Daniel R. McGowan

Phantom and Clinical Evaluation of the Bayesian Penalized Likelihood Reconstruction Algorithm Q.Clear on an LYSO PET/CT System

Q.Clear, a Bayesian penalized-likelihood reconstruction algorithm for PET, was recently introduced by GE Healthcare on their PET scanners to improve clinical image quality and quantification. In this work, we determined the optimum penalization factor (beta) for clinical use of Q.Clear and compared Q.Clear with standard PET reconstructions. Methods: A National Electrical Manufacturers Association image-quality phantom was scanned on a time-of-flight PET/CT scanner and reconstructed using ordered-subset expectation maximization (OSEM), OSEM with point-spread function (PSF) modeling, and the Q.Clear algorithm (which also includes PSF modeling). Q.Clear was investigated for β (B) values of 100–1,000. Contrast recovery (CR) and background variability (BV) were measured from 3 repeated scans, reconstructed with the different algorithms. Fifteen oncology body 18F-FDG PET/CT scans were reconstructed using OSEM, OSEM PSF, and Q.Clear using B values of 200, 300, 400, and 500. These were visually analyzed by 2 scorers and scored by rank against a panel of parameters (overall image quality; background liver, mediastinum, and marrow image quality; noise level; and lesion detectability). Results: As β is increased, the CR and BV decreases; Q.Clear generally gives a higher CR and lower BV than OSEM. For the smallest sphere reconstructed with Q.Clear B400, CR is 28.4% and BV 4.2%, with corresponding values for OSEM of 24.7% and 5.0%. For the largest hot sphere, Q.Clear B400 yields a CR of 75.2% and a BV of 3.8%, with corresponding values for OSEM of 64.4% and 4.0%. Scorer 1 and 2 ranked B400 as the preferred reconstruction in 13 of 15 (87%) and 10 of 15 (73%) cases. The least preferred reconstruction was OSEM PSF in all cases. In most cases, lesion detectability was highest ranked for B200, in 9 of 15 (67%) and 10 of 15 (73%), with OSEM PSF ranked lowest. Poor lesion detectability on OSEM PSF was seen in cases of mildly 18F-FDG–avid mediastinal nodes in lung cancer and small liver metastases due to background noise. Conversely, OSEM PSF was ranked second highest for lesion detectability in most pulmonary nodule evaluation cases. The combined scores confirmed B400 to be the preferred reconstruction. Conclusion: Our phantom measurement results demonstrate improved CR and reduced BV when using Q.Clear instead of OSEM. A β value of 400 is recommended for oncology body PET/CT using Q.Clear.

Does a novel penalized likelihood reconstruction of 18F-FDG PET-CT improve signal-to-background in colorectal liver metastases?

PURPOSE Iterative reconstruction algorithms are widely used to reconstruct positron emission tomography computerised tomography (PET/CT) data. Lesion detection in the liver by 18F-fluorodeoxyglucose PET/CT (18F-FDG-PET/CT) is hindered by 18F-FDG uptake in background liver parenchyma. The aim of this study was to compare semi-quantitative parameters of histologically-proven colorectal liver metastases detected by 18F-FDG-PET/CT using data based on a Bayesian penalised likelihood (BPL) reconstruction, with data based on a conventional time-of-flight (ToF) ordered subsets expectation maximisation (OSEM) reconstruction. METHODS A BPL reconstruction algorithm was used to retrospectively reconstruct sinogram PET data. This data was compared with OSEM reconstructions. A volume of interest was placed within normal background liver parenchyma. Lesions were segmented using automated thresholding. Lesion maximum standardised uptake value (SUVmax), standard deviation of background liver parenchyma SUV, signal-to-background ratio (SBR), and signal-to-noise ratio (SNR) were collated. Data was analysed using paired Students t-tests and the Pearson correlation. RESULTS Forty-two liver metastases from twenty-four patients were included in the analysis. The average lesion SUVmax increased from 8.8 to 11.6 (p<0.001) after application of the BPL algorithm, with no significant difference in background noise. SBR increased from 4.0 to 4.9 (p<0.001) and SNR increased from 10.6 to 13.1 (p<0.001) using BPL. There was a statistically significant negative correlation between lesion size and the percentage increase in lesion SUVmax (p=0.03). CONCLUSIONS This BPL reconstruction algorithm improved SNR and SBR for colorectal liver metastases detected by 18F-FDG-PET/CT, increasing the lesion SUVmax without increasing background liver SUV or image noise. This may improve the detection of FDG-avid focal liver lesions and the diagnostic performance of clinical 18F-FDG-PET/CT in this setting, with the largest impact for small foci.

Time to demand dosimetry for molecular radiotherapy

Molecular radiotherapy (MRT) has been used clinically for around 75 years. Despite this long history of clinical use, there is no established dosimetry practice for calculating the absorbed dose delivered to tumour targets or to organs at risk. As a result, treatment protocols have often evolved based on experience with relatively small numbers of patients, each receiving a similar administered activity but, potentially, widely varying doses. This is in stark contrast to modern external-beam radiotherapy practice. This commentary describes some of the barriers to MRT dosimetry and gives some opinions on the way forward.

18F‐fluoromisonidazole uptake in advanced stage non‐small cell lung cancer: A voxel‐by‐voxel PET kinetics study

Purpose The aim of this study was to determine the relative abilities of compartment models to describe time‐courses of 18F‐fluoromisonidazole (FMISO) uptake in tumor voxels of patients with non‐small cell lung cancer (NSCLC) imaged using dynamic positron emission tomography. Also to use fits of the best‐performing model to investigate changes in fitted rate‐constants with distance from the tumor edge. Methods Reversible and irreversible two‐ and three‐tissue compartment models were fitted to 24 662 individual voxel time activity curves (TACs) obtained from tumors in nine patients, each imaged twice. Descriptions of the TACs provided by the models were compared using the Akaike and Bayesian information criteria (AIC and BIC). Two different models (two‐ and three‐tissue) were fitted to 30 measured voxel TACs to provide ground‐truth TACs for a statistical simulation study. Appropriately scaled noise was added to each of the resulting ground‐truth TACs, generating 1000 simulated noisy TACs for each ground‐truth TAC. The simulation study was carried out to provide estimates of the accuracy and precision with which parameter values are determined, the estimates being obtained for both assumptions about the ground‐truth kinetics. A BIC clustering technique was used to group the fitted rate‐constants, taking into consideration the underlying uncertainties on the fitted rate‐constants. Voxels were also categorized according to their distance from the tumor edge. Results For uptake time‐courses of individual voxels an irreversible two‐tissue compartment model was found to be most precise. The simulation study indicated that this model had a one standard deviation precision of 39% for tumor fractional blood volumes and 37% for the FMISO binding rate‐constant. Weighted means of fitted FMISO binding rate‐constants of voxels in all tumors rose significantly with increasing distance from the tumor edge, whereas fitted fractional blood volumes fell significantly. When grouped using the BIC clustering, many centrally located voxels had high‐fitted FMISO binding rate‐constants and low rate‐constants for tracer flow between the vasculature and tumor, both indicative of hypoxia. Nevertheless, many of these voxels had tumor‐to‐blood (TBR) values lower than the 1.4 level commonly expected for hypoxic tissues, possibly due to the low rate‐constants for tracer flow between the vasculature and tumor cells in these voxels. Conclusions Time‐courses of FMISO uptake in NSCLC tumor voxels are best analyzed using an irreversible two‐tissue compartment model, fits of which provide more precise parameter values than those of a three‐tissue model. Changes in fitted model parameter values indicate that levels of hypoxia rise with increasing distance from tumor edges. The average FMISO binding rate‐constant is higher for voxels in tumor centers than in the next tumor layer out, but the average value of the more simplistic TBR metric is lower in tumor centers. For both metrics, higher values might be considered indicative of hypoxia, and the mismatch in this case is likely to be due to poor perfusion at the tumor center. Kinetics analysis of dynamic PET images may therefore provide more accurate measures of the hypoxic status of such regions than the simpler TBR metric, a hypothesis we are presently exploring in a study of tumor imaging versus histopathology.

18F-FDG PET/CT assessment of histopathologically confirmed mediastinal lymph nodes in non-small cell lung cancer using a penalised likelihood reconstruction

PurposeTo investigate whether using a Bayesian penalised likelihood reconstruction (BPL) improves signal-to-background (SBR), signal-to-noise (SNR) and SUVmax when evaluating mediastinal nodal disease in non-small cell lung cancer (NSCLC) compared to ordered subset expectation maximum (OSEM) reconstruction.Materials and methods18F-FDG PET/CT scans for NSCLC staging in 47 patients (112 nodal stations with histopathological confirmation) were reconstructed using BPL and compared to OSEM. Node and multiple background SUV parameters were analysed semi-quantitatively and visually.ResultsComparing BPL to OSEM, there were significant increases in SUVmax (mean 3.2–4.0, p<0.0001), SBR (mean 2.2–2.6, p<0.0001) and SNR (mean 27.7–40.9, p<0.0001). Mean background SNR on OSEM was 10.4 (range 7.6–14.0), increasing to 12.4 (range 8.2–16.7, p<0.0001). Changes in background SUVs were minimal (largest mean difference 0.17 for liver SUVmean, p<0.001). There was no significant difference between either algorithm on receiver operating characteristic analysis (p=0.26), although on visual analysis, there was an increase in sensitivity and small decrease in specificity and accuracy on BPL.ConclusionBPL increases SBR, SNR and SUVmax of mediastinal nodes in NSCLC compared to OSEM, but did not improve the accuracy for determining nodal involvement.Key Points• Penalised likelihood PET reconstruction was applied for assessing mediastinal nodes in NSCLC.• The new reconstruction generated significant increases in signal-to-background, signal-to-noise and SUVmax.• This led to an improvement in visual sensitivity using the new algorithm.• Higher SUVmaxthresholds may be appropriate for semi-quantitative analyses with penalised likelihood.

Eight years of growth and change in UK molecular radiotherapy with implications for the future: Internal Dosimetry Users Group survey results from 2007 to 2015.

Department of Nuclear Medicine, Royal Brompton Hospital, London, Department of Nuclear Medicine, Kent and Canterbury Hospital, Canterbury, Department of Radiation Physics and Protection, Oxford University Hospitals NHS Foundation Trust, Department of Oncology, University of Oxford, Oxford, Department of Medical Physics, University Hospital Southampton, Southampton, UK and Department of Nuclear Medicine, The Christie Hospital, Manchester, UK Correspondence to Bruno Rojas, MSc, CSci, Department of Nuclear Medicine, Royal Brompton Hospital, London SW3 6NP, UK Tel: + 44 207 351 8666; e-mail: bruno.rojasfisher@nhs.net

4D-PET reconstruction of dynamic non-small cell lung cancer [18-F]-FMISO-PET data using adaptive-knot cubic B-splines

4D-PET reconstruction has the potential to significantly increase the signal-to-noise ratio in dynamic PET by fitting smooth temporal functions during the reconstruction. However, the optimal choice of temporal function remains an open question. A 4D-PET reconstruction algorithm using adaptive-knot cubic B-splines is proposed. Using realistic Monte-Carlo simulated data from a digital patient phantom representing an [18-F]-FMISO-PET scan of a non-small cell lung cancer patient, this method was compared to a spectral model based 4D-PET reconstruction and the conventional MLEM and MAP algorithms. Within the entire patient region the proposed algorithm produced the best bias-noise trade-off, while within the tumor region the spline- and spectral model-based reconstructions gave comparable results.

Optimization of image reconstruction for yttrium-90 SIRT on a LYSO PET/CT system using a Bayesian penalized likelihood reconstruction algorithm

Imaging on a γ-camera with 90Y after selective internal radiotherapy (SIRT) may allow for verification of treatment delivery but suffers relatively poor spatial resolution and imprecise dosimetry calculation. 90Y PET/CT imaging is possible on 3-dimensional, time-of-flight machines; however, images are usually poor because of low count statistics and noise. A new PET reconstruction software using a Bayesian penalized likelihood (BPL) reconstruction algorithm (termed Q.Clear) was investigated using phantom and patient scans to optimize the reconstruction for post-SIRT imaging and clarify whether BPL leads to an improvement in clinical image quality using 90Y. Methods: Phantom studies over an activity range of 0.5–4.2 GBq were performed to assess the contrast recovery, background variability, and contrast-to-noise ratio for a range of BPL and ordered-subset expectation maximization (OSEM) reconstructions on a PET/CT scanner. Patient images after SIRT were reconstructed using the same parameters and were scored and ranked on the basis of image quality, as assessed by visual evaluation, with the corresponding SPECT/CT Bremsstrahlung images by 2 experienced radiologists. Results: Contrast-to-noise ratio was significantly better in BPL reconstructions when compared with OSEM in phantom studies. The patient-derived BPL and matching Bremsstrahlung images scored higher than OSEM reconstructions when scored by radiologists. BPL with a β value of 4,000 was ranked the highest of all images. Deadtime was apparent in the system above a total phantom activity of 3.3 GBq. Conclusion: BPL with a β value of 4,000 is the optimal image reconstruction in PET/CT for confident radiologic reading when compared with other reconstruction parameters for 90Y imaging after SIRT imaging. Activity in the field of view should be below 3.3 GBq at the time of PET imaging to avoid deadtime losses for this scanner.

Iodine-131 monitoring in sewage plant outflow

Three different hospital sites (Oxford, Sutton and Guildford) have performed sampling of their local sewage plant outflow to determine levels of radioactivity resulting from iodine-131 patients undergoing radionuclide therapies. It was found that a maximum of 20% of activity discharged from the hospitals was present in the sewage plant final effluent channel. This is significantly below the level predicted by mathematical models in current use. The results further show that abatement systems to reduce public exposure are unlikely to be warranted at hospital sites.

Fast Groupwise 4D Deformable Image Registration for Irregular Breathing Motion Estimation

Tumor heterogeneity can be assessed quantitatively by analyzing dynamic contrast-enhanced imaging modalities potentially leading to improvement in the diagnosis and treatment of cancer, for example of the lung. However, the acquisition of standard lung sequences is often compromised by irregular breathing motion artefacts, resulting in unsystematic errors when estimating tissue perfusion parameters. In this work, we illustrate implicit deformable image registration that integrates the Demons algorithm using the local correlation coefficient as a similarity measure, and locally adaptive regularization that enables incorporation of both spatial sliding motions and irregular temporal motion patterns. We also propose a practical numerical approximation of the regularization model to improve both computational time and registration accuracy, which are important when analyzing long clinical sequences. Our quantitative analysis of 4D lung Computed Tomography and Computed Tomography Perfusion scans from clinical lung trial shows significant improvement over state-of-the-art pairwise registration approaches.