Our previous design for a stationary tomographic molecular breast imaging (MBI) system was optimised for 140 keV photons (${ }^{99 \mathrm{~m}} \mathrm{Tc}$). The system consists of two planar CZT detectors and multipinhole collimators. The design allows for overlap of different pinhole projections on the detectors (multiplexing (MX)). This improves sensitivity and sampling but can lead to artefacts. This can be addressed by utilising depth-of-interaction (DoI) information from the detectors. The system performance depends on pinhole size, opening angle and separation. The photon energy ($E_{\gamma}$) also affects the performance, by pinhole-edge penetration and interaction depth in the detectors. In order to investigate the optimal system design for different radionuclides, including ${ }^{177} \mathrm{Lu}$ and 111 In, we have performed a series of computer simulations with a range of parameter values for $E_{\gamma}$ in the range $[113,245] \mathrm{keV}$. Projection data were simulated for a phantom with 6 spherical inserts, diam. 5-10 mm, TBR 10. The data were reconstructed using ML-EM, and the images were analysed in terms of contrast-to-noise ratio (CNR). Our results show that CNR decreases significantly at higher $E_{\gamma}$, with $\sim 30 \%$ reduction at 245 keV compared to 140 keV. With respect to parameter optimisation, the main effect is related to pinhole separation. At 140 keV, the optimum separation is 8 mm, while for higher energies it tends towards $\sim 10 \mathrm{~mm}$. This is due to a larger degree of MX due to greater detector penetration. However, the difference is quite small, suggesting that collimators optimised for lower $E_{\gamma}$ would also be appropriate at higher $E_{\gamma}$. In conclusion, we have shown that, as opposed to standard $\gamma$-cameras, it would not be necessary to change collimators on our novel MBI system for use with radionuclides with $E_{\gamma}$ up to 245 keV.

Parametric whole-body PET images can be obtained with a conventional PET scanner using multi-bed multi-pass dynamic whole-body (DWB) acquisition protocol. The scanning time can be shortened by starting later, as the early tissue data is not required for kinetic analysis of irreversible tracers. However, the initial peak of the image-derived input function (IDIF) will be missed. We previously proposed a protocol, with a low-activity 2nd injection for measuring the peak. Here we propose an improvement of this method by using an IDIF-library. 24 patients were scanned, with a dynamic acquisition over the thorax directly after tracer injection, a DWB scan and a standard-of-care (SoC) WB scan at 60 min. IDIFs were obtained from the aorta. In one additional subject, a 2nd injection (I2) of 10% was given and another dynamic acquisition over the thorax was performed for 5 min. IDIFs were generated by fitting the data from I2 together with a sub-set from the DWB scan, a) using an analytical function, and b) by selecting one scaled IDIF from the library, that best fit the data. The results were compared with the data from the 1st injection in terms of MSE. We simulated data corresponding to a reduced I2 activity by factors of up to 8, with 1000 noise replicates. Data were analysed with 3- or 5-min scans after I2 in combination with DWB data from 30 or 40 min p.i. The results showed that, with a 10% I2, the two models gave similar results. With reduced activity, for the analytical model fit, MSE increased by factors of up to 2.5 and SD by factors of up to 5. The IDIF-library approach was much more robust, with MSE increasing by factors of up to 1.25 and SD by factors of up to 1.5. In conclusion, based on data from one single patient study, our results suggest the activity in the 2nd injection can be reduced significantly using our proposed IDIF-library approach.

A prototype of an ultra-low-dose stationary tomographic Molecular Breast Imaging system (ULDMBI) has been developed and built by Kromek. The system combines superior energy, position, and depth of interaction (DOI) resolution of CZT detectors, wide range angular sampling of multi-pinhole (MPH) collimators and novel demultiplexing image reconstruction, to mitigate the MBI drawbacks (higher than mammography patient dose and long imaging time of 40 min) without compromising diagnostic content. The initial feasibility study of this new technology was conducted by Kromek and University College London. It indicated that it is possible to achieve a dose/time reduction of a factor of seven compared to the current MBI clinical practice. The new dual-head prototype has field-of-view of $132 \mathrm{~mm} \times 88 \mathrm{~mm}$ which is large enough to allow clinical evaluation with phantoms. Each head is made of a 6x4 array of Kromek’s $5 \mathbf{~ m m}$ thick CZT detectors with 2 mm pixels. The raw detector data is analysed using a novel machine-learning-based model which provides strong improvement in the sub-pixel position resolution and depth-of-interaction information. Preliminary evaluation of the model performance using planar imaging has demonstrated significant increase in the detector position resolution allowing achievement of $5 \times 5$ sub-pixelisation of 2 mm pixels producing images equivalent to $400 \mu \mathrm{m}$ pixel pattern. These results surpass any previous demonstration by a SPECT or MBI system with $\sim 2 \mathrm{~mm}$ pixel CZT detectors. This high-resolution data is used in tomographic image reconstruction allowing results of the prototype performance evaluation with $^{99 \mathrm{~m}} \mathrm{Tc}$-filled phantoms and further studies of the dose/time reduction to be presented and discussed.

Current harmonisation efforts in PET-CT focus on oncology and brain studies. However, there is a similar need for studies of diffuse lung disease, where voxel-wise air fractions (AF) determined from (smoothed) CT can be used to correct for variable air content in lung. Resolution mismatches between PET and CT can cause artefacts in the AF-corrected image. In this work, we investigate the utility of modified thorax phantoms for determining the kernel to smooth the CT for AF correction, and assess its applicability in regions that approximate diffuse lung disease, as well as a hot lesion. A matched kernel approach was used to determine the kernel for clinical reconstructions of 18 F scans on three different PET-CTs. The relative difference between the optimally smoothed ground truth and reconstructed image was within 15% for all regions and reconstructions investigated. This phantom design allows kernel applicability to be investigated in regions of low density, fine-scale structure, as seen in diffuse lung disease, as well as for hot lesions. It therefore has potential utility for determining scanner-specific reconstruction recommendations, with a view to harmonisation of PET-CT lung imaging.

Nuclear medicine imaging modalities like computed tomography (CT), single photon emission CT (SPECT) and positron emission tomography (PET) are employed in the field of theranostics to estimate and plan the dose delivered to tumors and the surrounding tissues and to monitor the effect of the therapy. However, therapeutic radionuclides often provide poor images, which translate to inaccurate treatment planning and inadequate monitoring images. Multimodality information can be exploited in the reconstruction to enhance image quality. Triple modality PET/SPECT/CT scanners are particularly useful in this context due to the easier registration process between images. In this study, we propose to include PET, SPECT and CT information in the reconstruction of PET data. The method is applied to Yttrium-90 ( ^90 Y) data. Data from a NEMA phantom filled with ^90 Y were used for validation. PET, SPECT and CT data from 10 patients treated with Selective Internal Radiation Therapy (SIRT) were used. Different combinations of prior images using the Hybrid kernelized expectation maximization were investigated in terms of VOI activity and noise suppression. Our results show that triple modality PET reconstruction provides significantly higher uptake when compared to the method used as standard in the hospital and OSEM. In particular, using CT-guided SPECT images, as guiding information in the PET reconstruction significantly increases uptake quantification on tumoral lesions. This work proposes the first triple modality reconstruction method and demonstrates up to 69 ^90 Y patient data. Promising results are expected for other radionuclide combination used in theranostic applications using PET and SPECT.

We are developing a stationary tomographic molecular breast imaging system consisting of two planar CZT detectors and multi-pinhole collimators. One downside is that the limited angular sampling results in a limited spatial resolution perpendicular to the detectors. Here we investigate the potential to improve both resolution and sensitivity over a reduced field-of-view (FOV) by introducing some flexibility in the detector orientation. By tilting the external parts of the detector arrays towards the centre, the angular sampling and sensitivity could be improved in the central FOV. The system would still be stationary during acquisition. We performed computer simulations, assuming 198 mm wide detector arrays, divided into 3 sections, such that the first and last one could be tilted by different angles. The activity distribution consisted of an elliptical cylinder containing three line-sources on the mid-plane between the detectors at different horizontal distances from the centre (0, 25 and 50 mm). On average for the three sources, the resolution was improved (reduced FWHM) by factors of 0.92, 0.81, and 0.76 in the x-direction and 0.88, 0.72 and 0.64 in the y-direction for detector tilt angles of 10°, 20° and 30°, respectively. The sensitivity was increased by factors of 1.19, 1.45 and 1.72, respectively, for the same angles. The improvements are due to both increased angular sampling and reduced detector distance. Our simulations show that, by introducing some flexibility in the detector configuration, it is possible to improve both spatial resolution and sensitivity by sacrificing part of the FOV. However, as the standard configuration of two opposing parallel planar detectors could be easier to integrate in a clinical system, we will explore different pinhole configurations as an alternative.

Abstract Background Increased pulmonary $$^{18}{}$$ 18 F-FDG metabolism in patients with idiopathic pulmonary fibrosis, and other forms of diffuse parenchymal lung disease, can predict measurements of health and lung physiology. To improve PET quantification, voxel-wise air fractions (AF) determined from CT can be used to correct for variable air content in lung PET/CT. However, resolution mismatches between PET and CT can cause artefacts in the AF-corrected image. Methods Three methodologies for determining the optimal kernel to smooth the CT are compared with noiseless simulations and non-TOF MLEM reconstructions of a patient-realistic digital phantom: (i) the point source insertion-and-subtraction method, $$h_{pts}$$ h pts ; (ii) AF-correcting with varyingly smoothed CT to achieve the lowest RMSE with respect to the ground truth (GT) AF-corrected volume of interest (VOI), $$h_{AFC}$$ h AFC ; iii) smoothing the GT image to match the reconstruction within the VOI, $$h_{PVC}$$ h PVC . The methods were evaluated both using VOI-specific kernels, and a single global kernel optimised for the six VOIs combined. Furthermore, $$h_{PVC}$$ h PVC was implemented on thorax phantom data measured on two clinical PET/CT scanners with various reconstruction protocols. Results The simulations demonstrated that at $$<200$$ < 200 iterations (200 i), the kernel width was dependent on iteration number and VOI position in the lung. The $$h_{pts}$$ h pts method estimated a lower, more uniform, kernel width in all parts of the lung investigated. However, all three methods resulted in approximately equivalent AF-corrected VOI RMSEs (<10%) at $$\ge$$ ≥ 200i. The insensitivity of AF-corrected quantification to kernel width suggests that a single global kernel could be used. For all three methodologies, the computed global kernel resulted in an AF-corrected lung RMSE <10% at $$\ge$$ ≥ 200i, while larger lung RMSEs were observed for the VOI–specific kernels. The global kernel approach was then employed with the $$h_{PVC}$$ h PVC method on measured data. The optimally smoothed GT emission matched the reconstructed image well, both within the VOI and the lung background. VOI RMSE was <10%, pre-AFC, for all reconstructions investigated. Conclusions Simulations for non-TOF PET indicated that around 200i were needed to approach image resolution stability in the lung. In addition, at this iteration number, a single global kernel, determined from several VOIs, for AFC, performed well over the whole lung. The $$h_{PVC}$$ h PVC method has the potential to be used to determine the kernel for AFC from scans of phantoms on clinical scanners.

Molecular breast imaging (MBI) has been shown to have high sensitivity for lesion detection, particularly in patients with dense breasts where conventional mammography is limited. However, relatively high radiation dose and long imaging time are limiting factors. Most current MBI systems are based on planar imaging. Improved performance can be achieved using tomographic techniques, which normally involve detector motion. Our goal is to develop a low-dose stationary tomographic MBI system with similar or better performance in terms of lesion detection compared to planar MBI. The proposed system utilizes two opposing cadmium zinc telluride detectors with high intrinsic resolution and depth of interaction (DOI) capability, combined with densely packed multipinhole collimators. This leads to improved efficiency and adequate angular sampling, but also to significant multiplexing (MX), which can result in artefacts. We have developed de-MX algorithms that take advantage of the DOI information. We have performed both analytic and Monte Carlo simulations to demonstrate the feasibility, optimize the design and investigate the expected performance of the proposed system. Lesion detectability was preserved with reduction of acquisition time (or radiation dose) by a factor of 2 compared to planar images at the lowest reported dose. The first prototype is under evaluation at Kromek.

A novel stationary tomographic Molecular Breast Imaging system (MBI) is being developed by Kromek and University College London. The new system combines the superior energy, position and depth of interaction (DOI) resolution of CZT detectors, wide range angular sampling of MPH collimators and novel de-multiplexing image reconstruction to mitigate the MBI drawbacks (higher than mammography patient dose and long imaging time of 40 min) without compromising diagnostic content.A new detector simulation model is being developed to facilitate the design and optimisation of a large MBI prototype. The existing simulation model, based on the principles described by Prettyman in 1999, has been extended to calculate a time-dependent solution, delivering charge pulse shapes, and using them to simulate the operation of a pulse shaping amplifier. The new model delivers more precise modelling of inter-pixel charge sharing and DOI effects, required for optimisation of the new MBI camera design and creation of large training data sets for machine learning algorithms for data analysis and image reconstruction.

The INSERT is the world’s first clinical SPECT-MRI brain imaging system based on scintillation detectors with a silicon photomultiplier readout. Here, we demonstrate its use within a clinical MRI environment for the first time. Using a standard transmit-receive head coil, and with an appropriate selection of a custom MRI sequence (GRE), we overcome mutual interference. The INSERT and its bulky 50kg tungsten collimator introduce magnetic field inhomogeneity. Due to the specific MRI-compatible collimator design, inhomogeneity is compensated by shimming, leading to simultaneous acquisition. We process the SPECT data acquired alongside the MRI sequence to evaluate the SPECT system performance and the impact of the MRI. Finally, we present a set of simultaneous SPECT-MRI acquisitions, demonstrating multimodal imaging capabilities, albeit with a limited MRI sequence.

Single-photon emission computed tomography (SPECT) systems with pinhole collimators are becoming increasingly important in clinical and preclinical nuclear medicine investigations as they can provide a superior resolution-sensitivity trade-off compared to conventional parallel-hole and fanbeam collimators. Previously, open-source software did not exist for reconstructing tomographic images from pinhole-SPECT datasets. A 3D SPECT system matrix modelling library specific for pinhole collimators has recently been integrated into STIR, an open-source software package for tomographic image reconstruction. The pinhole-SPECT library enables corrections for attenuation and the spatially variant collimator–detector response by incorporating their effects into the system matrix. Attenuation correction can be calculated with a simple single line of response or a full model. The spatially variant collimator–detector response can be modelled with a point spread function and depth of interaction corrections for increased system matrix accuracy. In addition, improvements to computational speed and memory requirements can be made with image masking. This work demonstrates the flexibility and accuracy of STIR’s support for pinhole-SPECT datasets using measured and simulated single-pinhole SPECT data from which reconstructed images were analysed quantitatively and qualitatively. The extension of the open-source STIR project with advanced pinhole-SPECT modelling will enable the research community to study the impact of pinhole collimators in several SPECT imaging scenarios and with different scanners.

Penalised PET image reconstruction algorithms are often accelerated during early iterations with the use of subsets. However, these methods may exhibit limit cycle behaviour at later iterations due to variations between subsets. Desirable converged images can be achieved for a subclass of these algorithms via the implementation of a relaxed step size sequence, but the heuristic selection of parameters will impact the quality of the image sequence and algorithm convergence rates. In this work, we demonstrate the adaption and application of a class of stochastic variance reduction gradient algorithms for PET image reconstruction using the relative difference penalty and numerically compare convergence performance to BSREM. The two investigated algorithms are: SAGA and SVRG. These algorithms require the retention in memory of recently computed subset gradients, which are utilised in subsequent updates. We present several numerical studies based on Monte Carlo simulated data and a patient data set for fully 3D PET acquisitions. The impact of the number of subsets, different preconditioners and step size methods on the convergence of regions of interest values within the reconstructed images is explored. We observe that when using constant preconditioning, SAGA and SVRG demonstrate reduced variations in voxel values between subsequent updates and are less reliant on step size hyper-parameter selection than BSREM reconstructions. Furthermore, SAGA and SVRG can converge significantly faster to the penalised maximum likelihood solution than BSREM, particularly in low count data.

We previously presented a dual injection protocol for dynamic whole-body (DWB) PET using multi-bed multi-pass acquisition. The purpose was to allow for a reduced scanning time, while still providing accurate parametric images for irreversible tracers, such as [ 18 F]-FDG. The protocol consisted of a DWB scan with a delayed start, followed by a standard SUV scan. A 2 nd injection (~10% of the dose) was given during the dynamic scan for estimating the initial part of the input function. However, this protocol caused complications to the kinetic modelling as well as possible effects of the 2 nd injection on the SUV scan. We therefore now propose an alternative protocol, with the 2 nd injection after the SUV scan. PET data were simulated based on the XCAT phantom, Poisson noise was added, and images were reconstructed with OSEM. The input function was obtained from the LV of the heart and fitted with an analytical function. The area under the curve (AUC, 0-60 min) was calculated and the effect of different injection fractions and scan start-times were investigated. A patient study was also performed, with data acquisition from 0 to 75 min p.i. A 2 nd injection (9%) was given after 70 min. The input function was estimated using different data sub-sets and the AUC was compared with the gold standard. The simulation results showed that a 2 nd injection fraction of 10% and a scan start-time of 30 min gave accurate results, while smaller fraction or later start gave increased bias and variance. The results of the patient study were in general agreement with the simulation results, with a bias of ~2.7% for scan start times of 13-43 min and 5.1% at 50 min p.i. In conclusion, we found that a scanning protocol with a late start and a second injection at the end is a practical way to obtain accurate whole-body parametric PET images with a reduced scanning time.

BACKGROUND:Selective internal radiation therapy with Yttrium-90 microspheres is an effective therapy for liver cancer and liver metastases. Yttrium-90 is mainly a high-energy beta particle emitter. These beta particles emit Bremsstrahlung radiation during their interaction with tissue making post-therapy imaging of the radioactivity distribution feasible. Nevertheless, image quality and quantification is difficult due to the continuous energy spectrum which makes resolution modelling, attenuation and scatter estimation challenging and therefore the dosimetry quantification is inaccurate. As a consequence a reconstruction algorithm able to improve resolution could be beneficial.METHODS:In this study, the hybrid kernelised expectation maximisation (HKEM) is used to improve resolution and contrast and reduce noise, in addition a modified HKEM called frozen HKEM (FHKEM) is investigated to further reduce noise. The iterative part of the FHKEM kernel was frozen at the 72nd sub-iteration. When using ordered subsets algorithms the data is divided in smaller subsets and the smallest algorithm iterative step is called sub-iteration. A NEMA phantom with spherical inserts was used for the optimisation and validation of the algorithm, and data from 5 patients treated with Selective internal radiation therapy were used as proof of clinical relevance of the method.RESULTS:The results suggest a maximum improvement of 56% for region of interest mean recovery coefficient at fixed coefficient of variation and better identification of the hot volumes in the NEMA phantom. Similar improvements were achieved with patient data, showing 47% mean value improvement over the gold standard used in hospitals.CONCLUSIONS:Such quantitative improvements could facilitate improved dosimetry calculations with SPECT when treating patients with Selective internal radiation therapy, as well as provide a more visible position of the cancerous lesions in the liver.

A novel stationary tomographic Molecular Breast Imaging (MBI) system is currently under development, with the aim of obtaining high image-quality with low dose and short scanning time. The system is based on dual opposing CZT detector arrays and multi-pinhole collimators. We have recently modified the iterative image reconstruction procedure by incorporating a novel relaxation scheme, in order to make the image contrast and noise properties more uniform throughout the field-of-view. In addition, we have introduced a post-reconstruction image denoising step based on the non-local means (NLM) filter. In view of the significant effect that these steps had on the image quality, we performed a new system parameter optimization. The parameters investigated were pinhole size, opening angle and separation, as well as the number of reconstruction iterations and the degree-of-smoothing parameter of the NLM filter. The optimization was performed based on simulated data, by maximizing the contrast-to-noise ratio (CNR) in the images. We found that the optimal system parameters were not so different with the new data processing steps as compared to previous results, while the CNR was improved by a factor > 3.

A prototype clinical brain single-photon emission computed tomography (SPECT) insert has been designed for use in simultaneous SPECT/MRI. The system utilizes novel slit-slat collimators which, like pinhole collimators, suffers from parallax errors due to the large incident angle of photons. A statistical algorithm has been developed to determine the depth-of-interaction (DOI) with a view to improving image performance. The importance of DOI correction was demonstrated using Monte Carlo simulation. This simulation also indicated that four DOI layers ( $3 \times 1.5\, \textrm {mm} + 3.5\, \textrm {mm}$ ) may be sufficient. The improvement in event localization was demonstrated on a single detector before implementing the algorithm on the full clinical prototype where some limitations in event localization in layers close to the readout plane were observed. Nevertheless, DOI enabled the rejection of poorly localized events with improved resolution in reconstructed line sources.