Robust real-time extraction of respiratory signals from PET list-mode data.

Respiratory motion, which typically cannot simply be suspended during PET image acquisition, affects lesions' detection and quantitative accuracy inside or in close vicinity to the lungs. Some motion compensation techniques address this issue via pre-sorting ('binning') of the acquired PET data into a set of temporal gates, where each gate is assumed to be minimally affected by respiratory motion. Tracking respiratory motion is typically realized using dedicated hardware (e.g. using respiratory belts and digital cameras). Extracting respiratory signals directly from the acquired PET data simplifies the clinical workflow as it avoids handling additional signal measurement equipment. We introduce a new data-driven method 'combined local motion detection' (CLMD). It uses the time-of-flight (TOF) information provided by state-of-the-art PET scanners in order to enable real-time respiratory signal extraction without additional hardware resources. CLMD applies center-of-mass detection in overlapping regions based on simple back-positioned TOF event sets acquired in short time frames. Following a signal filtering and quality-based pre-selection step, the remaining extracted individual position information over time is then combined to generate a global respiratory signal. The method is evaluated using seven measured FDG studies from single and multiple scan positions of the thorax region, and it is compared to other software-based methods regarding quantitative accuracy and statistical noise stability. Correlation coefficients around 90% between the reference and the extracted signal have been found for those PET scans where motion affected features such as tumors or hot regions were present in the PET field-of-view. For PET scans with a quarter of typically applied radiotracer doses, the CLMD method still provides similar high correlation coefficients which indicates its robustness to noise. Each CLMD processing needed less than 0.4 s in total on a standard multi-core CPU and thus provides a robust and accurate approach enabling real-time processing capabilities using standard PC hardware.

PET performance evaluation of the small-animal Hyperion IID PET/MRI insert based on the NEMA NU-4 standard.

The Hyperion IID PET insert is the first scanner using fully digital silicon photomultipliers for simultaneous PET/MR imaging of small animals up to rabbit size. In this work, we evaluate the PET performance based on the National Eletrical Manufacturers Association (NEMA) NU 4-2008 standard, whose standardized measurement protocols allow comparison of different small-animal PET scanners. The Hyperion IID small-animal PET/MR insert comprises three rings of 20 detector stacks with pixelated scintillator arrays with a crystal pitch of 1 mm, read out with digital silicon photomultipliers. The scanner has a large ring diameter of 209.6 mm and an axial field of view of 96.7 mm. We evaluated the spatial resolution, energy resolution, time resolution and sensitivity by scanning a 22Na point source. The count rates and scatter fractions were measured for a wide range of 18F activity inside a mouse-sized scatter phantom. We evaluated the image quality using the mouse-sized image quality phantom specified in the NEMA NU4 standard, filled with 18F. Additionally, we verified the in-vivo imaging capabilities by performing a simultaneous PET/MRI scan of a mouse injected with 18F-FDG. We processed all measurement data with an energy window of 250 keV to 625 keV and a coincidence time window of 2 ns. The filtered-backprojection reconstruction of the point source has a full width at half maximum (FWHM) of 1.7 mm near the isocenter and degrades to 2.5 mm at a radial distance of 50 mm. The scanner's average energy resolution is 12.7% (ΔE/E FWHM) and the coincidence resolution time is 609 ps. The peak absolute sensitivity is 4.0% and the true and noise-equivalent count rates reach their peak at an activity of 46 MBq with 483 kcps and 407 kcps, respectively, with a scatter fraction of 13%. The iterative reconstruction of the image quality phantom has a uniformity of 3.7%, and recovery coefficients from 0.29, 0.91 and 0.94 for rod diameters of 1 mm, 3 mm and 5 mm, respectively. After application of scatter and attenuation corrections, the air- and water-filled cold regions have spill-over ratios of 6.3% and 5.4%, respectively. The Hyperion IID PET/MR insert provides state-of-the-art PET performance while enabling simultaneous PET/MRI acquisition of small animals up to rabbit size.

Incorporating HYPR de-noising within iterative PET reconstruction (HYPR-OSEM).

HighlY constrained back-PRojection (HYPR) is a post-processing de-noising technique originally developed for time-resolved magnetic resonance imaging. It has been recently applied to dynamic imaging for positron emission tomography and shown promising results. In this work, we have developed an iterative reconstruction algorithm (HYPR-OSEM) which improves the signal-to-noise ratio (SNR) in static imaging (i.e. single frame reconstruction) by incorporating HYPR de-noising directly within the ordered subsets expectation maximization (OSEM) algorithm. The proposed HYPR operator in this work operates on the target image(s) from each subset of OSEM and uses the sum of the preceding subset images as the composite which is updated every iteration. Three strategies were used to apply the HYPR operator in OSEM: (i) within the image space modeling component of the system matrix in forward-projection only, (ii) within the image space modeling component in both forward-projection and back-projection, and (iii) on the image estimate after the OSEM update for each subset thus generating three forms: (i) HYPR-F-OSEM, (ii) HYPR-FB-OSEM, and (iii) HYPR-AU-OSEM. Resolution and contrast phantom simulations with various sizes of hot and cold regions as well as experimental phantom and patient data were used to evaluate the performance of the three forms of HYPR-OSEM, and the results were compared to OSEM with and without a post reconstruction filter. It was observed that the convergence in contrast recovery coefficients (CRC) obtained from all forms of HYPR-OSEM was slower than that obtained from OSEM. Nevertheless, HYPR-OSEM improved SNR without degrading accuracy in terms of resolution and contrast. It achieved better accuracy in CRC at equivalent noise level and better precision than OSEM and better accuracy than filtered OSEM in general. In addition, HYPR-AU-OSEM has been determined to be the more effective form of HYPR-OSEM in terms of accuracy and precision based on the studies conducted in this work.

Investigation of practical initial attenuation image estimates in TOF-MLAA reconstruction for PET/MR.

PURPOSE: Time-of-flight joint attenuation and activity positron emission tomography reconstruction requires additional calibration (scale factors) or constraints during or post-reconstruction to produce a quantitative µ-map. In this work, the impact of various initializations of the joint reconstruction was investigated, and the initial average mu-value (IAM) method was introduced such that the forward-projection of the initial µ-map is already very close to that of the reference µ-map, thus reducing/minimizing the offset (scale factor) during the early iterations of the joint reconstruction. Consequently, the accuracy and efficiency of unconstrained joint reconstruction such as time-of-flight maximum likelihood estimation of attenuation and activity (TOF-MLAA) can be improved by the proposed IAM method. METHODS: 2D simulations of brain and chest were used to evaluate TOF-MLAA with various initial estimates which include the object filled with water uniformly (conventional initial estimate), bone uniformly, the average µ-value uniformly (IAM magnitude initialization method), and the perfect spatial µ-distribution but with a wrong magnitude (initialization in terms of distribution). 3D gate simulation was also performed for the chest phantom under a typical clinical scanning condition, and the simulated data were reconstructed with a fully corrected list-mode TOF-MLAA algorithm with various initial estimates. The accuracy of the average µ-values within the brain, chest, and abdomen regions obtained from the MR derived µ-maps was also evaluated using computed tomography µ-maps as the gold-standard. RESULTS: The estimated µ-map with the initialization in terms of magnitude (i.e., average µ-value) was observed to reach the reference more quickly and naturally as compared to all other cases. Both 2D and 3D gate simulations produced similar results, and it was observed that the proposed IAM approach can produce quantitative µ-map/emission when the corrections for physical effects such as scatter and randoms were included. The average µ-value obtained from MR derived µ-map was accurate within 5% with corrections for bone, fat, and uniform lungs. CONCLUSIONS: The proposed IAM-TOF-MLAA can produce quantitative µ-map without any calibration provided that there are sufficient counts in the measured data. For low count data, noise reduction and additional regularization/rescaling techniques need to be applied and investigated. The average µ-value within the object is prior information which can be extracted from MR and patient database, and it is feasible to obtain accurate average µ-value using MR derived µ-map with corrections as demonstrated in this work.

Maximum likelihood positioning and energy correction for scintillation detectors.

An algorithm for determining the crystal pixel and the gamma ray energy with scintillation detectors for PET is presented. The algorithm uses Likelihood Maximisation (ML) and therefore is inherently robust to missing data caused by defect or paralysed photo detector pixels. We tested the algorithm on a highly integrated MRI compatible small animal PET insert. The scintillation detector blocks of the PET gantry were built with the newly developed digital Silicon Photomultiplier (SiPM) technology from Philips Digital Photon Counting and LYSO pixel arrays with a pitch of 1 mm and length of 12 mm. Light sharing was used to readout the scintillation light from the 30 × 30 scintillator pixel array with an 8 × 8 SiPM array. For the performance evaluation of the proposed algorithm, we measured the scanner's spatial resolution, energy resolution, singles and prompt count rate performance, and image noise. These values were compared to corresponding values obtained with Center of Gravity (CoG) based positioning methods for different scintillation light trigger thresholds and also for different energy windows. While all positioning algorithms showed similar spatial resolution, a clear advantage for the ML method was observed when comparing the PET scanner's overall single and prompt detection efficiency, image noise, and energy resolution to the CoG based methods. Further, ML positioning reduces the dependence of image quality on scanner configuration parameters and was the only method that allowed achieving highest energy resolution, count rate performance and spatial resolution at the same time.

Initial PET performance evaluation of a preclinical insert for PET/MRI with digital SiPM technology.

Hyperion-II(D) is a positron emission tomography (PET) insert which allows simultaneous operation in a clinical magnetic resonance imaging (MRI) scanner. To read out the scintillation light of the employed lutetium yttrium orthosilicate crystal arrays with a pitch of 1 mm and 12 mm in height, digital silicon photomultipliers (DPC 3200-22, Philips Digital Photon Counting) (DPC) are used. The basic PET performance in terms of energy resolution, coincidence resolution time (CRT) and sensitivity as a function of the operating parameters, such as the operating temperature, the applied overvoltage, activity and configuration parameters of the DPCs, has been evaluated at system level. The measured energy resolution did not show a large dependency on the selected parameters and is in the range of 12.4%-12.9% for low activity, degrading to ∼13.6% at an activity of ∼100 MBq. The CRT strongly depends on the selected trigger scheme (trig) of the DPCs, and we measured approximately 260 ps, 440 ps, 550 ps and 1300 ps for trig 1-4, respectively. The trues sensitivity for a NEMA NU 4 mouse-sized scatter phantom with a 70 mm long tube of activity was dependent on the operating parameters and was determined to be 0.4%-1.4% at low activity. The random fraction stayed below 5% at activity up to 100 MBq and the scatter fraction was evaluated as ∼6% for an energy window of 411 keV-561 keV and ∼16% for 250 keV-625 keV. Furthermore, we performed imaging experiments using a mouse-sized hot-rod phantom and a large rabbit-sized phantom. In 2D slices of the reconstructed mouse-sized hot-rod phantom (∅ = 28 mm), the rods were distinguishable from each other down to a rod size of 0.8 mm. There was no benefit from the better CRT of trig 1 over trig 3, where in the larger rabbit-sized phantom (∅ = 114 mm) we were able to show a clear improvement in image quality using the time-of-flight information. The findings will allow system architects-aiming at a similar detector design using DPCs-to make predictions about the design requirements and the performance that can be expected.

Software-Based Real-Time Acquisition and Processing of PET Detector Raw Data.

In modern positron emission tomography (PET) readout architectures, the position and energy estimation of scintillation events (singles) and the detection of coincident events (coincidences) are typically carried out on highly integrated, programmable printed circuit boards. The implementation of advanced singles and coincidence processing (SCP) algorithms for these architectures is often limited by the strict constraints of hardware-based data processing. In this paper, we present a software-based data acquisition and processing architecture (DAPA) that offers a high degree of flexibility for advanced SCP algorithms through relaxed real-time constraints and an easily extendible data processing framework. The DAPA is designed to acquire detector raw data from independent (but synchronized) detector modules and process the data for singles and coincidences in real-time using a center-of-gravity (COG)-based, a least-squares (LS)-based, or a maximum-likelihood (ML)-based crystal position and energy estimation approach (CPEEA). To test the DAPA, we adapted it to a preclinical PET detector that outputs detector raw data from 60 independent digital silicon photomultiplier (dSiPM)-based detector stacks and evaluated it with a [(18)F]-fluorodeoxyglucose-filled hot-rod phantom. The DAPA is highly reliable with less than 0.1% of all detector raw data lost or corrupted. For high validation thresholds (37.1 ± 12.8 photons per pixel) of the dSiPM detector tiles, the DAPA is real time capable up to 55 MBq for the COG-based CPEEA, up to 31 MBq for the LS-based CPEEA, and up to 28 MBq for the ML-based CPEEA. Compared to the COG-based CPEEA, the rods in the image reconstruction of the hot-rod phantom are only slightly better separable and less blurred for the LS- and ML-based CPEEA. While the coincidence time resolution (∼ 500 ps) and energy resolution (∼12.3%) are comparable for all three CPEEA, the system sensitivity is up to 2.5 × higher for the LS- and ML-based CPEEA.

PET performance and MRI compatibility evaluation of a digital, ToF-capable PET/MRI insert equipped with clinical scintillators.

We evaluate the MR compatibility of the Hyperion-II(D) positron emission tomography (PET) insert, which allows simultaneous operation in a clinical magnetic resonance imaging (MRI) scanner. In contrast to previous investigations, this work aims at the evaluation of a clinical crystal configuration. An imaging-capable demonstrator with an axial field-of-view of 32 mm and a crystal-to-crystal spacing of 217.6 mm was equipped with LYSO scintillators with a pitch of 4 mm which were read out in a one-to-one coupling scheme by sensor tiles composed of digital silicon photomultipliers from Philips Digital Photon Counting (DPC 3200-22). The PET performance degradation (energy resolution and coincidence resolution time (CRT)) was evaluated during simultaneous operation of the MRI scanner. We used clinically motivated imaging sequences as well as synthetic gradient stress test sequences. Without activity of the MRI scanner, we measured for trigger scheme 1 (first photon trigger) an energy resolution of 11.4% and a CRT of 213 ps for a narrow energy (NE) window using five (22)Na point-like sources. When applying the synthetic gradient sequences, we found worst-case relative degradations of the energy resolution by 5.1% and of the CRT by 33.9%. After identifying the origin of the degradations and implementing a fix to the read-out hardware, the same evaluation revealed no degradation of the PET performance anymore even when the most demanding gradient stress tests were applied. The PET performance of the insert was initially evaluated using the point sources, a high-activity phantom and hot-rod phantoms in order to assess the spatial resolution. Trigger schemes 2-4 delivered an energy resolution of 11.4% as well and CRTs of 279 ps, 333 ps and 557 ps for the NE window, respectively. An isocenter sensitivity of 0.41% using the NE window and 0.71% with a wide energy window was measured. Using a hot-rod phantom, a spatial resolution in the order of 2 mm was demonstrated and the benefit of time-of-flight PET was shown with a larger rabbit-sized phantom. In conclusion, the Hyperion architecture is an interesting platform for clinically driven hybrid PET/MRI systems.

A Digital Preclinical PET/MRI Insert and Initial Results.

Combining Positron Emission Tomography (PET) with Magnetic Resonance Imaging (MRI) results in a promising hybrid molecular imaging modality as it unifies the high sensitivity of PET for molecular and cellular processes with the functional and anatomical information from MRI. Digital Silicon Photomultipliers (dSiPMs) are the digital evolution in scintillation light detector technology and promise high PET SNR. DSiPMs from Philips Digital Photon Counting (PDPC) were used to develop a preclinical PET/RF gantry with 1-mm scintillation crystal pitch as an insert for clinical MRI scanners. With three exchangeable RF coils, the hybrid field of view has a maximum size of 160 mm × 96.6 mm (transaxial × axial). 0.1 ppm volume-root-mean-square B 0-homogeneity is kept within a spherical diameter of 96 mm (automatic volume shimming). Depending on the coil, MRI SNR is decreased by 13% or 5% by the PET system. PET count rates, energy resolution of 12.6% FWHM, and spatial resolution of 0.73 mm (3) (isometric volume resolution at isocenter) are not affected by applied MRI sequences. PET time resolution of 565 ps (FWHM) degraded by 6 ps during an EPI sequence. Timing-optimized settings yielded 260 ps time resolution. PET and MR images of a hot-rod phantom show no visible differences when the other modality was in operation and both resolve 0.8-mm rods. Versatility of the insert is shown by successfully combining multi-nuclei MRI ((1)H/(19)F) with simultaneously measured PET ((18)F-FDG). A longitudinal study of a tumor-bearing mouse verifies the operability, stability, and in vivo capabilities of the system. Cardiac- and respiratory-gated PET/MRI motion-capturing (CINE) images of the mouse heart demonstrate the advantage of simultaneous acquisition for temporal and spatial image registration.

A self-normalization reconstruction technique for PET scans using the positron emission data.

Positron emission tomography (PET) image quality in both clinical and preclinical environments highly depends on an accurate knowledge of the detector hardware to correct for image quality degrading effects like gain, temperature, and photon detection efficiency variations of the individual crystals. In conventional PET systems some of these variations are typically corrected using a dedicated calibration scan in which the scanner performance for a well-known activity source is measured. We propose an alternative method for estimating the relative sensitivity of each detector pixel using the coincidences as well as the singles emission data of each PET scan. The overall idea is to compare the total sum of all measured single photons before coincidence processing in each crystal with a steadily low-frequent distribution that can normally be expected. Both the estimated activity and the estimated detector sensitivity are simultaneously improved by using an extended iterative reconstruction scheme. This way we ensure the use of an optimal calibration correction (with the exception of a global factor) for each data set, even if the scanner performance has changed between two scans. Data measured with a preclinical PET scanner (HYPERIon-I) which uses analog silicon photomultipliers in combination with a custom-made ASIC shows a significant increase of image quality and homogeneity using the proposed method.

MRI-based attenuation correction for hybrid PET/MRI systems: a 4-class tissue segmentation technique using a combined ultrashort-echo-time/Dixon MRI sequence.

UNLABELLED: Accurate γ-photon attenuation correction (AC) is essential for quantitative PET/MRI as there is no simple relation between MR image intensity and attenuation coefficients. Attenuation maps (µ-maps) can be derived by segmenting MR images and assigning attenuation coefficients to the compartments. Ultrashort-echo-time (UTE) sequences have been used to separate cortical bone and air, and the Dixon technique has enabled differentiation between soft and adipose tissues. Unfortunately, sequential application of these sequences is time-consuming and complicates image registration. METHODS: A UTE triple-echo (UTILE) MRI sequence is proposed, combining UTE sampling for bone detection and gradient echoes for Dixon water-fat separation in a radial 3-dimensional acquisition (repetition time, 4.1 ms; echo times, 0.09/1.09/2.09 ms; field strength, 3 T). Air masks are derived mainly from the phase information of the first echo; cortical bone is segmented using a dual-echo technique. Soft-tissue and adipose-tissue decomposition is achieved using a 3-point Dixon-like decomposition. Predefined linear attenuation coefficients are assigned to classified voxels to generate MRI-based µ-maps. The results of 6 patients are obtained by comparing µ-maps, reciprocal sensitivity maps, reconstructed PET images, and brain region PET activities based on either CT AC, two 3-class MRI AC techniques, or the proposed 4-class UTILE AC. RESULTS: Using the UTILE MRI sequence, an acquisition time of 214 s was achieved for the head-and-neck region with 1.75-mm isotropic resolution, compared with 164 s for a single-echo UTE scan. MRI-based reciprocal sensitivity maps show a high correlation with those derived from CT scans (R(2) = 0.9920). The same is true for PET activities (R(2) = 0.9958). An overall voxel classification accuracy (compared with CT) of 81.1% was reached. Bone segmentation is inaccurate in complex regions such as the paranasal sinuses, but brain region activities in 48 regions across 6 patients show a high correlation after MRI-based and CT-based correction (R(2) = 0.9956), with a regression line slope of 0.960. All overall correlations are higher and brain region PET activities more accurate in terms of mean and maximum deviations for the 4-class technique than for 3-class techniques. CONCLUSION: The UTILE MRI sequence enables the generation of MRI-based 4-class µ-maps without anatomic priors, yielding results more similar to CT-based results than can be obtained with 3-class segmentation only.

Simultaneous reconstruction of activity and attenuation for PET/MR.

Medical investigations targeting a quantitative analysis of the position emission tomography (PET) images require the incorporation of additional knowledge about the photon attenuation distribution in the patient. Today, energy range adapted attenuation maps derived from computer tomography (CT) scans are used to effectively compensate for image quality degrading effects, such as attenuation and scatter. Replacing CT by magnetic resonance (MR) is considered as the next evolutionary step in the field of hybrid imaging systems. However, unlike CT, MR does not measure the photon attenuation and thus does not provide an easy access to this valuable information. Hence, many research groups currently investigate different technologies for MR-based attenuation correction (MR-AC). Typically, these approaches are based on techniques such as special acquisition sequences (alone or in combination with subsequent image processing), anatomical atlas registration, or pattern recognition techniques using a data base of MR and corresponding CT images. We propose a generic iterative reconstruction approach to simultaneously estimate the local tracer concentration and the attenuation distribution using the segmented MR image as anatomical reference. Instead of applying predefined attenuation values to specific anatomical regions or tissue types, the gamma attenuation at 511 keV is determined from the PET emission data. In particular, our approach uses a maximum-likelihood estimation for the activity and a gradient-ascent based algorithm for the attenuation distribution. The adverse effects of scattered and accidental gamma coincidences on the quantitative accuracy of PET, as well as artifacts caused by the inherent crosstalk between activity and attenuation estimation are efficiently reduced using enhanced decay event localization provided by time-of-flight PET, accurate correction for accidental coincidences, and a reduced number of unknown attenuation coefficients. First results achieved with measured whole body PET data and reference segmentation from CT showed an absolute mean difference of 0.005 cm⁻¹ (< 20%) in the lungs, 0.0009 cm⁻¹ (< 2%) in case of fat, and 0.0015 cm⁻¹ (< 2%) for muscles and blood. The proposed method indicates a robust and reliable alternative to other MR-AC approaches targeting patient specific quantitative analysis in time-of-flight PET/MR.