Benefit of Total-body PET Imaging: Effect of Time-of-Flight and Depth-of-Interaction on Quantification Performance (#3659)
X. Zhang1, K. Gong1, R. D. Badawi1, 2, S. R. Cherry1, 2, J. Qi1
1 University of California, Davis, Department of Biomedical Engineering, Davis, California, United States of America
There has been a steady trend of increasing the axial field-of-view of (AFOV) PET scanners. The aim of this study is to evaluate the benefit of the extended axial length of PET for quantification performance by incorporating the capability of time-of-flight (TOF) and depth-of-interaction (DOI) information. A theoretical framework was used to examine the contrast recovery coefficient (CRC) and variance tradeoff for region of interest (ROI) quantification based on the penalized maximum likelihood image reconstruction. A series of PET scanners with a ring diameter of 83.5 cm and an axial field-of-view (FOV) ranging from 22 cm to 220 cm were simulated. Multiple overlapping bed positions were used for scanners with a short AFOV to provide a relatively uniform sensitivity along a 110 cm axial range. A uniform cylinder of 20 cm diameter and 230 cm long was employed to model the attenuation and background activity. The CRC versus variance tradeoff curves were compared for quantifying activities in different ROIs. Different TOF resolutions (320 ps, 500ps, and non-TOF) and DOI resolutions (4 mm, 10 mm, and no DOI) are studied. The results show that even for a radially centered ROI, DOI is beneficial for scanners with long AFOV due to axial blurring at large oblique angles. Combining 320-ps TOF and 4-mm DOI, the 220-cm long scanner offers a 70-fold variance reduction over the 22-cm long scanner for imaging a 2-mm ROI and 90-fold variance reduction for imaging a 10-mm ROI. These theoretical results demonstrate that TOF and DOI can further improve the benefit of a long AFOV scanner in image quality compared to the current clinical PET scanners.
This work was funded by the National Institutes of Health under grant number R01-CA170874, R01-CA206187 (co-funded by the NCI, NIBIB and the Office of the Director) and a UC Davis Research Investment in Science and Engineering Program (RISE) award.
Keywords: EXPLORER, TOF, DOI, ROI Quantificaiton, Variance Reduction, Total-body PET
Demonstration of PET System Design Trade-offs Using Small Lesion Detectability as a Metric and Measured Phantom Data (#3719)
S. D. Wollenweber1, P. E. Kinahan2, A. M. Alessio2
1 GE Healthcare, MICT Engineering, Waukesha, Wisconsin, United States of America
PET system design involves trade-offs between cost and performance. It is a combination of capabilities that define the clinical utility of any given system design. System sensitivity and time-of-flight (TOF) timing resolution are two such parameters that influence system performance and cost. It is important to quantify the trade-offs using clinically-relevant measures prior to selecting a design. A lesion detection phantom has previously been demonstrated and is used in this work to determine the required change in timing resolution or system sensitivity that leads to a doubling of model observer SNR. Results: change in NEMA sensitivity produced more change in detectability than improvement in timing resolution from 550 to 385 ps.
Keywords: PET, image quality, lesion detectability, phantoms, system design
Impact of patient size on image quality in clinical PET with a convergent penalized likelihood image reconstruction algorithm (#2577)
S. Ahn1, K. A. Wangerin2, S. D. Wollenweber2, S. G. Ross2, C. W. Stearns2, P. E. Kinahan3
1 GE Global Research, Niskayuna, New York, United States of America
A penalized-likelihood (PL) image reconstruction algorithm using the relative difference penalty (RDP), called a PL-RDP algorithm, has recently been developed for clinical PET and has previously demonstrated improved lesion quantitation accuracy and lesion detectability over the current clinical standard PET image reconstruction algorithm, ordered subsets expectation maximization (OSEM). Patient size is one of the main factors affecting PET image quality but the effects of patient size on PL-RDP have not been studied yet. The purpose of this study is to evaluate the impact of patient size on image quality for PL-RDP compared to OSEM. For evaluation, we used clinical sinogram data with adding simulated liver lesion data. To evaluate lesion detectability, we used a channelized Hotelling observer and calculated the signal-to-noise ratio. To evaluate lesion quantitation accuracy, we compared recovery coefficients at matched image roughness (surrogate image noise measure). The lesion detectability and the quantitation accuracy in a small patient group were statistically significantly better than those in a large patient group. The detection and the quantitation performance of PL-RDP were statistically significantly better, or at least not worse, than those of OSEM in both patient groups. Interestingly, the improvements of PL-RDP in detection and quantitation over OSEM for the large patient group were statistically significantly larger, or at least not worse, than those for the small patient group.
Keywords: PET, image reconstruction, detection, quantitation, OSEM, penalized-likelihood, CHO
Task-based Optimization of In-Vivo Micro-CT Scan Protocols using Energy Integrating and Photon Counting Detectors (#1771)
C. Funck1, D. Prox1, J. Maier1, M. Kachelrieß1, J. Kuntz1, S. Sawall1
1 German Cancer Research Center, X-Ray Imaging and CT, Heidelberg, Germany
Small animal micro-CT is an important tool in preclinical research. Radiation dose, however, remains a concern particularly in longitudinal studies were exposure to the animal needs to be kept low to minimize metabolic interactions. Optimizing imaging protocols, i.e. maximizing image quality, on the basis of a limited applicable dose is necessary. We therefore seek to find the optimal number of projection angles over a full 360° rotation given a fixed dose as constraint. The trade–off to be found is between elevated image noise for projections approaching the quantum limit and streak artifacts due to sparse angular sampling. Scan protocols are simulated by forward projections of a high dose motion compensated in-vivo mouse reconstruction using different numbers of view angles and addition of an appropriate amount of noise to preserve a constant overall dose. Energy integrating (EI) and photon counting (PC) detectors are considered and compared. Evaluation is based on the performance of imaging tasks: The left ventricular volume is determined using the segmentation algorithm by Otsu and compared to a ground truth based on the high dose volume. A model observer with nonprewhitening matched filter template is used for the detection of a small lesion and its performance is evaluated to confirm the findings. Results for EI detectors indicate an optimal number of 180 view angles over 360° with increasing deviation from the ground truth for sparser or finer angular sampling. The absence of electronic noise in the case of a PC detector leads to increasing image quality for an increasing number of view angles as streak artifacts successively vanish. Confirming this dependence for PC detectors and identifying an optimum in the EI case allows for designing and improving future in-vivo imaging protocols. Phase correlated reconstruction will profit from the highest possible angular sampling if a PC detector is applied, removing streak artifacts caused by phase binning.
Keywords: Micro-CT, In-Vivo Imaging, Photon Counting Detectors, Task-based Image Quality
The Non-prewhitening and Hotelling Observers for Parameter Selection for Linear Iterative Reconstruction in Breast Tomosynthesis (#3417)
S. D. Rose1, A. A. Sanchez1, I. Reiser1, E. Y. Sidky1, X. Pan1
1 University of Chicago, Radiology, Chicago, Illinois, United States of America
Iterative image reconstruction for digital breast tomosynthesis (DBT) requires specification of a large number of parameters such as voxel size, aspect ratio, and regularization strength, which each alter the solution to the posed reconstruction optimization problem. The associated parameter spaces require investigation for each reconstruction optimization problem, task, and system design under consideration. Efficiently computable simulation-based image quality metrics are needed to facilitate this task. The purpose of this work is the development and comparsion of two task-based image quality metrics for assessing the effect of regularization strength on microcalcification detectability in DBT reconstruction. A region-of-interst (ROI) Hotelling observer and an ROI non-prewhitening observer are applied to a signal-known-exactly/background-known-exactly calcification detection task in simulation as regularization strength is varied for two linear iterative reconstruction algorithms. Trends in the ROI-HO and ROI-NPW metrics are compared with 3D reconstructions from ACR mammography accreditation phantom data acquired with a Hologic Selenia Dimensions DBT system.
Keywords: Model Observers, Breast Tomosynthesis, Image Reconstruction
A study on the impact of statistical weights on lesion detection performance in iterative CT reconstruction (#4192)
V. Haase2, A. Griffith1, Z. Guo1, K. Hahn2, H. Schoendube2, K. Stierstorfer2, F. Noo1
1 University of Utah, Radiology and Imaging Sciences, Salt Lake City, Utah, United States of America
Iterative CT reconstruction with the penalized least-square model may offer significant gains in terms of image quality at equal dose, and may thereby allow either dose reduction or improved diagnostic. In this work, we are interested in evaluating image quality improvements that result from using statistical weights in this model. Image quality is assessed in terms of lesion detection with unknown location, using the principles of LROC analysis with human observers. Reconstruction without and with statistical weights are compared for two penalties: a quadratic penalty, and an edge-preserving penalty. Interestingly, our study failed showing any major improvements due to the use of weights. Furthermore, it was even observed that performance with weights could be even worse, possibly due to the utilization of weights leading to disturbing discretization errors. Because there are a lot of degrees of freedom in our experimental set-up, it should not be concluded that statistical weights are not useful. However, we can state that improvements are not straightforward and may depend on many aspects including the task and also anatomical location and variability. This observation is valuable from a computational viewpoint since using statistical weights generally leads to long reconstruction times; if weights can be ignored or simplified in some settings, reconstruction times can be largely improved for these settings.
Keywords: LROC, image quality, statistical weights, CT, iterative reconstruction, MBIR