JNC-10-076-OA.R1
On-Line Supplement
Image Acquisition and Processing – Additional Details
The basal four slices were not used for quantitative analysis due to low counts in the membranous inter-ventricular septum. Two apical slices were not used for quantitative analysis due to potential much larger partial volume errors caused by small partial thickness slices through the tip of LV.
For the phantom studies, a fresh Rb-82 generator and some practice were required to complete an infusion of 50mCi of Rb-82 into a beaker containing the exact volume of water after infusion that was mixed and poured into the tree phantom to fill it exactly by tilting or shaking until each branch of the tree phantom was filled without bubbles followed by positioning in the scanner for imaging.
Partial volume correction should have little effect on the stress/rest ratio measurements of activity concentration, since the correction factors are the same at rest and stress. However, for absolute perfusion in cc/min/gm that is of substantial clinical and research interest, the partial volume correction is essential. As validated experimentally, in either our multi-compartmental or simple flow models for calculating absolute myocardial perfusion in cc/min/gm using either Rb-82 or N-13 ammonia, the partial volume corrections for the aortic arterial input and myocardial uptake are different due to their different sizes (see equations on page 1705 of J Nuc Med 37:1701-1712, 1996.). As a clinical example, the partial volume corrections are essential for measuring the absolute threshold of stress perfusion in cc/min/gm for an observed pressure rate product causing angina or ischemia at PET stress imaging as a basis for deciding on revascularization procedures or not.
Our method is applicable for any PET scanner whose counts are proportional to the actual activity in any arbitrary units. The recovery coefficient is a unit-less scalar that accounts for relative underestimation of signal due to finite spatial resolution. This systolic/diastolic method is valid regardless of whether measured counts are scaled into Bq/cc (as in most scanners), left as raw counts, or converted into other units of count or activity density.
The details of header information for the DICOM data are a separate issue from the partial volume correction. We report what we did to prove that exporting data from the GE system to the quantitative software on the Positron workstation did not distort or lose data or alter the scaling factors. The GE–to-Positron program exports the following GE header elements into corresponding elements of the Positron software: (0028,1053) ci_slope; (0018,1242) the pwidth x1000 to convert to seconds; (0028,1052) ci_intercept as 0 and the amplitude_scale_factor of 1.0; the GE DICOM voxel sizes; the conversion factor for Bq/ml to µCi/ml as the ci_slope (divide DICOM element (0028x, 1053x) by 37037).
Three Phantom Details
The tree phantom was constructed as 4cm deep chambers machine out of a large plastic block such that there was 2 to 3 cm of plastic around all borders of the “tree branches”. The Z dimension of the tree phantom was 4 cm so that this dimension was large enough to neglect PV loses in that direction.
The tree phantom was not imaged in water since the lateral plastic walls were 2 cm to 3 cm thick plastic depending on which branch of the tree phantom is considered. In order to determine whether Rb range was different in plastic compared to water, we constructed two additional phantoms each comprised of 1cm internal diameter cylinders 6 cm long.
One phantom (thick walled) was made by drilling a 1cm hole longitudinally into a 2.54cm diameter plastic rod. The other phantom (thin walled) was made by molding 0.0762mm (0.003 inch) thick acrylic sheet around a brass rod with outside diameter of 1cm using mild heat, sealed with plastic weld and removed from the rod to make a thin walled cylindrical container. Both ends were stoppered with wood, plastic or cork, both filled with the same source of Rb-82 eluted into water and imaged simultaneously in a beaker filled with water. For the thin and thick walled cylindrical phantoms of identical physical internal diameter, the cross sectional PET images and activity profiles were identical indicating that by scanner imaging, there was no differential range effects of plastic compared to water on positron range.
The mild visual variability of tree images in color is due to several factors: (i) at edges of phantom branches, partial volume losses do occur interacting with reconstruction and background noise to make edges appear mildly irregular (ii) color scaling enhances contrast or density variation not apparent in black and white, such that a visually prominent red to white difference may correspond to small quantitative differences not apparent on black and white images (iii) absolute and relative scales are set automatically and do not always match visually so that paired displays are approximate and illustrative but do not substitute for, or exactly match, the true quantitative numbers.
Rubidium Range Reported In The Literature and Model Simulation
There is considerable variability in the reported positron ranges for PET isotopes as shown
in Table 1, quoted from several standard nuclear medicine/cardiology texts that differs by over
2-fold. Almost uniformly these texts do not cite literature to support their stated values. The differences arise since positron range can be measured in various media (air, water, foam) by various methods (experimental studies, empiric range formula from theortetical considerations, Monte Carlo simulations) and reported using various ranges (RMS, FWHM, FW-tenth-max [FWTM]). Table 1 highlights these important complexities of positron range that are not discussed in standard reference texts.
We chose to reference (#33) with the equations for positron range derived by Derenzo. He fit experimental data to a weighted sum of two exponential decays (see equation 1 on page 565 of ref 33):
range(x) = A*exp(-x/B) + (1-A)*exp(-x/C)
where constants A,B,C were derived from best fits to projection data. His Table 1 (on top of page 566 of ref 33) gives A,B,C for F-18 and Rb-82 (in addition to other isotopes) and lists RMS, FWHM, FWTM for point and line spread functions.
Our simulations were carried out in Matlab using the range equations from Derenzo that were further convolved with a Gaussian. A standard non-linear fitting method (Levenberg-Marquardt) was used to solve for the best FWHM of the Gaussian given an ideal phantom shape (one-dimensional step
function of a specified mm width) combined with the Derenzo range function for the tracer of interest. We took the average FWHM from our experimentally observed PET F18 profiles in 30mm (FWHM=9.1mm), 20mm (FWHM=9.3mm), and 15mm (FWHM=8.6mm) targets. This FWHM=9mm was used for the simulations seen in Figures 1 and 6.
Time Resolution of 8 Bin ECG Gate
The time resolution for diastolic-systolic images is limited to one of the 8 gates. The scanner cannot acquire more gates. In addition, the data density with more, shorter gates becomes so poor that noise in the shorter lower count images degrades the quantitative signal recovery. Therefore, our systolic diastolic images are approximate averages of systole and diastole that are practically useful to minimize wall motion effects while still retaining adequate data density and acceptable signal to noise ratio.
Physics Background of the Authors As the Basis for This Paper
The first author has a degree in Computer Sciences And Physics, a Masters in Biostatistics, an MD degree, four years of Cardiology training and substantial experience with publications in broad areas of signal processing. The senior author also has a degree in Physics and headed the team that designed and built the first University of Texas multi-ring PET scanner for imaging the whole heart without indexing (Wong et al to Gould, Proceedings of the IEEE Workshop on Time-of-Flight Tomography, May 16-19, 1982. Wong et al to Gould, J Nucl Med 1983 24:52-60. Mullani et al to Gould, Dynamic Imaging with High Resolution TOFPET Camera - TOFPET I. IEEE Trans. on Nucl Sci 1984 Vol. NS-31, 609-613. Mullani et al to Gould, Design and Performance of Posicam 6.5 BGO Positron Camera. J Nucl Med 1990 31:610-616.), the first SAC method of attenuation correction for cardiac PET (Xu et al to Gould, J Nucl Med 1991 32:161-165) and the first large definitive clinical reports on the consequences, causes and the solution for attenuation-emission misregistration artifacts in both classical pure PET (Loghin et to Gould J Nuc Med 2004; 45:1029-1039) and in PET-CT (Gould et al J Nucl Med 2007;48:1112-1121).
4