Ida Häggström

Dynamic PET simulator via tomographic emission projection for kinetic modeling and parametric image studies

PURPOSE To develop and evaluate a fast and simple tool called dpetstep (Dynamic PET Simulator of Tracers via Emission Projection), for dynamic PET simulations as an alternative to Monte Carlo (MC), useful for educational purposes and evaluation of the effects of the clinical environment, postprocessing choices, etc., on dynamic and parametric images. METHODS The tool was developed in matlab using both new and previously reported modules of petstep (PET Simulator of Tracers via Emission Projection). Time activity curves are generated for each voxel of the input parametric image, whereby effects of imaging system blurring, counting noise, scatters, randoms, and attenuation are simulated for each frame. Each frame is then reconstructed into images according to the user specified method, settings, and corrections. Reconstructed images were compared to MC data, and simple Gaussian noised time activity curves (GAUSS). RESULTS dpetstep was 8000 times faster than MC. Dynamic images from dpetstep had a root mean square error that was within 4% on average of that of MC images, whereas the GAUSS images were within 11%. The average bias in dpetstep and MC images was the same, while GAUSS differed by 3% points. Noise profiles in dpetstep images conformed well to MC images, confirmed visually by scatter plot histograms, and statistically by tumor region of interest histogram comparisons that showed no significant differences (p < 0.01). Compared to GAUSS, dpetstep images and noise properties agreed better with MC. CONCLUSIONS The authors have developed a fast and easy one-stop solution for simulations of dynamic PET and parametric images, and demonstrated that it generates both images and subsequent parametric images with very similar noise properties to those of MC images, in a fraction of the time. They believe dpetstep to be very useful for generating fast, simple, and realistic results, however since it uses simple scatter and random models it may not be suitable for studies investigating these phenomena. dpetstep can be downloaded free of cost from https://github.com/CRossSchmidtlein/dPETSTEP.

A Monte Carlo Study of the Dependence of Early Frame Sampling on Uncertainty and Bias in Pharmacokinetic Parameters from Dynamic PET

Compartmental modeling of dynamic PET data enables quantification of tracer kinetics in vivo, through the calculated model parameters. In this study, we aimed to investigate the effect of early frame sampling and reconstruction method on pharmacokinetic parameters obtained from a 2-tissue model, in terms of bias and uncertainty (SD). Methods: The GATE Monte Carlo software was used to simulate 2 × 15 dynamic 3′-deoxy-3′-18F-fluorothymidine (18F-FLT) brain PET studies, typical in terms of noise level and kinetic parameters. The data were reconstructed by both 3-dimensional (3D) filtered backprojection with reprojection (3DRP) and 3D ordered-subset expectation maximization (OSEM) into 6 dynamic image sets with different early frame durations of 1, 2, 4, 6, 10, and 15 s. Bias and SD were evaluated for fitted parameter estimates, calculated from regions of interest. Results: The 2-tissue-model parameter estimates K1, k2, and fraction of arterial blood in tissue depended on early frame sampling, and a sampling of 6–15 s generally minimized bias and SD. The shortest sampling of 1 s yielded a 25% and 42% larger bias than the other schemes, for 3DRP and OSEM, respectively, and a parameter uncertainty that was 10%–70% higher. The schemes from 4 to 15 s were generally not significantly different in regards to bias and SD. Typically, the reconstruction method 3DRP yielded less frame-sampling dependence and less uncertain results, compared with OSEM, but was on average more biased. Conclusion: Of the 6 sampling schemes investigated in this study, an early frame duration of 6–15 s generally kept both bias and uncertainty to a minimum, for both 3DRP and OSEM reconstructions. Very-short frames of 1 s should be avoided because they typically resulted in the largest parameter bias and uncertainty. Furthermore, 3DRP may be preferred over OSEM for short frames with poor statistics.

The influence of time sampling on parameters in the Logan plot

The Logan plot is a graphical method for reversible tracer bindings. The bias and uncertainties of this method have previously been analyzed with respect to noise, but little is known about the direct effects from varying the time sampling scheme. This study aims to investigate the effect of time sampling on the binding potential from the reference Logan plot. Image data from seven healthy subjects imaged with [11C]raclopride was reconstructed into six dynamic series of equal length time frames with frame times between 15 s and 480 s. Images were reconstructed using both filtered back projection (FBP) and a resolution enhanced ordered subset expectation maximization (OSEM) algorithm, SharpIR. For each sampling scheme, the nondisplaceable binding potential (BPND) parameter was calculated from the reference Logan plot with cerebellum as a reference region. The variation in BPND was analyzed as percentage deviations from the BPND for the 480 s scheme. R2 of the linear fit was also analyzed. Comparison between all sampling schemes showed that the largest deviation in BPND was 7.4% between the 15 s sampling scheme and the 480 s sampling scheme reconstructed with SharpIR. The corresponding deviation for FBP images was 1.6%. R2 was highest for long time frames, but all R2 values were above 0.997 in this study.

The influence of time sampling scheme on kinetic parameters obtained from compartmental modeling of a dynamic PET study - A Monte Carlo study

Compartmental modeling of dynamic PET data enables quantification of tracer kinetics in vivo, through the obtained model parameters. The dynamic data is sorted into frames during or after the acquisition, with a sampling interval usually ranging from 10 s to 300 s. In this study we wanted to investigate the effect of the chosen sampling interval on kinetic parameters obtained from a 2-tissue model, in terms of bias and standard deviation, using a complete Monte Carlo simulated dynamic 18F_FLT PET study. The results show that the bias and standard deviation in parameter Kl is small regardless of sampling scheme or noise in the time-activity curves (TACs), and that the bias and standard deviation in k4 is large for all cases. The bias in Va is clearly dependent on sampling scheme, increasing for increased sampling interval. In general, a too short sampling interval results in very noisy images and a large bias of the parameter estimate, and a too long sampling interval also increases bias. Noise in the TACs is the largest source of bias.

Survival Modeling of Pancreatic Cancer with Radiology Using Convolutional Neural Networks.

No reliable biomarkers for early detection of pancreatic cancer are known to date but morphological signatures from non-invasive imaging might be able to close this gap. In this paper, we present a convolutional neural network-based survival model trained directly from computed tomography (CT) images. 159 CT images with associated survival data, and 3D segmentations of organ and tumor were provided by the Pancreatic Cancer Survival Prediction MICCAI grand challenge. A simple, yet novel, approach was used to convert CT slices into RGB-channel images in order to utilize pre-training of the model’s convolutional layers. The proposed model achieves a concordance index of 0.85, indicating a relationship between high-level features in CT imaging and disease progression. The ultimate hope is that these promising results translate to more personalized treatment decisions and better cancer care for patients.

Performance modeling of a wearable brain PET (BET) camera

Purpose: To explore, by means of analytical and Monte Carlo modeling, performance of a novel lightweight and low-cost wearable helmet-shaped Brain PET (BET) camera based on thin-film digital Geiger Avalanche Photo Diode (dGAPD) with LSO and LaBr3 scintillators for imaging in vivo human brain processes for freely moving and acting subjects responding to various stimuli in any environment. Methods: We performed analytical and Monte Carlo modeling PET performance of a spherical cap BET device and cylindrical brain PET (CYL) device, both with 25 cm diameter and the same total mass of LSO scintillator. Total mass of LSO in both the BET and CYL systems is about 32 kg for a 25 mm thick scintillator, and 13 kg for 10 mm thick scintillator (assuming an LSO density of 7.3 g/ml). We also investigated a similar system using an LaBr3 scintillator corresponding to 22 kg and 9 kg for the 25 mm and 10 mm thick systems (assuming an LaBr3 density of 5.08 g/ml). In addition, we considered a clinical whole body (WB) LSO PET/CT scanner with 82 cm ring diameter and 15.8 cm axial length to represent a reference system. BET consisted of distributed Autonomous Detector Arrays (ADAs) integrated into Intelligent Autonomous Detector Blocks (IADBs). The ADA comprised of an array of small LYSO scintillator volumes (voxels with base a×a: 1.0 ≤ a ≤ 2.0 mm and length c: 3.0 ≤ c ≤ 6.0 mm) with 5–65 μm thick reflective layers on its five sides and sixth side optically coupled to the matching array of dGAPDs and processing electronics with total thickness of 50 μm. Simulated energy resolution was 10.8% and 3.3% for LSO and LaBr3 respectively and the coincidence window was set at 2 ns. The brain was simulated as a sphere of uniform F-18 activity with diameter of 10 cm embedded in a center of water sphere with diameter of 10 cm. Results: Analytical and Monte Carlo models showed similar results for lower energy window values (458 keV versus 445 keV for LSO, and 492 keV versus 485 keV for LaBr3), and for the relative performance of system sensitivity. Monte Carlo results further showed that the BET geometry had >50% better noise equivalent count (NEC) performance relative to the CYL geometry, and >1100% better performance than a WB geometry for 25 mm thick LSO and LaBr3. For 10 mm thick LaBr3 equivalent mass systems LSO (7 mm thick) performed ~40% higher NEC than LaBr3. Analytic and Monte Carlo simulations also showed that 1×1×3 mm scintillator crystals can achieve ~1.2 mm FWHM spatial resolution. Conclusions: This study shows that a spherical cap brain PET system can provide improved NEC while preserving spatial resolution when compared to an equivalent dedicated cylindrical PET brain camera and shows greatly improved PET performance relative to a conventional whole body PET/CT. In addition, our simulations show that LSO will generally outperform LaBr3 for NEC unless the timing resolution for LaBr3 is considerably smaller than presently used for LSO, i.e. well below 300 ps.

Initial performance studies of a wearable brain positron emission tomography camera based on autonomous thin-film digital Geiger avalanche photodiode arrays

Abstract. Using analytical and Monte Carlo modeling, we explored performance of a lightweight wearable helmet-shaped brain positron emission tomography (PET), or BET camera, based on thin-film digital Geiger avalanche photodiode arrays with Lutetium-yttrium oxyorthosilicate (LYSO) or LaBr3 scintillators for imaging in vivo human brain function of freely moving and acting subjects. We investigated a spherical cap BET and cylindrical brain PET (CYL) geometries with 250-mm diameter. We also considered a clinical whole-body (WB) LYSO PET/CT scanner. The simulated energy resolutions were 10.8% (LYSO) and 3.3% (LaBr3), and the coincidence window was set at 2 ns. The brain was simulated as a water sphere of uniform F-18 activity with a radius of 100 mm. We found that BET achieved >40% better noise equivalent count (NEC) performance relative to the CYL and >800% than WB. For 10-mm-thick LaBr3 equivalent mass systems, LYSO (7-mm thick) had ∼40% higher NEC than LaBr3. We found that 1×1×3  mm scintillator crystals achieved ∼1.1  mm full-width-half-maximum spatial resolution without parallax errors. Additionally, our simulations showed that LYSO generally outperformed LaBr3 for NEC unless the timing resolution for LaBr3 was considerably smaller than that presently used for LYSO, i.e., well below 300 ps.

SU-E-QI-03: Compartment Modeling of Dynamic Brain PET - The Effect of Scatter and Random Corrections On Parameter Errors.

PURPOSE To investigate the effects of corrections for random and scattered coincidences on kinetic parameters in brain tumors, by using ten Monte Carlo (MC) simulated dynamic FLT-PET brain scans. METHODS The GATE MC software was used to simulate ten repetitions of a 1 hour dynamic FLT-PET scan of a voxelized head phantom. The phantom comprised six normal head tissues, plus inserted regions for blood and tumor tissue. Different time-activity-curves (TACs) for all eight tissue types were used in the simulation and were generated in Matlab using a 2-tissue model with preset parameter values (K1,k2,k3,k4,Va,Ki). The PET data was reconstructed into 28 frames by both ordered-subset expectation maximization (OSEM) and 3D filtered back-projection (3DFBP). Five image sets were reconstructed, all with normalization and different additional corrections C (A=attenuation, R=random, S=scatter): Trues (AC), trues+randoms (ARC), trues+scatters (ASC), total counts (ARSC) and total counts (AC). Corrections for randoms and scatters were based on real random and scatter sinograms that were back-projected, blurred and then forward projected and scaled to match the real counts. Weighted non-linearleast- squares fitting of TACs from the blood and tumor regions was used to obtain parameter estimates. RESULTS The bias was not significantly different for trues (AC), trues+randoms (ARC), trues+scatters (ASC) and total counts (ARSC) for either 3DFBP or OSEM (p<0.05). Total counts with only AC stood out however, with an up to 160% larger bias. In general, there was no difference in bias found between 3DFBP and OSEM, except in parameter Va and Ki. CONCLUSION According to our results, the methodology of correcting the PET data for randoms and scatters performed well for the dynamic images where frames have much lower counts compared to static images. Generally, no bias was introduced by the corrections and their importance was emphasized since omitting them increased bias extensively.

Do scatter and random corrections affect the errors in kinetic parameters in dynamic PET? - a Monte Carlo study

Dynamic positron emission tomography (PET) data can be evaluated by compartmental models, yielding model specific kinetic parameters. For the parameters to be of quantitative use however, understanding and estimation of errors and uncertainties associated with them are crucial. The aim in this study was to investigate the effects of the inclusion of scattered and random counts and their respective corrections on kinetic parameter errors. The MC software GATE was used to simulate two dynamic PET scans of a phantom containing three regions; blood, tissue and a static background. The two sets of time-activity-curves (TACs) used were generated for a 2-tissue compartment model with preset parameter values (K1, k2, k3, k4 and Va). The PET data was reconstructed into 19 frames by both ordered-subset expectation maximization (OSEM) and 3D filtered backprojection with reprojection (3DFBPRP) with normalization and additional corrections (A=attenuation, R=random, S=scatter, C=correction): True counts (AC), true+random counts (ARC), true+scattered counts (ASC) and total counts (ARSC). The results show that parameter estimates from true counts (AC), true+random counts (ARC), true+scattered counts (ASC) and total counts (ARSC) were not significantly different, with the exception of Va where the bias increased with added corrections. Thus, the inclusion of and correction for scattered and random counts did not affect the bias in parameter estimates K1, k2, k3, k4 and Ki. Uncorrected total counts (only AC) resulted in biases of hundreds or even thousands of percent, emphasizing the need for proper corrections. Reconstructions with 3DFBPRP resulted in overall 20-40% less biased estimates compared to OSEM.