A new method for inter-crystal event classification for PET detectors with light sharing design (#3046)
M. S. Lee1, S. K. Kang2, J. S. Lee2
1 Seoul National University, Interdisciplinary Program in Radiation Applied Life Science, Seoul, Republic of Korea
We propose a new method to classify ICS events that can be used for PET detectors with light sharing design. Consider that when an ICS event occurred at crystal i and j with energies of Ei and Ej (Etotal=Ei+Ej). Then observation y of an ICS event can be expressed as y=yi×Ei/Etotal + yj×Ej/Etotal. By converting into a matrix formation: y=Ax; where y is [m×1] observation, A is [m×n] characteristic m observations for n crystals, and x is [n×1] energy ratios for n crystals. Consequently, by solving a linear problem, finding x, we can extract ICS event positions and energies. Three methods were suggested to find x: 1) maximum peak detection, 2) matrix pseudoinverse calculation, x=A-1y, and 3) convex constrained optimization, min||y-Ax||2, subject to x>0, sum(x)=1.
A simulation study was initially conducted to see ICS classification and energy estimation performance. For experimental validation, digital SiPMs were coupled with of 8×8 (1-to-1 coupling) and 10×10 (light sharing) arrays of 3×3×20 mm3 LGSO crystals. Intrinsic spatial resolutions were measured for 1-to-1 and light sharing detector pairs. Three ICS classification methods were implemented, and intrinsic spatial resolutions in FWHMs and FWTMs were compared with and without ICS recovery. For ICS recovery, the proportional method was used (Lage et al., Med Phys, 2015).
The simulation study showed that the proposed Method 3 yielded robust energy estimation and high classification accuracies of 0.918 and 0.852 for 1-to-1 and light sharing detectors, respectively. The experimental study showed the resolution improvement after recovering ICS event that was classified with the proposed method. Average intrinsic resolutions in FWHM and FWTM for 1-to-1 detector were 1.92 and 3.49 mm without ICS recovery and improved to 1.74 and 3.17 mm after ICS recovery. Average intrinsic resolutions in FWHM and FWTM for light sharing detector were 2.58 and 4.69 mm without ICS recovery and improved to 1.95 and 3.56 mm after ICS recovery.
Keywords: Inter-crystal scattering, PET detector, Light sharing detector, Convex optimization
Improved discrimination between benign and malignant screening-detected lung nodules with dynamic FDG PET (#3075)
Q. Ye1, 2, J. Wu1, Y. Lu1, J. - D. Gallezot1, T. Ma2, 3, Y. Liu2, 3, L. Tanoue4, F. Detterbeck5, J. Blasberg5, M. - K. Chen1, M. Casey6, R. E. Carson1, C. Liu1
1 Yale University, Department of Radiology and Biomedical Imaging, New Haven, Connecticut, United States of America
Introduction: Lung cancer mortality rate can be significantly reduced by up to 20% through the routine low-dose computed tomography (LDCT) screening, which, however, has high sensitivity but low specificity, resulting in a high rate of indeterminate and false-positive nodules. Combining PET with CT is expected to provide the most accurate diagnosis for indeterminate nodules detected by initial screening LDCT. In this work, we investigated the classification capability of static and dynamic PET using a virtual clinical trial simulation based on patient study and ground truth.
Methods: Patients with initial LDCT screening-detected lung nodules received 90-min single-bed dynamic FDG PET scans. Static and ROI-based dynamic analysis were used to estimate standard uptake value (SUV) and net uptake parameter Ki in patient studies, based on which, a virtual clinical trial simulation was performed by adding nodule population variation, measurement noise, and SUV start time variation to the patient data. 600 virtual patients each with 100 repeated PET scans were simulated in total. We used non-prewhitening matched filter signal-to-noise ratio (NPW SNR) to estimate the classification capability of SUV and Ki from their simulated benign and malignant distributions. The scan durations and start time (t*) in dynamic Patlak plots were optimized in terms of NPW SNR.
Results: NPW SNRs were higher with increased scan durations and decreased t* in Patlak plots, and were further improved using population-based input functions (PBIF). NPW SNRs of Ki using various combinations of t* and scan durations were mostly higher than that of SUV, representing superior classification capability.
Conclusion: It is feasible to perform dynamic PET studies with short scan time and PBIF. Ki is superior to SUV in discrimination between benign and malignant LDCT screening-detected nodules.
Keywords: FDG PET, lung cancer screening, kinetic modeling, low-dose
Experimental Design Study for MRC-SPECT-II System: A Hyperspectral Simultaneous SPECT-MR system based on Inverted Compound Gamma Camera (#1565)
E. M. Zannoni1, X. Lai2, M. D. Wilson3, L. - J. Meng2, 4
1 Universiy of Illinois at Urbana-Champaign, Bioengineering, Urbana, Illinois, United States of America
Simultaneous nuclear (SPECT/PET) and MR imaging offers a tremendously powerful tool for in vivo molecular imaging studies thanks to the synergic functional and anatomical information acquired.
In this presentation we present: (a) the detailed design of the second generation MRI compatible SPECT system, the MRC-SPECT-II, a hyperspectral compound-eye imaging system bioinspired by invertebrate vision, (b) the performance benefit from the ultrahigh energy resolution provided by the HEXITEC detectors for multi-isotope SPECT imaging, and (c) a preliminary evaluation of the MRC-SPECT system design using a single module prototype system.
As the natural compound eye is a dense structure of thousands of separate photoreceptors, the Synthetic Inverted Compound Eye (S-ICE) Gamma Camera is composed of multiple closely packed micro-pinhole-camera-elements (MPCEs). Each MCPE has a pinhole with a narrow open-angle that projects a small fraction of the object volume onto a designated area on the high-resolution gamma-ray detector. The S-ICE Gamma Camera geometry guarantees an unprecedented density of angular sampling and a dramatically improved sensitivity in the field of view (FOV), in combination with a compact system design that can be placed inside small-bore pre-clinical MRI scanners. Finally, the S-ICE camera design offers a great imaging flexibility, by incorporating different types of MCPEs with different pinhole sizes, angular coverages and packing densities, for achieving the optimal tradeoff among spatial resolution, sensitivity and FOV.
The Monte Carlo studies show that the proposed MRC-SPECT-II design can achieve peak sensitivity of 0.4% in comparison with the typical 0.1%-0.01% found in modern pre-clinical SPECT systems while maintaining a spatial resolution <400 μm. In addition, the adopted gamma-ray detector shows superior energy performance (<0.8 keV at 35.5 keV for I-125 in a standard pixel spectrum) that makes it attractive for enhanced multi-tracer SPECT imaging.
Keywords: gamma-ray detector, hyperspectral, MRI-SPECT
An Optimized Feature Detector for Markerless Motion Tracking in Motion-Compensated Neuroimaging (#2582)
D. Henry1, Y. Yao3, R. Fulton1, A. Kyme2
1 The University of Sydney, Faculty of Health Sciences, Brain and Mind Centre, Sydney, NSW, Australia
Head movements during PET and MRI scans can have a detrimental effect on image quality and quantitative measurements. For both of these modalities, motion correction methods exist that rely on accurate characterization of head motion. In the case of prospective correction in MRI, the motion estimates also need to be delivered in real-time. Motion tracking methods that rely on attached markers are susceptible to decoupling of the head and marker, as well as being a hindrance to clinical workflow. In this study, we aim to optimize a methodology that measures head motion by detecting and tracking features native to the forehead. These features can be extracted and described in a number of ways, with different algorithms offering varying levels of computational efficiency and robustness to scene changes. A phantom study was performed to assess the accuracy and speed performance of five different feature detectors: SIFT, SURF, ORB, BRISK and AKAZE. With the exception of ORB, motion estimates obtained using the different feature detectors showed similar agreement (error <0.2 mm) with the ground-truth robot measurements. Processing time varied, with SURF, BRISK and AKAZE offering a substantial speed increase over SIFT while maintaining similar accuracy. We conclude that any of the latter feature detectors are an excellent choice for prospective motion correction in MRI and MRI-PET.
Keywords: feature detector, prospective motion correction, markerless motion tracking, neuroimaging
The Importance of Accurate X-ray Energy Spectra for Modelling Dose Deposition with Monte Carlo Techniques (#1789)
L. Forth1, R. Speller1, R. Moss1
1 UCL, Medical Physics and Biomedical Engineering, London, United Kingdom of Great Britain and Northern Ireland
Knowledge of the incident X-ray spectrum is essential to accurately modelling dose to bodies and patients in an X-ray beam. The X-ray spectrum of a SAXG 1701 tungsten target X-ray generator operating at 120 kVp was estimated through use of SpekCalc, GEANT4 simulation and direct spectral measurements with a CdTe spectrometer. Using these independently determined spectra the dose deposited in sets of thermoluminescent dosimeters (TLDs) were modelled through use of GEANT4 Monte Carlo code, under varying conditions and experimentally measured for comparison. The simulations were run using all estimated spectra to determine which spectrum estimation method was most accurate. Exposures using cumulative layers of filtration show the degrading accuracy of each spectrum as filtration is increased. The spectrum taken from measurement with the spectrometer proves to be the most reliable for modelling dose in reality. To investigate how dose prediction is affected from a clinical perspective, models were run to simulate the dose to skin, bone and lung tissue (chest) and the dose to each material was calculated. The results showed, depending on the source spectrum selected, that dose might be overestimated at the surface by up to 20% and underestimating in deep tissue by up to 12%, with inaccuray increasing with penetrative depth. This highlights the potential for inaccurate dose prediction for patients undergoing treatment involving X-rays.
Keywords: Dose Determination, GEANT4 Simulation
Performance of a SiPM based semi-monolithic scintillator PET detector (#1423)
X. Zhang1, 2, X. Wang1, N. Ren1, Z. Kuang1, X. Deng1, X. Fu1, Y. Yang1
1 Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Institute of Biomedical and Health Engineering, Shenzhen, China
A semi-monolithic scintillator crystal PET detector was proposed and a SiPM based prototype detector was measured to demonstrate its feasibility. The semi-monolithic scintillator crystal consists of 11 polished LYSO slices measuring 1×11.6×10 mm3. The slices were glued together with ESR reflector in between and outside of the slices. The bottom surface of the crystals is coupled to a SiPM array with a 1 mm light guide and silicon grease between them. No reflector was used on the top surface and two sides of the crystal. The signals of a 4×4 SiPM array are grouped along rows and columns separately into 8 signals. Four column signals are used to identify the slices. Four row signals are used to estimate y and z (DOI) positions. The detector was measured with 1 mm sampling interval in both the y and z directions with electric collimation. The results show that y positions calculated with the center of gravity method are different for interactions happening at the same y, but different z positions due to depth dependent edge effects. A better positioning accuracy can be obtained as compared to the monolithic detector, since the calibration data can be obtained for any specific interaction points. A mean absolute error (MAE) which is defined as the probability-weighted mean of the absolute value of the positioning error is used to evaluate the spatial resolution. The least-squares minimization and the maximum likelihood positioning algorithms were developed and both methods improved the spatial resolution at the edges of the detector as compared with the center of gravity method. An average MAE spatial resolution of ~ 1.2 mm was obtained in both y and z directions. All slices of the detector can be clearly resolved and the average energy resolution is 12%. In the next step, longer detector will be built to reduce the edge effects of the semi-monolithic detector, thick detector up to 20 mm will be explored and the positioning algorithms will be further optimized.
Keywords: PET, semi-monolithic scintillator, calibration, Positioning algorirhms