Synopsis
We
present a fully automated data management workflow and quality
assurance, which is set up for large, multicentric cohort studies
including whole-body MR imaging. The workflow includes a modality
worklist, exam-synchronous DICOM transfer to centralized storage,
quality control of MR acquisition, various image-based quality
measures, web-based radiological image review for incidental
findings, visual quality scores, as well as long-term archiving. This
workflow, implemented in the MRI
Study of the German National Cohort, enables to acquire and process
more than 30 whole-body MRI
scans per day, available for IF reading within 4 hours. Deviations,
outliers, technical failures are pointed out on-the-fly.Introduction
The
German National Cohort is a population-based, longitudinal cohort study with joint
interdisciplinary endeavour to investigate causes, development and
distribution of major chronic diseases [1]. A total of 200,000 volunteers
(between 20 and 69 years) will be
recruited during 2014-2018 across 18 study centers in Germany.
A subset of 30,000 subjects will undergo in
addition a whole-body, one hour MR imaging program to build-up a comprehensive
morphologic and functional image biobank [2].
Highly standardized imaging, efficient workflow and
data handling, continuous quality control (QC) and assurance (QA), as well as
management of incidental findings (IF) are essential in this long-term,
multicentric MRI study. We set up a fully automated MRI data management
workflow offering centralized MRI data transfer, QA, web-based thin client radiologic
IF review, long-term archiving, as well as an accompanying weekly phantom QC
procedure.
Methods
The
MR imaging is conducted
at 5 MR centers distributed across Germany; all sites are equipped with new,
identical, dedicated scanners (Skyra 3T, Siemens Healthcare,
Erlangen). Each site will recruit 6,000 subjects (randomly drawn from the German population) within 4 years. The MRI program
is strongly standardized and includes advanced protocols of 4 research foci:
neurological (3DT1w, 2DFLAIR, RestingState); musculoskeletal (PD fatsat hip, T2w whole
spine); thoracoabdominal (ssT2w, 3DVibe two-point Dixon, multiecho 3DVibe); cardiovascular
(native MRA, Cine-LAX/SAX, Molli). Examination time is restricted to
approximately 15 min per focus; further details
are described in [2]. In addition, 4 Imaging Cores have been established for Coordination and Training (Munich), IFs (Heidelberg), QA (Greifswald), and IT (Bremen).
DICOM Modality Worklist: The fully automated workflow is
initiated by submitting an e-form capturing the subject's informed consent and
already pseudomized information; subject data for the MRI exam are provided by a
Modality Worklist server.
DICOM data transfer: Synchronously during image acquisition, all data are sent via
VPN-secured DICOM transfer and stored centrally on a dedicated server.
Procedure tests, QC, QA: Shortly after exam finish, various
automated quality control units check for completeness and uniqueness of the
acquired data; deviations in the protocols order and MRI parameters are
monitored and saved. Additionally, various image-based quality measures (e.g.
UIQI [3], sharpness [4], SNR [5], drift, noise correlation, N/2 correlation)
are performed. Values are compared on-the-fly to monthly cohort statistics, and
visualized graphically using ElasticSearch/Kibana [6].
Web-based upload approval: The image data upload is automatically
checked versus an independent electronic case report, and in case of repeated
protocols, the local radiographers select appropriate volumes and approve the
upload. A web-based viewing application UploadApprovalViewer
(developed within the framework of MeVisLab
[7]) does support the radiographers by indicating missing uploads, repeated
protocols, and displaying the QC measures.
Web-based IF review: The IF imaging core supervises specifically
trained radiologists at the local MRI sites who review all 12 main protocols
through a thin-client, web-based viewing application NAKOViewer (based on MeVisLab [7]), mark IFs (if applicable) and score visual quality
according to an agreed criteria catalogue. The NAKOViewer allows for efficient reading by dedicated focus
hangings and specific features (e.g. linked locations across contrasts, cine
modi, standardized reporting with protocol specific IF catalogue).
IF and Visual Quality Assurance: Coordinated by the IF and QA imaging cores, additional blinded second reads
(>10% of all subjects) are performed by IF and QA supervisors through the same web-viewing
application and forwarded to the QS team.
Monthly quality reports by the QS team inform study contributors and project leaders
about the current study status. Further administrative tasks, e.g. response to
study participants, are covered by the central NAKO data management together
with local recruiting centers.
Long-term archive: All original images, as well as associated
metadata are transferred to the NAKO's central database.
Results and Discussion
To date, more than 3,500 subjects have been included, daily acquisition
rate is close to 30 scans (~80GB). Data transfer and processing is smooth and fully
automated; ambiguities and conspicuities are signalized through meaningful
email reports. Data import and image-based QC is processed in less than 60 min;
subjects are generally ready for IF reading the same day. Loading times of
image volumes in NAKOViewer are around 3 - 5 sec.
Due to highly
standardized protocols on identical scanners and the continuous QA value
surveillance, data comparability of this long-term multicentric study can be
assured. Outliers, noticeable technical failures, as well as trends in study
changes ring an immediate alarming bell. Furthermore, all QA will support data
evaluation later on. After finishing
of baseline assessments in 2018, collected data will be available for research
projects according to the Use and Access Policy of the German National Cohort
[8].
Acknowledgements
The German
National Cohort is an interdisciplinary project conducted by more than 30
research groups within the Helmholtz and the Leibniz Association, German
Universities, and the Fraunhofer Association. The embedded MR Imaging Study is
organized by the NAKO MR Imaging Cores with colleagues from the University
Hospital Großhadern Munich (project lead), University Medicine Greifswald
(central data management, quality assurance), University Hospital Heidelberg
(incidental findings committee), and the Fraunhofer MEVIS
Bremen (MR data management and quality control). Further information on
involved people is found in references [1, 2]. The study was funded by the German Ministry for
Education and Research (BMBF), the Federal States, and the
Helmholtz
Association.
References
[1] German National Cohort (GNC) Consortium. The
German National Cohort: aims, study design and organization. Eur J
Epidemiol, 2014 May;29(5):371-82.
[2] Bamberg F, et al. Whole-Body MR Imaging in the
German National Cohort: Rationale, Design, and Technical
Background. Radiology, 2015 Oct;277(1):206-20.
[3] Wang Z, Bovik C. A Universal Image Quality Index. IEEE Signal Proc Letters, 2002
March;9(3):81-4.
[4] Henkelman RM. Measurement
of signal intensities in the presence of noise in MR images. Med Phys, 1985
12(2):232-3.
[5] Firbank MJ, et al. A comparison of two methods for measuring the signal to noise ratio on
MR images. Phys Med Biol, 1999 Dec;44(12):N261-4.
[6] ElasticSearch & Kibana,
https://www.elastic.co/products/kibana.
[7] Link F, et al. A
Flexible Research and Development Platform for Medical Image Processing and
Visualization. Proc Radiological Society of North America; 2004 Dec;
Chicago.
[8]
http://www.nationale-kohorte.de/nutzungsordnung.html.