Half-time myocardial perfusion SPECT imaging with attenuation and Monte Carlo-based scatter correction

Application of Monte Carlo Algorithms to Cardiac Imaging Reconstruction

Monte Carlo algorithms have a growing impact on nuclear medicine reconstruction processes. One of the main limitations of myocardial perfusion imaging (MPI) is the effective mitigation of the scattering component, which is particularly challenging in Single Photon Emission Computed Tomography (SPECT). In SPECT, no timing information can be retrieved to locate the primary source photons. Monte Carlo methods allow an event-by-event simulation of the scattering kinematics, which can be incorporated into a model of the imaging system response. This approach was adopted in the late Nineties by several authors, and recently took advantage of the increased computational power made available by high-performance CPUs and GPUs. These recent developments enable a fast image reconstruction with improved image quality, compared to deterministic approaches. Deterministic approaches are based on energy-windowing of the detector response, and on the cumulative estimate and subtraction of the scattering component. In this paper, we review the main strategies and algorithms to correct the scattering effect in SPECT and focus on Monte Carlo developments, which nowadays allow the threedimensional reconstruction of SPECT cardiac images in a few seconds.

Analysis of quantitative [I-123] mIBG SPECT/CT in a phantom and in patients with neuroblastoma

Purpose

To determine the accuracy of quantitative SPECT, intersystem and interpatient standardized uptake value (SUV) calculation consistency for a manufacturer-independent quantitative SPECT/CT reconstruction algorithm, and the range of SUVs of normal and neoplastic tissue.

Methods

A NEMA body phantom with 6 spheres (ranging 10–37 mm) was filled with a known activity-to-volume ratio and used to determine the contrast recovery coefficient (CRC) for each visible sphere, and the measured SUV accuracy of those spheres and background water solution. One hundred eleven 123I-metaiodobenzylguanidine ([I-123] mIBG) SPECT/CT examinations from 43 patients were reconstructed using SUV SPECT® (HERMES Medical Solutions Inc.); 42 examinations were acquired using a GE Infinia Hawkeye 4 SPECT/CT, and 69 were acquired on a Siemens Symbia Intevo SPECT/CT. Inter scanner SUV analysis of 9 regions of normal [I-123] mIBG tissue uptake was conducted. Intrapatient mean SUV variability was calculated by measuring normal liver uptake within patients scanned on both cameras. The intensity of uptake by neoplastic tissue in the images was quantified using maximum SUV and, if present, compared over time.

Results

The phantom results of the visible spheres and background resulted in accuracy calculations better than 5–10% with CRC correction. Interscanner SUV variability showed no statistical difference (average p value 0.559; range 0.066–1.0) among the 9 normal tissues analyzed. Intrapatient liver mean SUV varied ≤ 16% as calculated for 28 patients (87 examinations) studied on both scanners. In one patient, a thoracic tumor evaluated over 10 time points (18 months) underwent a 74% (3.1/12.0) reduction in maximum SUV with treatment.

Conclusion

The results demonstrate quantitative accuracy to better than 10%, and both consistent SUV calculation between 2 different SPECT/CT scanners for 9 tissues, and low intrapatient measurement variability for quantitative SPECT/CT analysis in a pediatric population with neuroblastoma. Quantitative SPECT/CT offers the opportunity for objective analysis of tumor response using [I-123] mIBG by normalizing the uptake to injected dose and patient weight, as is done for PET.

Cardiac 123I-MIBG planar heart to mediastinum ratios depend on patient size; phantom studies suggest SPECT-CT could improve quantification

INTRODUCTION

Planar ¹²³I-MIBG (meta-iodobenzylguanidine) cardiac imaging is listed as an indicative biomarker in the 2017 international consensus criteria for the diagnosis of dementia with Lewy bodies. There has been very little research into the relationship between apparent cardiac uptake and patient size, or in the possible advantage of attenuation and scatter corrected SPECT-CT compared to planar imaging. We aimed to evaluate this in both a chest phantom and in older adults with normal cognition.

MATERIALS AND METHODS

An anthropomorphic chest phantom was filled with ¹²³I solution using activities typical of healthy subjects. The phantom was scanned on a Siemens Intevo gamma camera with MELP collimators using both planar and SPECT-CT techniques. Further scans were acquired with a PMMA chest plate added, then water filled plastic breasts. The SPECT-CT images were reconstructed using a resolution recovery OSEM method with and without attenuation and scatter correction (ACSC) applied. Twenty-nine adults over 60 years of age (mean 75.2 ± 8.3 years) underwent planar cardiac MIBG imaging, followed by SPECT-CT. SPECT images were reconstructed as above. Heart-to-mediastinum ratios (HMRs) were calculated for planar and SPECT images.

RESULTS

Phantom planar HMR decreased by 20% with the PMMA chest plate added; 39% with plate and breasts. ACSC SPECT cardiac counts showed less dependence on phantom size than SPECT without ACSC (3% versus 37%). The body mass indices (BMI) of the older adults ranged from 22 to 38. There was a significant linear relationship between planar HMR and BMI (R² = 0.44, p<0.01), but not for ACSC SPECT. However, there was no significant difference between the slopes for planar and ACSC SPECT (p = 0.11).

CONCLUSION

Planar cardiac ¹²³I-MIBG HMR results are correlated with BMI. Phantom results suggest that ACSC SPECT can correct for patient size. A large patient population or clinical database would be required to demonstrate a clinical effect.

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.

Phantom and clinical evaluation of the effect of full Monte Carlo collimator modelling in post-SIRT yttrium-90 Bremsstrahlung SPECT imaging

Background

Post-therapy SPECT/CT imaging of ⁹⁰Y microspheres delivered to hepatic malignancies is difficult, owing to the continuous, high-energy Bremsstrahlung spectrum emitted by ⁹⁰Y. This study aimed to evaluate the utility of a commercially available software package (HybridRecon, Hermes Medical Solutions AB) which incorporates full Monte Carlo collimator modelling. Analysis of image quality was performed on both phantom and clinical images in order to ultimately provide a recommendation of an optimum reconstruction for post-therapy ⁹⁰Y microsphere SPECT/CT imaging.

A 3D-printed anthropomorphic liver phantom was filled with ⁹⁰Y with a sphere-to-background ratio of 4:1 and imaged on a GE Discovery 670 SPECT/CT camera. Datasets were reconstructed using ordered-subsets expectation maximization (OSEM) 1–7 iterations in order to identify the optimal OSEM reconstruction (5 iterations, 15 subsets). Quantitative analysis was subsequently carried out on phantom datasets obtained using four reconstruction algorithms: the default OSEM protocol (2 iterations, 10 subsets) and the optimised OSEM protocol, both with and without full Monte Carlo collimator modelling. The quantitative metrics contrast recovery (CR) and background variability (BV) were calculated.

The four algorithms were then used to retrospectively reconstruct 10 selective internal radiation therapy (SIRT) patient datasets which were subsequently blind scored for image quality by a consultant radiologist.

Results

The optimised OSEM reconstruction (5 iterations, 15 subsets with full MC collimator modelling) increased the CR by 42% (p < 0.001) compared to the default OSEM protocol (2 iterations, 10 subsets). The use of full Monte Carlo collimator modelling was shown to further improve CR by 14% (30 mm sphere, CR = 90%, p < 0.05).

The consultant radiologist had a significant preference for the optimised OSEM over the default OSEM protocol (p < 0.001), with the optimised OSEM being the favoured reconstruction in every one of the 10 clinical cases presented.

Conclusions

OSEM (5 iterations, 15 subsets) with full Monte Carlo collimator modelling is quantitatively the optimal image reconstruction for post-SIRT ⁹⁰Y Bremsstrahlung SPECT/CT imaging. The use of full Monte Carlo collimator modelling for correction of image-degrading effects significantly increases contrast recovery without degrading clinical image quality.

Ejection fraction in myocardial perfusion imaging assessed with a dynamic phantom: comparison between IQ-SPECT and LEHR

Background

Developments in single photon emission tomography instrumentation and reconstruction methods present a potential for decreasing acquisition times. One of such recent options for myocardial perfusion imaging (MPI) is IQ-SPECT. This study was motivated by the inconsistency in the reported ejection fraction (EF) and left ventricular (LV) volume results between IQ-SPECT and more conventional low-energy high-resolution (LEHR) collimation protocols. IQ-SPECT and LEHR quantitative results were compared while the equivalent number of iterations (EI) was varied. The end-diastolic (EDV) and end-systolic volumes (ESV) and the derived EF values were investigated.

A dynamic heart phantom was used to produce repeatable ESVs, EDVs and EFs. Phantom performance was verified by comparing the set EF values to those measured from a gated multi-slice X-ray computed tomography (CT) scan (EF_True). The phantom with an EF setting of 45, 55, 65 and 70% was imaged with both IQ-SPECT and LEHR protocols. The data were reconstructed with different EI, and two commonly used clinical myocardium delineation software were used to evaluate the LV volumes.

Results

The CT verification showed that the phantom EF settings were repeatable and accurate with the EF_True being within 1% point from the manufacture’s nominal value. Depending on EI both MPI protocols can be made to produce correct EF estimates, but IQ-SPECT protocol produced on average 41 and 42% smaller EDV and ESV when compared to the phantom’s volumes, while LEHR protocol underestimated volumes by 24 and 21%, respectively. The volume results were largely similar between the delineation methods used.

Conclusions

The reconstruction parameters can greatly affect the volume estimates obtained from perfusion studies. IQ-SPECT produces systematically smaller LV volumes than the conventional LEHR MPI protocol. The volume estimates are also software dependent.

Performance of 3DOSEM and MAP algorithms for reconstructing low count SPECT acquisitions

PURPOSE

Low count single photon emission computed tomography (SPECT) is becoming more important in view of whole body SPECT and reduction of radiation dose. In this study, we investigated the performance of several 3D ordered subset expectation maximization (3DOSEM) and maximum a posteriori (MAP) algorithms for reconstructing low count SPECT images.

MATERIALS AND METHODS

Phantom experiments were conducted using the National Electrical Manufacturers Association (NEMA) NU2 image quality (IQ) phantom. The background compartment of the phantom was filled with varying concentrations of pertechnetate and indiumchloride, simulating various clinical imaging conditions. Images were acquired using a hybrid SPECT/CT scanner and reconstructed with 3DOSEM and MAP reconstruction algorithms implemented in Siemens Syngo MI.SPECT (Flash3D) and Hermes Hybrid Recon Oncology (Hyrid Recon 3DOSEM and MAP). Image analysis was performed by calculating the contrast recovery coefficient (CRC),percentage background variability (N%), and contrast-to-noise ratio (CNR), defined as the ratio between CRC and N%. Furthermore, image distortion is characterized by calculating the aspect ratio (AR) of ellipses fitted to the hot spheres. Additionally, the performance of these algorithms to reconstruct clinical images was investigated.

RESULTS

Images reconstructed with 3DOSEM algorithms demonstrated superior image quality in terms of contrast and resolution recovery when compared to images reconstructed with filtered-back-projection (FBP), OSEM and 2DOSEM. However, occurrence of correlated noise patterns and image distortions significantly deteriorated the quality of 3DOSEM reconstructed images. The mean AR for the 37, 28, 22, and 17mm spheres was 1.3, 1.3, 1.6, and 1.7 respectively. The mean N% increase in high and low count Flash3D and Hybrid Recon 3DOSEM from 5.9% and 4.0% to 11.1% and 9.0%, respectively. Similarly, the mean CNR decreased in high and low count Flash3D and Hybrid Recon 3DOSEM from 8.7 and 8.8 to 3.6 and 4.2, respectively. Regularization with smoothing priors could suppress these noise patterns at the cost of reduced image contrast. The mean N% was 6.4% and 6.8% for low count QSP and MRP MAP reconstructed images. Alternatively, regularization with an anatomical Bowhser prior resulted in sharp images with high contrast, limited image distortion, and low N% of 8.3% in low count images, although some image artifacts did occur. Analysis of clinical images suggested that the same effects occur in clinical imaging.

CONCLUSION

Image quality of low count SPECT acquisitions reconstructed with modern 3DOSEM algorithms is deteriorated by the occurrence of correlated noise patterns and image distortions. The artifacts observed in the phantom experiments can also occur in clinical imaging.

Adaptive Autoregressive Model for Reduction of Noise in SPECT

This paper presents improved autoregressive modelling (AR) to reduce noise in SPECT images. An AR filter was applied to prefilter projection images and postfilter ordered subset expectation maximisation (OSEM) reconstruction images (AR-OSEM-AR method). The performance of this method was compared with filtered back projection (FBP) preceded by Butterworth filtering (BW-FBP method) and the OSEM reconstruction method followed by Butterworth filtering (OSEM-BW method). A mathematical cylinder phantom was used for the study. It consisted of hot and cold objects. The tests were performed using three simulated SPECT datasets. Image quality was assessed by means of the percentage contrast resolution (CR%) and the full width at half maximum (FWHM) of the line spread functions of the cylinders. The BW-FBP method showed the highest CR% values and the AR-OSEM-AR method gave the lowest CR% values for cold stacks. In the analysis of hot stacks, the BW-FBP method had higher CR% values than the OSEM-BW method. The BW-FBP method exhibited the lowest FWHM values for cold stacks and the AR-OSEM-AR method for hot stacks. In conclusion, the AR-OSEM-AR method is a feasible way to remove noise from SPECT images. It has good spatial resolution for hot objects.

Comparison of attenuation, dual-energy-window, and model-based scatter correction of low-count SPECT to 82Rb PET/CT quantified myocardial perfusion scores

BACKGROUND

New reconstruction algorithms allow reduction in acquisition times or the amount of injected radioactivity. We examined the impact of different corrections on low-count clinical SPECT myocardial perfusion images (MPI) and compared to (82)Rb PET/CT. We compared no corrections (NC) to attenuation correction (AC) with and without scatter correction by either a dual-energy-window (AC-DEW) or model-based (AC-ESSE) approach. All reconstructions included resolution recovery.

METHODS

56 patients were imaged using a standard rest/stress Tc-99m-tetrofosmin MPI SPECT/CT protocol with an additional half-time acquisition. A (82)Rb-rest/stress PET/CT MPI was acquired within 4 weeks. Reconstruction methods were compared using summed rest/stress/difference scores from an objective algorithm (SRS/SSS/SDS).

RESULTS

The SRS and SSS for NC were significantly (P < .01) higher than for AC, but well correlated (r ≥ 0.87). The correlation in SRS/SSS among AC, AC-DEW, and AC-ESSE was excellent (r ≥ 0.98). AC-ESSE and AC-DEW had higher SRS (P ≤ .05) than AC, but the SDS values were not significantly different. Concordance with PET normal/abnormal classification was 76% for NC and ≥85% for the AC methods.

CONCLUSION

AC significantly improves the accuracy of low-count myocardial perfusion SPECT half-time imaging for the detection of disease compared to NC. Compared to PET, there was no significant difference among AC, AC-DEW, and AC-ESSE.

Optimisation of simultaneous tl-201/tc-99m dual isotope reconstruction with monte-carlo-based scatter correction

Simultaneous Tl-201/Tc-99m dual isotope myocardial perfusion SPECT is seriously hampered by down-scatter from Tc-99m into the Tl-201 energy window. This paper presents and optimises the ordered-subsets-expectation-maximisation-(OS-EM-) based reconstruction algorithm, which corrects the down-scatter using an efficient Monte Carlo (MC) simulator. The algorithm starts by first reconstructing the Tc-99m image with attenuation, collimator response, and MC-based scatter correction. The reconstructed Tc-99m image is then used as an input for an efficient MC-based down-scatter simulation of Tc-99m photons into the Tl-201 window. This down-scatter estimate is finally used in the Tl-201 reconstruction to correct the crosstalk between the two isotopes. The mathematical 4D NCAT phantom and physical cardiac phantoms were used to optimise the number of OS-EM iterations where the scatter estimate is updated and the number of MC simulated photons. The results showed that two scatter update iterations and 10(5) simulated photons are enough for the Tc-99m and Tl-201 reconstructions, whereas 10(6) simulated photons are needed to generate good quality down-scatter estimates. With these parameters, the entire Tl-201/Tc-99m dual isotope reconstruction can be accomplished in less than 3 minutes.