Thomas Koesters

Whole-Body PET/MR Imaging: Quantitative Evaluation of a Novel Model-Based MR Attenuation Correction Method Including Bone

In routine whole-body PET/MR hybrid imaging, attenuation correction (AC) is usually performed by segmentation methods based on a Dixon MR sequence providing up to 4 different tissue classes. Because of the lack of bone information with the Dixon-based MR sequence, bone is currently considered as soft tissue. Thus, the aim of this study was to evaluate a novel model-based AC method that considers bone in whole-body PET/MR imaging. Methods: The new method (“Model”) is based on a regular 4-compartment segmentation from a Dixon sequence (“Dixon”). Bone information is added using a model-based bone segmentation algorithm, which includes a set of prealigned MR image and bone mask pairs for each major body bone individually. Model was quantitatively evaluated on 20 patients who underwent whole-body PET/MR imaging. As a standard of reference, CT-based μ-maps were generated for each patient individually by nonrigid registration to the MR images based on PET/CT data. This step allowed for a quantitative comparison of all μ-maps based on a single PET emission raw dataset of the PET/MR system. Volumes of interest were drawn on normal tissue, soft-tissue lesions, and bone lesions; standardized uptake values were quantitatively compared. Results: In soft-tissue regions with background uptake, the average bias of SUVs in background volumes of interest was 2.4% ± 2.5% and 2.7% ± 2.7% for Dixon and Model, respectively, compared with CT-based AC. For bony tissue, the −25.5% ± 7.9% underestimation observed with Dixon was reduced to −4.9% ± 6.7% with Model. In bone lesions, the average underestimation was −7.4% ± 5.3% and −2.9% ± 5.8% for Dixon and Model, respectively. For soft-tissue lesions, the biases were 5.1% ± 5.1% for Dixon and 5.2% ± 5.2% for Model. Conclusion: The novel MR-based AC method for whole-body PET/MR imaging, combining Dixon-based soft-tissue segmentation and model-based bone estimation, improves PET quantification in whole-body hybrid PET/MR imaging, especially in bony tissue and nearby soft tissue.

Joint MR-PET Reconstruction Using a Multi-Channel Image Regularizer

While current state of the art MR-PET scanners enable simultaneous MR and PET measurements, the acquired data sets are still usually reconstructed separately. We propose a new multi-modality reconstruction framework using second order Total Generalized Variation (TGV) as a dedicated multi-channel regularization functional that jointly reconstructs images from both modalities. In this way, information about the underlying anatomy is shared during the image reconstruction process while unique differences are preserved. Results from numerical simulations and in-vivo experiments using a range of accelerated MR acquisitions and different MR image contrasts demonstrate improved PET image quality, resolution, and quantitative accuracy.

Dixon Sequence with Superimposed Model-Based Bone Compartment Provides Highly Accurate PET/MR Attenuation Correction of the Brain.

Simultaneous PET/MR of the brain is a promising technology for characterizing patients with suspected cognitive impairment or epilepsy. Unlike CT, however, MR signal intensities do not correlate directly with PET photon attenuation correction (AC), and inaccurate radiotracer SUV estimation can limit future PET/MR clinical applications. We tested a novel AC method that supplements standard Dixon-based tissue segmentation with a superimposed model-based bone compartment. Methods: We directly compared SUV estimation between MR-based AC and reference CT AC in 16 patients undergoing same-day PET/CT and PET/MR with a single 18F-FDG dose for suspected neurodegeneration. Three Dixon-based MR AC methods were compared with CT: standard Dixon 4-compartment segmentation alone, Dixon with a superimposed model-based bone compartment, and Dixon with a superimposed bone compartment and linear AC optimized specifically for brain tissue. The brain was segmented using a 3-dimensional T1-weighted volumetric MR sequence, and SUV estimations were compared with CT AC for whole-image, whole-brain, and 91 FreeSurfer-based regions of interest. Results: Modifying the linear AC value specifically for brain and superimposing a model-based bone compartment reduced the whole-brain SUV estimation bias of Dixon-based PET/MR AC by 95% compared with reference CT AC (P < 0.05), resulting in a residual −0.3% whole-brain SUVmean bias. Further, brain regional analysis demonstrated only 3 frontal lobe regions with an SUV estimation bias of 5% or greater (P < 0.05). These biases appeared to correlate with high individual variability in frontal bone thickness and pneumatization. Conclusion: Bone compartment and linear AC modifications result in a highly accurate MR AC method in subjects with suspected neurodegeneration. This prototype MR AC solution appears equivalent to other recently proposed solutions and does not require additional MR sequences and scanning time. These data also suggest that exclusively model-based MR AC approaches may be adversely affected by common individual variations in skull anatomy.

Quantitative graphical analysis of simultaneous dynamic PET/MRI for assessment of prostate cancer.

Purpose Dynamic FDG imaging for prostate cancer characterization is limited by generally small size and low uptake in prostate tumors. Our aim in this pilot study was to explore feasibility of simultaneous PET/MRI to guide localization of prostate lesions for dynamic FDG analysis using a graphical approach. Methods Three patients with biopsy-proven prostate cancer underwent simultaneous FDG PET/MRI, incorporating dynamic prostate imaging. Histology and multiparametric MRI findings were used to localize tumors, which in turn guided identification of tumors on FDG images. Regions of interest were manually placed on tumor and benign prostate tissue. Blood activity was extracted from a region of interest placed on the femoral artery on PET images. FDG data were analyzed by graphical analysis using the influx constant Ki (Patlak analysis) when FDG binding seemed irreversible and distribution volume VT (reversible graphical analysis) when FDG binding seemed reversible given the presence of washout. Results Given inherent coregistration, simultaneous acquisition facilitated use of MRI data to localize small lesions on PET and subsequent graphical analysis in all cases. In 2 cases with irreversible binding, tumor had higher Ki than benign using Patlak analysis (0.023 vs 0.006 and 0.019 vs 0.008 mL/cm3 per minute). In 1 case appearing reversible, tumor had higher VT than benign using reversible graphical analysis (0.68 vs 0.52 mL/cm3). Conclusions Simultaneous PET/MRI allows localization of small prostate tumors for dynamic PET analysis. By taking advantage of inclusion of the femoral arteries in the FOV, we applied advanced PET data analysis methods beyond conventional static measures and without blood sampling.

Joint reconstruction of simultaneously acquired MR-PET data with multi sensor compressed sensing based on a joint sparsity constraint

State-of-the-art MR-PET scanners allow simultaneous data acquisition. However, image reconstruction is performed separately and results are only combined at the visualization stage. PET images are reconstructed using a variant of EM and MR data are reconstructed using an inverse Fourier transform or iterative algorithms for parallel imaging or compressed sensing. We propose a new iterative joint reconstruction framework based on multi-sensor compressed sensing that exploits anatomical correlations between MR and PET. Joint reconstruction is motivated by the fact that MR and PET are based on the same anatomy. High resolution MR information can be used to enhance the PET reconstruction and MR artifacts are not present in the PET image. Therefore a dedicated reconstruction can exploit the incoherence of artifacts in the joint space. Our approach uses a nonlinear constrained optimization problem. In each iteration OSEM enforces data consistency of the current solution with measured PET rawdata. An l1-norm based joint sparsity term exploits anatomical correlations. MR data consistency is enforced with the MR forward operator, consisting of coil sensitivity modulation and a (non-uniform) Fourier transform. Data were acquired on a clinical 3T MR-PET unit (Siemens Biograph mMR). 10 mCi 18F-FDG were injected followed by a 60min list mode scan. 3D MP-RAGE was used for MR data acquisition: TR=2300ms, TE=2.98ms, TI=900ms, FA=9°, acceleration factor 2, 24 ACS lines, 256 matrix, voxel size=1×1×1mm3, 192 slices. Joint MR-PET reconstruction improves resolution in PET images when structures are aligned with MR. PET signal information cannot be improved in regions showing no distinctive MR contrast, but it is also not influenced falsely. The availability of simultaneously-acquired MR and PET data will also enable incorporation of dynamic correlations and motion correction into the joint reconstruction framework. We expect that this provides additional enhancements to the information content of multimodality studies.

PET reconstruction with non-smooth gradient-based priors

Quantitative PET imaging is hindered by limited spatial resolution and high Poisson noise. A way to overcome those limitations, especially in brain PET/MR examinations, is the inclusion of anatomical prior knowledge in the image reconstruction process. Recently, the concept of Parallel Level Sets (PLS) [1]–[3] has been introduced as a promising anatomical prior for PET reconstruction. This prior relies on the exchange of gradient information. However, PLS is a non-smooth function which hampers the optimization process in iterative reconstruction. In this proceeding we show how to use the EM-TV algorithm by Sawatzki et al. [4] to solve the non-smooth PLS-regularized PET reconstruction problem efficiently which enables application to 3D clinical PET data.

Plane-dependent ML scatter scaling: 3D extension of the 2D simulated single scatter estimate

In this work, we propose a plane-dependent maximum likelihood (ML) scatter scale estimation from the emission measurements. The scatter scales obtained are validated using a Monte Carlo simulation of a NEMA-like phantom, and results are shown from two whole-body patient scans.

Simultaneous PET-MRI reconstruction with vectorial second order total generalized variation

State of the art PET-MR systems are capable of performing both PET and MR measurements simultaneously. However, the resulting data sets are usually processed in two separate reconstruction pipelines. The goal of this work is to complement simultaneous data acquisition with a new multi-modality reconstruction framework based on second order total generalized variation that simultaneously reconstructs both PET and MR images. Information of the underlying anatomy is shared during the image reconstruction process with a dedicated multi-channel regularization functional. Results of numerical simulations and in-vivo data are presented that demonstrate improved PET resolution and reduced noise in MR in comparison to conventional reconstructions.

Multi-bed elastic motion correction for whole body MR-PET

Single bed motion correction algorithms are increasingly being used to model and compensate for the respiratory motion in clinical PET images. In this work we develop a multi-bed elastic motion correction algorithm and workflow for the MR-PET (Siemens biograph mMR) scanner. For validation of the proposed multi-bed motion algorithm we performed multi-bed NCAT phantom simulations as well as physical multi-bed rod phantom studies using the clinical scanner. The NCAT and rod phantoms were used as the motion vectors are exactly know and any bias in the final reconstructed image can be easily identified. Finally four patient scans were performed using a two bed protocol. A self-gated radial VIBE MRI sequence was used to generate the gated MR images. The motion vectors were calculated from the gated MR images using demons algorithm. In the mMR system the clinical PET data can be acquired simultaneously along with the MR diagnostic scans of the patient. The list-mode PET data for each bed is divided into pre-determined frames based on the amplitude of the respiratory waveform. The mu-maps corresponding to each frame was generated and used to reconstruct each PET frame separately. A post reconstruction motion correction algorithm was used to compensate for the motion in the different frame of the multi-bed acquisition. Each frame of the multi-bed reconstructed images was warped to the whole body reference image taking into account the acquisition time of each frame as well as the possibility of miss-match in the motion vectors between the beds. The phantom as well as patient studies were motion compensated to the reference image and found to give a faithful whole body reconstruction.

PET motion correction using MR-derived motion parameters

With the improving resolution of modern PET scanners, any slight motion during the scan can cause significant blurring and loss of resolution. MRI scanners have the capacity to perform quick successive scans and thus provide a means to track motion during a scan. Hence, with the advent of simultaneous PET-MR scanners, it has become possible to use the MR scanner to track the motion and thereby provide the necessary motion parameters to correct the PET data. Using a suitable segmentation approach a separate MR scan can provide the attenuation map to produce quantitative PET images. An FDG brain scan was acquired on a Siemens Biograph mMR PET-MR scanner. The MR scan was acquired using the Golden Angle Radial Sparse Parallel (GRASP) sequence [1], simultaneously with a 5 minute PET scan, while the patient was asked to move his head repetitively from side to side for proof-of-principle purposes. A separate static scan was also acquired prior to the motion scan, to be used as a control. The MR data were divided into a series of 268 images with a frequency of approximately 1 Hz. The motion parameters were derived by performing a rigid (6 degrees-of-freedom) registration of the masked MR images to a chosen reference image. The PET list-mode data were corrected on an event-by-event basis [2, 3]. List-mode maximum likelihood expectation-maximisation (accelerated with ordered subsets [4]) was used for the reconstruction, incorporating the attenuation correction (as a pre-correction to the data) as well as weighted-average sensitivity [2] to achieve a quantitative reconstruction. Motion correction successfully removed almost all motion artefacts, recovered the resolution and allowed for quantitative images to be produced. Techniques to improve upon the coarse sampling of the MR images, such as interpolating between motion data points, will be investigated.