A Digital Reference Object for Testing FDG-PET/CT Display Software By Larry A. Pierce,PhD, and Paul E. Kinahan, PhD
Once a CT, MR or other scanner generates a medical image file in DICOM format, this file is typically viewed on multiple display stations with different viewing software packages [See Figure 1].
Figure 1. Data flow for PET/CT DICOM images. The proposed role of the PET/CT digital reference object (DRO) as a reference standard is also shown.
In PET/CT imaging, the standardized uptake value (SUV) is the most common quantitative imaging biomarker for diagnosis, staging, treatment planning, and assessing therapy response. When examining a region of interest (ROI) in a PET image volume, either the maximum or mean SUV voxel value within the ROI is reported. The DICOM standard lists hundreds of modality-specific data fields that are stored as part of the digital medical image files [www.dicom.org]. PET/CT viewing software should be capable of reading and interpreting the fields as needed to ensure the fidelity of the image display and to calculate SUVs correctly.
Based on these concepts, we identified the following knowledge gap: For an identical physical test object, do different PET/CT image display systems produce the same results? This is illustrated in Figure 1 by the proposed insertion of a PET/CT digital reference object (DRO) as a reference standard with known truth.
We created a 3D PET/CT DRO from a set of DICOM PET and CT image files with known voxel values and DICOM data fields. The PET/CT DRO was then viewed on 22 different combinations of medical image viewing workstations and software packages by QIBA FDG-PET/CT Technical Committee members. Interestingly, there was no identical combination of the software, version, or display station base system in the 22 tests. Six circular ROIs were specified for use with the DRO and the maximum, minimum, mean, and standard deviation of the SUV values within each ROI was recorded. The color-coded map [See Figure 2] classifies the reported SUV values as either acceptable, borderline, or outside the acceptable range.
Figure 2. Top: Results from user-reported SUV values from the PET/CT DRO. Each column represents a single site/system and each row corresponds to a metric from one of the six ROIs indicated. Bottom: ROIs overlaid on the primary slice of the PET component of the DRO and an illustration of the threshold definitions for one ROI metric.
Results generally indicate that common metrics (e.g., SUVmax for ROI 1 and SUVmean for ROI 2) are correctly reported. But it is also evident that there are multiple failure modes for relatively basic metrics. In one case, the maximum SUV value was under-reported by 38% when analyzing the single hot voxel (ROI 3). In the checkerboard region, four software packages over-reported the max SUV values by 11% and over-reported the mean SUV values for that region by as much as 100%. Other anomalies included misalignment of the PET and CT images, artifacts appearing on zoomed images, and the inability to see the checkerboard regions on the monitor.
These results illustrate the potential role of the PET/CT DRO to help ensure that SUV values are computed correctly. The DRO has been included as a component of the QIBA FDG-PET/CT Profile  and results have been presented at scientific meetings and discussed with vendors.
The authors acknowledge the contributions of David Clunie, MBBS, Dennis Nelson, PhD, and Brian Elston to this project as well as the DRO testing and comments provided by QIBA FDG-PET/CT Technical Committee members.
 Quantitative Imaging Biomarkers Alliance. QIBA Profile. FDG-PET/CT as an Imaging Biomarker Measuring Response to Cancer Therapy. 2013; http://qibawiki.rsna.org/index.php?title=FDG-PET_tech_ctte.
|Larry Pierce, PhD, is a research scientist in the Imaging Research Laboratory of the Department of Radiology at the University of Washington in Seattle. He has concentrated on PET image reconstruction and analysis for several years.
|Paul Kinahan, PhD, is a professor of radiology at the University of Washington. He is chair of the American Association of Physicists in Medicine (AAPM)/SNMMI Task Group on Quantitative PET/CT Imaging and participates in SNMMI, AAPM, and RSNA initiatives on quantitative medical imaging as a biomarker.
[ BACK TO TOP ]