Joseph Meier

MO-DE-207B-07: Assessment of Reproducibility Of FDG-PET-Based Radiomics Features Across Scanners Using Phantom Imaging

PURPOSE Use a NEMA-IEC PET phantom to assess the robustness of FDG-PET-based radiomics features to changes in reconstruction parameters across different scanners. METHODS We scanned a NEMA-IEC PET phantom on 3 different scanners (GE Discovery VCT, GE Discovery 710, and Siemens mCT) using a FDG source-to-background ratio of 10:1. Images were retrospectively reconstructed using different iterations (2-3), subsets (21-24), Gaussian filter widths (2, 4, 6mm), and matrix sizes (128,192,256). The 710 and mCT used time-of-flight and point-spread-functions in reconstruction. The axial-image through the center of the 6 active spheres was used for analysis. A region-of-interest containing all spheres was able to simulate a heterogeneous lesion due to partial volume effects. Maximum voxel deviations from all retrospectively reconstructed images (18 per scanner) was compared to our standard clinical protocol. PET Images from 195 non-small cell lung cancer patients were used to compare feature variation. The ratio of a features standard deviation from the patient cohort versus the phantom images was calculated to assess for feature robustness. RESULTS Across all images, the percentage of voxels differing by <1SUV and <2SUV ranged from 61-92% and 88-99%, respectively. Voxel-voxel similarity decreased when using higher resolution image matrices (192/256 versus 128) and was comparable across scanners. Taking the ratio of patient and phantom feature standard deviation was able to identify features that were not robust to changes in reconstruction parameters (e.g. co-occurrence correlation). Metrics found to be reasonably robust (standard deviation ratios > 3) were observed for routinely used SUV metrics (e.g. SUVmean and SUVmax) as well as some radiomics features (e.g. co-occurrence contrast, co-occurrence energy, standard deviation, and uniformity). Similar standard deviation ratios were observed across scanners. CONCLUSIONS Our method enabled a comparison of feature variability across scanners and was able to identify features that were not robust to changes in reconstruction parameters.

Evaluation of a novel elastic respiratory motion correction algorithm on quantification and image quality in abdomino-thoracic PET/CT

Our aim was to evaluate in phantom and patient studies a recently developed elastic motion deblurring (EMDB) technique that makes use of all the acquired PET data and compare its performance with other conventional techniques such as phase-based gating (PBG) and HD⋅Chest (HDC), both of which use fractions of the acquired data. Comparisons were made with respect to static whole-body (SWB) images with no motion correction. Methods: A phantom simulating respiratory motion of the thorax with lung lesions (5 spheres with internal diameters of 10–28 mm) was scanned with 0, 1, 2, and 3 cm of motion. Four reconstructions were performed: SWB, PBG, HDC, and EMDB. For PBG, the average (PBGave) and maximum bin (PBGmax) were used. To compare the reconstructions, the ratios of SUVmax, SUVpeak, and contrast-to-noise ratio (CNR) were calculated with respect to SWB. Additionally, 46 patients with lung or liver tumors less than 3 cm in diameter were studied. Measurements of SUVmax, SUVpeak, and CNR were made for 46 lung and 19 liver lesions. To evaluate image noise, the SUV SD was measured in healthy lung and liver tissue and in the phantom background. Finally, the subjective image quality of patient examinations was scored on a 5-point scale by 4 radiologists. Results: In the phantom, EMDB increased SUVmax and SUVpeak over SWB but to a lesser extent than the other reconstruction methodologies. The ratio of CNR with respect to SWB for EMDB, however, was higher than all other reconstructions (0.68 with EMDB > 0.54 with HDC > 0.41 with PBGmax > 0.31 with PBGave). Similar results were seen in patient studies. SUVmax and SUVpeak were higher by, respectively, 19.3% and 11.1% with EMDB, 21.6% and 13.9% with HDC, 22.8% and 12.8% with PBGave, and 45.6% and 26.8% with PBGmax, compared with SWB. Lung and liver noise increased with EMDB by, respectively, 3% and 15%, with HDC by 35% and 56%, with PBGave by 100% and 170%, and with PBGmax by 146% and 219%. CNR increased in lung and liver tumors only with EMDB (18% and 13%, respectively) and decreased with HDC (−14% and −23%), PBGave (−39% and −63%), and PBGmax (−18% and −46%). The average radiologist scores of image quality were 4.0 ± 0.8 with SWB, 3.7 ± 1.0 with EMDB, 3.1 ± 1.0 with HDC, and 1.5 ± 0.7 with PBG. Conclusion: The EMDB algorithm had the least increase in image noise, improved lesion CNR, and had the highest overall image quality score.

HU deviation in lung and bone tissues: Characterization and a corrective strategy

INTRODUCTION In the era of precision medicine, quantitative applications of x-ray Computed Tomography (CT) are on the rise. These require accurate measurement of the CT number, also known as the Hounsfield Unit. In this study, we evaluated the effect of patient attenuation-induced beam hardening of the x-ray spectrum on the accuracy of the HU values and a strategy to correct for the resulting deviations in the measured HU values. MATERIALS AND METHODS A CIRS electron density phantom was scanned on a Siemens Biograph mCT Flow CT scanner and a GE Discovery 710 CT scanner using standard techniques that are employed in the clinic to assess the HU deviation caused by beam hardening in different tissue types. In addition, an anthropomorphic ATOM adult male upper torso phantom was scanned on the GE Discovery 710 scanner. Various amounts of Superflab bolus material were wrapped around the phantoms to simulate different patient sizes. The mean HU values that were measured in the phantoms were evaluated as a function of the water-equivalent area (Aw ), a parameter that is described in the report of AAPM Task Group 220. A strategy by which to correct the HU values was developed and tested. The variation in the HU values in the anthropomorphic ATOM phantom under different simulated body sizes, both before and after correction, were compared, with a focus on the lung and bone tissues. RESULTS Significant HU deviations that depended on the simulated patient size were observed. A positive correlation between HU and Aw was observed for tissue types that have an HU of less than zero, while a negative correlation was observed for tissue types with HU values that are greater than zero. The magnitude of the difference increases as the underlying attenuation property deviates further away from that of water. In the electron density phantom study, the maximum observed HU differences between the measured and reference values in the cortical bone and lung materials were 426 and 94 HU, respectively. In the anthropomorphic phantom study, the HU difference was as much as -136.7 ± 8.2 HU (or -7.6% ± 0.5% of the attenuation coefficient, AC) in the spine region, and up to 37.6 ± 1.6 HU (or 17.3% ± 0.8% of AC) in the lung region between scenarios that simulated normal and obese patients. Our HU correction method reduced the HU deviations to 8.5 ± 9.1 HU (or 0.5% ± 0.5%) for bone and to -6.4 ± 1.7 HU (or -3.0% ± 0.8%) for lung. The HU differences in the soft tissue materials before and after the correction were insignificant. Visual improvement of the tissue contrast was also achieved in the data of the simulated obese patient. CONCLUSIONS The effect of a patients size on the HU values of lung and bone tissues can be significant. The accuracy of those HU values was substantially improved by the correction method that was developed for and employed in this study.

A measurement of the attenuation of radiation from F-18 by a PET/MR scanner

Abstract The attenuation of 511 keV photons by the structure of a PET/MR scanner was measured prior to energizing the magnet. The exposure rate from a source of fluorine‐18 was measured in air and, with the source placed at the isocenter of the instrument, at various points outside of the scanner. In an arc from 45 to 135 degrees relative to the long axis of the scanner and at a distance of 1.5 m from the isocenter, the attenuation by the scanner is at least 5.6 half‐value layers from the MR component alone and at least 6.6 half‐value layers with the PET insert installed. This information could inform better design of the radiation shielding for PET/MR scanners.

SU-F-J-224: Impact of 4D PET/CT On PERCIST Classification of Lung and Liver Metastases in NSLC and Colorectal Cancer

PURPOSE To quantify the impact of 4D PET/CT on PERCIST metrics in lung and liver tumors in NSCLC and colorectal cancer patients. METHODS 32 patients presenting lung or liver tumors of 1-3 cm size affected by respiratory motion were scanned on a GE Discovery 690 PET/CT. The bed position with lesion(s) affected by motion was acquired in a 12 minute PET LIST mode and unlisted into 8 bins with respiratory gating. Three different CT maps were used for attenuation correction: a clinical helical CT (CT_clin), an average CT (CT_ave), and an 8-phase 4D CINE CT (CT_cine). All reconstructions were 3D OSEM, 2 iterations, 24 subsets, 6.4 Gaussian filtration, 192×192 matrix, non-TOF, and non-PSF. Reconstructions using CT_clin and CT_ave used only 3 out of the 12 minutes of the data (clinical protocol); all 12 minutes were used for the CT_cine reconstruction. The percent change of SUVbw_peak and SUVbw_max was calculated between PET_CTclin and PET_CTave. The same percent change was also calculated between PET_CTclin and PET_CTcine in each of the 8 bins and in the average of all bins. A 30% difference from PET_CTclin classified lesions as progressive metabolic disease (PMD) using maximum bin value and the average of eight bin values. RESULTS 30 lesions in 25 patients were evaluated. Using the bin with maximum SUVbw_peak and SUVbw_max difference, 4 and 13 lesions were classified as PMD, respectively. Using the average bin values for SUVbw_peak and SUVbw_max, 3 and 6 lesions were classified as PMD, respectively. Using PET_CTave values for SUVbw_peak and SUVbw_max, 4 and 3 lesions were classified as PMD, respectively. CONCLUSION These results suggest that response evaluation in 4D PET/CT is dependent on SUV measurement (SUVpeak vs. SUVmax), number of bins (single or average), and the CT map used for attenuation correction.

TU-H-206-07: Assessment of Geometric Distortion in EPI with a SPAMM Tagged Acquisition

PURPOSE Echo planar imaging (EPI) is known to exhibit gross geometric distortion caused by multiple factors, including B0 inhomgeneity and transient eddy currents. However, diffusion weighted (DW) EPI has become indispensable for diagnosis and therapy assessment. We propose a methodology for quantifying distortion in EPI sequences that does not require the use of dedicated spatial accuracy phantoms, enabling flexibility in phantom design for QA of distortion effects in EPI protocols. METHODS The proposed methodology utilizes a saturation technique known as Spatial Modulation of Magnetization (SPAMM) that tags the imaging subject with saturated grid lines. Originally intended for tracking cardiac motion, these grids are applied to assess differences between diffusion weighting directions and b-values, or against a more geometrically robust sequence such as fast spin echo (FSE). The saturation preparation sequence consists of binomially weighted (e.g. 1-3-3-1) pulses interleaved with gradient blips along the frequency encode direction, followed by the same sequence with gradient blips in the phase encode direction. Three phantoms were assessed with these sequences: a spherical head-sized phantom, a large shimming phantom, and a modified PET ACR phantom that included compartments of water, air, oil, and Teflon. Each phantom was acquired with three sequences using parameters from a clinically appropriate protocol (22 cm head or 46 cm abdomen): a conventional DW-EPI sequence (3 DW directions), and both the DW-EPI and FSE sequences with tagging. Differences in grid locations were visualized with minimum intensity projection between images, and measured using intersecting locations on the grids. RESULTS Grid lines were clearly visualized on tagged images and enabled quantification of distortions. Maximum eddy current induced errors of 10.8 to 14.8 mm were observed in areas away from isocenter with DW gradients applied in various directions. CONCLUSION SPAMM tagging provides a promising mechanism for assessing spatial distribution of distortion in EPI sequences.

Design and development of novel and practical PET detectors for advanced imaging applications

New DOI-measurable PET detectors have been designed, developed and evaluated with advanced silicon photomultiplier (SiPM) and readout technologies. The detector consists of an 8×8 array of 1.5×1.5×20 or 1.5×1.5×30 mm3 LYSO scintillators which is optically coupled to a 4×4 array of 3×3 mm2 SiPM array at each crystal array end through a 2 mm thick optical plate. Scintillator surfaces, reflectors and coupling were designed and fabricated to reserve the air-gap to achieve high depth-of-interaction (DOI) resolution and other detection performance. The insensitive edges around each detector is less than 0.2 mm, making it practical to seamlessly tile detectors together to assemble a large size detector panel for developing a PET system. A 16-ch ASIC based PCB readout electronics was developed to solve the challenging SiPM array readout problem. Each compact PCB contains 4 ASIC chips and one detector-level FPGA, with analog signal being inputted from each SiPM array through a flexible printed circuit cable, converted to digital timing pulses, processed online by FPGA to record interaction information (energy, timing, and position), and transferred through a fast LVDS connection to system FPGA for event selection and data acquisition. Initial measurements showed excellent crystal identification with all crystals were clearly separated from each other in a flood source image, with resolutions of energy, timing and DOI were around 17%, 2.7 ns and 2.0 mm (mean value), respectively. Overall, comparing to the previous prototype DOI detectors we developed, the new detector is simplified in design without using complicated light guide yet with significantly improved DOI resolution (from ~5 to ~2 mm), more compact packaging for making a large size flat-panel detector, and more integrated and fast readout electronics. The new detector and readout is expected leading to an advanced PET with leapfrog imaging performance improvement.

SU‐E‐T‐69: Energy Response Characterization and Calibration of Electronic Personal Dosimeters

PURPOSE To characterize the personal dose equivalent energy response, Hp(10), for electronic personal dosimeters (EPDs) with commonly used radionuclides in nuclear medicine. METHODS Rados-60R with an energy compensated PIN diode and the SAIC Pd-10i with a miniature energy compensated Geiger-Muller tube EPDs were characterized. The experimental setup and calculation of EPD energy responses were based on ANSI/HPS N13.11-2009. 15 Rados-60R and 2 SAIC Pd-10i units were irradiated using 99mTc, 131I, and 18F radionuclides corresponding to emission energies at 140 keV, 364 keV, and 511 keV, respectively. EPDs output in Hp(10) [mrem] were recorded for free-in-air and with 15-in. thick PMMA to simulate backscatter form the torso. Simultaneous exposure rate measurements were also performed using 2 Victoreen 451-B ionization survey meters to serve as gold standard measurements. The expected EPD Hp(10) values were calculated from exposure (from the Victoreen 451-B survey meters) to Hp(10) as specified in ANSI/HPS N13.11-2009 and ICRU Report 57. These measurements were repeated for both setups at all energies. RESULTS On average, in the presence of acrylic, the reported Rados-60R values increased by 27%, 12%, and 13% and those reported by the SAIC Pd-10i increased by 23%, 19%, and 12% at 140 keV, 364 keV, and 511 keV, respectively. The Rados-60R EPDs were observed to under-respond at 140 keV by ∼16%, and agreed to within 5% and 10% of the expected values at 364 and 511 keV, respectively. The SAIC Pd-10i EPDs were observed to over-respond at 140 keV by ∼20%, and agreed to within 5% of the expected values 364 and 511 keV. CONCLUSION Both Rados-60R and SAICPD-10i EPDs displayed Hp(10) values that were accurate to within 10% at energies above 364 keV. However, their accuracies degraded to 15-20% at lower energies (140 keV) suggesting the need to calculate energy dependent correction factors.