Respiratory motion correction for PET oncology applications using affine transformation of list mode data

Respiratory motion is a source of artefacts and reduced image quality in PET. Proposed methodology for correction of respiratory effects involves the use of gated frames, which are however of low signal-to-noise ratio. Therefore a method accounting for respiratory motion effects without affecting the statistical quality of the reconstructed images is necessary. We have implemented an affine transformation of list mode data for the correction of respiratory motion over the thorax. The study was performed using datasets of the NCAT phantom at different points throughout the respiratory cycle. List mode data based PET simulated frames were produced by combining the NCAT datasets with a Monte Carlo simulation. Transformation parameters accounting for respiratory motion were estimated according to an affine registration and were subsequently applied on the original list mode data. The corrected and uncorrected list mode datasets were subsequently reconstructed using the one-pass list mode EM (OPL-EM) algorithm. Comparison of corrected and uncorrected respiratory motion average frames suggests that an affine transformation in the list mode data prior to reconstruction can produce significant improvements in accounting for respiratory motion artefacts in the lungs and heart. However, the application of a common set of transformation parameters across the imaging field of view does not significantly correct the respiratory effects on organs such as the stomach, liver or spleen.     

Single‐voxel MRS with prospective motion correction and retrospective frequency correction

Subject motion during MRS investigations is a factor limiting the quality and the diagnostic value of the spectra. The possibility of using external motion tracking data to correct for artefacts in MR imaging has been demonstrated previously. In this paper the utility of prospective motion correction for single‐voxel proton MRS is investigated. The object motion data are used in real time to update the position of the spectroscopy voxel during the acquisition prior to every sequence repetition cycle. It is not, however, sufficient to update the voxel position alone due to shim changes accompanying subject motion. Adverse effects of frequency shifts induced by subject motion are effectively suppressed by the interleaved reference scan method. Copyright © 2010 John Wiley & Sons, Ltd.     

List-mode-based reconstruction for respiratory motion correction in PET using non-rigid body transformations

Respiratory motion in emission tomography leads to reduced image quality. Developed correction methodology has been concentrating on the use of respiratory synchronized acquisitions leading to gated frames. Such frames, however, are of low signal-to-noise ratio as a result of containing reduced statistics. In this work, we describe the implementation of an elastic transformation within a list-mode-based reconstruction for the correction of respiratory motion over the thorax, allowing the use of all data available throughout a respiratory motion average acquisition. The developed algorithm was evaluated using datasets of the NCAT phantom generated at different points throughout the respiratory cycle. List-mode-data-based PET-simulated frames were subsequently produced by combining the NCAT datasets with Monte Carlo simulation. A non-rigid registration algorithm based on B-spline basis functions was employed to derive transformation parameters accounting for the respiratory motion using the NCAT dynamic CT images. The displacement matrices derived were subsequently applied during the image reconstruction of the original emission list mode data. Two different implementations for the incorporation of the elastic transformations within the one-pass list mode EM (OPL-EM) algorithm were developed and evaluated. The corrected images were compared with those produced using an affine transformation of list mode data prior to reconstruction, as well as with uncorrected respiratory motion average images. Results demonstrate that although both correction techniques considered lead to significant improvements in accounting for respiratory motion artefacts in the lung fields, the elastic-transformation-based correction leads to a more uniform improvement across the lungs for different lesion sizes and locations.     

Tomographic motion detection and correction directly in sinogram space

Patient motion, especially respiratory motion, results in various artefacts such as blurring and streaks in tomographic images. The interplay of the movement of the beam aperture and variations of organ anatomy during delivery can create 'hot' and 'cold' spots throughout the field in intensity-modulated radiation therapy (IMRT). Detection and correction of patient motion is extremely important in tomographic imaging and IMRT. Tomographic projection data (sinogram) encode not only the patient anatomy information, but also the intra-scanning motion information. In this paper, we developed an algorithm to detect and correct the in-plane respiratory motion directly in sinogram space. The respiratory motion is modelled as time-varying scaling along the x and y directions. Its effects on the sinogram are discussed. Based on the traces of some nodal points in the sinogram, the intra-scanning motion is determined. The motion correction is also implemented in sinogram space. The motion-corrected sinogram is used for reconstruction by the filtered back-projection (FBP) method. Computer simulations validate the motion detection and correction algorithm. The reconstructed images from the motion-corrected sinogram eliminate the majority of the artefacts. The method could be applied to projection data used in CT and ECT, as well as in tomotherapy delivery modification and dose reconstruction.     

Reduced respiratory motion artifacts using structural similarity in fast 2D dynamic contrast enhanced MRI of liver lesions

The purpose of this work was to improve dynamic contrast enhanced MRI (DCE‐MRI) of liver lesions by removing motion corrupted images as identified by a structural similarity (SSIM) algorithm, and to assess the effect of this correction on the pharmacokinetic parameter K^trans using automatically determined arterial input functions (AIFs). Fifteen patients with colorectal liver metastases were measured twice with a T₁ weighted multislice 2D FLASH sequence for DCE‐MRI (time resolution 1.2 s). AIFs were automatically derived from contrast inflow in the aorta of each patient. Thereafter, SSIM identified motion corrupted images of the liver were removed from the DCE dataset. From this corrected data set K^trans and its reproducibility were determined. Using the SSIM algorithm a median fraction of 46% (range 37–50%) of the liver images in DCE time series was labeled as motion distorted. Rejection of these images resulted in a significantly lower median K^trans (p < 0.05) and lower coefficient of repeatability of K^trans in liver metastases compared with an analysis without correction. SSIM correction improves the reproducibility of the DCE‐MRI parameter K^trans in liver metastasis and reduces contamination of K^trans values of lesions by that of surrounding normal liver tissue.     

Real-time 3D motion tracking for small animal brain PET

High-resolution positron emission tomography (PET) imaging of conscious, unrestrained laboratory animals presents many challenges. Some form of motion correction will normally be necessary to avoid motion artefacts in the reconstruction. The aim of the current work was to develop and evaluate a motion tracking system potentially suitable for use in small animal PET. This system is based on the commercially available stereo-optical MicronTracker S60 which we have integrated with a Siemens Focus-220 microPET scanner. We present measured performance limits of the tracker and the technical details of our implementation, including calibration and synchronization of the system. A phantom study demonstrating motion tracking and correction was also performed. The system can be calibrated with sub-millimetre accuracy, and small lightweight markers can be constructed to provide accurate 3D motion data. A marked reduction in motion artefacts was demonstrated in the phantom study. The techniques and results described here represent a step towards a practical method for rigid-body motion correction in small animal PET. There is scope to achieve further improvements in the accuracy of synchronization and pose measurements in future work.     

Registration of dynamic contrast-enhanced MRI using a progressive principal component registration (PPCR)

Registration of dynamic contrast-enhanced magnetic resonance images (DCE-MRI) of soft tissue is difficult. Conventional registration cost functions that depend on information content are compromised by the changing intensity profile, leading to misregistration. We present a new data-driven model of uptake patterns formed from a principal components analysis (PCA) of time-series data, avoiding the need for a physiological model. We term this process progressive principal component registration (PPCR). Registration is performed repeatedly to an artificial time series of target images generated using the principal components of the current best-registered time-series data. The aim is to produce a dataset that has had random motion artefacts removed but long-term contrast enhancement implicitly preserved. The procedure is tested on 22 DCE-MRI datasets of the liver. Preliminary assessment of the images is by expert observer comparison with registration to the first image in the sequence. The PPCR is preferred in all cases where a preference exists. The method requires neither segmentation nor a pharmacokinetic uptake model and can allow successful registration in the presence of contrast enhancement.     

A novel method for incorporating respiratory-matched attenuation correction in the motion correction of cardiac PET-CT studies

Mismatches between PET and CT datasets due to respiratory effects can lead to artefactual perfusion defects. To overcome this, we have proposed a method of aligning a single CT with each frame of a gated PET study in a semi-automatic manner, incorporating a statistical shape model of the diaphragm and a rigid registration of the heart. This ensures that the structures that could influence the appearance of the reconstructed cardiac activity are correctly matched between emission and transmission datasets. When tested on two patient studies, it was found in both cases that attenuation correction using the proposed technique resulted in PET images that were closer to the gold standard of attenuation correction with a gated CT, compared with scenarios where only heart matching was considered (and not the diaphragm) or where no transformation was performed (i.e. where a single CT frame was used to attenuation-correct all PET frames). These preliminary results suggest that diaphragm matching between PET and CT improves the quantitative accuracy of reconstructed PET images and that the proposed method of using a statistical shape model to describe the diaphragm shape and motion, in combination with a rigid registration to determine respiratory-induced heart motion, is a feasible method of achieving this.     

Event-based motion correction for PET transmission measurements with a rotating point source

Accurate attenuation correction is important for quantitative positron emission tomography (PET) studies. When performing transmission measurements using an external rotating radioactive source, object motion during the transmission scan can distort the attenuation correction factors computed as the ratio of the blank to transmission counts, and cause errors and artefacts in reconstructed PET images. In this paper we report a compensation method for rigid body motion during PET transmission measurements, in which list mode transmission data are motion corrected event-by-event, based on known motion, to ensure that all events which traverse the same path through the object are recorded on a common line of response (LOR). As a result, the motion-corrected transmission LOR may record a combination of events originally detected on different LORs. To ensure that the corresponding blank LOR records events from the same combination of contributing LORs, the list mode blank data are spatially transformed event-by-event based on the same motion information. The number of counts recorded on the resulting blank LOR is then equivalent to the number of counts that would have been recorded on the corresponding motion-corrected transmission LOR in the absence of any attenuating object. The proposed method has been verified in phantom studies with both stepwise movements and continuous motion. We found that attenuation maps derived from motion-corrected transmission and blank data agree well with those of the stationary phantom and are significantly better than uncorrected attenuation data.     

Towards optimized MR thermometry of the human heart at 3T

Catheter ablation using radio frequency (RF) has been used increasingly for the treatment of cardiac arrhythmias and may be combined with proton resonance frequency shift (PRFS) ?based MR thermometry to determine the therapy endpoint. We evaluated the suitability of two different MR thermometry sequences (TFE and TFE‐EPI) and three blood suppression techniques. Experiments were performed without heating, using an optimized imaging protocol including navigator respiratory compensation, cardiac triggering, and image processing for the compensation of motion and susceptibility artefacts. Blood suppression performance and its effect on temperature stability were evaluated in the ventricular septum of eight healthy volunteers using multislice double inversion recovery (MDIR), motion sensitized driven equilibrium (MSDE), and inflow saturation by saturation slabs (IS). It was shown that blood suppression during MR thermometry improves the contrast‐to‐noise ratio (CNR), the robustness of the applied motion correction algorithm as well as the temperature stability. A gradient echo sequence accelerated by an EPI readout and parallel imaging (SENSE) and using inflow saturation blood suppression was shown to achieve the best results. Temperature stabilities of 2 °C or better in the ventricular septum with a spatial resolution of 3.5 × 3.5 × 8mm³ and a temporal resolution corresponding to the heart rate of the volunteer, were observed. Our results indicate that blood suppression improves the temperature stability when performing cardiac MR thermometry. The proposed MR thermometry protocol, which optimizes temperature stability in the ventricular septum, represents a step towards PRFS‐based MR thermometry of the heart at 3 T. Copyright © 2011 John Wiley & Sons, Ltd.     

Fast generation of 4D PET-MR data from real dynamic MR acquisitions

We have implemented and evaluated a framework for simulating simultaneous dynamic PET-MR data using the anatomic and dynamic information from real MR acquisitions. PET radiotracer distribution is simulated by assigning typical FDG uptake values to segmented MR images with manually inserted additional virtual lesions. PET projection data and images are simulated using analytic forward projections (including attenuation and Poisson statistics) implemented within the image reconstruction package STIR. PET image reconstructions are also performed with STIR. The simulation is validated with numerical simulation based on Monte Carlo (GATE) which uses more accurate physical modelling, but has 150× slower computation time compared to the analytic method for ten respiratory positions and is 7000× slower when performing multiple realizations. Results are validated in terms of region of interest mean values and coefficients of variation for 65 million coincidences including scattered events. Although some discrepancy is observed, agreement between the two different simulation methods is good given the statistical noise in the data. In particular, the percentage difference of the mean values is 3.1% for tissue, 17% for the lungs and 18% for a small lesion. The utility of the procedure is demonstrated by simulating realistic PET-MR datasets from multiple volunteers with different breathing patterns. The usefulness of the toolkit will be shown for performance investigations of the reconstruction, motion correction and attenuation correction algorithms for dynamic PET-MR data.     

Robust-Deep: A Method for Increasing Brain Imaging Datasets to Improve Deep Learning Models’ Performance and Robustness

A small dataset commonly affects generalization, robustness, and overall performance of deep neural networks (DNNs) in medical imaging research. Since gathering large clinical databases is always difficult, we proposed an analytical method for producing a large realistic/diverse dataset. Clinical brain PET/CT/MR images including full-dose (FD), low-dose (LD) corresponding to only 5 % of events acquired in the FD scan, non-attenuated correction (NAC) and CT-based measured attenuation correction (MAC) PET images, CT images and T1 and T2 MR sequences of 35 patients were included. All images were registered to the Montreal Neurological Institute (MNI) template. Laplacian blending was used to make a natural presentation using information in the frequency domain of images from two separate patients, as well as the blending mask. This classical technique from the computer vision and image processing communities is still widely used and unlike modern DNNs, does not require the availability of training data. A modified ResNet DNN was implemented to evaluate four image-to-image translation tasks, including LD to FD, LD+MR to FD, NAC to MAC, and MRI to CT, with and without using the synthesized images. Quantitative analysis using established metrics, including the peak signal-to-noise ratio (PSNR), structural similarity index metric (SSIM), and joint histogram analysis was performed for quantitative evaluation. The quantitative comparison between the registered small dataset containing 35 patients and the large dataset containing 350 synthesized plus 35 real dataset demonstrated improvement of the RMSE and SSIM by 29% and 8% for LD to FD, 40% and 7% for LD+MRI to FD, 16% and 8% for NAC to MAC, and 24% and 11% for MRI to CT mapping task, respectively. The qualitative/quantitative analysis demonstrated that the proposed model improved the performance of all four DNN models through producing images of higher quality and lower quantitative bias and variance compared to reference images.     

3D tomographic reconstruction of coronary arteries using a precomputed 4D motion field

In this paper, we present a new method to perform 3D tomographic reconstruction of coronary arteries from cone-beam rotational x-ray angiography acquisitions. We take advantage of the precomputation of the coronary artery motion, modelled as a parametric 4D motion field. Contrary to data gating or data triggering approaches, we homogeneously use all available frames, independently of the cardiac phase. In addition, we artificially subtract angiograms from their background structures. Our method significantly improves the reconstruction, by removing both motion and background artefacts. We have successfully tested it on the datasets from a synthetic phantom and 10 patients.     

基于薄板样条的MRI图像与脑图谱的配准方法

为了将CT、MRI、PET或SPECT等断层扫描图像用于疾病的辅助诊断、放射治疗、手术计划和引导 ,就必须知道图像中感兴趣区 (ROI)是什么解剖组织 ,即解决医学图像的解剖标识问题。医生通常是从解剖书籍、图谱及自身经验来对ROI做出判断。这些书籍和图谱往往给出的是文字描述和有限数目的 ,固定位置和方向的断层图片 ,很难与患者的实际图像联系起来。对于缺乏临床经验的医生来说尤为困难。数字化 3D人脑解剖图谱使医生对人脑深部组织全方位可视化。因此可以将其通过一定的空间变换 ,与MR体积数据集中的ROI进行比较 ,从而得到ROI的解剖标识。但是 ,没有两个人的大脑是完全一样的。人脑的解剖个体差异较大 ,这就要求利用非线性变形的方法做解剖标识。本研究介绍通过薄板样条变换用Talirach脑图谱对MR图像做解剖标识的方法。     

The design and implementation of a motion correction scheme for neurological PET

A method is described to monitor the motion of the head during neurological positron emission tomography (PET) acquisitions and to correct the data post acquisition for the recorded motion prior to image reconstruction. The technique uses an optical tracking system, Polaris, to accurately monitor the position of the head during the PET acquisition. The PET data are acquired in list mode where the events are written directly to disk during acquisition. The motion tracking information is aligned to the PET data using a sequence of pseudo-random numbers, which are inserted into the time tags in the list mode event stream through the gating input interface on the tomograph. The position of the head is monitored during the transmission acquisition, and it is assumed that there is minimal head motion during this measurement. Each event, prompt and delayed, in the list mode event stream is corrected for motion and transformed into the transmission space. For a given line of response, normalization, including corrections for detector efficiency, geometry and crystal interference and dead time are applied prior to motion correction and rebinning in the sinogram. A series of phantom experiments were performed to confirm the accuracy of the method: (a) a point source located in three discrete axial positions in the tomograph field of view, 0 mm, 10 mm and 20 mm from a reference point, (b) a multi-line source phantom rotated in both discrete and gradual rotations through +/- 5 degrees and +/- 15 degrees, including a vertical and horizontal movement in the plane. For both phantom experiments images were reconstructed for both the fixed and motion corrected data. Measurements for resolution, full width at half maximum (FWHM) and full width at tenth maximum (FWTM), were calculated from these images and a comparison made between the fixedand motion corrected datasets. From the point source measurements, the FWHM at each axial position was 7.1 mm in the horizontal direction, and increasing from 4.7 mm at the 0 mm position, to 4.8 mm, 20 mm offset, in the vertical direction. The results from the multi-line source phantom with +/- 5 degrees rotations showed a maximum degradation in FWHM, when compared with the stationary phantom, of 0.6 mm, in the horizontal direction, and 0.3 mm in the vertical direction. The corresponding values for the larger rotation, +/- 15 degrees, were 0.7 mm and 1.1 mm, respectively. The performance of the method was confirmed with a Hoffman brain phantom moved continuously, and a clinical acquisition using [11C]raclopride (normal volunteer). A visual comparison of both the motion and non-motion corrected images of the Hoffman brain phantom clearly demonstrated the efficacy of the method. A sample time-activity curve extracted from the clinical study showed irregularities prior to motion correction, which were removed after correction. A method has been developed to accurately monitor the motion of the head during a neurological PET acquisition, and correct for this motion prior to image reconstruction. The method has been demonstrated to be accurate and does not add significantly to either the acquisition or the subsequent data processing.     

The effect of motion correction on pharmacokinetic parameter estimation in dynamic-contrast-enhanced MRI

A dynamic-contrast-enhanced magnetic resonance imaging (DCE-MRI) dataset consists of many imaging frames, often acquired both before and after contrast injection. Due to the length of time spent acquiring images, patient motion is likely and image re-alignment or registration is required before further analysis such as pharmacokinetic model fitting. Non-rigid image registration procedures may be used to correct motion artefacts; however, a careful choice of registration strategy is required to reduce misregistration artefacts associated with enhancing features. This work investigates the effect of registration on the results of model-fitting algorithms for 52 DCE-MR mammography cases for 14 patients. Results are divided into two sections: a comparison of registration strategies in which a DCE-MRI-specific algorithm is preferred in 50% of cases, followed by an investigation of parameter changes with known applied deformations, inspecting the effect of magnitude and timing of motion artefacts. Increased motion magnitude correlates with increased model-fit residual and is seen to have a strong influence on the visibility of strongly enhancing features. Motion artefacts in images close to the contrast agent arrival have a disproportionate effect on discrepancies in parameter estimation. The choice of algorithm, magnitude of motion and timing of the motion are each shown to influence estimated pharmacokinetic parameters even when motion magnitude is small.     

DIMENSION: Dynamic MR imaging with both k-space and spatial prior knowledge obtained via multi-supervised network training

Dynamic MR image reconstruction from incomplete k-space data has generated great research interest due to its capability in reducing scan time. Nevertheless, the reconstruction problem is still challenging due to its ill-posed nature. Most existing methods either suffer from long iterative reconstruction time or explore limited prior knowledge. This paper proposes a dynamic MR imaging method with both k-space and spatial prior knowledge integrated via multi-supervised network training, dubbed as DIMENSION. Specifically, the DIMENSION architecture consists of a frequential prior network for updating the k-space with its network prediction and a spatial prior network for capturing image structures and details. Furthermore, a multi-supervised network training technique is developed to constrain the frequency domain information and the spatial domain information. The comparisons with classical k-t FOCUSS, k-t SLR, L+S and the state-of-the-art CNN-based method on in vivo datasets show our method can achieve improved reconstruction results in shorter time.     

A study of artefacts in simultaneous PET and MR imaging using a prototype MR compatible PET scanner.

We have assessed the possibility of artefacts that can arise in attempting to perform simultaneous positron emission tomography (PET) and magnetic resonance imaging (MRI) using a small prototype MR compatible PET scanner (McPET). In these experiments, we examine MR images for any major artefacts or loss in image quality due to inhomogeneities in the magnetic field, radiofrequency interference or susceptibility effects caused by operation of the PET system inside the MR scanner. In addition, possible artefacts in the PET images caused by the static and time-varying magnetic fields or radiofrequency interference from the MR system were investigated. Biological tissue and a T2-weighted spin echo sequence were used to examine susceptibility artefacts due to components of the McPET scanner (scintillator, optical fibres) situated in the MR field of view. A range of commonly used MR pulse sequences was studied while acquiring PET data to look for possible artefacts in either the PET or MR images. Other than a small loss in signal-to-noise using gradient echo sequences, there was no significant interaction between the two imaging systems. Simultaneous PET and MR imaging of simple phantoms was also carried out in different MR systems with field strengths ranging from 0.2 to 4.7 T. The results of these studies demonstrate that it is possible to acquire PET and MR images simultaneously, without any significant artefacts or loss in image quality, using our prototype MR compatible PET scanner.     

Spectral trimming technique: a new approach for suppressing motion artefacts in stress electrocardiography

Abstract

Motion artefacts in electrocardiographic (ECG) signal are suppressed mainly by adaptive noise cancellation and wavelet denoising. While the former requires a motion sensor in addition to ECG electrodes, the latter removes some of the desired low-frequency components in the signal. In this paper spectral trimming technique is being introduced for suppressing the motion artefacts in stress electrocardiography. In this method, Fourier spectral coefficients up to 1.221?Hz of noisy signal are trimmed on the basis of template derived from resting ECG signal in the same subject. The proposed spectral trimming technique has yielded the lowest value of mean?±?standard deviation for root mean square error (18.92?±?8.71) and highest value of the signal to noise ratio (6.439?±?4.266) as compared to other three methods, namely adaptive noise cancellation, wavelet decomposition and adaptive line enhancement with compatible value of correlation coefficient. Subsequently, spectral trimming technique has been implemented in real-time (deferred by 8.2?s) application for stress electrocardiography. Spectral trimming technique thus offers a method of choice for motion artefact suppression in offline as well as deferred online applications. This method takes care of the limitations of conventional methods such as adaptive noise cancellation or wavelet denoising for suppressing motion artefacts in stress electrocardiography.     

Extension of a data-driven gating technique to 3D, whole body PET studies

Respiratory gating can be used to separate a PET acquisition into a series of near motion-free bins. This is typically done using additional gating hardware; however, software-based methods can derive the respiratory signal from the acquired data itself. The aim of this work was to extend a data-driven respiratory gating method to acquire gated, 3D, whole body PET images of clinical patients. The existing method, previously demonstrated with 2D, single bed-position data, uses a spectral analysis to find regions in raw PET data which are subject to respiratory motion. The change in counts over time within these regions is then used to estimate the respiratory signal of the patient. In this work, the gating method was adapted to only accept lines of response from a reduced set of axial angles, and the respiratory frequency derived from the lung bed position was used to help identify the respiratory frequency in all other bed positions. As the respiratory signal does not identify the direction of motion, a registration-based technique was developed to align the direction for all bed positions. Data from 11 clinical FDG PET patients were acquired, and an optical respiratory monitor was used to provide a hardware-based signal for comparison. All data were gated using both the data-driven and hardware methods, and reconstructed. The centre of mass of manually defined regions on gated images was calculated, and the overall displacement was defined as the change in the centre of mass between the first and last gates. The mean displacement was 10.3 mm for the data-driven gated images and 9.1 mm for the hardware gated images. No significant difference was found between the two gating methods when comparing the displacement values. The adapted data-driven gating method was demonstrated to successfully produce respiratory gated, 3D, whole body, clinical PET acquisitions.