Mark W. Lenox

Motion compensation in histogram-mode and list-mode EM reconstructions: beyond the event-driven approach

With continuous improvements in spatial resolution of positron emission tomography (PET) scanners, small patient movements during PET imaging become a significant source of resolution degradation. This work explores incorporation of motion information into expectation-maximization (EM) reconstruction algorithms. An important issue addressed is the existence of lines-of-response (LORs) corresponding to no actual pairs of detectors and their motion-induced interaction with the detectable LORs. An example of this is a scanner design with gaps existing in between the detector heads. It is shown that to properly account for such LORs in histogram-mode and list-mode EM reconstructions, in addition to motion correction of the events, the algorithms themselves must be modified. This modification is implemented by including motion-compensated sensitivity correction factors. We are able to demonstrate experimentally that the proposed approach resolves image artifacts that can appear when the conventional purely event-driven motion correction technique is used. An alternate image-space-based method for calculation of motion-compensated sensitivity factors is also derived, applicable in both histogram-mode and list-mode reconstruction tasks, which has the potential of being considerably faster in presence of frequent motion, especially in high-resolution tomographs.

Statistical list-mode image reconstruction for the high resolution research tomograph

We have investigated statistical list-mode reconstruction applicable to a depth-encoding high resolution research tomograph. An image non-negativity constraint has been employed in the reconstructions and is shown to effectively remove the overestimation bias introduced by the sinogram non-negativity constraint. We have furthermore implemented a convergent subsetized (CS) list-mode reconstruction algorithm, based on previous work (Hsiao et al 2002 Conf. Rec. SPIE Med. Imaging 4684 10-19; Hsiao et al 2002 Conf. Rec. IEEE Int. Symp. Biomed. Imaging 409-12) on convergent histogram OSEM reconstruction. We have demonstrated that the first step of the convergent algorithm is exactly equivalent (unlike the histogram-mode case) to the regular subsetized list-mode EM algorithm, while the second and final step takes the form of additive updates in image space. We have shown that in terms of contrast, noise as well as FWHM width behaviour, the CS algorithm is robust and does not result in limit cycles. A hybrid algorithm based on the ordinary and the convergent algorithms is also proposed, and is shown to combine the advantages of the two algorithms (i.e. it is able to reach a higher image quality in fewer iterations while maintaining the convergent behaviour), making the hybrid approach a good alternative to the ordinary subsetized list-mode EM algorithm.

NEMA NU 2-2007 performance measurements of the Siemens Inveon preclinical small animal PET system.

National Electrical Manufacturers Association (NEMA) NU 2-2007 performance measurements were conducted on the Inveon preclinical small animal PET system developed by Siemens Medical Solutions. The scanner uses 1.51 x 1.51 x 10 mm LSO crystals grouped in 20 x 20 blocks; a tapered light guide couples the LSO crystals of a block to a position-sensitive photomultiplier tube. There are 80 rings with 320 crystals per ring and the ring diameter is 161 mm. The transaxial and axial fields of view (FOVs) are 100 and 127 mm, respectively. The scanner can be docked to a CT scanner; the performance characteristics of the CT component are not included herein. Performance measurements of spatial resolution, sensitivity, scatter fraction and count rate performance were obtained for different energy windows and coincidence timing window widths. For brevity, the results described here are for an energy window of 350-650 keV and a coincidence timing window of 3.43 ns. The spatial resolution at the center of the transaxial and axial FOVs was 1.56, 1.62 and 2.12 mm in the tangential, radial and axial directions, respectively, and the system sensitivity was 36.2 cps kBq(-1) for a line source (7.2% for a point source). For mouse- and rat-sized phantoms, the scatter fraction was 5.7% and 14.6%, respectively. The peak noise equivalent count rate with a noisy randoms estimate was 1475 kcps at 130 MBq for the mouse-sized phantom and 583 kcps at 74 MBq for the rat-sized phantom. The performance measurements indicate that the Inveon PET scanner is a high-resolution tomograph with excellent sensitivity that is capable of imaging at a high count rate.

QuickSilver: A Flexible, Extensible, and High-Speed Architecture for Multi-Modality Imaging

A flexible, extensible, high-speed architecture, called Quicksilvertrade and specifically geared to the requirements of small animal imaging, has been developed. The architecture is composed of ring-based event processing modules (EPMs) with nearest neighbor, high-speed digital communication transmitting event packets via a store and forward concept. Each EPM is capable of transmitting up to 15.6M events/sec to other EPMs. Coincidence determination is performed at the EPM level around the ring. This distributes the load and eliminates the need for a separate coincidence processor. Each EPM is capable of transmitting up to 1.9M coincidence events/sec to an event routing subsystem (ERS) for acquisition and processing. The ERS has 2 transport interfaces for acquiring events: an IEEE 1394A interface and a PCI interface. The IEEE 1394A interface can support up to 5.3M events/sec and the PCI interface can support up to 16.7M events/sec. Thus this architecture provides a new level of capability for small animal PET imaging, but is also extremely well suited for PET research, single photon emission computed tomography (SPECT) imaging, and use with X-ray CT and magnetic resonance imaging (MRI).

A Data Acquisition, Event Processing and Coincidence Determination Module for a Distributed Parallel Processing Architecture for PET and SPECT Imaging

The QuickSilver Event Processing Module (EPM), an electronics module used in Siemens Inveon PET and SPECT systems, provides data acquisition, event processing and coincidence determination functions. Custom mixed-signal CMOS ASICs and high speed ADCs are utilized to provide the front-end analog portion of the data acquisition. A high performance FPGA provides the digital portion of the data acquisition running at 100 MHz, the subsequent event processing, and the coincidence determination. The high performance FPGA also provides multiple high speed serial data channels for external interconnection. This interconnection allows the module to be replicated as needed in a distributed parallel processing architecture to provide flexible, high performance, PET and SPECT imaging. The module also provides controllable high voltage needed to bias detectors. A 64 channel, LSO based PET system built using 16 EPMs yielded 1.22 ns FWHM system timing and better than 14% energy resolution.

Space-variant and anisotropic resolution modeling in list-mode EM reconstruction

One issue common to PET scanners is the space- variance of the point spread function (PSF): manifesting itself as resolution degradation as one moves away from the center of the field-of-view (FOV). This effect occurs due to a higher probability of inter-crystal penetration with higher angles of radiation incident on crystal fronts. Depth-of-interaction (DOI) encoding is known to improve this problem, but has not reached complete space-invariance. In this work, a space-variant PSF has been incorporated into the system matrix of a list-mode EM algorithm. Furthermore, in an effort to further extend generality and accuracy of the model, anisotropicity of the PSF has also been considered: finite resolution effects at any position in the FOV are allowed to have distinct values along the axial and the two transaxial directions, and are allowed to degrade differently with increasing distance from the center of the FOV. The spatial distribution of image resolution has been measured and fit using exponential and inverse-Gaussian functions. It is shown that the proposed modeling of the PSF, compared to space-invariant and isotropic modeling, improves resolution recovery across the FOV.

A Neural Network Based Algorithm for Building Crystal Look-up Table of PET Block Detector

Crystal look-up table (CUT) used in the Siemens Inveon dedicated PET scanner defines the matching relation between signal position of a detected event to a corresponding detector pixel location. It is the result of the first stage scanner calibration and brings significant influence to the gantry overall performance. The currently used method involves a lot of human interaction for CLT corrections, and can not be implemented as an on-line process due to its complexity. This paper introduces a neural network based algorithm for crystal identification. A modified unsupervised self-organizing feature map (SOFM) is trained by the incoming events to construct a CLT. The algorithm is implemented in a field programmable gate array (FPGA) chip in the Inveon event processing module (EPM) electronics, which significantly reduces the training time and brings feasibility to detector on-line monitoring. The preliminary training result shows that SOFM can be used effectively in CLT construction with excellent accuracy.

Evaluation of single photon transmission for the HRRT

A dedicated whole human brain positron emission tomograph (PET), the High Resolution Research Tomograph (ECAT HRRT) is utilized to evaluate attenuation correction using single photon based transmission scanning. The patented transmission procedure uses a 740 MBq Cs-137 point source, which is extended into the FOV and collimated to flood the opposing detectors only. An attenuation map is then calculated iteratively using a blank and transmission scan, scaled to 511 keV, and re-projected using inverse Fourier rebinning to estimate the 3D attenuation correction. We have evaluated the accuracy of the single-based transmission procedure and attenuation correction process. In particular, we compare variance weighted OSEM and a dedicated TR algorithm with regularization (MAP-TR) for the reconstruction of the /spl mu/-image. Contamination from emission is estimated from a mock scan without moving the source. Results of a measurement of the patient dose during HRRT transmission scans show a 4 times lower dose compared to patient transmission scans on the ECAT HR.

Continuously sampled digital pulse processing for inveon small animal PET scanner

The quicksilver event processing module (EPM) is a key component of a high performance data acquisition platform from Siemens Molecular Imaging (Knoxville, TN) for use in the Inveontrade line of multimodal PET and SPECT preclinical imaging systems. The cards main purpose is to condition, digitize and process incoming analog pulses from PMT or APD based PET or SPECT detectors. Analog pulses from a detector are digitized using a 100 MHz continuous sampling ADC and read into a Xilinx Virtex II Pro FPGA for processing. The FPGA performs digital integration, baseline offset correction and pileup rejection. Because these functions are done in the digital domain, different algorithms can be quickly re-implemented and tested. The EPM has the ability to capture raw event ADC samples, allowing for the quick development and comparison of new algorithms in software on actual event samples. The Inveontrade small animal PET scanner uses a larger LSO block detector and new analog front end than previous generation scanners which increases the likelihood of pileup events. The digital pulse processing methods presented here have been evaluated to obtain the best energy and positioning performance from the high pixel count Inveontrade detectors while maintaining high stability across countrates.

Digital Time Alignment of High Resolution PET Inveon Block Detectors

Time alignment is critical to the operation of high performance PET systems. Optimizing the countrate performance of a given system requires proper timing to minimize the effects of random events as early in the datastream as possible to reduce electronic deadtime. Siemens Inveon PET systems high countrate performance is made possible by their exceedingly tight timing tolerance of 1.22 ns FWHM, allowing the use of an optimal 3.5 ns coincidence window. This paper describes the hardware capabilities used to align the system as well as the processing techniques used to perform the alignment.