Ian S. Armstrong

Impact of point spread function modeling and time-of-flight on myocardial blood flow and myocardial flow reserve measurements for rubidium-82 cardiac PET.

BackgroundMyocardial flow reserve (MFR) obtained from dynamic cardiac positron emission tomography (PET) with rubidium-82 (Rb-82) has been shown to be a useful measurement in assessing coronary artery disease. Advanced PET reconstructions with point spread function modeling and time-of-flight have been shown to improve image quality but also have an impact on kinetic analysis of dynamic data. This study aims to determine the impact of these algorithms on MFR data.MethodsDynamic Rb-82 cardiac PET images from 37 patients were reconstructed with standard and advanced reconstructions. Area under curve (AUC) of the blood input function (BIF), myocardial blood flow (MBF) and MFR were compared with each reconstruction.ResultsNo significant differences were seen in MFR for the two reconstructions. A relatively small mean difference in MBF data of +11.9% was observed with advanced reconstruction compared with the standard reconstruction but there was considerable variability in the degree of change (95% confidence intervals of −16.2% to +40.0%). Small systematic relative differences were seen for AUC BIF (mean difference of −6.3%; 95% CI −17.5% to +5.4%).Conclusion:MFR results from Rb-82 dynamic PET appear to be robust when generated by standard or advanced PET reconstructions. Considerable increases in MBF values may occur with advanced reconstructions, and further work is required to fully understand this.

Impact of point spread function modelling and time of flight on FDG uptake measurements in lung lesions using alternative filtering strategies

BackgroundThe use of maximum standardised uptake value (SUVmax) is commonplace in oncology positron emission tomography (PET). Point spread function (PSF) modelling and time-of-flight (TOF) reconstructions have a significant impact on SUVmax, presenting a challenge for centres with defined protocols for lesion classification based on SUVmax thresholds. This has perhaps led to the slow adoption of these reconstructions. This work evaluated the impact of PSF and/or TOF reconstructions on SUVmax, SUVpeak and total lesion glycolysis (TLG) under two different schemes of post-filtering.MethodsPost-filters to match voxel variance or SUVmax were determined using a NEMA NU-2 phantom. Images from 68 consecutive lung cancer patients were reconstructed with the standard iterative algorithm along with TOF; PSF modelling - Siemens HD·PET (HD); and combined PSF modelling and TOF - Siemens ultraHD·PET (UHD) with the two post-filter sets. SUVmax, SUVpeak, TLG and signal-to-noise ratio of tumour relative to liver (SNR(T-L)) were measured in 74 lesions for each reconstruction. Relative differences in uptake measures were calculated, and the clinical impact of any changes was assessed using published guidelines and local practice.ResultsWhen matching voxel variance, SUVmax increased substantially (mean increase +32% and +49% for HD and UHD, respectively), potentially impacting outcome in the majority of patients. Increases in SUVpeak were less notable (mean increase +17% and +23% for HD and UHD, respectively). Increases with TOF alone were far less for both measures. Mean changes to TLG were <10% for all algorithms for either set of post-filters. SNR(T-L) were greater than ordered subset expectation maximisation (OSEM) in all reconstructions using both post-filtering sets.ConclusionsMatching image voxel variance with PSF and/or TOF reconstructions, particularly with PSF modelling and in small lesions, resulted in considerable increases in SUVmax, inhibiting the use of defined protocols for lesion classification based on SUVmax. However, reduced partial volume effects may increase lesion detectability. Matching SUVmax in phantoms translated well to patient studies for PSF reconstruction but less well with TOF, where a small positive bias was observed in patient images. Matching SUVmax significantly reduced voxel variance and potential variability of uptake measures. Finally, TLG may be less sensitive to reconstruction methods compared with either SUVmax or SUVpeak.

Accurate markerless respiratory tracking for gated whole body PET using the Microsoft Kinect

The motion due to respiration is responsible for greatly reducing image quality of whole body positron emission tomography, PET. A simple method to produce a respiratory signal to enable gating of PET listmode data using the Microsoft Kinect, a consumer grade 3D scanner, is presented. Phantom data produced by a sinusoidally oscillating phantom being tracked by an existing commercially available respiratory monitoring system and a Kinect based contactless tracking system shows that the Kinect system outperforms the Varian RPM tracking by producing higher resolution traces. When testing the Kinect using human subjects, the collected surface trace is sensitive enough to detect the patients heart rate. This has potential for improving motion correction of whole body imaging by including extra surface information provided by the Kinect, in addition to the basic respiratory signal.

Reduced-count myocardial perfusion SPECT with resolution recovery.

AimThe aim of this study was to determine whether resolution recovery (RR) iterative reconstruction algorithms can consistently produce diagnostic quality myocardial perfusion SPECT images for the patient population routinely scanned in this department. Reduced-count data were compared with full-count data without RR according to our established protocol. The desired outcome would be to implement the software to allow a reduction in the administered activity for routine myocardial SPECT. MethodsHalf-count SPECT data were derived from full-count datasets for 53 stress and rest routine myocardial SPECT scans on a GE Infinia camera. Full-count data were reconstructed using standard non-RR ordered subset expectation maximization reconstruction, whereas half-count data were reconstructed using Evolution RR software. Myocardial functional values, image quality and report outcomes of the full-count and half-count reports were compared. Sequential full-time and half-time myocardial SPECT acquisitions were performed for 15 stress and rest studies on a Siemens c.cam dedicated cardiac camera. Half-count data were reconstructed using Siemens Flash 3D RR. ResultsNo degradation in image quality was found when comparing full-count and half-count studies from the Infinia. Ten percent of the half-count studies from the c.cam were considered slightly worse than full-count data. Statistically significant differences in some full-count versus half-count functional values were found but the actual mean differences were not considered clinically significant. No difference was found for 44 out of 53 full-count versus half-count reports, a minor difference for seven out of 53 and a significant difference in two cases. ConclusionRR was found to produce diagnostic image quality for nearly all scans, but it was felt that a reduction of 50% from our standard protocol was too great. A proposed reduction of 33% administered activity was considered acceptable to produce consistently adequate diagnostic images for both manufacturers.

The assessment of time-of-flight on image quality and quantification with reduced administered activity and scan times in 18F-FDG PET.

ObjectivesThe last decade has seen considerable technological innovations in PET detectors with the availability, among other advances, of time-of-flight (TOF). TOF has been shown to increase the signal-to-noise ratio (SNR), which should allow for a reduction in acquired counts while maintaining image quality. MethodsFifty-eight patients referred for routine 18F-flurodeoxyglucose (18F-FDG) oncology PET studies were included in this study. Patients with weight below or above 100 kg were prescribed 350 or 400 MBq of 18F-FDG, respectively. Listmode data were acquired for 2.5 min per bed position and reconstructed with ordered-subset expectation maximization (OSEM) reconstruction. TOF reconstruction was performed on reduced-count data, with two levels of reduction (−20 and −40% for patients <100 kg and −16 and −30% for patients >100 kg) achieved by clipping the listmode data. Liver SNR, mediastinum mean standardized uptake value (SUVmean), and lesion maximum standardized uptake value (SUVmax) were measured in all images. All images were visually assessed as adequate or suboptimal. ResultsNo significant difference was seen in mediastinum SUVmean or lesion SUVmax when comparing reduced-count TOF with full-count OSEM images. Compared with the original OSEM images, liver SNR was higher for TOF images using the more conservative −20% reduction of counts (P<0.001, Wilcoxon’s signed-rank test), whereas no significant statistical difference was seen with −40% reductions. ConclusionIncorporation of TOF allows for a reduction in acquired counts; this method has been implemented at our institution, with administered activity reduced for all patients to 280 MBq and a reduction in scan times for all but the largest patients. This has significantly reduced the patient radiation dose and improved scanner flexibility and throughput.

Apical thinning: real or artefact?

Background and objectiveApical thinning is a well-known phenomenon in myocardial perfusion SPECT, often attributed to reduced myocardial thickness at the apex of the left ventricle. Attenuation correction processing appears to exaggerate this effect. Although currently there is agreement that reduced apical counts are not a diagnostic indicator, opinions differ over the cause of this effect; we sought to clarify this using results from a phantom study. MethodsA commercially available anthropomorphic torso phantom was expanded using attachments mimicking tissue and bone to create three phantoms of increasing size. These were imaged using a dual-headed gamma camera and low-dose CT-based attenuation correction. The data were processed using iterative reconstruction, with and without attenuation correction. ResultsThe cardiac insert had a uniform wall thickness and yet defects characteristic of apical thinning appeared after attenuation correction, increasing in severity with phantom size. Before attenuation correction, a flare of activity was seen at the apex corresponding in position and size to the defect after attenuation correction. Further investigations showed the following: depth-dependent resolution was not responsible; the severity of the defect was more noticeably dependent on the addition of breast activity than the addition of attenuating material; the artefact was not unique to one particular algorithm; increasing the number of iterations reduced the severity of the artefact. ConclusionData acquisition and processing methods are thought to be responsible for the apparent apical defect. This phantom study therefore demonstrates that apical thinning is not simply an anatomical feature but can also be an artefact introduced by the use of attenuation correction.

Activity concentration measurements using a conjugate gradient (Siemens xSPECT) reconstruction algorithm in SPECT/CT.

The interest in quantitative single photon emission computer tomography (SPECT) shows potential in a number of clinical applications and now several vendors are providing software and hardware solutions to allow ‘SUV-SPECT’ to mirror metrics used in PET imaging. This brief technical report assesses the accuracy of activity concentration measurements using a new algorithm ‘xSPECT’ from Siemens Healthcare. SPECT/CT data were acquired from a uniform cylinder with 5, 10, 15 and 20 s/projection and NEMA image quality phantom with 25 s/projection. The NEMA phantom had hot spheres filled with an 8 : 1 activity concentration relative to the background compartment. Reconstructions were performed using parameters defined by manufacturer presets available with the algorithm. The accuracy of activity concentration measurements was assessed. A dose calibrator–camera cross-calibration factor (CCF) was derived from the uniform phantom data. In uniform phantom images, a positive bias was observed, ranging from ∼6% in the lower count images to ∼4% in the higher-count images. On the basis of the higher-count data, a CCF of 0.96 was derived. As expected, considerable negative bias was measured in the NEMA spheres using region mean values whereas positive bias was measured in the four largest NEMA spheres. Nonmonotonically increasing recovery curves for the hot spheres suggested the presence of Gibbs edge enhancement from resolution modelling. Sufficiently accurate activity concentration measurements can easily be measured on images reconstructed with the xSPECT algorithm without a CCF. However, the use of a CCF is likely to improve accuracy further. A manual conversion of voxel values into SUV should be possible, provided that the patient weight, injected activity and time between injection and imaging are all known accurately.

Accuracy and variability of quantitative measurements using PET with time-of-flight information and resolution modelling

Quantitative measurements are increasingly used in cancer staging and therapy monitoring. This work investigates the impact of resolution modeling and time-of-flight reconstruction on the accuracy and variability of two types of activity concentration (AC) measurements, mean and maximum, in spheres of different sizes and contrast. Results are compared with conventional OSEM reconstruction that does not include these additions. Positive bias was seen in nearly all maximum AC measurements, particularly with OSEM in larger, low contrast spheres with no or small amounts of post-filtering. Maximum AC was far less dependent on sphere size in high contrast with combined resolution modeling and time-of-flight. Resolution modeling and time-of-flight were shown to reduce the variability of maximum AC measurements for all sphere sizes and post-filters. Negative bias was seen in all mean AC measurements, with mean AC recovery being greatest in smaller spheres with resolution modeling and time-of-flight. Mean AC recovery was comparable in the larger spheres for all reconstruction algorithms. Either no or very small reductions in the variability of mean AC measurements were seen with resolution modeling and time-of-flight.

Single Photon Emission Computed Tomography (SPECT) Myocardial Perfusion Imaging Guidelines: Instrumentation, Acquisition, Processing, and Interpretation

Single Photon Emission Computed Tomography (SPECT) Myocardial Perfusion Imaging Guidelines: Instrumentation, Acquisition, Processing, and Interpretation Sharmila Dorbala, MD, MPH, Karthik Ananthasubramaniam, MD, Ian S. Armstrong, PhD, MIPEM, Panithaya Chareonthaitawee, MD, E. Gordon DePuey, MD, Andrew J. Einstein, MD, PhD, Robert J. Gropler, MD, Thomas A. Holly, MD, John J. Mahmarian, MD, Mi-Ae Park, PhD, Donna M. Polk, MD, MPH, Raymond Russell III, MD, PhD, Piotr J. Slomka, PhD, Randall C. Thompson, MD, and R. Glenn Wells, PhD

Evaluation of general-purpose collimators against high-resolution collimators with resolution recovery with a view to reducing radiation dose in myocardial perfusion SPECT: A preliminary phantom study.

BackgroundThere is a growing focus on reducing radiation dose to patients undergoing myocardial perfusion imaging. This preliminary phantom study aims to evaluate the use of general-purpose collimators with resolution recovery (RR) to allow a reduction in patient radiation dose.MethodsImages of a cardiac torso phantom with inferior and anterior wall defects were acquired on a GE Infinia and Siemens Symbia T6 using both high-resolution and general-purpose collimators. Imaging time, a surrogate for administered activity, was reduced between 35% and 40% with general-purpose collimators to match the counts acquired with high-resolution collimators. Images were reconstructed with RR with and without attenuation correction. Two pixel sizes were also investigated. Defect contrast was measured.ResultsDefect contrast on general-purpose images was superior or comparable to the high-resolution collimators on both systems despite the reduced imaging time. Infinia general-purpose images required a smaller pixel size to be used to maintain defect contrast, while Symbia T6 general-purpose images did not require a change in pixel size to that used for standard myocardial perfusion SPECT.ConclusionThis study suggests that general-purpose collimators with RR offer a potential for substantial dose reductions while providing similar or better image quality to images acquired using high-resolution collimators.