As part of a NIST-wide effort to develop a metrological infrastructure for quantitative medical imaging, the Radioactivity Group established a dedicated facility to support its efforts to develop the necessary standards and measurement techniques for quantitative PET-CT.
The cornerstone of the facility is a Philips Gemini-TF clinical PET-CT scanner that was acquired under the America Recovery and Reinvestment Act (ARRA) during FY10. The scanner is used for characterizing new phantoms that the Group is develops for both PET and CT and acts as a testbed for the development of data acquisition and analysis methods, as well as acts as a reference for interlaboratory phantom comparisons.
The first project was calibration of the scanner for 18F activity measurement on an absolute basis with traceability to National standards for that radionuclide. Additional studies are ongoing to cross-calibrate the scanner for other positron emitters, such as 68Ge, and investigate its use as a transfer standard for calibrating various phantoms for contained activity. A much larger, long-term project is being carried out to identify and quantify the principal components of uncertainty in the imaging process and how they affect the activity determination. Basic imaging physics studies, such as partial volume correction methods and translation of absolute image quantification to the clinic are an important part of the program.
In addition to the work on PET imaging standards, a similar program is being carried out in the area of CT to provide measurement infrastructure for that modality as well. Some of the tasks carried out include:
CT lung density reference. NIST has developed a lung density reference (SRM-2088) based on commercially available polyurethane foams, with SI traceable physical densities through lot certification using the CT scanner. The effort was collaboration between the staff members in the PET/CT Imaging team and members from other NIST groups: Zachary Levine, Technical champion; Adam Pintar, statistical analysis; Dan Sawyer, Dimension metrology; and Michelle O’Brien, x-ray transmission measurements. Batches of 8 units of the 5-density foam suite were scanned and volume histograms generated; curve fitting to the peaks determines the centroid and SD of the densities and uncertainties. SI-traceable physical measurements (length and mass) were performed on fifteen randomly selected units, the density vs CT number calibration was used to certify the density of each sample in the entire lot. This effort is closely aligned with the monitoring of emphysema/COPD disease progression using lung density as a biomarker established by quantitative CT imaging, an objective of RSNA/QIBA Lung Density Subcommittee. In order to assess the consistency and reproducibility of CT lung density quantitatively, the Subcommittee formed by vendors, radiologists and medical physicist developed a lung phantom using similar foams. The first round of phantom scanning carried out have revealed large variations of the HU values from scanner to scanner. The scanners can be cross calibrated by using the internal foams by using the assumed nominal density. With the NIST SRM for which the density is certified for each of the 5-foam block, a linear relationship between the HU value and density can be obtained for each scanner protocol without making assumptions. The next round of vendor scanning will incorporate the SRM foam to build a reliable and robust calibration procedure. In addition, various aspects of the SRM are being further characterized. For example, the SRM is certified by in-air measurements. When placed inside a chest phantom where attenuation and scattering of the x-ray beams may affect the reconstructions, would the HU value shift? A set of measurements was carried out inside a anthropomorphic phantom at a clinical CT machine in Mt. Sinai hospital, in collaboration with Rick Avila (Accumetra) and David Yankelevitz (Mt. Sinai). Repeat scans of two current setting (high/low dose) and two recon thicknesses (high/low noise) were acquired, and volume histogram obtained by the same image analysis as before. The in-air density peaks are much narrower than the in-phantom ones, as expected. The standard deviation of the peaks depends on the measurement parameters, however the centroid (mean density) values are quite constant. Therefore, calibration of the CT machine can be carried out reliably within experimental uncertainty. Statistical analysis is needed to quantify the errors to specify the accuracy and precision from these measurements which can serve as a lower limit for real biological systems. Collaboration with John Lu is continuing to assess the effect of the different factors. After more rigorous statistical analysis, we hope to be able to answer the question of the smallest measureable changes in lung density attainable.
Direct realization of absorbed dose to phantom in CT beams. The air kerma is the current diagnostics x-ray standard (maintained by the NIST Dosimetry Group) which requires extensive modeling to be converted to dose to patients. In the therapy world, the standard is calorimetry which direct realizes the absorbed dose in material by measuring the temperature change in the material due to the absorption of the radiation. It is well-established for absolute dosimetry at the MeV range and doses in the order of Gys. However, for diagnostic CT beams typically in the order of 100 keV and mGy, calorimetry methods had not been attempted. We have embarked on the calorimetry measurement of absorbed dose to phantom in the CT beam by using the recently implanted AAPM TG200 high density polyethylene phantom fitted with a thermistor-imbedded solid polystyrene detector core, and migrating the detection circuitry and calibration schemes previously employed in the NIST water calorimeter. Together with Ron Tosh and Fred Bateman of the Dosimetry Group, this system was first tested in therapy-level radiation in the NIST Clinac 2100C medical accelerator, and yielded results found to be in excellent agreement with results from an ionization chamber calibrated for absorbed dose to water in the same beam, after converting the latter to dose to polystyrene via Monte Carlo simulation of radiation transport done by another Dosimetry Group colleague, Paul Bergstrom). The system was then moved to the PET/CT scanner where consecutive axial scans (a single rotation of the x-ray source in 2 s) were performed and time waveform recorded. The same measurement was performed using a calibrated ionization based on air kerma standard. While excellent linearity over a factor of 3 increased x-ray tube current was observed between the calorimetry signal demonstrated and the machine output (CTDIvol) or the air kerma value reported by the ionization chamber, several technical challenges remain to obtain the true dose to phantom, due to the much lower doses compared to therapy beams, and large amount of excess-heat due to the high-Z materials in the thermistor circuitry with a higher absorption cross section amplified at lower x-ray energy that are negligible in therapy beams. Another challenge involves converting absorbed dose to phantom to measured air kerma that involves modeling and numerical calculations. Three approaches have been tested in correcting the calorimeter response for excess heat: 1) Least square fit of the temperature waveform acquired at various dose rates to determine the ratio of contributions of heating due to non-PS materials to that due to PS (true dose component); the verification of this by comparing the corrected dose to the converted results of the air kerma requires monte carlo simulation of the latter conversion, which is under continuing investigation. 2) Finite-element simulations of heat transfer coupled with Monte Carlo simulations of relative dose distributions within elements of the calorimeter core. This also requires a precise knowledge of the geometry and thermal property of the materials. A non-Monte Carlo radiation transport (Linearized Boltzmann Transport Equations) has been investigated for a simplified model and compared with Monte Carlo results (Matt Mille, NRC Postdoc, Dosimetry Group) to adapt deterministic dose distribution assessment for the subsequent heat transport model. 3) Further refinements of the hardware have been planned to reduce the effect of excess heat, thereby reducing the associated uncertainty in the correction factors applied to measured values to obtain the true dose to phantom.