skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Image blurring due to light-sharing in PET block detectors

Abstract

The spatial resolution in PET is poorer than that of CT or MRI. All modern PET scanners use block detectors, i.e., clusters of scintillation crystals coupled to four photomultiplier tubes. Some of the loss of spatial resolution in PET is attributed to the use of block detectors, because a photon that interacts with one crystal in the cluster may be incorrectly positioned, resulting in blurring of the reconstructed image. This is called the ''block effect.'' The block effect was measured for detectors from the CTI HR+ scanner, and the GE Advance scanner; two popular clinical PET scanners. The effect of changing the depth of first interaction of a gamma ray in the scintillation crystals was also studied to determine if it may be a contributor to the block effect. The block effect was found to be 1.2{+-}0.5 mm for the central crystals and negligible for the edge crystals in the CTI HR+ block. It was 0.9{+-}0.3 mm in the central crystals of the GE Advance detector, and 0.7{+-}0.3 mm in the edge crystals of the GE Advance detector. In the CTI HR+ detector, a depth dependence on the positioning of the event was observed, as was a dependence on themore » crystal location (edge versus center). In the GE Advance detector events that occurred at different interaction depths were positioned consistently. The percentage of events that may be positioned inaccurately was also calculated for both detectors. In the CTI HR+ detector as many as 16% of all events in the block detector may be positioned incorrectly. In the GE Advance detector as many as 13% of all events in the block detector may be positioned inaccurately. These results suggest that the depth of interaction of an annihilation photon may contribute to the block effect in detectors that use crystals cut from a single scintillation crystal (pseudodiscrete crystals), and is less dominant a factor for detectors that use discrete crystals with light sharing between the crystals. Investigating the effect of changing photon interaction depth in PET detectors can lead to better detector design, and an intuitive explanation of what sources of blurring may exist in the detector examined.« less

Authors:
;  [1]
  1. Montreal Neurological Institute, McGill University, Montreal, Quebec H3A 2B4 (Canada)
Publication Date:
OSTI Identifier:
20775066
Resource Type:
Journal Article
Resource Relation:
Journal Name: Medical Physics; Journal Volume: 33; Journal Issue: 2; Other Information: DOI: 10.1118/1.2161406; (c) 2006 American Association of Physicists in Medicine; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
62 RADIOLOGY AND NUCLEAR MEDICINE; ANNIHILATION; GAMMA DETECTION; GAMMA RADIATION; IMAGE SCANNERS; IMAGES; NMR IMAGING; PHOTOMULTIPLIERS; POSITRON COMPUTED TOMOGRAPHY; SCINTILLATIONS; SOLID SCINTILLATION DETECTORS; SPATIAL RESOLUTION

Citation Formats

St James, Sara, and Thompson, Christopher J. Image blurring due to light-sharing in PET block detectors. United States: N. p., 2006. Web. doi:10.1118/1.2161406.
St James, Sara, & Thompson, Christopher J. Image blurring due to light-sharing in PET block detectors. United States. doi:10.1118/1.2161406.
St James, Sara, and Thompson, Christopher J. Wed . "Image blurring due to light-sharing in PET block detectors". United States. doi:10.1118/1.2161406.
@article{osti_20775066,
title = {Image blurring due to light-sharing in PET block detectors},
author = {St James, Sara and Thompson, Christopher J.},
abstractNote = {The spatial resolution in PET is poorer than that of CT or MRI. All modern PET scanners use block detectors, i.e., clusters of scintillation crystals coupled to four photomultiplier tubes. Some of the loss of spatial resolution in PET is attributed to the use of block detectors, because a photon that interacts with one crystal in the cluster may be incorrectly positioned, resulting in blurring of the reconstructed image. This is called the ''block effect.'' The block effect was measured for detectors from the CTI HR+ scanner, and the GE Advance scanner; two popular clinical PET scanners. The effect of changing the depth of first interaction of a gamma ray in the scintillation crystals was also studied to determine if it may be a contributor to the block effect. The block effect was found to be 1.2{+-}0.5 mm for the central crystals and negligible for the edge crystals in the CTI HR+ block. It was 0.9{+-}0.3 mm in the central crystals of the GE Advance detector, and 0.7{+-}0.3 mm in the edge crystals of the GE Advance detector. In the CTI HR+ detector, a depth dependence on the positioning of the event was observed, as was a dependence on the crystal location (edge versus center). In the GE Advance detector events that occurred at different interaction depths were positioned consistently. The percentage of events that may be positioned inaccurately was also calculated for both detectors. In the CTI HR+ detector as many as 16% of all events in the block detector may be positioned incorrectly. In the GE Advance detector as many as 13% of all events in the block detector may be positioned inaccurately. These results suggest that the depth of interaction of an annihilation photon may contribute to the block effect in detectors that use crystals cut from a single scintillation crystal (pseudodiscrete crystals), and is less dominant a factor for detectors that use discrete crystals with light sharing between the crystals. Investigating the effect of changing photon interaction depth in PET detectors can lead to better detector design, and an intuitive explanation of what sources of blurring may exist in the detector examined.},
doi = {10.1118/1.2161406},
journal = {Medical Physics},
number = 2,
volume = 33,
place = {United States},
year = {Wed Feb 15 00:00:00 EST 2006},
month = {Wed Feb 15 00:00:00 EST 2006}
}
  • Two state-of-the-art modular PET block detectors using either discrete or pseudo-discrete BGO crystals coupled to two dual PMTs, utilizing different light sharing schemes, were evaluated. Both detectors were approximately 30 mm thick, which each element of the GE 6 x 6 block detector was 4 mm x 8.4 mm, and each element of the CTI 8 x 7 block detector was 2.8 mm x 5.8 mm. In addition to measurements with gamma sources, the detectors were also irradiated with a Sr/Y-90 beta source to evaluate the performance of the block detectors without inter-element scatter of annihilation photons. The best andmore » worst energy resolutions of individual elements at 511 keV were 21.8% and 44.2% for the GE detector, and 22.7% and 42.5% for the CTI detector. The peak to valley ratios in the detector identification ratio histograms were generally better than 3 to 1. Measurements with beta sources indicate that light sharing in both detectors is a large component of even mispositioning along with inter-detector scatter of the annihilation photons.« less
  • In multiwire proportional chambers used with honeycomb lead converters for detecting 511 KeV ..gamma.. rays from positron annihilation, a source of image blurring is generated by multiple interaction events due to the escape photoelectric x-ray or from the Compton scattered photon. Using the delay line readout method the majority of these double events are eliminated by using the fact that the sum of the time intervals from the prompt anode signal to the signal arrival at each end of the delay line is a constant to within the timing accuracy for a single interaction. Double interaction events produce a timemore » sum which is shorter. Good improvement in image quality is obtained. The observed number of multiple events is larger than calculations would predict.« less
  • In this work, the authors studied the artifacts generated by the application of methods of crystal efficiency calculation based on the fan-sum, to situations of high variability of efficiency, namely the cases of a ring with defective blocks and acquisitions using narrow energy windows. They propose two efficiency calculation methods to reduce these artifacts: one method discriminates the crystals with an efficiency very different from the average, and then uses a modification of the fan-sum to estimate the efficiency; the other method calculates iteratively the fan-sum of each crystal, weighted by the efficiencies of the opposed detectors. The effect ofmore » using an off-center uniform cylinder is also discussed, since it influences the performance of the algorithms. The proposed algorithms are compared with the more conventional fan-sum method and with an algorithm used in a commercial tomography, using measurements and simulations. The results show a significant reduction of the efficiency estimation errors, in severe situations where detectors in a ring have very different detection efficiency.« less
  • In a conventional PET system with block detectors, a timing estimator is created by generating the analog sum of the signals from the four photomultiplier tubes (PMT) in a module and discriminating the sum with a single constant fraction discriminator (CFD). The differences in the propagation time between the PMTs in the module can potentially degrade the timing resolution of the module. While this degradation is probably too small to affect performance in conventional PET imaging, it may impact the timing inaccuracy for time-of-flight PET systems (which have higher timing resolution requirements). Using a separate CFD for each PMT wouldmore » allow for propagation time differences to be removed through calibration and correction in software. In this paper we investigate and quantify the timing resolution achievable when the signal from each of the 4 PMTs is digitized by a separate CFD. Several methods are explored for both obtaining values for the propagation time differences between the PMTs and combining the four arrival times to form a single timing estimator. We find that the propagation time correction factors are best derived through an exhaustive search, and that the ''weighted average'' method provides the best timing estimator. Using these methods, the timing resolution achieved with 4 CFDs (1052 {+-} 82 ps) is equivalent to the timing resolution with the conventional single CFD setup (1067 {+-} 158 ps).« less
  • Purpose: To assess uncertainties in voxel-wise kinetic modeling (KM) of dynamic 18F-FMISO PET (dPET) due to image noise for different lengths of dPET acquisitions. Methods: 12 tumor time activity curves (TACs) deduced from 6 head and neck cancer patient dPET datasets (45min dPET + 10min frames at ∼95min and ∼160min respectively) were modeled using an irreversible two-tissue compartment model to estimate kinetic rate constants (KRCs) k3, vB, K1, and K1/k2 (reference standard). For each modeled TAC, 1000 noisy TACs were simulated by adding Gaussian noise with standard deviation equivalent to that observed on a voxel level in GE DSTE PET-CT.more » KM was conducted for each set of noisy TACs using (i) full dataset (170min), and repeated using (ii) 105min and (iii) 45min shortened subsets, with an input function that was image-derived from the dPET dataset of matching length. Absolute value of percent difference between KRCs from noisy TACs as estimated from either of 3 datasets and reference standard KRCs was used to represent bias. Results: For all KRCs, bias was higher for shortened datasets. For k3, bias was lower when actual k3 was higher: from 7%±5% (170min), 9%±7% (105min), and 14%±11% (45min) (actual k3=0.0098), to 32%±25% (170min), 45%±34% (95min), and 79%±48% (45min) (actual k3=0.0008). Similar relationship was observed for vB, with bias being between ∼7% (actual vB=0.28) and ∼26–34% (actual vB=0.08). For K1 [K1/k2], bias was ∼6–10% [∼2–3%] (dependence on actual value of K1 [K1/k2] was not observed). Conclusion: Uncertainties in voxel-wise KM of shortened 18F-FMISO dPET due to image noisy only are larger for shorter acquisitions and, for k3 [vB], the bias was found to be inversely correlated with the actual values of k3 [vB]. Using 45min subset, k3, vB, K1, and K1/k2 could be estimated to within ∼15–80%, ∼10–35%, ∼10%, and ∼3%, respectively.« less