Artikel
Reduction of inter-institutional variability in MRI readings by consensus training in the Competence Network Congenital Heart Defects
Suche in Medline nach
Autoren
-
Philipp Beerbaum
- King's College, Division of Imaging Sciences, St Thomas' Hospital, London
-
Siegfried Kropf
- Otto-von-Guericke-Universität, Institut für Biometrie und Medizinische Informatik, Magdeburg
-
Peter Barth
- Klinik für Angeborene Herzfehler, Ruhr Universität, Herz- und Diabetes-Zentrum, Bad Oeynhausen
-
Samir Sarikouch
- Medizinische Hochschule Hannover, Klinik für Herz-, Thorax-, Transplantations- und Gefäßchirurgie, Hannover
-
Ulrich Sax
- University Medical Center Goettingen, Information technology, Goettingen
-
Andrea Kelter-Kloepping
- Klinik für Angeborene Herzfehler, Ruhr Universität, Herz- und Diabetes-Zentrum, Bad Oeynhausen
-
Matthias Gutberlet
- Abteilung für Diagnostische und Interventionelle Radiologie, Universität / Herzzentrum, Leipzig
-
Titus Kuehne
- Abteilung für Angeborene Herzfehler - Kinderkardiologie, Deutsches Herzzentrum, Berlin
| Veröffentlicht: | 2. September 2009 |
|---|
Gliederung
Text
In the German "Competence Network Congenital Heart Defects", clinical trials using cardiac MRI are carried out in a multicentre setting. A central MRI corelab was implemented involving three centres with highly experienced MRI readers connected via teleradiological infrastructure. This expert group retrieves image data from a central data base into which the various study centres submit their MRI data. We sought to investigate to what degree the different institutions would disagree in their independent image interpretation, and what impact a consensus training would have.
In an initial pilot study, a total of 32 adults with repaired Tetralogy of Fallot and 23 controls underwent cardiac MRI using standardized protocols for ventricular volumes/mass and pulmonary/aortic stroke volumes. Data were blinded and analyzed independently in each of three institutions by experienced readers using validated, semi-automated segmentation software based on threshold-adjustments. Observer variance between the institutions was determined before and after training.
The initial variability of right ventricular parameters was substantial in patients, not so much in the control group. The training measures succeeded and brought a decrease. Expressed as coefficients of variation, the variability for the end diastolic volume, the end systolic volume and the ejection fraction on transverse planes for the right ventricle was reduced from 22%/19%/19% to 7%/10%/8%, respectively (p<0.025 for each). Reductions of variability were also found in the left ventricle of patients and for short axis volumetric MRI. There was low interobserver variability (2%–4%) for velocity-encoded MRI in patients and controls.
Interobserver variability is an important issue in MRI techniques and it can dilute outcome effects in clinical trials. In our network, we established a core lab for MRI readings on a high level, which is offered also for external researchers meanwhile.
This work was funded by the Federal Ministry of Education and Research (BMBF), FKZ 01G10210.
