung der Ergebnisse zum („ “) ein. OBX-Feld Beschreibung OBX-1 fortlaufende Nummer bei mehreren Einzelbefunden pro Nachricht („1“); OBX-2 Ergebnisformat, in diesem Fall das String-Format („ST“); OBX-3 Bezeichnung der Untersuchungsmethode („002^KALIUM“); OBX -5 Messwert des Ergebnisses („91“); OBX -6 Maßeinheit des Ergebnisses („/min“); OBX -7 Referenzbereich („60-100“), in dem die Werte als normal anzusehen sind; OBX -8 Bewertung des Ergebnisses; „N“ steht für Normalbefund; OBX -10 Art des Referenzbereichs; „N“ steht für „genereller Normwert“; OBX -11 Ergebnisstatus; „F“ steht für Endbefund; OBX -14 Zeitpunkt der Untersuchung, hier der um 7.00 Uhr („ “) - MSA|AA|CDB22222|P (als Segment in Antwort für Acknowledgement)

13 IRMACON Viewer It shows preliminary results from our IRMACON viewer. The IRMACON viewer is adapted for representing CBIR results based on our CBIR SR template; layout adjusted for CADx. The advantage in contrast to DICOMscope or any general SR viewer is that we utilized our knowledge about the workflow and setting in which the CBIR results are used. In the top area the query image is represented. Below, the determined similar images are given in descending order of similarity. Each downscaled image can be increased by double click. In future: information on reference database and CBIR system can be received by clicking on a menu symbol. Allows insight into identified similar examinations for convenient comparison by accessing patient’s information and diagnostic findings

14 IRMACON Viewer (Detail)

15 IRMACON Viewer (Findings)

16 Summary Aim: Integration of CBIR results into radiological clinical routine for CADx Solution: Novel IRMACON viewer CBIR results (DICOM SR documents) diagnostic findings (HL7 results of observation) Achievement standardized and comprehensive representation evidences sourced from CBIR, PACS and HIS convenient integration into clinical environments improvement of radiologist’s workflow and medical diagnosis IRMACON(general viewer) establishes an essential link between CBIR systems and applied CADx. We hope that our proposal will contribute to the widespread use of CBIR for CADx in daily radiological routine.

17 Outlook Extension of test settings: HIS
Processing region of interest: restricting CBIR to a certain phenomenon

18 Acknowledgments Fund
German Research Foundation (DFG) grants Le 1108/9


20 Possible Format Solutions
Standard file formats (e.g., PDF) include layout directives no automatic processing HL7 Clinical Document Architecture (CDA) XML format supports automatic processing no layout directives DICOM Structured Reporting (DICOM-SR) supports automatic processing DICOM is standard for communication in PACS used in integration profiles of the Integrating the Healthcare Enterprise (IHE) Portable Document Format (PDF) or MS Word are widely-used; CDA documents are encoded in XML format. Layout formatting is realized by XSLT for example. Hence, the responsibility of a meaningful representation lies with the viewer. DICOM SR is a mean of encoding structured information. It does not provide any specification regarding the layout of the representation. DICOM guarantees functional interoperability, but no semantic interoperability. utilizing established clinical Infrastructure Templates define valid contents and value types to restrain different encoding of the same content in order to simplify the processing of SR documents. They facilitate an automated processing and interpretation of SR documents. DICOM standard templates do not cover the CBIR application for CADx. We proposed a template adapted to the special requirements of CBIR. It comprises the query image, identified similar images along with their similarity score, and a description of the applied CBIR system. Our template is applicable by any CBIR system and the resulting SR document may be stored to PACS. Welter P, Deserno TM, Gülpers R et al. (2010) Exemplary Design of a DICOM Structured Report Template for CBIR Integration into Radiological Routine. Procs SPIE Medical Imaging 7628.

Download ppt "Combined DICOM and HL7 viewer in support of a bridge from content-based image retrieval to computer-aided diagnosis - ITiB - June 7, 2010 Welter P1,"

Similar presentations

Ads by Google