Internal dosimetry study of [82Rb]Cl using a long axial field-of-view PET/CT

PURPOSE

Long axial field-of-view (LAFOV) positron emission tomography (PET) systems allow to image all major organs with one bed position, which is particularly useful for acquiring whole-body dynamic data using short-lived radioisotopes like 82Rb.

METHODS

We determined the absorbed dose in target organs of three subjects (29, 40, and 57 years old) using two different methods, i.e., MIRD and voxel dosimetry. The subjects were injected with 407.0 to 419.61 MBq of [82Rb]Cl and were scanned dynamically for 7 min with a LAFOV PET/CT scanner.

RESULTS

Using the MIRD formalism and voxel dosimetry, the absorbed dose ranged from 1.84 to 2.78 μGy/MBq (1.57 to 3.92 μGy/MBq for voxel dosimetry) for the heart wall, 2.76 to 5.73 μGy/MBq (3.22 to 5.37 μGy/MBq for voxel dosimetry) for the kidneys, and 0.94 to 1.88 μGy/MBq (0.98 to 1.92 μGy/MBq for voxel dosimetry) for the lungs. The total body effective dose lied between 0.50 and 0.76 μSv/MBq.

CONCLUSION

Our study suggests that the radiation dose associated with [82Rb]Cl PET/CT can be assessed by means of dynamic LAFOV PET and that it is lower compared to literature values.

Comparison of post reconstruction- and reconstruction-based deep learning denoising methods in cardiac SPECT

Objective. The quality of myocardial perfusion SPECT (MPS) images is often hampered by low count statistics. Poor image quality might hinder reporting the studies and in the worst case lead to erroneous diagnosis. Deep learning (DL)-based methods can be used to improve the quality of the low count studies. DL can be applied in several different methods, which might affect the outcome. The aim of this study was to investigate the differences between post reconstruction- and reconstruction-based denoising methods.Approach. A UNET-type network was trained using ordered subsets expectation maximization (OSEM) reconstructed MPS studies acquired with half, quarter and eighth of full-activity. The trained network was applied as a post reconstruction denoiser (OSEM+DL) and it was incorporated into a regularized reconstruction algorithm as a deep learning penalty (DLP). OSEM+DL and DLP were compared against each other and against OSEM images without DL denoising in terms of noise level, myocardium-ventricle contrast and defect detection performance with signal-to-noise ratio of a non-prewhitening matched filter (NPWMF-SNR) applied to artificial perfusion defects inserted into defect-free clinical MPS scans. Comparisons were made using half-, quarter- and eighth-activity data.Main results. OSEM+DL provided lower noise level at all activities than other methods. DLP's noise level was also always lower than matching activity OSEM's. In addition, OSEM+DL and DLP outperformed OSEM in defect detection performance, but contrary to noise level ranking DLP had higher NPWMF-SNR overall than OSEM+DL. The myocardium-ventricle contrast was highest with DLP and lowest with OSEM+DL. Both OSEM+DL and DLP offered better image quality than OSEM, but visually perfusion defects were deeper in OSEM images at low activities.Significance. Both post reconstruction- and reconstruction-based DL denoising methods have great potential for MPS. The preference between these methods is a trade-off between smoother images and better defect detection performance.

Variability in PET image quality and quantification measured with a permanently filled 68Ge-phantom: a multi-center study

BACKGROUND

This study evaluated, as a snapshot, the variability in quantification and image quality (IQ) of the clinically utilized PET [¹⁸F]FDG whole-body protocols in Finland using a NEMA/IEC IQ phantom permanently filled with ⁶⁸Ge.

METHODS

The phantom was imaged on 14 PET-CT scanners, including a variety of models from two major vendors. The variability of the recovery coefficients (RC_max, RC_mean and RC_peak) of the hot spheres as well as percent background variability (PBV), coefficient of variation of the background (COV_BG) and accuracy of corrections (AOC) were studied using images from clinical and standardized protocols with 20 repeated measurements. The ranges of the RCs were also compared to the limits of the EARL ¹⁸F standards 2 accreditation (EARL2). The impact of image noise on these parameters was studied using averaged images (AVIs).

RESULTS

The largest variability in RC values of the routine protocols was found for the RC_max with a range of 68% and with 10% intra-scanner variability, decreasing to 36% when excluding protocols with suspected cross-calibration failure or without point-spread-function (PSF) correction. The RC ranges of individual hot spheres in routine or standardized protocols or AVIs fulfilled the EARL2 ranges with two minor exceptions, but fulfilling the exact EARL2 limits for all hot spheres was variable. RC_peak was less dependent on averaging and reconstruction parameters than RC_max and RC_mean. The PBV, COV_BG and AOC varied between 2.3-11.8%, 9.6-17.8% and 4.8-32.0%, respectively, for the routine protocols. The RC ranges, PBV and COV_BG were decreased when using AVIs. With AOC, when excluding routine protocols without PSF correction, the maximum value dropped to 15.5%.

CONCLUSION

The maximum variability of the RC values for the [¹⁸F]FDG whole-body protocols was about 60%. The RC ranges of properly cross-calibrated scanners with PSF correction fitted to the EARL2 RC ranges for individual sphere sizes, but fulfilling the exact RC limits would have needed further optimization. RC_peak was the most robust RC measure. Besides COV_BG, also RCs and PVB were sensitive to image noise.

Comparison of deep learning-based denoising methods in cardiac SPECT

BACKGROUND

Myocardial perfusion SPECT (MPS) images often suffer from artefacts caused by low-count statistics. Poor-quality images can lead to misinterpretations of perfusion defects. Deep learning (DL)-based methods have been proposed to overcome the noise artefacts. The aim of this study was to investigate the differences among several DL denoising models.

METHODS

Convolution neural network (CNN), residual neural network (RES), UNET and conditional generative adversarial neural network (cGAN) were generated and trained using ordered subsets expectation maximization (OSEM) reconstructed MPS studies acquired with full, half, three-eighths and quarter acquisition time. All DL methods were compared against each other and also against images without DL-based denoising. Comparisons were made using half and quarter time acquisition data. The methods were evaluated in terms of noise level (coefficient of variation of counts, CoV), structural similarity index measure (SSIM) in the myocardium of normal patients and receiver operating characteristic (ROC) analysis of realistic artificial perfusion defects inserted into normal MPS scans. Total perfusion deficit scores were used as observer rating for the presence of a perfusion defect.

RESULTS

All the DL denoising methods tested provided statistically significantly lower noise level than OSEM without DL-based denoising with the same acquisition time. CoV of the myocardium counts with the different DL noising methods was on average 7% (CNN), 8% (RES), 7% (UNET) and 14% (cGAN) lower than with OSEM. All DL methods also outperformed full time OSEM without DL-based denoising in terms of noise level with both half and quarter acquisition time, but this difference was not statistically significant. cGAN had the lowest CoV of the DL methods at all noise levels. Image quality and polar map uniformity of DL-denoised images were also better than reduced acquisition time OSEM's. SSIM of the reduced acquisition time OSEM was overall higher than with the DL methods. The defect detection performance of full time OSEM measured as area under the ROC curve (AUC) was on average 0.97. Half time OSEM, CNN, RES and UNET provided equal or nearly equal AUC. However, with quarter time data CNN, RES and UNET had an average AUC of 0.93, which was lower than full time OSEM's AUC, but equal to quarter acquisition time OSEM. cGAN did not achieve the defect detection performance of the other DL methods. Its average AUC with half time data was 0.94 and 0.91 with quarter time data.

CONCLUSIONS

DL-based denoising effectively improved noise level with slightly lower perfusion defect detection performance than full time reconstruction. cGAN achieved the lowest noise level, but at the same time the poorest defect detection performance among the studied DL methods.

Edge-preserving adaptive autoregressive model for Poisson noise reduction

Autoregressive models in image processing are linear prediction models that split an image into a predicted (i.e. filtered) image and a prediction error image, which extracts data on the image edges. Edge separation is a crucial feature of an autoregressive model. Data on the edges can be processed in different ways and then added to the filtered image. Another basic feature of our method is spatially varying modelling. In this short article, we propose an improved autoregressive model that preserves image sharpness around the edges of the image and focus on the reduction of Poisson noise, which degrades nuclear medicine images and presents a special challenge in medical imaging.

Comparing calculated and experimental activity and dose values obtained from image-based quantification of 90Y SPECT/CT Data

PURPOSE

Selective internal radiation therapy (SIRT) is a treatment for various kinds of liver tumours by injecting ⁹⁰Y bearing microspheres into the liver vessels. To perform meaningful post-treatment dosimetry, quantitative imaging is performed.

METHODS

This work uses a Monte-Carlo based reconstruction software with scatter and attenuation correction and collimator modelling that allows the quantification of ⁹⁰Y bremsstrahlung SPECT/CT data. A dataset comprising 17 patients and measurements on a Jaszczak phantom, a NEMA IEC Body phantom and an anthropomorphic liver phantom are analysed and activities and dose values are acquired. These measured values are compared with applied activities and pre-treatment calculations, allowing to assess the quality of the SPECT reconstruction. A detailed uncertainty budget is presented, including uncertainties of the dose calibrator, the count rate, non-included interactions and other factors.

RESULTS

The applied method is validated by finding measurements repeatable within the given uncertainty, and it is shown the influence of various parameters on the reconstruction process is negligible. Furthermore, activities and doses measured in the phantoms show good agreement with calculated values, if they are corrected for partial volume effects.

CONCLUSIONS

The strict observation of metrological requirements and the creation of an uncertainty budget increase the reliability and traceability of this novel approach to ⁹⁰Y dosimetry. It gives an example of successful voxel-based dosimetry based on quantitative ⁹⁰Y SPECT/CT image data.

Quantitative Monte Carlo-based brain dopamine transporter SPECT imaging

OBJECTIVE

Brain dopamine transporter imaging with I-123-labeled radioligands is technically demanding due to the small size of the imaging target relative to the spatial resolution of most SPECT systems. In addition, I-123 has high-energy peaks which can penetrate or scatter in the collimator and be detected in the imaging energy window. The aim of this study was to implement Monte Carlo (MC)-based full collimator-detector response (CDR) compensation algorithm for I-123 into a third-party commercial SPECT reconstruction software package and to evaluate its effect on the quantitative accuracy of dopaminergic-image analysis compared to a method where only the geometric component of the CDR is compensated.

METHODS

In this work, we utilized a full Monte Carlo collimator-detector model and incorporated it into an iterative SPECT reconstruction algorithm. The full Monte Carlo model reconstruction was compared to standard reconstruction using an anthropomorphic striatal phantom filled with different I-123 striatal/cortex uptake ratios and with clinical I-123 Ioflupane DaTScan studies.

RESULTS

Reconstruction with the full model yielded higher (13-25%) striatal uptake ratios than the conventional reconstruction, but the uptake ratios were still much lower than the true ratios due to partial volume effect. Visually, images reconstructed with the full Monte Carlo model had better contrast and resolution than the conventional images, with both phantom and patient studies.

CONCLUSIONS

Reconstruction with full Monte Carlo collimator-detector model yields higher quantitative accuracy than conventional reconstruction. Additional work to reduce the partial volume effect related errors would improve the accuracy further.

Implementation of GPU accelerated SPECT reconstruction with Monte Carlo-based scatter correction

OBJECTIVE

Statistical SPECT reconstruction can be very time-consuming especially when compensations for collimator and detector response, attenuation, and scatter are included in the reconstruction. This work proposes an accelerated SPECT reconstruction algorithm based on graphics processing unit (GPU) processing.

METHODS

Ordered subset expectation maximization (OSEM) algorithm with CT-based attenuation modelling, depth-dependent Gaussian convolution-based collimator-detector response modelling, and Monte Carlo-based scatter compensation was implemented using OpenCL. The OpenCL implementation was compared against the existing multi-threaded OSEM implementation running on a central processing unit (CPU) in terms of scatter-to-primary ratios, standardized uptake values (SUVs), and processing speed using mathematical phantoms and clinical multi-bed bone SPECT/CT studies.

RESULTS

The difference in scatter-to-primary ratios, visual appearance, and SUVs between GPU and CPU implementations was minor. On the other hand, at its best, the GPU implementation was noticed to be 24 times faster than the multi-threaded CPU version on a normal 128 × 128 matrix size 3 bed bone SPECT/CT data set when compensations for collimator and detector response, attenuation, and scatter were included.

CONCLUSIONS

GPU SPECT reconstructions show great promise as an every day clinical reconstruction tool.

Quantitative accuracy of (177)Lu SPECT reconstruction using different compensation methods: phantom and patient studies

Background

In targeted radionuclide therapy (TRT), accurate quantification using SPECT/CT images is important for optimizing radiation dose delivered to both the tumour and healthy tissue. Quantitative SPECT images are regularly reconstructed using the ordered subset expectation maximization (OSEM) algorithm with various compensation methods such as attenuation (A), scatter (S) and detector and collimator response (R). In this study, different combinations of the compensation methods are applied during OSEM reconstruction and the effect on the ¹⁷⁷Lu quantification accuracy is studied in an anthropomorphic torso phantom. In addition, the phantom results are reflected to (177)Lu-DOTA-Tyr3-octreotate (¹⁷⁷Lu-DOTATATE)-treated patient data and kidney absorbed dose estimates.

Methods

The torso phantom was imaged with nine various sized (0.4–104.4 cm³) spherical inserts, filled with known ¹⁷⁷Lu activity ranging from 0.5 to 105.5 MBq. Images were reconstructed using OSEM algorithm using A, AR and ARS compensation method combinations. The compensation method combinations were compared by calculating the concentration recovery coefficient (cRC) for each insert. In addition, ten ¹⁷⁷Lu-DOTATATE-treated patient’s post-therapy dosimetry acquisitions were reconstructed, and the absorbed dose to kidneys was estimated.

Results

cRC values depend on the insert size for all compensation methods. AR and ARS produced significantly higher cRC values than attenuation correction alone. There were no cRC value differences between the methods for the smallest 1-cm-diameter insert, cRC being 0.18. However, the collimator and detector response compensation method (R) made the 1.3-cm-diameter insert clearly visible and improved cRC estimate from 0.19 to 0.43. ARS produced slightly higher cRC values for small- and medium-sized inserts than AR. On the patient data, a similar trend could be seen. AR and ARS produced higher kidney activities than using attenuation correction alone; the total absorbed doses to the right and left kidneys were on average 15 and 20 % higher for AR and 19 and 25 % higher for ARS, respectively. The effective half-life decay estimated from time-activity curves however showed no notable difference between the compensation methods.

Conclusions

The highest cRC values were achieved by applying ARS compensation during reconstruction. The results were notably higher than those using attenuation correction alone. Similarly, higher activity estimates and thus higher absorbed dose estimates were found in patient data when all compensation methods were applied. ARS improved cRC especially in small-sized sources, and it thus might aid tumour dosimetry for ¹⁷⁷Lu PRRT treatments.

Fast voxel-level dosimetry for (177)Lu labelled peptide treatments

In peptide receptor radionuclide therapy (PRRT), voxel-level radiation absorbed dose calculations can be performed using several different methods. Each method has it strengths and weaknesses; however, Monte Carlo (MC) simulation is presently considered the most accurate method at providing absorbed dose distributions. Unfortunately MC simulation is time-consuming and often impractical to carry out in a clinical practice. In this work, a fast semi-Monte Carlo (sMC) absorbed dose calculation method for (177)Lu PRRT dosimetry is presented. The sMC method is based on a local electron absorption assumption and fast photon MC simulations. The sMC method is compared against full MC simulation code built on PENELOPE (vxlPen) using digital phantoms to assess the accuracy of these assumptions.Due to the local electron absorption assumption of sMC, the potential errors in cross-fire dose from electrons and photons emitted by (177)Lu were first evaluated using an ellipsoidal kidney model by comparing vxlPen and sMC. The photon cross-fire dose from background to kidney and kidney to background with varying kidney-to-background activity concentration ratios were calculated. In addition, kidney to kidney photon and electron cross-dose with different kidney to kidney distances were studied. Second, extended cardiac-torso (XCAT) phantoms were created with liver lesions and with realistic activity distributions and tissue densities. The XCAT phantoms were used to simulate SPECT projections and 3D activity distribution images were reconstructed using an OSEM algorithm. Image-based dose rate distributions were calculated using vxlPen and sMC. Total doses and dose rate volume histograms (DrVH) produced by the two methods were compared.The photon cross-fire dose from the kidney increased the background's absorbed dose by 5% or more up to 5.8 cm distance with 20 : 1 kidney to background activity concentration ratio. On the other hand, the photon cross-fire dose from the background to the kidney volume was negligible. The vxlPen results showed that the cross fire dose between two similar kidney volumes relative to the source kidney's self-dose were 0.5% and 0.02% for photon and electrons, respectively, when source and target kidneys were modelled next to each other. The photon cross-dose decreased as function of distance, and electron doses were zero at distances larger than 4 mm. The difference between sMC and vxlPen kidney total doses in the XCAT phantom study was -0.4% while the electron dose DrVHs were identical between the methods. There was a systematic 5% difference in photon doses in soft tissue between the codes due to different simulations parameters. However, the photons produced only 4% of the kidney's total dose, thus the difference was not considered significant for total dose calculations.The comparison studies show that the absorbed doses calculated using the sMC differ only slightly from dedicated MC simulator results, while the dose estimates can be obtained in a fraction of the dedicated simulator's calculation time. Results imply that there is no need for electron MC simulation for (177)Lu absorption calculations with current SPECT systems. However, the photon cross-fire dose should be taken into account in healthy tissues, which have a relatively low uptake especially in cases where there are high uptake volumes are nearby.

Comparison of multi-ray and point-spread function based resolution recovery methods in pinhole SPECT reconstruction

BACKGROUND AND OBJECTIVES

Statistical reconstruction methods allow resolution recovery in tomographic reconstruction. Even though resolution recovery has the potential to improve overall image quality, pinhole SPECT images are still often reconstructed using simplified models of the acquisition geometry in order to reduce reconstruction time. This paper investigates the benefits of two resolution recovery methods, multi-ray and point-spread function based, in pinhole SPECT by comparing them to uncorrected reconstruction.

METHODS

Resolution recovery was incorporated into ordered subsets expectation maximization reconstruction algorithm. The first of the correction methods used a simple but very fast multiple projection ray approach, whereas the second, much slower, method modelled the acquisition geometry more accurately using the analytical point-spread function of the pinhole collimator. Line source, Jaszczak and contrast phantom studies were performed and used for comparison.

RESULTS

Resolution recovery improved resolution, contrast and visual quality of the images when compared to reconstructions without it. The method based on the point-spread function performed slightly better, but was almost 50 times slower than the much simpler multi-ray approach.

CONCLUSION

The multiple projection ray approach is a promising method for very fast and easy resolution recovery in pinhole SPECT. It has a profound effect on image quality and can markedly improve the resolution-sensitivity trade-off.

A new reconstruction strategy for image improvement in pinhole SPECT

Pinhole single-photon emission computed tomography (SPECT) is able to provide information on the biodistribution of several radioligands in small laboratory animals, but has limitations associated with non-uniform spatial resolution or axial blurring. We have hypothesised that this blurring is due to incompleteness of the projection data acquired by a single circular pinhole orbit, and have evaluated a new strategy for accurate image reconstruction with better spatial resolution uniformity. A pinhole SPECT system using two circular orbits and a dedicated three-dimensional ordered subsets expectation maximisation (3D-OSEM) reconstruction method were developed. In this system, not the camera but the object rotates, and the two orbits are at 90 degrees and 45 degrees relative to the object's axis. This system satisfies Tuy's condition, and is thus able to provide complete data for 3D pinhole SPECT reconstruction within the whole field of view (FOV). To evaluate this system, a series of experiments was carried out using a multiple-disk phantom filled with 99mTc solution. The feasibility of the proposed method for small animal imaging was tested with a mouse bone study using 99mTc-hydroxymethylene diphosphonate. Feldkamp's filtered back-projection (FBP) method and the 3D-OSEM method were applied to these data sets, and the visual and statistical properties were examined. Axial blurring, which was still visible at the edge of the FOV even after applying the conventional 3D-OSEM instead of FBP for single-orbit data, was not visible after application of 3D-OSEM using two-orbit data. 3D-OSEM using two-orbit data dramatically reduced the resolution non-uniformity and statistical noise, and also demonstrated considerably better image quality in the mouse scan. This system may be of use in quantitative assessment of bio-physiological functions in small animals.

Improving the quality of small animal brain pinhole SPECT imaging by Bayesian reconstruction

The possibility of using existing hardware makes pinhole single-photon emission computed tomography (SPECT) attractive when pursuing the ultra-high resolution required for small animal brain imaging. Unfortunately, the poor sensitivity and the heavy weight of the collimator hamper the use of pinhole SPECT in animal studies by generating noisy and misaligned projections. To improve the image quality we have developed a new Bayesian reconstruction method, pinhole median root prior (PH-MRP), which prevents the excessive noise accumulation from the projections to the reconstructed image. The PH-MRP algorithm was used to reconstruct data acquired with our small animal rotating device, which was designed to reduce the rotation orbit misalignments. Phantom experiments were performed to test the device and compare the PH-MRP with the conventional Feldkamp-Davis-Kress (FDK) and pinhole ordered subsets maximum likelihood expectation maximisation (PH-OSEM) reconstruction algorithms. The feasibility of the system for small animal brain imaging was studied with Han-Wistar rats injected with (123)I-epidepride and (99m)Tc-hydroxy methylene diphosphonate. Considering all the experiments, no shape distortions due to orbit misalignments were encountered and remarkable improvements in noise characteristics and also in overall image quality were observed when the PH-MRP was applied instead of the FDK or PH-OSEM. In addition, the proposed methods utilise existing hardware and require only a certain amount of construction and programming work, making them easy to implement.

Accelerated median root prior reconstruction for pinhole single-photon emission tomography (SPET)

Pinhole collimation can be used to improve spatial resolution in SPET. However, the resolution improvement is achieved at the cost of reduced sensitivity, which leads to projection images with poor statistics. Images reconstructed from these projections using the maximum likelihood expectation maximization (ML-EM) algorithms, which have been used to reduce the artefacts generated by the filtered backprojection (FBP) based reconstruction, suffer from noise/bias trade-off: noise contaminates the images at high iteration numbers, whereas early abortion of the algorithm produces images that are excessively smooth and biased towards the initial estimate of the algorithm. To limit the noise accumulation we propose the use of the pinhole median root prior (PH-MRP) reconstruction algorithm. MRP is a Bayesian reconstruction method that has already been used in PET imaging and shown to possess good noise reduction and edge preservation properties. In this study the PH-MRP algorithm was accelerated with the ordered subsets (OS) procedure and compared to the FBP, OS-EM and conventional Bayesian reconstruction methods in terms of noise reduction, quantitative accuracy, edge preservation and visual quality. The results showed that the accelerated PH-MRP algorithm was very robust. It provided visually pleasing images with lower noise level than the FBP or OS-EM and with smaller bias and sharper edges than the conventional Bayesian methods.

Pinhole single-photon emission tomography reconstruction based on median root prior

The maximum likelihood expectation maximisation (ML-EM) algorithm can be used to reduce reconstruction artefacts produced by filtered backprojection (FBP) methods in pinhole single-photon emission tomography (SPET). However, ML-EM suffers from noise propagation along iterations, which leads to quantitatively unpleasant reconstruction results. To avoid this increase in noise, the median root prior (MRP) algorithm for pinhole SPET was implemented. Projection data of a line source and Picker's thyroid phantom were collected using a single-head gamma camera with a pinhole collimator. MRP was added to existing pinhole ML-EM reconstruction algorithm and the phantom studies were reconstructed using MRP, ML-EM and FBP for comparison. Coefficients of variation, contrasts and full-widths at half-maximum were calculated and showed a clear reduction in noise without significant loss of resolution or decrease in contrast when MRP was applied. MRP also produced visually pleasing images even with high iteration numbers, free of the checkerboard-type noise patterns which are typical of ML-EM images.

PET imaging using a triple-head gamma camera

Interest in clinical fluorodeoxyglucose (FDG) imaging with multiple-head gamma cameras is growing. To improve sensitivity, triple-head coincidence imaging has been proposed. We report our initial experiences with a triple-head coincidence gamma camera with 19 mm sodium iodide crystal thickness. Several positron emission tomography-image quality parameters were evaluated using a Carlson and line source phantom. The system sensitivity with two-dimensional axial shields was 830 cps kBq-1 ml-1 and maximum noise equivalent count rate 1900 cps for an 18F-activity of 50 MBq. The imaging resolution was in central axial 7.0 mm and in central transaxial 7.6 mm, respectively. The average scatter fraction in scattered media was 29%. Clinical brain, heart and whole body images studies with [18F]FDG were acquired and they show good correlation with the phantom image quality. As a conclusion, triple-head coincidence gamma camera provides relatively high-count rate imaging with good contrast and resolution.