M3D2  Data Corrections and Quantitative Imaging Techniques I

Wednesday, Nov. 4  16:30-18:30  Pacific Salon 1&2

Session Chair:  Irene Buvat, IMNC France, ; Roger Fulton, University of Sydney, Australia

Show/Hide All Abstracts

(16:30) M3D2-1, Attenuation Map Estimation Using Scatter Window Projection for SPECT

Y. Du, E. C. Frey

Department of Radiology, Johns Hopkins University, Baltimore, MD, United States
In SPECT imaging, it is important to have an accurate attenuation map to improve image quality and quantitative accuracy. However, acquiring an attenuation map typically requires a transmission scan. Several methods have been proposed to estimate an attenuation map directly from emission data. However, many of these methods used unscattered photons only and the problem is thus ill-posed, resulting in an attenuation map that is affected by crosstalk from activity images. We have developed a method to estimate attenuation map using data recorded in a scatter energy window below the main photopeak energy window based on the effective source scatter estimation (ESSE) method. In the method, the scatter window data are reconstructed to generate an image of the effective scatter source (ESS). The photopeak window data were also reconstructed to estimate the activity distribution. The ratio between the ESS and a convolution of the estimated activity distribution with a scatter kernel produces an estimate of the normalized electron density map of the object. This can then be transformed into an attenuation map by scaling with the linear attenuation coefficient of water at the photopeak energy. The method was validated using simulated SPECT data of a brain phantom by comparing the estimated attenuation map with truth. Results showed there was a very good agreement between the estimates and the truth. The estimated attenuation coefficient for cerebral region was 0.1456/cm. Compared to the true value of 0.1475/cm, a difference of 1.3%. The skull and bones were also clearly visible in the estimated attenuation map. There was mismatch in regions outside the brain. This is mainly due to the lack of activity distribution in those regions. In conclusion, we have developed a method that can estimate the attenuation map directly from image data. The method appears accurate and can be useful in cases where transmission scan is not available.

(16:45) M3D2-2, Investigation of Foam as a Lung Equivalent Material for Quantitative SPECT/CT Phantom Studies

R. S. Miyaoka, W. A. McDougald, L. Zhang, R. L. Harrison, H. J. Vesselle, T. K. Lewellen

Radiology/Nuclear Medicine, University of Washington, Seattle, WA, USA
Objectives: Styrofoam beads with added water have traditionally been used to simulate lung in SPECT phantoms. However, this may not be appropriate for SPECT/CT using CT-based attenuation and scatter correction as the CT number at 120 kVp is typically -640 HU as opposed to approximately -850 HU for healthy lung and -900 HU for diseased lung. We evaluate low density open cell flexible polyurethane foam (LDOCFPF) as a material to model lung for quantitative SPECT/CT imaging studies. Methods: We used three 20 cm diam cylinders of LDOCFPF with different densities (1.04, 1.236, and 1.50 lbs/ft3 ). We measured the average density of the foam and the average CT number for the LDOCFPF after it was saturated with water and then after squeezing to remove a majority of the water. We next added 99mTc to the water to investigate the quantitative imaging capabilities of a current SPECT/CT system. The system was calibrated using a 4 liter cylindrical phantom. Finally, we assessed the correlation between the SPECT emission image and the CT image to eavluate the ability to image heterogeneity. Results: All three LDOCFPF cylinders had around the same CT number when they were damp. The lowest density foam had the most homogeneous density distribution after squeezing, ranging from -750 to -900 HU. After practicing the squeezing technique the average HU was consistently close to -850 HU. Initial studies reported absolute errors of < 6.5% for imaging of aqueous 99mTc in LDOCFPF for different source and attenuation distributions. There was also a strong correlation between CT numbers and emission activity counts. Conclusions: Low density open cell flexible polyurethane foam (LDOCFPF) can be an appropriate material to simulate lung tissue for SPECT/CT imaging studies. Initial evaluation of the quantitative accuracy of SPECT/CT lung imaging using corrections (attenuation and scatter correction) demonstrates that biases < 6.5% can be achieved.

(17:00) M3D2-3, Beta-Gamma Imager for Quantitative Plant Imaging

H. Ranjbar1, J. Wen1, A. J. Mathews1, Q. Wang2, S. Komarov2, K. Li1, J. A. O'Sullivan1, Y.-C. Tai2

1Electrical and Systems Engineering, Washington University in St. Louis, Saint Louis, United States
2Radiological Sciences, Mallinckrodt Institute of Radiology, Saint Louis, United States
Application of nuclear imaging technology in plant sciences and research is advancing. Positron emitting isotopes, such as 11C, 13N, and 18F, can label plants to study their biological processes particularly metabolism and photosynthesis; which could contribute to higher crops yield and biomass. Measurements and resulting images from PET scanners are not quantitative in young plant structures or in plant leaves; due to poor positron annihilation in thin objects. We have designed, built, modeled, and tested a nuclear imaging system that can simultaneously detect positrons (ß) and coincidence gamma rays (?). This system employs two planar detectors; one is a phoswich detector using a BC-404 plastic scintillator for beta detection and LYSO array for gamma detection. A forward model for positron is proposed along with a joint image reconstruction formulation to utilize ß and ? measurements for estimating radioactivity distribution in plant leaves. More specifically, the algorithm first reconstructs ß and ? images independently to estimate the thickness component of the beta forward model, and subsequently jointly estimates the radioactivity distribution in the object. We have validated the Physics model and reconstruction framework through phantom imaging studies and imaging a plant labeled with 11CO2. The results demonstrate that the simultaneously acquired ß and coincidence ? data, combined with our proposed joint reconstruction algorithm improved the quantitative accuracy of estimating radioactivity distribution in thin objects such as leaves.

(17:15) M3D2-4, Assessment of a Nonparametric Sinogram-Based Bootstrap Resampling Performance in Volumes of Interest for SPECT Data

E. M. Vicente, A. K. Jha, E. C. Frey

Div. of Medical Imaging Physics, Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University, Baltimore, MD 21287, USA
Characterizing the statistical properties of quantitative parameters estimated from nuclear medicine images is important for a variety of clinical applications. While repeated scans are the gold standard way to obtain this information, this is often difficult in practice. Bootstrap resampling methods have been proposed as an alternative, allowing the assessment of the statistical properties of the data for a particular setting. These methods usually employ list data, typically not available in SPECT, or use parametric methods that assume a statistical distribution of the projection data, potentially resulting in bias of the statistical estimates. In this work we propose a nonparametric bootstrap method based on random sampling of events in an original sinogram, thus avoiding need for list data or assumptions about the distribution. The proposed method was validated and evaluated, using realistic SPECT simulations assuming a Poisson distribution, by comparison to the gold standard of repeated acquisitions and against a parametric method, which has been previously validated. A total of 1000 sinograms were obtained for each method. The sinograms were reconstructed using a 3D-OSEM reconstruction method including attenuation, scatter and the full collimator-detector response compensations. Comparisons of statistical properties of the images showed that both bootstrap methods achieved similar performance and were suitable for estimating standard deviations of the mean activity in organ volumes-of-interest (VOIs) with biases less than 7% for sufficiently large VOIs.

(17:30) M3D2-5, Evaluating the Effect of Incremental Dose Reduction on Perfusion Defect Detection Employing Hybrid Cardiac Perfusion SPECT Slices

P. H. Pretorius1, M. A. King1, K. Johnson1, Y. Yang2, M. N. Wernick2

1Department of Radiology/Division of Nuclear Medicine, University of Massachusetts Medical School, Worcester,MA, USA
2Department of Electrical Engineering/Medical Imaging Research Center, Illinois Institute of Technology, Chicago, Il, USA
Accruing enough patient studies with a gold standard diagnosis to evaluate reconstruction strategies is costly and cumbersome. We recently improved a method developed earlier for inserting cardiac perfusion defects into SPECT slices of hearts deemed normal by experienced physicians. In this study we employed our improved methodology to study the effect of incremental dose reduction on perfusion defect detection. All patients included in this study were under IRB approval with written consent. Imaging consisted of list-mode acquisitions during stress Tc-99m sestamibi perfusion SPECT on a BrightView SPECT/CT (Philips, Cleveland, OH) employing the standard clinical protocol. A method was developed to randomly select fractions of list-mode events for inclusion in decreased dose acquisitions spanning the total acquisition time. The original data were then reconstructed with a combination of respiratory and rigid-body motion compensation and perfusion defects inserted to reflect a 50% reduction in wall counts. The same defects were subsequently inserted into the lower dose acquisitions. Finally, the hybrid projections were reconstructed with and without respiratory motion compensation and evaluated using polar map quantitation comparing the percentage reduction of the counts in the lower dose defects with that in the original hybrid acquisitions. Also, success was assumed when the polar map software flagged the inserted defect as a region with disease according to the ASNC guidelines. In our preliminary evaluation, we were able to reduce the dose to 25% of the original counts and obtain similar decreases in defect counts while still been flagged as diseased. However, other distractors (noise, respiratory motion when not corrected) now clearly influence the accuracy in location and size of the defects. Therefore, other processing methods not yet clinically available should be explored, especially for dose reduction below 50% of that used in standard acquisition protocols.

(17:45) M3D2-6, Simultaneous Multiple Kinect V2 for Extended Field of View Motion Tracking

P. Noonan1, J. Howard2, J. Ma2, D. Cole3, W. Hallett2, B. Glocker1, R. Gunn1,2

1Division of Brain Sciences, Imperial College London, London, UK
2Imanova Imaging Sciences, London, UK
3University of Sussex, Brighton, UK
Markerless motion tracking is an attractive technique to monitor subject motion during Positron Emission Tomography, PET, imaging. Consumer grade depth sensors such as the Microsoft Kinect v2 offer a low cost solution to obtain the 3D position and orientation of a subject's head during brain PET. This tracking can be performed in real time at 30 Hz by registering of all the depth points measured by the Kinect v2 to an initial template, as implemented in the KinectFusion iterative closest point, ICP, algorithm. Due to USB 3.0 bandwidth constraints, it is only possible to connect a single Kinect v2 to a PC. In this work, depth data from a second Kinect v2 is sent over a gigabit ethernet connection, and is incorporated in parallel with the KinectFusion tracking application. This enables tracking to be performed from the two viewpoints of the two Kinect v2 cameras. We demonstrate the application of concurrent multiple Kinects for real time head tracking and for whole body rapid shape measuring which has potential uses in PET and radiotherapy. The Kinects are modified to work in the near mode (170 mm) and are positioned inside the PET/CT scanner gantry.

(18:00) M3D2-7, Markerless Head Tracking Evaluation with Human Subjects for a Dedicated Brain PET Scanner

S. Anishchenko1,2, D. Beylin3, P. Stepanov3, A. Stepanov3, I. N. Weinberg3, S. Schaeffer3, V. Zavarzin3, D. Shaposhnikov2, M. F. Smith1

1University of Maryland School of Medicine, Baltimore, Maryland, USA
2A.B. Kogan Research Institute for Neurocybernetics, Southern Federal University, Rostov-on-Don, Russia
3Brain Biosciences, Inc., Rockville, Maryland, USA
Objectives. The goal of this work is to evaluate a markerless head motion tracking system for a dedicated brain PET scanner on human subjects. Methods. A mockup of dedicated brain PET (Brain Biosciences Inc., USA) was created and equipped with four off-the-shelf wide angle web cameras (640x480 px) and a magnetic tracking device (Polhemus Inc., USA). Calibration was done to form two stereo pairs as well as to find the transformations from both stereo coordinate systems to the magnetic tracking one. Volunteers were asked to perform a mock PET scan. Two sensors of a magnetic tracking device were fixed to the head with a headband. Head motion data were collected for each subject in three tests runs. In the first run the subjects were asked to remain still for 15 minutes, in the second run they were asked to stay 5 minutes in the scanner without any restriction of motion; in the third run they were asked to perform certain motions and facial expressions. Head pose was tracked with the magnetic motion tracking device and the four calibrated cameras. Results were compared to evaluate the markerless system tracking performance for test points inside the head. Results. The maximum displacement of the test-points was 11.4 mm, 36.2 mm and 35.8 mm, respectively, for each type of study. Ground truth average error due to headband movement was 0.8 mm, 2.0 mm, 1.4 mm, respectively. Average error±standard deviation of the optical tracking system compared with the magnetic tracking system (ground truth) was 1.6±0.8 mm, 3.8±2.5 mm and 2.8±4.4 mm, respectively. Summary. A markerless motion tracking system based on 4 calibrated off-the-shelf web cameras has been developed and evaluated for a dedicated brain PET scanner.

(18:15) M3D2-8, Structural Analysis of Solid Tumors Based on Regularized Tubular Modeling

E. Wolsztynski, J. O'Sullivan, M. P. Kennedy, K. O'Regan, J. F. Eary, F. O'Sullivan

School of Mathematical Sciences, UCC, Cork, Ireland, Cork, Ireland
Intratumoral heterogeneity has been associated with treatment outcome in several cancer types, and statistical analyses of PET imaging data allow for its assessment non-invasively. Objective characterizations of heterogeneity have been found to correlate with other standard prognostic factors, for example in lung cancer, and are most commonly performed through texture analysis. The success of image-based tumor characterization relies on our capacity to derive analytical summaries that are biologically meaningful and therefore useful in making critical treatment-adaptive decisions. Here we consider a fully-automatic modeling of tumor development profiles derived at voxel level, from which we can assess intratumoral heterogeneity as well as other quantitative features. A detailed, non-robust initial tubular representation of the spatial uptake data is obtained from a crude input volume using a methodology presented recently. This automatically delineated tubular volume is then further adjusted by thin-plate smoothing splines, which allows us to control for subregions of the 3D volume with sparse sampling, and smoothed across via model regularization to yield 3D-coherent and robut tumor descriptors. The result is a localized assessment of intratumoral texture, development status and heterogeneity. We present a numerical analysis of the concept on simulated data as well as demonstration on clinical studies from two medical centers (located in the USA and Ireland). The features derived from this approach create an opportunity both for improved patient prognosis and for advanced image-guided treatment (e.g. by guiding biopsy or radiotherapy). This modeling technique can be applied to FDG uptake information or any other PET tracer, and is also applicable to PET/CT and PET/MR data. The concept also applies to a range of solid tumors.