William F. Jones

Design of a super fast three-dimensional projection system for positron emission tomography

A hardware architecture for rapid three-dimensional reconstruction is considered for positron emission tomography (PET). For possibly improved PET performance, obliquely oriented lines of response (LORs) are to be collected and properly utilized by one of several experimental 3-D reconstruction algorithms. Image signal-to-noise may improve. Septa removal increases the signal by allowing extra LOR collection but also increases the noise due to reduced shielding against out-of-plane events. The primary utility for all LOR collection and 3-D reconstruction algorithms may lie with count starved applications. A VLSI-based architecture is described which will support forward- and back-projection for a 3-D image (128*128*32 voxels) and 4096 2-D views (each from 192*32 LOR) totaling over 25 million lines of response projected into 0.5 million voxels. Varying the number of processing engines allows the time for either forward- or back-projection to be under 60 s. The custom VLSI chip for 3-D projections should require 45000 transistors. >

Positron emission tomographic images and expectation maximization: a VLSI architecture for multiple iterations per second

A digital electronic architecture for parallel processing of the expectation maximization (EM) algorithm for positron-emission-tomography (PET) image reconstruction is proposed. Rapid (0.2-s) EM iterations on high-resolution (256*256) images are supported. Arrays of two VLSI chips perform forward and back projection calculations. The architecture is described, including data flow and partitioning relevant to EM and parallel processing. EM images are shown that are produced with software simulating the proposed hardware reconstruction algorithm. Projected cost of the system is estimated to be small in comparison to the cost of current PET scanners. >

SPMD cluster-based parallel 3D OSEM

This study empirically compares two approaches to parallel 3D OSEM that differ as to whether calculations are assigned to nodes by projection number or by transaxial plane number. For projection space decomposition (PSD), the forward projection is completely parallel, but backprojection requires a slow image synchronization. For image space decomposition (ISD), the communication associated with forward projection can be overlapped with calculation, and the communication associated with backprojection is more efficient. To compare these methods, an implementation of 3D OSEM for three PET scanners is developed that runs on an experimental, 9-node, 18-processor cluster computer. For selected benchmarks, both methods exhibit speedups in excess of 8 for 9 nodes, and comparable performance for the tested range of cluster sizes.

Design and performance of a single photon transmission measurement for the ECAT ART

The authors have implemented dual multi-slit collimator singles transmission sources on the ECAT ART comprising two sets of 12 slits having an axial pitch of twice the axial crystal ring pitch. The slits illuminating opposing detector banks are interleaved, achieving full coverage of all crystal rings as two /sup 137/Cs point sources are simultaneously rastered axially behind the collimators. Transmission data are normalized to a slab phantom scan and corrected for scatter and cross section variation using a log-linear transformation of the attenuation factors. The value of the linear attenuation coefficient in water phantoms has been measured as 0.0946/spl plusmn/0.0044 cm/sup -1/ for 2 min scans employing two 6 mCi sources. A quasi-simultaneous mock scan capability permits direct acquisition and subtraction of the emission background in post-injection transmission acquisitions.

Performance evaluation of a new LSO high resolution research tomograph-HRRT

In order to improve the capability for investigating the living human brain using positron emission tomography with regard to blood flow, metabolism and receptor characteristics for small structures such as cortical sublayers and nuclei, the spatial resolution has to be improved relative to what is available today. A spatial resolution of 2 mm or less in all three dimensions may be necessary to reach these research goals. In order to meet this goal a new next generation high resolution 3D-only LSO brain PET has been developed at CTI. The HRRT (High Resolution Research Tomograph) is not only the first LSO PET for human studies it is also the first PET with full DOI capability over an extended FOV. The HRRT has been delivered to the MPI Cologne in February 1999 and is presently being optimized in terms of energy discrimination, crystal positioning, coincidence timing and PSD to achieve optimal system performance with respect to resolution, count rate efficiency and scatter. The panel detector setup with its new setup challenges is still under development and investigations. First evaluation measurements are presently obtained. The measurements promise an excellent high resolution performance with a high count rate capability of the HRRT. The high random rates and the SF fraction measurements underline the necessity for a short coincidence time window and an improved energy resolution for such an in the images reveal a not yet optimized system setup and reconstruction parameters. The authors believe the results demonstrate the high capability of the new brain tomograph and justify the excitement for the first LSO tomograph with DOI capability for human brain investigations and the new scintillator LSO, superior for coincidence timing and energy resolution.

Design and performance of collimated coincidence point sources for simultaneous transmission measurements in 3-D PET

The authors have implemented a simultaneous emission-transmission measurement for three-dimensional positron emission tomography (3-D PET) using a collimated coincidence point source design employing a fast, dedicated, reference detector close to the transmission source. This design reduces the effects of randoms, scatter, dead time, and sensitivity loss on the emission data compared to previous implementations. It also greatly reduces the effect of emission contamination of the transmission data compared to the use of rod sources. Here, the authors present performance characterizations of this measurement technique on both the Siemens/CTI ECAT ART and PET/SPECT tomographs. The main effect of the transmission sources on the emission measurement is an increased randoms rate, which lends to a 10-25% reduction in NECR at specific activities >2 kBq/mL in a 21-cm-diameter phantom on the PET/SPECT. Emission contamination effects on the transmission measurement are estimated to be less than 1% for up to 20 kBq/mL in a 21-cm phantom on the PET/SPECT. Both the emission and transmission NECR are dominated by the effects of randoms. Considering the effects of both emission and transmission noise on the final corrected image, it appears that 3-6 kBq/mL of emitter concentration is an optimal imaging range for simultaneous acquisitions. The authors present the first images of a normal volunteer using this system on a Siemens/CTI PET/SPECT tomograph.

A VMEBUS based, real time sorter design for positron emission tomography

Important design aspecrs of a real-time sorting system for PET have been discussed. A general functional description was offered. Principal data flows witnin the design wrere listed. Processing algcrithms were detailed. A critical performance goal of 1MHz throughput and measured compliance were given.

Strategies for accurate attenuation correction with single photon transmission measurements in 3D PET

3D PET systems benefit from using single photon sources for transmission measurements as this reduces dead time and improves the photon flux, and therefore the signal-to-noise ratio, in the attenuation (CL) correction data. However, uncollimated point source measurements will include a large (/spl sim/50%) scatter component which biases the /spl mu/ values. No single approach will necessarily be able to mitigate against this for different scanning scenarios. The authors have investigated five approaches to scatter removal and/or correction using a /sup 137/Cs source for transmission data: (i) rescaling the measured data, (ii) energy window methods, (iii) restricting the polar acceptance angle (/spl theta/), (iv) collimating the source (ECAT ART only), and (v) segmentation and reassignment with expected /spl mu/ values. The results indicate that physical collimation of the ECAT ART point source and segmentation on the fixed ring EXACT 3D produce quantitatively useful data.

Optimizing rod window width in positron emission tomography

A technique that determines optimal rod window width for rotating rod transmission studies in positron emission tomography (PET) is discussed. Rod windowing reduces noise in rotating rod transmission studies. Lines-of-response (LOR) which intersect the rods generate primarily true coincidence events. LORs which pass far from the rods generate random and scatter events. Since the angular position of the rotating rods is known in real-time, LORs, which produce mostly noise are gated off.<<ETX>>

LSO PET/SPECT spatial resolution: critical on-line DOI rebinning methods and results

This article reports on PET-mode transaxial spatial resolution for a new hybrid PET/SPECT system. This system is based on a rotating dual-head tomograph using LSO/NaI scintillators and is capable of depth-of-interaction (DOI) measurement. For this LSO PET/SPECT, each detector head contains two 84/spl times/120 planar arrays of crystals-one array NaI, the other LSO. LSO and NaI crystals are each 0.44/spl times/0.44/spl times/1 cm. In PET mode, each head is radially offset 36 cm from FOV center. Relative orientation of the two heads is 158 degrees. Dual-level DOI discriminates LSO versus NaI by scintillation pulse decay time. While significant for good image resolution, dual-level DOI isolates gamma detection only to the (1 cm) long axis depth of each individual crystal. Critical to delivery of excellent spatial resolution are additional methods to precisely characterize the depth of the centroid of the probability of gamma interaction along the crystal long axis. This depth is measured to /spl plusmn/0.01 cm for each of 4 planar arrays and applied for more accurate LOR positioning in super-fast real-time rebinning hardware. With DOI and measured centroid depth applied. FWHM transaxial spatial resolution is 0.4 cm at FOV center and 0.6 cm at 10 cm off center. Without DOI, resolution degrades as much as 33%.