Off-line motion correction methods for multi-frame PET data

The impact of a method for MR-based respiratory motion correction of PET data on lesion visibility and quantification in patients with oncologic findings in the lung was evaluated. Twenty patients with one or more lesions in the lung were included. Hybrid imaging was performed on an integrated PET/MR system using 18F-FDG as radiotracer. The standard thoracic imaging protocol was extended by a free-breathing self-gated acquisition of MR data for motion modelling. PET data was acquired simultaneously in list-mode for 5-10 mins. One experienced radiologist and one experienced nuclear medicine specialist evaluated and compared the post-processed data in consensus regarding lesion visibility (scores 1–4, 4 being best), image noise levels (scores 1–3, 3 being lowest noise), SUVmean and SUVmax. Motion-corrected (MoCo) images were additionally compared with gated images. Non-motion-corrected free-breathing data served as standard of reference in this study. Motion correction generally improved lesion visibility (3.19 ± 0.63) and noise ratings (2.95 ± 0.22) compared to uncorrected (2.81 ± 0.66 and 2.95 ± 0.22, respectively) or gated PET data (2.47 ± 0.93 and 1.30 ± 0.47, respectively). Furthermore, SUVs (mean and max) were compared for all methods to estimate their respective impact on the quantification. Deviations of SUVmax were smallest between the uncorrected and the MoCo lesion data (average increase of 9.1% of MoCo SUVs), while SUVmean agreed best for gated and MoCo reconstructions (MoCo SUVs increased by 1.2%). The studied method for MR-based respiratory motion correction of PET data combines increased lesion sharpness and improved lesion activity quantification with high signal-to-noise ratio in a clinical setting. In particular, the detection of small lesions in moving organs such as the lung and liver may thus be facilitated. These advantages justify the extension of the PET/MR imaging protocol by 5–10 minutes for motion correction.

Subject motion in positron emission tomography (PET) studies leads to image blurring and artifacts; simultaneously acquired magnetic resonance imaging (MRI) data provides a means for motion correction (MC) in integrated PET/MRI scanners. To assess the effect of realistic head motion and MR-based MC on static [18 F]-fluorodeoxyglucose (FDG) PET images in dementia patients. Observational study. Thirty dementia subjects were recruited. 3T hybrid PET/MR scanner where EPI-based and T1 -weighted sequences were acquired simultaneously with the PET data. Head motion parameters estimated from high temporal resolution MR volumes were used for PET MC. The MR-based MC method was compared to PET frame-based MC methods in which motion parameters were estimated by coregistering 5-minute frames before and after accounting for the attenuation-emission mismatch. The relative changes in standardized uptake value ratios (SUVRs) between the PET volumes processed with the various MC methods, without MC, and the PET volumes with simulated motion were compared in relevant brain regions. The absolute value of the regional SUVR relative change was assessed with pairwise paired t-tests testing at the P = 0.05 level, comparing the values obtained through different MR-based MC processing methods as well as across different motion groups. The intraregion voxelwise variability of regional SUVRs obtained through different MR-based MC processing methods was also assessed with pairwise paired t-tests testing at the P = 0.05 level. MC had a greater impact on PET data quantification in subjects with larger amplitude motion (higher than 18% in the medial orbitofrontal cortex) and greater changes were generally observed for the MR-based MC method compared to the frame-based methods. Furthermore, a mean relative change of ∼4% was observed after MC even at the group level, suggesting the importance of routinely applying this correction. The intraregion voxelwise variability of regional SUVRs was also decreased using MR-based MC. All comparisons were significant at the P = 0.05 level. Incorporating temporally correlated MR data to account for intraframe motion has a positive impact on the FDG PET image quality and data quantification in dementia patients. 3 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2018;47:1288-1296.

BackgroundRespiratory motion in PET/CT leads to well-known image degrading effects commonly compensated using elastic motion correction approaches. Gate-to-gate motion correction techniques are promising tools for improving clinical PET data but suffer from relatively long reconstruction times. In this study, the performance of a fast elastic motion compensation approach based on motion deblurring (DEB-MC) was evaluated on patient and phantom data and compared to an EM-based fully 3D gate-to-gate motion correction method (G2G-MC) which was considered the gold standard.MethodsTwenty-eight patients were included in this study with suspected or confirmed malignancies in the thorax or abdomen. All patients underwent whole-body [18F]FDG PET/CT examinations applying hardware-based respiratory gating. In addition, a dynamic anthropomorphic thorax phantom was studied with PET/CT simulating tumour motion under controlled but realistic conditions. PET signal recovery values were calculated from phantom scans by comparing lesion activities after motion correction to static ground truth data. Differences in standardized uptake values (SUV) and metabolic volume (MV) between both reconstruction methods as well as between motion-corrected (MC) and non motion-corrected (NOMC) results were statistically analyzed using a Wilcoxon signed-rank test.ResultsPhantom data analysis showed high lesion recovery values of 91% (2 cm motion) and 98% (1 cm) for G2G-MC and 83% (2 cm) and 90% (1 cm) for DEB-MC. The statistical analysis of patient data found significant differences between NOMC and MC reconstructions for SUV max, SUV mean, MV, and contrast-to-noise ratio (CNR) for both reconstruction algorithms. Furthermore, both methods showed similar increases of 11–12% in SUV max and SUV mean after MC. The statistical analysis of the MC/NOMC ratio found no significant differences between the methods.ConclusionBoth motion correction techniques deliver comparable improvements of SUV max, SUV mean, and CNR after MC on clinical and phantom data. The fast elastic motion compensation technique DEB-MC may thereby be a valuable alternative to state-of-the art motion correction techniques.

Recently published methods for motion correction in neurological PET include the multiple acquisition frame (MAF) and LOR rebinning methods. The aim of the present work was to compare the accuracy of reconstructions obtained with these methods when multiple, arbitrary movements were applied to a Hoffman brain phantom during 3D list mode acquisition. A reflective target attached to the phantom enabled a Polaris optical motion tracking system to monitor the phantom position and orientation in the scanner coordinate frame. The motion information was used in the motion correction algorithms. The MAF method was applied to the list-mode data after sorting them into a series of dynamic frames, while the LOR rebinning method was applied directly to the list-mode data. A proportion of the list mode events had to be discarded during rebinning because the application of the corrective spatial transformation removed them from the 3D projection space. A correction for these 'lost' events was implemented as a global post-reconstruction scale factor, based on the overall fraction of lost events. Reconstructions from both motion correction methods were compared with a motion-free reference scan of the same phantom. Motion correction produced a marked improvement in image clarity and reduced errors with respect to the reference scan. LOR rebinning with global loss correction was found to be more accurate than the MAF method

Motion of the head during brain positron emission tomography (PET) acquisitions has been identified as a source of artifact in the reconstructed image. A number of techniques have been proposed to correct for this motion artifact, but they are unable to correct for a motion during an acquisition. The aim of this study was to develop a sinogram-based motion correction (SBMC) technique to correct directly the head motion during a PET scan using a motion tracking system and list-mode data acquisition. This method uses a rebinning procedure whereby the lines of response (LOR) are geometrically transformed according to the current values of six-dimensional motion data. A Michelogram was recomposed using the rebinned LOR, and the motion-corrected sinogram was generated. In the motion corrected image, the blurring artifact due to the motion was reduced by the SBMC technique. This technique was applied to actual PET data acquired in the list-mode, and demonstrated the potential for real-time motion correction of head movements during a PET acquisition.

Advances in PET technology have led to improved resolution capability in modern PET scanners, andhence the impact of motion has become increasingly significant in order to produce high resolutionPET images. This thesis aims to advance the utility of motion correction in combined PET and MR byimproving the capabilities of motion correction, and by easing the effort required in order to applythem. In particular, PET-based data-driven motion tracking methods, which are able to track motionwithout requiring additional hardware, are investigated and applied to the head. First, existing techniques for motion correction, especially in the context of PET/MR, are reviewedand a general framework for PET motion correction is presented. The framework provides a top-levelperspective on PET motion correction, simplifying the process of comparing and contrasting variousmotion correction techniques. Next, the implementation of a patient-based orientation system withinthe open-source PET reconstruction packages the Software for Tomographic Image Reconstruction(STIR) and the Synergistic Image Reconstruction Software (SIRF) is described. Additionally, animplementation of the general PET motion correction framework platform is introduced. To investigate current technological limitations in combined PET and MR motion correction, theefficacy of a commercial PET/MR motion correction product, Brain COMPASS (herein just COMPASSfor brevity), was assessed. An experiment introduced a methodology for measuring motion blur invivo and uses this to measure reduction in motion blur after applying COMPASS. N=10 subjects worea head-mounted apparatus with fiducial markers affixed. COMPASS reduced blur in the directionof motion, measured by the principal component of the full-width at half-maximum (FWHM) ofthe source point-spread-function (PSF) distribution, only when the motion was sufficient to causethat blur to exceed 1.9 mm. However, blur across all directions almost always increased (mean=0.3mm) when COMPASS was used. Results suggested that any error in estimating the patient posewas low and was not the cause of the blur. Increases in overall blur were therefore suspected to becaused by interpolation when resampling transformed reconstructions. A second experiment measuredneocortical standardised uptake value ratio (SUVr) in an 18F-Florbetaben (18F-FBB) study of N=42young, healthy, asymptomatic volunteers with high polygenetic risk of developing Alzheimer’s Disease.No significant changes in neocortical SUVr were observed. This cohort exhibited low motion, andthe neocortex is a large region-of-interest, so is less sensitive to motion corruption than other cohorts.Hence, it was concluded that COMPASS did not aid this investigation, but also does not have adiscernible negative impact despite low motion. During the COMPASS investigation, MR acquisitions were the time-limiting factor of the session,meaning that the MR acquisitions introduced for PET motion correction increased the overall sessiontime. Because MR is not always available for motion tracking, alternative PET-alone motion correctiontechniques were investigated. Several data-driven tracking techniques are available; however, it is notclear which are most appropriate. A phantom investigation was conducted on a Biograph mCT Flowto determine which data-driven head motion tracking techniques were most appropriate. Of the fourmethodologies investigated principal component analysis (PCA) was found to be the most promisingtechnique for PET data-driven motion tracking of head motion. The PET-alone motion correction technique outlined above had two key limitations. Firstly, theyprovide frame-based motion correction with low temporal resolution. However, the type of motionseen in the COMPASS investigation were continuous “drooping” of the head, is not modelled in adiscrete, frame-based scheme. Secondly, the PET reconstructions without attenuation provide noisypose estimates compared to those from MR volume registration. Therefore, a scheme was proposedthat, unlike previous techniques, models head motion directly from motion surrogates. Here, a radialbasis function (RBF) regression relates the endogenous tracking signal to the exogenous rigid poseparameters, which are supplied by temporal-sparse MR acquisitions. A proof-of-concept is presentedin which low temporal resolution MR pose estimates are upsampled. The technique is compared tolinear and nearest-neighbour interpolation, and for the regression, both linear and RBF are investigated.The proposed technique was found to introduce some negligible blurring when no motion is present (<0.1 mm mean-of-maximum displacement (MMD)) but reduce blurring by 0.7 mm MMD when motionis present. This demonstrates the capability of the technique to reduce error in motion estimation byimproving the temporal resolution. In a clinical setting, motion correction is likely to become increasingly important in the future. Thework in this thesis will make motion correction easier to apply in a clinical setting, so that it may beapplied more regularly, providing wider benefit.

Evaluation of functional magnetic resonance imaging (fMRI) as a reliable clinical imaging tool requires accurate assessment and correction of head motion artifacts. As the correction of bulk head motion is vital, the loss of signal strength from the confounding effect of head motion on spin magnetization may be an additional factor in activation analysis error. This study focuses on the evaluation and correction of the spin saturation artifact that occurs when parts of adjacent slices are selected due to changing head positions in single-shot multislice acquisitions. As a consequence of head movement, the acquired slices constituting a fMRI volume are no longer parallel to each other and the spin magnetization in fMRI voxels becomes dependent on head motion history. Motion corrections applying the same rigid motion estimates to all the slices in a volume may not be a reasonable approximation in cases where the magnitude of head motion exceeds a subvoxel range. For realistic ranges of motion in fMRI, an accurate estimate of rigid motion parameters for each echo planar imaging (EPI) slice is essential to correctly register voxel intensities. Previously we have implemented the map-slice-to-volume (MSV) motion correction method that maps each slice in a time series onto a reference anatomical volume, which proved to be effective in improving activation detection. To correctly evaluate the motion dependence of spin magnetization, each voxel is tracked with movement history that is available from MSV motion estimates. Relatively low in resolution, EPI voxels are composed of varying mixtures of white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) and variations in the tissue composition give rise to voxel intensities that are functions of tissue T1 properties. We have developed a weighted-average spin saturation (WASS) correction method that can handle full rigid motion and account for the melange of different brain tissue isochromats at each EPI voxel location. We evaluated the effect of spin saturation artifacts and the performance of the WASS correction using simulated fMRI time series synthesized with known true activation, motion, and the associated spin saturation artifact. Two different ranges of head rotations, [-5,5] and [-2,2] deg, were introduced and the effect of the spin saturation artifact was quantified to show 18% and 13% reduction in activation detection rate, respectively. Following the MSV motion and WASS correction, results indicate that WASS correction can improve activation detection by 17% relative to MSV only correction.

Previously, we developed the method for non-rigid motion correction in the quasi-image space of histo-projections. The TOF locality property is used to locally perform motion correction; that is, the approximation of motion as locally rigid on a scale of TOF resolution. The motion vector field (MFV) is estimated from motion-sampled histo-images, which represents the most likely annihilation locations and is a combination of all histo-projection data. The histo-image is relatively noisy, compared with a full statistical reconstruction image from the same data set. Excessive noise leads to a suboptimal estimation of MVF when a standard optical flow method is applied. Nevertheless, histo-images may be denoised (and deblurred) based on target images of histo-projection data reconstruction, resulting in improved quality of histo-image generation.Deep Learning methodology is used in histo-image denoising, where computationally expensive statistical reconstruction from the same data provides training images. U-Net CNN, found to be effective in denoising PET images, is utilized.Biograph Vision Quadra (Siemens Healthineers) patient data was employed in investigations. This PET/CT scanner of long axial field of view (LAFOV) offers a very good time-of-flight (TOF) resolution of about 215 ps, suitable for quasi-image motion correction techniques. Data-driven motion detection is particularly attractive in LAFOV scanners since the entire patient is visible at once.The initial results showcased that DL histo-image denoising and deblurring improve MFV estimation. This procedure leads to less motion-blurry images in the case of respiratory motion correction, performed entirely in the TOF projection space.

Respiratory motion (RM) and cardiac motion (CM) degrade the quality and resolution in cardiac PET scans. We have developed non-rigid motion estimation methods to estimate both RM and CM based on 4D cardiac gated PET data alone, and compensate the dual respiratory and cardiac (R&C) motions after (MCAR), during (MCDR), and before (MCBR) image reconstruction. In all three R&C motion correction methods, attenuation-activity mismatch effect was modeled by using transformed attenuation maps using the estimated RM. The difference of using activity preserving and non-activity preserving models in R&C correction was also studied. Realistic Monte Carlo simulated 4D cardiac PET data using the 4D XCAT phantom and accurate models of the scanner design parameters and performance characteristics at different noise levels were employed as the known truth and for method development and evaluation. Results from the simulation study suggested that all three dual R&C motion correction methods provide substantial improvement in the quality of 4D cardiac gated PET images as compared with no motion correction. Specifically, the MCDR method yields the best performance for all different noise levels compared with the MCAR and MCBR methods. While MCBR reduces computational time dramatically but the resultant 4D cardiac gated PET images has overall inferior image quality when compared to that from the MCAR and MCDR approaches in the ‘almost’ noise free case. Also, the MCBR method has better noise handling properties when compared with MCAR and provides better quantitative results in high noise cases. When the goal is to reduce scan time or patient radiation dose, MCDR and MCBR provide a good compromise between image quality and computational times.

Patient motion artifacts created in cardiac SPECT imaging can lead to misinterpretation of the images, resulting in false diagnoses. This simulation study proposes a new technique for patient motion correction (MC), where we utilize a modified template projection/reconstruction (TPR) algorithm to perform a voxel-by-voxel correction to the original image. Using NCAT, we developed two female phantoms with large breasts containing a non-beating heart (heart: background=5:1). Phantom 1 had a healthy heart, and phantom 2 had a heart with a small (10%) perfusion defect in the lateral wall (severity=50%). The SimSET code was used to perform simulations for both phantoms modeling cardiac SPECT acquisitions with Tc-99m, 128×128 matrix, and 60 camera stops. In addition to two standard (no motion) acquisitions (ST) for each phantom, seven acquisitions with different degrees of phantom motion were created by manually shifting a selected number of projections in a given direction (motion ranged from 8 to 22 mm). MC images (MCI) were created using a modified TPR, where the projected template was adjusted to match the motion detected in the experimental projections by aligning the center of mass in each projection. All reconstructions were performed using OSEM with resolution recovery and attenuation correction. For all simulated movements, the MCI images exhibited improvements in both standard deviation (SD) and mean accuracy relative to the uncorrected experimental reconstructions (ER). On average, the accuracies calculated for ER, MCI and the ST reconstructions were 68%, 77%, and 76%, respectively. The average SD for ER, MCI and the ST reconstructions were 5.2, 4.0, and 4.0, respectively. Our proposed technique offers a voxel-by-voxel motion correction, which provides improved image accuracy and standard deviation of counts relative to the uncorrected images and, in many cases, the images created without patient motion.

Dynamic contrast-enhanced MRI (DCE-MRI) time series data are subject to unavoidable physiological motion during acquisition (e.g., due to breathing) and this motion causes significant errors when fitting tracer kinetic models to the data, particularly with voxel-by-voxel fitting approaches. Motion correction is problematic, as contrast enhancement introduces new features into postcontrast images and conventional registration similarity measures cannot fully account for the increased image information content. A methodology is presented for tracer kinetic model-driven registration that addresses these problems by explicitly including a model of contrast enhancement in the registration process. The iterative registration procedure is focused on a tumor volume of interest (VOI), employing a three-dimensional (3D) translational transformation that follows only tumor motion. The implementation accurately removes motion corruption in a DCE-MRI software phantom and it is able to reduce model fitting errors and improve localization in 3D parameter maps in patient data sets that were selected for significant motion problems. Sufficient improvement was observed in the modeling results to salvage clinical trial DCE-MRI data sets that would otherwise have to be rejected due to motion corruption.

Motion correction of neurological 3D PET data by the multiple acquisition frame (MAF) method presents some challenges that do not apply to the same extent in 2D. One of the most significant is that 3D PET scans have a much higher scatter component, and the inclusion of valid correction for scatter is essential for accurate quantification. This paper describes a technique for motion correction of PET data acquired in 3D that uses information provided by an optical motion tracking system, and includes corrections for temporal variations in attenuation and scatter. The technique is demonstrated in 3D dynamic and list mode scans of a Hoffman 3D brain phantom performed on a Siemens/CTI ECAT EXACT HR+ scanner in which multiple arbitrary movements were applied to the phantom during data acquisition. The MAF method cannot correct for motion that occurs during the acquisition of an individual frame. The feasibility of avoiding this limitation by acquiring the data in list mode, and re-framing into a dynamic scan in which a new frame is commenced whenever significant motion occurs was investigated. Both approaches provided a major reduction in motion distortion when compared to a conventional reconstruction without motion correction. It appears that effective motion compensation is achievable in 3D neurological PET.

Limitations from commercial software applications prevent the implementation of a robust and cost-efficient high-throughput cancer imaging radiomic feature extraction and perfusion analysis workflow. This study aimed to develop and validate a cancer research computational solution using open-source software for vendor- and sequence-neutral high-throughput image processing and feature extraction. The Cancer Radiomic and Perfusion Imaging (CARPI) automated framework is a Python-based software application that is vendor- and sequence-neutral. CARPI uses contour files generated using an application of the user's choice and performs automated radiomic feature extraction and perfusion analysis. This workflow solution was validated using two clinical data sets, one consisted of 40 pelvic chondrosarcomas and 42 sacral chordomas with a total of 82 patients, and a second data set consisted of 26 patients with undifferentiated pleomorphic sarcoma (UPS) imaged at multiple points during presurgical treatment. Three hundred sixteen volumetric contour files were processed using CARPI. The application automatically extracted 107 radiomic features from multiple magnetic resonance imaging sequences and seven semiquantitative perfusion parameters from time-intensity curves. Statistically significant differences (P < .00047) were found in 18 of 107 radiomic features in chordoma versus chondrosarcoma, including six first-order and 12 high-order features. In UPS postradiation, the apparent diffusion coefficient mean increased 41% in good responders (P = .0017), while firstorder_10Percentile (P = .0312) was statistically significant between good and partial/nonresponders. The CARPI processing of two clinical validation data sets confirmed the software application's ability to differentiate between different types of tumors and help predict patient response to treatment on the basis of radiomic features. Benchmark comparison with five similar open-source solutions demonstrated the advantages of CARPI in the automated perfusion feature extraction, relational database generation, and graphic report export features, although lacking a user-friendly graphical user interface and predictive model building.

Hybrid PET-MRI Applications in Movement Disorders.

BackgroundLarge curated data sets are required to leverage speech-based tools in health care. These are costly to produce, resulting in increased interest in data sharing. As speech can potentially identify speakers (ie, voiceprints), sharing recordings raises privacy concerns. This is especially relevant when working with patient data protected under the Health Insurance Portability and Accountability Act.ObjectiveWe aimed to determine the reidentification risk for speech recordings, without reference to demographics or metadata, in clinical data sets considering both the size of the search space (ie, the number of comparisons that must be considered when reidentifying) and the nature of the speech recording (ie, the type of speech task).MethodsUsing a state-of-the-art speaker identification model, we modeled an adversarial attack scenario in which an adversary uses a large data set of identified speech (hereafter, the known set) to reidentify as many unknown speakers in a shared data set (hereafter, the unknown set) as possible. We first considered the effect of search space size by attempting reidentification with various sizes of known and unknown sets using VoxCeleb, a data set with recordings of natural, connected speech from >7000 healthy speakers. We then repeated these tests with different types of recordings in each set to examine whether the nature of a speech recording influences reidentification risk. For these tests, we used our clinical data set composed of recordings of elicited speech tasks from 941 speakers.ResultsWe found that the risk was inversely related to the number of comparisons an adversary must consider (ie, the search space), with a positive linear correlation between the number of false acceptances (FAs) and the number of comparisons (r=0.69; P<.001). The true acceptances (TAs) stayed relatively stable, and the ratio between FAs and TAs rose from 0.02 at 1 × 105 comparisons to 1.41 at 6 × 106 comparisons, with a near 1:1 ratio at the midpoint of 3 × 106 comparisons. In effect, risk was high for a small search space but dropped as the search space grew. We also found that the nature of a speech recording influenced reidentification risk, with nonconnected speech (eg, vowel prolongation: FA/TA=98.5; alternating motion rate: FA/TA=8) being harder to identify than connected speech (eg, sentence repetition: FA/TA=0.54) in cross-task conditions. The inverse was mostly true in within-task conditions, with the FA/TA ratio for vowel prolongation and alternating motion rate dropping to 0.39 and 1.17, respectively.ConclusionsOur findings suggest that speaker identification models can be used to reidentify participants in specific circumstances, but in practice, the reidentification risk appears small. The variation in risk due to search space size and type of speech task provides actionable recommendations to further increase participant privacy and considerations for policy regarding public release of speech recordings.