PET Reconstruction Research Articles

Reconstruction, analysis and interpretation of posterior probability distributions of PET images, using the posterior bootstrap

The uncertainty of reconstructed PET images remains difficult to assess and to interpret for the use in diagnostic and quantification tasks. Here we provide (1) an easy-to-use methodology for uncertainty assessment for almost any Bayesian model in PET reconstruction from single datasets and (2) a detailed analysis and interpretation of produced posterior image distributions. We apply a recent posterior bootstrap framework to the PET image reconstruction inverse problem and obtain simple parallelizable algorithms based on random weights and on existing maximum a posteriori (MAP) (posterior maximum) optimization-based algorithms. Posterior distributions are produced, analyzed and interpreted for several common Bayesian models. Their relationship with the distribution of the MAP image estimate over multiple dataset realizations is exposed. The coverage properties of posterior distributions are validated. More insight is obtained for the interpretation of posterior distributions in order to open the way for including uncertainty information into diagnostic and quantification tasks.

Low activity [11C]raclopride kinetic modeling in the mouse brain using the spatiotemporal kernel method

Depending on the molar activity of the tracer, the maximal allowable injected activity in mouse brain PET studies can be extremely low in order to avoid receptor saturation. Therefore, a high level of noise can be present in the image. We investigate several dynamic PET reconstruction methods in reduced counts, or equivalently in reduced injected activity, data exemplified in [11C]raclopride BP ND and quantification using the simplified reference tissue model (SRTM). We compared independent frame reconstruction (IFR), post-reconstruction HYPR denoising (IFR + HYPR), direct reconstruction using the SRTM model (DIR-SRTM), and the spatial (KERS) and spatiotemporal kernel reconstruction (KERST). Additionally, HYPR denoising of the frames used as features for the calculation of the spatial kernel matrix, was investigated (KERS-HYPR and KERST-HYPR). In vivo data of 11 mice, was used to generate list-mode data for five reduced count levels corresponding to reductions by a factor 4, 8, 12, 16 and 32 (equivalently 2.07, 1.04, 0.691, 0.518, and 0.260 MBq). Correlation of regional BP ND and values (reduced versus full counts reconstructions) was high (r > 0.94) for all methods, with KERS-HYPR and KERST-HYPR reaching the highest correlation (r > 0.96). Among methods with regularization, DIR-SRTM showed the largest variability in BP ND (Bland–Altman SD from 3.0% to 12%), while IFR showed it for (5.1%–14.6%). KERST and KERST-HYPR were the only methods with Bland–Altman bias and SD below 5% for noise level up to a reduction factor of 16. At the voxel level, BP ND and correlation was gradually decreased with increasing noise, with the largest correlation (BP ND r > 0.88, r > 0.62) for KERS-HYPR and KERST-HYPR. The spatial and the spatiotemporal kernel methods performed similarly, while using only temporal regularization with direct reconstruction showed more variability. Although R1 values present noise, using the spatiotemporal kernel reconstruction, accurate estimates of binding potential could be obtained with mouse injected activities as low as 0.26–0.518 MBq. This is desirable in order to maintain the tracer kinetics principle in mouse studies.

Artificial intelligence in molecular imaging.

AI has, to varying degrees, affected all aspects of molecular imaging, from image acquisition to diagnosis. During the last decade, the advent of deep learning in particular has transformed medical image analysis. Although the majority of recent advances have resulted from neural-network models applied to image segmentation, a broad range of techniques has shown promise for image reconstruction, image synthesis, differential-diagnosis generation, and treatment guidance. Applications of AI for drug design indicate the way forward for using AI to facilitate molecular-probe design, which is still in its early stages. Deep-learning models have demonstrated increased efficiency and image quality for PET reconstruction from sinogram data. Generative adversarial networks (GANs), which are paired neural networks that are jointly trained to generate and classify images, have found applications in modality transformation, artifact reduction, and synthetic-PET-image generation. Some AI applications, based either partly or completely on neural-network approaches, have demonstrated superior differential-diagnosis generation relative to radiologists. However, AI models have a history of brittleness, and physicians and patients may not trust AI applications that cannot explain their reasoning. To date, the majority of molecular-imaging applications of AI have been confined to research projects, and are only beginning to find their ways into routine clinical workflows via commercialization and, in some cases, integration into scanner hardware. Evaluation of actual clinical products will yield more realistic assessments of AI’s utility in molecular imaging.

Imitation learning for improved 3D PET/MR attenuation correction.

The assessment of the quality of synthesised/pseudo Computed Tomography (pCT) images is commonly measured by an intensity-wise similarity between the ground truth CT and the pCT. However, when using the pCT as an attenuation map (μ-map) for PET reconstruction in Positron Emission Tomography Magnetic Resonance Imaging (PET/MRI) minimising the error between pCT and CT neglects the main objective of predicting a pCT that when used as μ-map reconstructs a pseudo PET (pPET) which is as similar as possible to the gold standard CT-derived PET reconstruction. This observation motivated us to propose a novel multi-hypothesis deep learning framework explicitly aimed at PET reconstruction application. A convolutional neural network (CNN) synthesises pCTs by minimising a combination of the pixel-wise error between pCT and CT and a novel metric-loss that itself is defined by a CNN and aims to minimise consequent PET residuals. Training is performed on a database of twenty 3D MR/CT/PET brain image pairs. Quantitative results on a fully independent dataset of twenty-three 3D MR/CT/PET image pairs show that the network is able to synthesise more accurate pCTs. The Mean Absolute Error on the pCT (110.98HU ± 19.22HU) compared to a baseline CNN (172.12HU ± 19.61HU) and a multi-atlas propagation approach (153.40HU ± 18.68HU), and subsequently lead to a significant improvement in the PET reconstruction error (4.74% ± 1.52% compared to baseline 13.72% ± 2.48% and multi-atlas propagation 6.68% ± 2.06%).

Comparison of pre- and post-contrast-enhanced attenuation correction using a CAIPI-accelerated T1-weighted Dixon 3D-VIBE sequence in 68Ga-DOTATOC PET/MRI

ObjectivesTo investigate the influence of contrast agent administration on attenuation correction (AC) based on a CAIPIRINHA (CAIPI)-accelerated T1-weighted Dixon 3D-VIBE sequence in 68Ga-DOTATOC PET/MRI. Material and methodsFifty-one patients with neuroendocrine tumors underwent whole-body 68Ga-DOTATOC PET/MRI for tumor staging. Two PET reconstructions were performed using AC-maps that were created using a high-resolution CAIPI-accelerated Dixon-VIBE sequence with an additional bone atlas and truncation correction using the HUGE (B0 homogenization using gradient enhancement) method before and after application of Gadolinium (Gd)-based contrast agent. Standardized uptake values (SUVs) of 21 volumes of interest (VOIs) were compared between in both PET data sets per patient. A student’s t-test for paired samples was performed to test for potential differences between both AC-maps and both reconstructed PET data sets. Bonferroni correction was performed to prevent α-error accumulation, p < 0.0024 was considered to indicate statistical significance. ResultsSignificant quantitative differences between SUVmax were found in the perirenal fat (19.65 ± 48.03 %, p < 0.0001), in the axillary fat (17.46 ± 63.67 %, p < 0.0001) and in the dorsal subcutaneous fat on level of lumbar vertebral body L4 (10.26 ± 25.29 %, p < 0.0001). Significant differences were also evident in the lungs apical (5.80 ± 10.53 %, p < 0.0001), dorsal at the level of the pulmonary trunk (15.04 ± 19.09 %, p < 0.0001) and dorsal in the basal lung (51.27 ± 147.61 %, p < 0.0001). ConclusionThe administration of (Gd)-contrast agents in this study has shown a considerable influence on the AC-maps in PET/MRI and, consequently impacted quantification in the reconstructed PET data. Therefore, dedicated PET/MRI staging protocols have to be adjusted so that AC-map acquisition is performed prior to contrast agent administration.

Validation of CARE kV automated tube voltage selection for PET-CT: PET quantification and CT radiation dose reduction in phantoms

BackgroundApplied tube voltage (kilovolts, kV) and tube current (milliampere seconds, mAs) affect CT radiation dose and image quality and should be optimised for the individual patient. CARE kV determines the kV and mAs providing the lowest dose to the patient, whilst maintaining user-defined reference image quality. Given that kV changes affect CT values which are used to obtain attenuation maps, the aim was to evaluate the effect of kV changes on PET quantification and CT radiation dose using phantoms.MethodFour phantoms (‘Lungman’, ‘Lungman plus fat’, ‘Esser’ and ‘NEMA image quality’ (NEMA IQ)) containing F-18 sources underwent 1 PET and 5 CT scans, with CARE kV on (automatic kV selection and mAs modulation) and in semi mode with specified tube voltages of 140, 120, 100 and 80 kV (mAs modulation only). A CARE kV image quality reference of 120 kV/50 mAs was used. Impact on PET quantification was determined by comparing measured activity concentrations for PET reconstructions from different CT scans with the reconstruction using the 120 kV reference, and dose (DLP, CTDIvol) differences calculated by comparing doses from all kV settings with the 120 kV reference.ResultsCARE kV-determined optimal tube voltage and CARE kV ‘on’ dose (DLP) savings compared with the 120 kV reference were: Lungman, 100 kV, 2.0%; Lungman plus fat, 120 kV, 0%; Esser, 100 kV, 9.3%; NEMA IQ, 100 kV, 3.4%. Using tube voltages in CARE kV ‘semi’ mode which were not advised by CARE kV ‘on’ resulted in dose increases ≤ 65% compared with the 120 kV reference (greatest difference Lungman plus fat, 80 kV). Clinically insignificant differences in PET activity quantification of up to 0.7% (Lungman, 100 kV, mean measured activity concentration) were observed when using the optimal tube voltage advised by CARE kV. Differences in PET quantification of up to 4.0% (Lungman, 140 kV, maximum measured activity concentration) were found over the full selection of tube voltages in semi mode, with the greatest differences seen at the most suboptimal kV for each phantom. However, most differences were within 1%.ConclusionsCARE kV on can provide CT radiation dose savings without concern over changes in PET quantification.

Clinical Evaluation of MR-Gated Respiratory Motion Correction in Simultaneous PET/MRI.

The recently available gated T1-weighted imaging with the Dixon technique enables the synchronized gating signal for both MR acquisition and PET reconstruction. Herein, we evaluated the clinical value of this MR-gated PET reconstruction in the thoracic-abdominal PET/MRI compared with non-MR-gated method. Twenty patients (28 hypermetabolic target lesions) underwent PET/MRI. Four types of PET images were reconstructed: non-MR-gating + gated attenuation correction (AC) (group A), MR-gating + gated AC (group B), non-MR-gating + breath-hold (BH) AC (group C), and MR-gating + BH AC (group D). A 4-point objective scale (from well match to obvious mismatch was scored from 3 to 0) was proposed to evaluate the mismatch. The detection rate and quantitative metrics were also evaluated. In the patient-based analysis, for groups A through D, the detection rates were 90%, 100%, 85%, and 90% as well as 95%, 100%, 85%, and 85%, assessed by readers 1 and 2, respectively, and significant difference of mismatch score was observed with the highest proportion of 3 points in group B (85%, 90%, 35%, and 40%, and 80%, 90%, 35%, and 20%, assessed by readers 1 and 2, respectively). The lesion-based analysis demonstrated significant differences in quantitative metrics for groups A through D (all P's < 0.05), with the highest quantitative metrics in group B (SUVmax: 7.49 ± 3.37, 8.45 ± 3.82, 6.90 ± 3.24, and 7.69 ± 3.50; SUVmean: 3.90 ± 1.60, 4.34 ± 1.84, 3.67 ± 1.61, and 4.03 ± 1.81; SUVpeak: 5.60 ± 2.50, 6.10 ± 2.80, 5.22 ± 2.40, and 5.65 ± 2.68; signal-to-noise ratio: 136.06 ± 90.58, 136.24 ± 81.63, 99.52 ± 53.16, and 107.57 ± 69.05). The MR-gated reconstruction using gated AC reduced the mismatch between MR and PET images and improved the thoracic-abdominal PET image quality in simultaneous PET/MRI systems.

Stress-associated neurobiological activity is linked with acute plaque instability via enhanced macrophage activity: a prospective serial 18F-FDG-PET/CT imaging assessment.

Emotional stress is associated with future cardiovascular events. However, the mechanistic linkage of brain emotional neural activity with acute plaque instability is not fully elucidated. We aimed to prospectively estimate the relationship between brain amygdalar activity (AmygA), arterial inflammation (AI), and macrophage haematopoiesis (HEMA) in acute myocardial infarction (AMI) as compared with controls. 18F-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG-PET/CT) imaging was performed within 45 days of the index episode in 62 patients (45 with AMI, mean 60.0 years, 84.4% male; 17 controls, mean 59.6 years, 76.4% male). In 10 patients of the AMI group, serial 18F-FDG-PET/CT imaging was performed after 6 months to estimate the temporal changes. The signals were compared using a customized 3D-rendered PET reconstruction. AmygA [target-to-background ratio (TBR), mean ± standard deviation: 0.65 ± 0.05 vs. 0.60 ± 0.05; P = 0.004], carotid AI (TBR: 2.04 ± 0.39 vs. 1.81 ± 0.25; P = 0.026), and HEMA (TBR: 2.60 ± 0.38 vs. 2.22 ± 0.28; P < 0.001) were significantly higher in AMI patients compared with controls. AmygA correlated significantly with those of the carotid artery (r = 0.350; P = 0.005), aorta (r = 0.471; P < 0.001), and bone marrow (r = 0.356; P = 0.005). Psychological stress scales (PHQ-9 and PSS-10) and AmygA assessed by PET/CT imaging correlated well (P < 0.001). Six-month after AMI, AmygA, carotid AI, and HEMA decreased to a level comparable with the controls. AmygA, AI, and HEMA were concordantly enhanced in patients with AMI, showing concurrent dynamic changes over time. These results raise the possibility that stress-associated neurobiological activity is linked with acute plaque instability via augmented macrophage activity and could be a potential therapeutic target for plaque inflammation in AMI.

Quantitative bone SPECT/CT reconstruction utilizing anatomical information

BackgroundBone SPECT/CT has been shown to offer superior sensitivity and specificity compared to conventional whole-body planar scanning. Furthermore, bone SPECT/CT allows quantitative imaging, which is challenging with planar methods. In order to gain better quantitative accuracy, Bayesian reconstruction algorithms, including both image derived and anatomically guided priors, have been utilized in reconstruction in PET/CT scanning, but they have not been widely used in SPECT/CT studies. Therefore, the aim of this work was to evaluate the performance of CT-guided reconstruction in quantitative bone SPECT.MethodsThree Bayesian reconstruction methods were evaluated against the conventional ordered subsets expectation maximization (OSEM) reconstruction method. One of the studied Bayesian methods was the relative difference prior (RDP), which has recently gained popularity in PET reconstruction. The other two methods, anatomically guided smoothing prior (AMAP-S) and anatomically guided relative difference prior (AMAP-R), utilized anatomical information from the CT scan. The reconstruction methods were evaluated in terms of quantitative accuracy with artificial lesions inserted in clinical patient studies and with 20 real clinical patients. Maximum and mean standardized uptake values (SUVs) of the lesions were defined.ResultsThe analyses showed that all studied Bayesian methods performed better than OSEM and the anatomical priors also outperformed RDP. The average relative error in mean SUV for the artificial lesion study for OSEM, RDP, AMAP-S, and AMAP-R was − 53%, − 35%, − 15%, and − 10%, when the CT study had matching lesions. In the patient study, the RDP method gave 16 ± 9% higher maximum SUV values than OSEM, while AMAP-S and AMAP-R offered increases of 36 ± 8% and 36 ± 9%, respectively. Mean SUV increased for RDP, AMAP-S, and AMAP-R by 18 ± 9%, 26 ± 5%, and 33 ± 5% when compared to OSEM.ConclusionsThe Bayesian methods with anatomical prior, especially the relative difference prior-based method (AMAP-R), outperformed OSEM and reconstruction without anatomical prior in terms of quantitative accuracy.

Representation Disentanglement for Multi-modal Brain MRI Analysis.

Multi-modal MRIs are widely used in neuroimaging applications since different MR sequences provide complementary information about brain structures. Recent works have suggested that multi-modal deep learning analysis can benefit from explicitly disentangling anatomical (shape) and modality (appearance) information into separate image presentations. In this work, we challenge mainstream strategies by showing that they do not naturally lead to representation disentanglement both in theory and in practice. To address this issue, we propose a margin loss that regularizes the similarity in relationships of the representations across subjects and modalities. To enable robust training, we further use a conditional convolution to design a single model for encoding images of all modalities. Lastly, we propose a fusion function to combine the disentangled anatomical representations as a set of modality-invariant features for downstream tasks. We evaluate the proposed method on three multi-modal neuroimaging datasets. Experiments show that our proposed method can achieve superior disentangled representations compared to existing disentanglement strategies. Results also indicate that the fused anatomical representation has potential in the downstream task of zero-dose PET reconstruction and brain tumor segmentation.

Impact of MR‐guided PET reconstruction on tau detection and quantification with [18F]‐MK‐6240

AbstractBackgroundTau PET offers the potential of in vivo imaging neurofibrillary tangles (NFT) in the brain to assess Alzheimer’s disease (AD) progression and help guide the development of disease‐modifying treatments. In this study, we evaluate visualization and quantification of NFT using simultaneous acquisition of MRI and PET using [18F]‐MK‐6240 [1]. To overcome limitations in PET spatial resolution resulting in partial volume effects (PVE), MRI‐guided PET reconstruction using an asymmetrical Bowsher prior [2,3] was tested.MethodCognitively normal volunteers from the NYU Center Sleep and Brain Health Center (n=18, 15 female, age=65.4±5.6) underwent examination on a 3‐T integrated PET‐MRI system (Biograph mMR (Siemens Healthcare, Erlangen, Germany). MRI and PET were performed simultaneously; MRI contrasts included MPRAGE (1‐mm isotropic) for anatomy and ultrashort echo‐time (UTE) (1.6‐mm isotropic) for attenuation correction. [18F]‐MK‐6240 was injected intravenously and PET data was reconstructed 70‐90 min post‐injection. Standard iterative reconstruction (OSEM) (2‐mm isotropic) was compared to asymmetrical Bowsher [2,3] reconstruction with MPRAGE prior regularization weight β=30 (1x1x2mm3). PET standard uptake value (SUV) maps were co‐registered to respective MPRAGE, and (cerebellum or pons normalized) SUVR values were extracted from FreeSurfer‐derived regions of interest (ROIs).ResultsExample SUVR maps of standard PET and MR‐guided PET (Fig.1) reveal the Bowsher reconstruction to have more localized and increased focal uptake in the amygdala and cortical regions including the entorhinal, posterior cingulate and precuneus, with regional SUVR‐increases of 39%, 53%, 32%, and 16% respectively. Over all subjects (Fig.2), Bowsher reconstruction results in 30% or 48% higher SUVR, and an 64% or 84%‐increase of dynamic SUVR‐range (Fig.3) when normalized versus cerebellum or pons, respectively. These trends are emphasized for entorhinal cortex (Fig.4).ConclusionsThese initial results demonstrate that MR‐guided PET reconstruction of tau data may improve NFT localization and quantification. Study limitations include small number of high uptake tau‐cases and lack of standard such as histology for comparison. Future work will compare against CSF tau‐markers, cognition and comparison to other PVE correction methods.

Non-local mean denoising using multiple PET reconstructions

ObjectivesNon-local mean (NLM) filtering has been broadly used for denoising of natural and medical images. The NLM filter relies on the redundant information, in the form of repeated patterns/textures, in the target image to discriminate the underlying structures/signals from noise. In PET (or SPECT) imaging, the raw data could be reconstructed using different parameters and settings, leading to different representations of the target image, which contain highly similar structures/signals to the target image contaminated with different noise levels (or properties). In this light, multiple-reconstruction NLM filtering (MR-NLM) is proposed, which relies on the redundant information provided by the different reconstructions of the same PET data (referred to as auxiliary images) to conduct the denoising process.MethodsImplementation of the MR-NLM approach involved the use of twelve auxiliary PET images (in addition to the target image) reconstructed using the same iterative reconstruction algorithm with different numbers of iterations and subsets. For each target voxel, the patches of voxels at the same location are extracted from the auxiliary PET images based on which the NLM denoising process is conducted. Through this, the exhaustive search scheme performed in the conventional NLM method to find similar patches of voxels is bypassed. The performance evaluation of the MR-NLM filter was carried out against the conventional NLM, Gaussian and bilateral post-reconstruction approaches using the experimental Jaszczak phantom and 25 whole-body PET/CT clinical studies.ResultsThe signal-to-noise ratio (SNR) in the experimental Jaszczak phantom study improved from 25.1 when using Gaussian filtering to 27.9 and 28.8 when the conventional NLM and MR-NLM methods were applied (p value < 0.05), respectively. Conversely, the Gaussian filter led to quantification bias of 35.4%, while NLM and MR-NLM approaches resulted in a bias of 32.0% and 31.1% (p value < 0.05), respectively. The clinical studies further confirm the superior performance of the MR-NLM method, wherein the quantitative bias measured in malignant lesions (hot spots) decreased from − 12.3 ± 2.3% when using the Gaussian filter to − 3.5 ± 1.3% and − 2.2 ± 1.2% when using the NLM and MR-NLM approaches (p value < 0.05), respectively.ConclusionThe MR-NLM approach exhibited promising performance in terms of noise suppression and signal preservation for PET images, thus translating into higher SNR compared to the conventional NLM approach. Despite the promising performance of the MR-NLM approach, the additional computational burden owing to the requirement of multiple PET reconstruction still needs to be addressed.

Can alternative PET reconstruction schemes improve the prognostic value of radiomic features in non-small cell lung cancer?

To evaluate the potential benefit of using alternative reconstruction schemes of PET images for the prognostic value of radiomic features. Patients (n=91) with non-small cell lung cancer were prospectively included. All had a PET/CT examination before treatment. Three different PET images were reconstructed for each patient: the standard clinical protocol (i.e., 4×4×4 mm3 voxels, 5mm Gaussian filter, denoted '200G5'), as well as using smaller voxels (i.e., 2×2×2 mm3 with a larger reconstruction matrix, denoted 400G1) and/or 1mm post-reconstruction Gaussian filter, denoted 200G1). Metabolic volumes of the primary tumors were semi-automatically delineated on the PET images and IBSI compliant radiomic features (intensity, shape, textural) were extracted. First, the distributions of 200G1 and 400G1 features were compared to the reference clinical protocol (200G5) through Bland-Altman tests and the use of linear mixed models. Then, the prognostic value of the features from each of the 3 reconstructions was evaluated in a univariate analysis, through their stratification power in Kaplan-Meier curves through a threshold set at the median. The 3 reconstructions led to different distributions for most of the features. The larger shifts and standard deviations of differences was observed between 200G5 and 400G1, which was also confirmed through linear mixed models. However, these relatively important differences in distributions did not translate into a significant impact on the stratification power of the features in terms of prognosis, although a trend in decreasing prognostic value could be observed (smaller number of features with HR above 2, overall lower HR values). Most prognostic features displayed high correlation with either volume or SUVmax, although there was great variability of prognostic value for similar levels of correlation with these basic metrics. Using smaller voxels or less strong filtering options in the reconstruction settings of PET images compared to the standard clinical protocols led to different distributions of the resulting radiomic features. However, the hierarchy between patients according to these distributions remained overall the same and therefore the resulting stratification power of the radiomic features was not significantly altered. These results should be compared to those obtained in the context of other pathologies where radiomic features displaying lower correlation with volume or SUVmax may have predictive value, such as in cervical cancer.

Whole-body 18-F-FDG-PET in patients with leptomeningeal disease and correlation with MRI.

Studies evaluating leptomeningeal disease on whole-body 18F-FDG PET are lacking. The purpose was to evaluate PET imaging of leptomeningeal disease and investigate the incremental utility of newer PET reconstructions in leptomeningeal disease. PET imaging of 56 patients with leptomeningeal disease detected initially on MRI (n = 53) or cytopathology (n = 35) were retrospectively reviewed. Regular 3-dimensional iterative reconstruction (3D IR, n = 56) and advanced reconstruction (AdvRecon, n = 41) PET images were evaluated by readers blinded to clinical and MRI findings for uptake involving cauda equina, posterior fossa and spinal cord. Spinal cord uptake pattern was classified as normal (uptake < liver), uptake = liver, conus uptake > liver, conus and cervical cord uptake > liver and multifocal/diffuse uptake > liver. SUVmax ratios of conus/liver, conus/left atrium and conus/cervical cord were compared between 3D IR and AdvRecon datasets. Cauda equina uptake was seen in 64% and 78% on 3D IR and AdvRecon; posterior fossa uptake was seen in 52% and 54% on 3D IR and AdvRecon, respectively. Twelve percent had cauda equina or posterior fossa uptake visible only on AdvRecon. On 3D IR, normal spinal cord uptake was most common (27%); on AdvRecon, conus and cervical cord uptake > liver was most common (32%). Seven of 11 patients with normal spinal cord uptake on 3D IR were upgraded to increased uptake on AdvRecon. AdvRecon showed significantly higher conus/liver, conus/blood pool and conus/cervical cord SUVmax ratios (P < 0.0001). Abnormal uptake in cauda equina, posterior fossa and spinal cord uptake are visible on FDG PET in leptomeningeal disease with increased conspicuity advanced PET reconstructions.

Influence of X-ray computed tomography (CT) exposure and reconstruction parameters on positron emission tomography (PET) quantitation

BackgroundThe CT of PET CT provides diagnostic information, anatomic localisation and attenuation correction (AC). When only AC is required, very lose dose CT is desirable. CT iterative reconstruction (IR) improves image quality with lower exposures however there is little data on very low dose IR CT for AC of PET. This work assesses the impact of CT exposure and reconstruction algorithm on PET voxel values.MethodAn anthropomorphic torso phantom was filled with physiologically typical [18]F concentrations in heart, liver and background compartments. A 17-mm-diameter right lung “tumour” filled with [18]F was included (surrounding lung contained no 18[F]). PET was acquired followed by 24 CT acquisitions with varying CT exposures (15–50 mAs, 80–120 kVp, pitch 0.671 or 0.828). Each CT was reconstructed twice using filtered back projection (FBP) or IR and these used for AC of PET. The reference PET reconstruction (RR) used CT acquired at 50 mAs, 120 kVp, pitch 0.828, IR, all others were test PET reconstructions (TR). Regions of interest (ROIs) were drawn in the liver, soft tissue and over “tumour” on each TR and compared with the RR. Voxel values in each TR were compared to the RR using a paired t test and by calculating which and what proportion of voxels in each TR differed by a quantitatively significant difference (QSD) from the RR.ResultsTRs reconstructed using lower dose CTs underestimated mean and maximum ROI activity relative to the RR; greater with IR than FBP. Once CT dose index (CTDI) increased to 1 mGy, differences were less than QSD. On voxel analysis, all TRs were significantly different to the RR (p < 0.0001). TRs reconstructed at the lowest CT exposure with IR had 6% of voxels that differed by greater than QSD. Differences were reduced with increasing CTDI and FBP reconstruction. Voxels which exceeded the QSD were spatially localised to regions of high activity, interfaces between different attenuation and areas of CT beam hardening.ConclusionsVery low dose CT exposures are feasible for accurate PET AC. Scanner- and reconstruction-specific validation should be employed prior very low dose CT AC for PET.

Approximating anatomically-guided PET reconstruction in image space using a convolutional neural network

In the last two decades, it has been shown that anatomically-guided PET reconstruction can lead to improved bias-noise characteristics in brain PET imaging. However, despite promising results in simulations and first studies, anatomically-guided PET reconstructions are not yet available for use in routine clinical because of several reasons. In light of this, we investigate whether the improvements of anatomically-guided PET reconstruction methods can be achieved entirely in the image domain with a convolutional neural network (CNN). An entirely image-based CNN post-reconstruction approach has the advantage that no access to PET raw data is needed and, moreover, that the prediction times of trained CNNs are extremely fast on state of the art GPUs which will substantially facilitate the evaluation, fine-tuning and application of anatomically-guided PET reconstruction in real-world clinical settings.In this work, we demonstrate that anatomically-guided PET reconstruction using the asymmetric Bowsher prior can be well-approximated by a purely shift-invariant convolutional neural network in image space allowing the generation of anatomically-guided PET images in almost real-time. We show that by applying dedicated data augmentation techniques in the training phase, in which 16 [18F]FDG and 10 [18F]PE2I data sets were used, lead to a CNN that is robust against the used PET tracer, the noise level of the input PET images and the input MRI contrast. A detailed analysis of our CNN in 36 [18F]FDG, 18 [18F]PE2I, and 7 [18F]FET test data sets demonstrates that the image quality of our trained CNN is very close to the one of the target reconstructions in terms of regional mean recovery and regional structural similarity.

Impact of the Q.Clear reconstruction algorithm on the interpretation of PET/CT images in patients with lymphoma

BackgroundQ.Clear is a new Bayesian penalized-likelihood PET reconstruction algorithm. It has been documented that Q.Clear increases the SUVmax values of different malignant lesions.PurposeSUVmax values are crucial for the interpretation of PET/CT images in patients with lymphoma, particularly when the early and final responses to treatment are evaluated. The aim of the study was to systematically analyse the impact of the use of Q.Clear on the interpretation of PET/CT in patients with lymphoma.MethodsA total of 280 18F-FDG PET/CT scans in patients with lymphoma were performed for staging (sPET), for early treatment response (iPET), after the end of treatment (ePET) and when a relapse of lymphoma was suspected (rPET). Scans were separately reconstructed with two algorithms, Q.Clear and OSEM, and further compared.ResultsThe stage of lymphoma was concordantly diagnosed in 69/70 patients with both algorithms on sPET. Discordant assessment of the Deauville score (p < 0.001) was found in 11 cases (15.7%) of 70 iPET scans and in 11 cases of 70 ePET scans. An upgrade from a negative to a positive scan by Q.Clear occurred in 3 cases (4.3%) of iPET scans and 7 cases (10.0%) of ePET scans. The results of all 70 rPET scans were concordant. The SUVmax values of the target lymphoma lesions measured with Q.Clear were higher than those measured with OSEM in 88.8% of scans.ConclusionAlthough the Q.Clear algorithm may alter the interpretations of PET/CT in only a small proportion of patients, we recommend using standard OSEM reconstruction for the assessment of treatment response.

Detection of sub-centimeter lesions using digital TOF-PET/CT system combined with Bayesian penalized likelihood reconstruction algorithm.

Many advances in PET/CT technology can potentially improve image quality and the ability to detect small lesions. A new digital TOF-PET/CT scanner based on silicon photomultipliers (SiPM) integrated with a Bayesian penalized likelihood (BPL) PET reconstruction algorithm (Q.Clear; GE Healthcare) has been introduced into clinical practice. The present study aimed to quantify the ability of a digital TOF-PET/CT scanner combined with BPL reconstruction to detect small lesions, and to determine the optimal penalization factor (β) in BPL to accurately detect such lesions. All PET data were acquired from a NEMA body phantom using a Discovery MI (DMI) PET/CT system (GE Healthcare). The phantom included six spheres with diameters of 4, 5, 6, 8, 10, and 13mm, and contained a background activity level of 5.3kBq/mL, with target-to-background ratios (TBR) of 4:1 and 8:1. Images were reconstructed using a baseline OSEM algorithm, with OSEM + PSF, OSEM + TOF, OSEM + PSF + TOF, and BPL + PSF + TOF (β: 50-400). The matrix size was 192 × 192 and 384 × 384. Data acquired in 100-min list mode were re-binned into acquisition times ranging from 2 to 100min. The quantitative accuracy and detectability of small hot spheres were evaluated by physical assessment of a recovery coefficient (RC) and a detectability index (DI), as well as visual assessment of PET images at each acquisition time. The RC and DI of sub-centimeter spheres were improved, because the digital TOF-PET/CT scanner has a larger TOF performance gain due to better timing resolution. The RC and DI were higher with BPL in sub-centimeter spheres, than with other OSEM-based types of reconstruction. The BPL for an 8-mm sphere overestimated uptake due to edge artifact overshoot induced by PSF modeling. The variability of RC and DI for acquisition times and TBR differed considerably according to β values. The RC for ~ 8-mm spheres were > 1 at β values between 50 and 100, but were close to 1 at β value of 200. The visual scores for β = 200 in BPL were maximal, whereas those for spheres that were ≥ 6mm exceeded the criterion of 3. The BPL in the digital TOF-PET/CT scanner improved the quantitation and detectability of sub-centimeter spheres compared with OSEM-based reconstruction. Optimization of the β value in BPL might allow the detection of lesions ≤ 6mm, although detectability depended on the TBR of lesions. A β value of 200 seemed optimal for detecting sub-centimeter lesions.

Coronary artery 18F-NaF PET analysis with the use of an elastic motion correction software

Introduction18F-Sodium Fluoride Positron Emission Tomography (18F-NaF PET) is a novel molecular imaging modality with promise for use as a risk stratification tool in cardiovascular disease. There are limitations in the analysis of small and rapidly moving coronary arteries using traditional PET technology. We aimed to validate the use of a motion correction algorithm (eMoco) on coronary 18F-NaF PET outcome parameters. MethodsPatients admitted with an acute coronary syndrome underwent 18F-NaF PET and computed tomography coronary angiography. 18F-NaF PET data were analyzed using a diastolic reconstruction, an ungated reconstruction and the eMoco reconstruction. ResultsTwenty patients underwent 18F-NaF PET imaging and 17 patients had at least one positive lesion that could be used to compare PET reconstruction datasets. eMoco improved noise (the coefficient of variation of the blood pool radiotracer activity) compared to the diastolic dataset (0.09 [0.07 to 0.12] vs 0.14[0.11 to 0.17], p < .001) and marginally improved coronary lesion maximum tissue-to-background ratios compared to the ungated dataset (1.33 [1.05 to 1.48]vs 1.29 [1.04 to 1.40], p = .011). ConclusionIn this pilot dataset, the eMoco reconstruction algorithm for motion correction appears to have potential in improving coronary analysis of 18F-NaF PET by reducing noise and increasing maximum counts. Further testing in a larger patient dataset is warranted.

Improved Low-Count Quantitative PET Reconstruction With an Iterative Neural Network.

Image reconstruction in low-count PET is particularly challenging because gammas from natural radioactivity in Lu-based crystals cause high random fractions that lower the measurement signal-to-noise-ratio (SNR). In model-based image reconstruction (MBIR), using more iterations of an unregularized method may increase the noise, so incorporating regularization into the image reconstruction is desirable to control the noise. New regularization methods based on learned convolutional operators are emerging in MBIR. We modify the architecture of an iterative neural network, BCD-Net, for PET MBIR, and demonstrate the efficacy of the trained BCD-Net using XCAT phantom data that simulates the low true coincidence count-rates with high random fractions typical for Y-90 PET patient imaging after Y-90 microsphere radioembolization. Numerical results show that the proposed BCD-Net significantly improves CNR and RMSE of the reconstructed images compared to MBIR methods using non-trained regularizers, total variation (TV) and non-local means (NLM). Moreover, BCD-Net successfully generalizes to test data that differs from the training data. Improvements were also demonstrated for the clinically relevant phantom measurement data where we used training and testing datasets having very different activity distributions and count-levels.