Small-animal PET and SPECT are essential tools in the development of pharmaceutical drugs, radionuclide therapies, and diagnostic radiotracers. In particular, theranostic tracers, which combine predictive imaging biomarkers with therapeutic agents, have become increasingly important in preclinical research. Imaging these tracers is challenging because their gamma emissions often span a wide energy range, sometimes extending beyond conventional energies, and because activity concentrations can be low. These challenges highlight the need for imaging systems capable of operating over a broad energy range while maintaining high sensitivity and quantitative accuracy.
To address these challenges, the Biomedical Imaging group at TU Delft, in close collaboration with MILabs B.V., has focused on extending the maximum imageable gamma energy in preclinical imaging. This work resulted in the development of the Versatile Emission Computed Tomography (VECTor) system, a fully integrated preclinical PET/SPECT platform employing pinhole collimation for both modalities. VECTor has demonstrated high performance across an extended gamma energy range up to 1 MeV, achieving 0.4 mm resolution in collimated ⁹⁹ᵐTc-SPECT and 0.6 mm resolution in ¹⁸F-PET. Building on this foundation, this simulation-based study investigates software and hardware optimizations to improve the system’s quantitative imaging performance and sensitivity across this wide energy range.
On the software side, several joint reconstruction techniques were evaluated to improve image quality under low-count conditions. Three approaches, Single-Band Joint Reconstruction (SB-JR), Mixed Multi-Band Joint Reconstruction (mMB-JR), and Multi-Band Joint Reconstruction (MB-JR), were assessed using Monte Carlo simulations of resolution phantoms filled with ²²⁵Ac, ²²⁶Ac, or ⁸⁹Zr. Among these methods, MB-JR consistently provided the best image resolution and highest contrast-to-noise ratio across all isotopes and activity levels.
In parallel, hardware optimizations of the gamma camera were explored to improve performance at high energies. Simulations examined the effect of increasing the NaI(Tl) scintillation crystal thickness from the conventional 9.5 mm to 20 mm and 40 mm, combined with optimized light guides and four photomultiplier tube (PMT) geometries. For 511 keV photons, increased crystal thickness yielded substantial sensitivity gains (27% for 20 mm and 57% for 40 mm), with only modest spatial resolution losses when using cost-effective PMTs and potential resolution improvements when smaller PMTs were employed.
Finally, two novel collimator designs optimized for high-energy gamma emissions were evaluated. The Twisted Clustered Pinhole (TCP) collimator retains the clustered geometry of the standard clustered pinhole (CP) design while enabling narrower pinhole opening angles by twisting pinholes around their cluster central axis. For 511 keV (¹⁸F) and 909 keV (⁸⁹Zr) gamma emissions, TCP improved both sensitivity (15.6% for ¹⁸F and 29.4% for ⁸⁹Zr) and spatial resolution compared to CP.
The Super-Cluster (SC) collimator employs a simpler geometry with uniformly distributed pinholes, allowing larger pinhole diameters and a more flexible adjustment of the resolution–sensitivity trade-off. Relative to CP, SC achieved sensitivity gains of up to threefold for ¹⁸F and twofold for ⁸⁹Zr, particularly benefiting low-activity imaging through improved resolution and contrast recovery.
Together, these results demonstrate effective targeted strategies to extend VECTor applicability for high-energy and low-activity preclinical PET/SPECT imaging.
... Read more

Proton radiotherapy is a cancer therapy that uses ionising radiation in the form of protons, offering an alternative and complement to conventional radiotherapy with high energy X-rays. In comparison to X-rays, proton beams have a finite range in tissue and deposit dose more locally. As a consequence, proton radiotherapy has the potential to deliver a higher dose to the tumour while better sparing surrounding healthy tissue. However, the delivered dose distribution in the case of protons is much more sensitive to changes in patient anatomy compared to X-rays, which therefore requires an accurate knowledge of proton stopping power ratio (SPR) values of the tissues to be treated..... ... Read more

By implementing cone-beam computed tomography (CBCT) into proton therapy radiationunits, a predetermined treatment plan could be updated prior to each treatment fraction ac-cording to the changing anatomy of the patient for better dose distribution. However, CBCTdoes not produce high enough image quality compared to fan-beam CT (FBCT), which nowa-days is used to build a treatment plan based on the stopping power ratio (SPR) of the objectivetissue in the body. Spectral CBCT is a promising method to potentially increase the imagequality of conventional CBCT. A provided joint reconstruction spectral CBCT algorithm inMATLAB is used to determine whether the low image quality of CBCT can be improved, as jointreconstruction algorithms have been proven to improve image quality for FBCT in practicalexperiments. The provided code is converted to Python, after which equivalent results areensured using comparative analysis. A phantom with multiple biological materials is thenimplemented in this acquired Python code to investigate the quality of the reconstruction im-ages. Moreover, SPR maps and a VMI are constructed from these images and their qualitydetermined.The results show that for 10 to 12 iterations, the used reconstruction provides the reconstructedimages with the lowest mean squared error (MSE). For higher iterations, the image becomesoversmoothed and loses quality. In future research, the used joint reconstruction algorithmshould be compared to non-joint reconstruction algorithms to investigate the impact of thistechnique on the image quality, after which it could be applied to more realistic data. ... Read more

Computed Tomography (CT) imaging is an important step in treatment planning for proton therapy. The conversion from Hounsfield Unit (HU) to stopping power ratio (SPR) accounts for more than half of the 3.5% error that is considered in treatment planning. The aim of the study is to investigate the efficacy of photon-counting CT (PCCT) for predicting the SPR in proton therapy. Its performance will be compared to single-energy CT (SECT) and dual-energy CT (DECT).

Fresh tissue samples—steak, ground beef, and bone marrow—were secured in 3D-printed holders to minimize air inclusion and maintain tissue stability during measurements. These
samples were secured in a head and a body configuration of a phantom and CT scans were conducted using standard clinical protocols. SPR values were derived from the CT data
using automated DirectSPR software for the DECT and PCCT scans, and converted from HUs to SPR using a Hounsfield look up table (HLUT) for the SECT scans. Results were validated
against actual SPR measurements obtained via proton beam measurements.

Two measurement series have been performed. For the validation of the first measurement series, proton beam energies of 150 and 175 MeV both resulted in an SPR value of 1.00 for the
ground beef sample. For the steak sample, SPR values of 1.02 and 1.03 were found. The CT-based SPR predictions resulted in errors up to 1.5% for ground beef and 3.5% for steak. For
the validation of the second measurement series, SPR values for ground beef (0.89-0.92), steak (0.96-0.98), and bone marrow (0.87-0.91) were obtained. The percentage error between the CT-based SPR predictions and the validation measurements showed no substantial differences between the CT-modalities, except for SECT providing lower SPRs for the bone marrow samples than DECT and PCCT. The validation measurements of the second series resulted in lower SPR values than found in literature and the first
series, indicating possible methodological inconsistencies. Factors such as sample heterogeneities and measurement errors might have contributed to these inconsistencies, but do not account for these inconsistencies completely. These results highlight the need
for further research with standardized phantoms and animal tissue samples to evaluate PCCT’s true potential in combination with DirectSPR prediction.

From this study it cannot be concluded that PCCT demon-
strates an advantage over SECT and DECT as CT-modality for
the HU-SPR conversion in proton therapy treatment planning ... Read more

Motivation: Effective management of liver tumors through thermal ablation requires precise monitoring of the ablation zone to ensure successful treatment outcomes. Computed tomography (CT) thermometry offers a promising non-invasive solution to monitor if tumor cells have been heated to the lethal temperature threshold. However, achieving reproducible, precise, and accurate temperature measurements remains a challenge, particularly due to metal artifacts introduced by the ablation equipment.

Purpose: This study investigates the applicability of spectral CT thermometry in monitoring liver microwave ablation. It compares the reproducibility, precision and accuracy of CT thermometry on attenuation value images, with CT thermometry on physical density maps using spectral CT. Furthermore, it identifies the optimal metal artifact reduction (MAR) method — among O-MAR, deep learning-MAR, spectral CT, or a combination — to reduce needle artifacts and improve CT thermometry precision.

Materials and Methods: Four liver-mimicking gel phantoms embedded with temperature sensors underwent a 10-minute, 60W microwave ablation imaged by dual-layer spectral CT using a Philips CT7500 scanner. Each scan was processed to reconstruct standard 120 kVp images alongside physical density maps, which were derived from virtual monochromatic imaging (70 - 150 keV) and effective atomic number maps. During each procedure, 23 CT scans were acquired to monitor attenuation and physical density values in proximity of the ablation antenna over time. Attenuation-based and physical density-based thermometry models were tested for reproducibility (coefficient of variation) over three repetitions; a fourth repetition focused on accuracy (Bland-Altman analysis). MAR techniques were applied to a single repetition to evaluate temperature precision in artifact-corrupted slices.

Results: The correlation between CT value and temperature was highly linear with an R-squared value exceeding 96% for both attenuation and physical density-based thermometry. Model parameters for attenuation-based and physical density-based thermometry were -0.38 HU/ºC and 0.00039 ºC-1, with coefficients of variation of 0.023 and 0.067, respectively, indicating a high reproducibility. CT thermometry precision increased with distance from the ablation antenna, the use of attenuation maps and deep learning-MAR. Physical density maps generated at 150 keV alone and in combination with O-MAR and deep learning-MAR reduced needle artifacts by 73% on average (p=0.003) compared to attenuation images. Bland-Altman analysis reveals limits-of-agreement of -7.7°C to 5.3°C and -9.5°C to 8.1°C for attenuation and physical density-based thermometry, respectively.

Conclusion: Spectral CT has the potential to make CT thermometry more universally applicable. This study demonstrates the effectiveness of spectral CT thermometry for non-invasive temperature monitoring during liver microwave ablation. It shows that using spectral physical density maps at 150 keV, alongside deep learning-MAR and O-MAR, enhances temperature accuracy and minimizes metal artifacts. However, standardizing thermometry parameters across different patient conditions remains a challenge. Future enhancements in photon counting CT and deep learning technologies could further refine this method, ultimately reducing the risk of local tumor recurrence. ... Read more

Purpose: Organ motion may have an impact on the radiation dose administered in radiotherapy. Motion can occur during treatment (intrafractional motion) or between fractions of treatment (interfractional motion). A clinically relevant dosimetric impact on OAR would mean that additional measures such as 4D-CT planning or a PRV are needed. In this study, the dosimetric impact of intra- and interfractional motion is studied for TBI delivered with VMAT is studied.
Method: For the intrafractional part, motion magnitude was determined with 4D-CT data, and this magnitude was used to expand and reduce the volumes of the OAR in 27 VMAT-TBI plans, dose was analyzed in these volumes. For the interfractional part, the CBCT images taken right before treatment are registered with the planning CT. The planning CT is then deformed to match the anatomy of the CBCT for each fraction. The dose is recalculated on this deformed CT and the mean dose in OAR volumes is evaluated for each fraction.
Results: For the intrafractional part, on average, an expansion with a typical respiratory motion magnitude results in a 0.3 Gy higher mean dose in the kidneys. This expansion results in a 0.6 Gy higher mean dose in the left lung and a 0.8 Gy higher dose in the right lung. For the interfractional part, the mean dose difference between the fraction with the lowest mean dose and the one with the highest mean dose was 0.04 Gy on average.
Conclusion: After discussing these dose differences with a radiooncologist, it was concluded that these differences were not clinically relevant. Therefore, 4D-CT planning and a PRV are not necessary. ... Read more

Radiotherapy is one of the main treatments for cancer and relies heavily on CT images to calculate radiation dose. With research on radiotherapy moving to adaptive treatments aiming to calculate these doses at real-time speeds while maintaining high precision, a need for accurate CT imaging at comparable real-time speeds has emerged. Currently, the best performing CT image reconstruction methods are iterative reconstruction (IR) methods, which suffer from slow reconstruction speed. Faster methods are accompanied by artifacts due to the implementation of simplified physics models.

Recently, the Dose Transformer Algorithm (DoTA) [47], [48] and improved DoTA (iDoTA) [49] have shown to successfully calculate radiation therapy dose by modelling particle transport in 3D with the use of a neural network. By implementing a Transformer architecture [62], DoTA is able to capture the relationship between elements in a 3D CT volume while processing it as an input sequence. This results in an accurate prediction of particle transport, while significantly reducing computation times compared to other methods.

A neural network based on the DoTA-architecture is presented. It predicts projection data from CT input, modelling the x-ray photon transport. The network processes 2D CT images as a sequence of 1D lines. The ground truth data contains Monte Carlo projections of cylindrical water phantoms with inserts composed of five different materials.

The predictions are compared to Monte Carlo projections and raytracing projections generated with Astra Toolbox [45], as well as a Two-Angle Convolution (TAC) network [11]. The average NRMSE of the Transformer predictions was 0.725% compared to 2.20% and 1.09% respectively for the raytracer and TAC. The Transformer showed the ability to predict from unseen types of geometries and intensity values. Due to bias in the training data, it does not generalize well to input phantoms with an unseen outer shape.

Two phantoms were reconstructed using the network within an IR algorithm. For the Transformer and raytracer, the highest achieved CNR values are similar for low-contrast regions (6.88 and 8.28 for the raytracer compared to 7.10 and 7.35 for the Transformer) as well as high-contrast regions (37.40 and 41.94 for the raytracer compared to 39.01 and 39.80 for the Transformer). Convergence rates based on low-contrast CNR are higher for the raytracer (39 and 34 iterations compared to 41 and 41 iterations for the Transformer, respectively). The Transformer performs significantly better than the raytracer with respect to beam-hardening artefacts. The IR algorithm has not been tuned for use with the Transformer, suggesting that a higher performance is obtainable with adjustments such as the implementation of a different backprojector or a different value for correction factors used in the algorithm.

Limitations in prediction quality are likely related to factors outside of the model predictions, such as biases in the input data and resolution loss due to interpolation of the input data. When its prediction speed is optimised, the CT Transformer model has potential to replace conventional forward projections in IR methods, achieving Monte Carlo-level accuracy with a fraction of the computation time.
... Read more

Background Dynamic SPECT scanning provides a non-invasive way to image the time-dependent distribution of radio-labelled tracers inside living tissue. Beside human medicine, dynamic SPECT also finds its applications in pre-clinical research on small animals. In pre-clinical research, multi-pinhole collimators are used to enable high-resolution sub-millimeter imaging. Conventional iterative reconstruction methods, such as Maximum Likelihood Expectation Maximisation (MLEM) perform poorly in reconstructing the noisy and low-count scans in dynamic SPECT. This limits the temporal resolution that can be achieved.
Method The reconstruction of noisy, low-count time-frames can be aided by incorporating information from earlier and later time-frames. Wang and Qi (2015) published the paper 'PET Image Reconstruction Using Kernel Method', with a proposed method entitled Kernelised Expectation Maximisation (KEM) for dynamic PET, a method that uses principles from Machine Learning, such as Support Vector Machines and the 'kernel trick' to incorporate prior information in the reconstruction algorithm. This method is highly adjustable due to a number of input parameters of the method. In this paper, KEM is implemented for dynamic multi-pinhole SPECT. The effects of the KEM parameters are explored in computer simulations. Two different dynamic phantoms are used, one of the striata in a mouse brain which were adapted from a paper by Vastenhouw et al. (2007) and one of the hepatobiliary system adapted from Vaissier et al. (2012). The results of KEM are benchmarked against conventional MLEM with a Gaussian post-filter. 
Results In high-count simulations, the MLEM reconstructions had a lower mean-squared-error than the KEM image, while the signal-to-noise ratio of KEM was better than MLEM. The images produced after 200 iterations were indistinguishable, however. In the low-count regime, KEM was shown to be more resistant to noise than MLEM. Varying the input parameters of KEM gave rise to differences in performance, such as (over-)smoothing effects and a different level of noise-suppresion in the reconstructed image. 
Conclusions In the simulations used in this paper, KEM was shown to outperform conventional MLEM with a Gaussian post-filter from low-count projections. The optimal input parameters of the KEM algorithm, however, need to be found ad hoc by searching the parameter space. Further research should look into finding a set of rules or guidelines for finding the optimal parameters. ... Read more

CT is a versatile medical imaging method to diagnose and monitor patient diseases. However, varying patient characteristics and scan settings create challenges in maintaining consistent image quality, complicating image comparisons, especially across different sources. The reconstruction kernel in CT image reconstruction is a key parameter in the reconstruction process. It affects image characteristics such as sharpness, contrast, and noise. There is an urgent need for a method that effectively compares and categorises reconstruction kernels from different vendors using real patient scans. Therefore, this thesis focuses on extracting features from real patient images to facilitate kernel comparisons within and across manufacturers.

This research aims to create a machine learning (ML) method that categorises reconstruction kernels from various vendors into groups based on their sharpness. This categorisation relies on image features extracted directly from real patient scans with diverse scan parameters.

Two methods were explored using CT datasets from the National Lung Screening Trial (NLST) and the Lung Image Database Consortium (LIDC-IDRI). The first method uses noise features, specifically the standard deviation of homogeneous regions and the central frequency derived from the noise power spectrum. These features were used in a linear support vector machine (SVC_noise). The second method uses radiomic features extracted from selected homogeneous regions and is trained using a random forest classifier (RFC_radiomics).

Both models were evaluated using accuracy and ROC AUC. McNemar’s test was used to assess statistical differences. Due to the lack of ground truth, a subset of smooth and sharp kernels was used for training and validation, and remaining kernels were classified to establish a reference ground truth.

Both models performed strongly on 270 cases with 37 reconstruction kernels. The SVC_noise model achieved a ROC AUC of 0.97 with eight misclassifications, while the RFC_radiomics model achieved 0.96 with ten misclassifications. McNemar’s test showed no significant difference between the models. Only one discrepancy in ground truth assignment was observed for kernel “B50s”.

In conclusion, both models demonstrate strong and comparable performance in distinguishing kernel sharpness while being robust to variations in scan parameters and patient characteristics. However, results are preliminary and should be interpreted with caution when extending to other reconstruction kernels or imaging settings. ... Read more

This thesis aimed to validate Beekman's patent on how collimation could virtually reduce the focal spot size of an existing small-animal cone beam Computed Tomography (CT) system to diminish the penumbra effect. The ever-remaining drive to improve the spatial resolution in CT for enhanced image quality introduces the need for small focal spot sizes, as the focal spot size is directly related to the geometric unsharpness in the image. Therefore, a collimation method was proposed to virtually reduce the focal spot of existing systems as an alternative to fully replace the current X-ray tube.

A collimator was designed, consisting of numerous tiny hexagonal-shaped holes focused at the center of the focal spot. Theoretical derivations were formulated for its dimensions, and the collimator's efficacy was validated using Monte Carlo simulations. It was concluded that it is theoretically achievable to use collimation to virtually reduce the focal spot size to an arbitrarily chosen smaller virtual focal spot for existing CT systems, significantly reducing the penumbra effect, without requiring any integral changes to the X-ray tube. However, the collimator's practical suitability and manufacturing feasibility were problematic due to its significantly low collimator sensitivity and exceptionally tiny dimensions.

Future work could build on this thesis by obtaining reconstructions from multiple projections of the virtual focal spot to quantitatively assess their theoretically improved spatial resolution. The quantitative confirmation could further establish the theoretical effectiveness of focal spot collimation for future work to enhance reconstructions to uncover valuable information previously hidden.
... Read more

Photon-counting detectors (PCD) for medical X-ray computed tomography (CT) are designed to measure the number of X-ray photons incident on a detector pixel as well as the energy of the individual X-rays. They are expected to yield improvements in image quality for a given radiation dose, and to offer opportunities for spectral imaging beyond dual-energy techniques. However, the fluence rate incident on the detector can exceed 108 mm-2 s-1 in CT, so that the detector pulses generated by the X-rays likely pile up on each other, which distorts the measurement. The semiconductors CdTe and Cd1-xZnxTe (CZT, x ≈ 0.1-0.2) are commonly considered efficient X-ray absorbers that provide sufficient rate capability (fast pulses in the order of 101 ns and a high pixel density ≥ 4 mm-2) and energy resolution (8-20% FWHM at 60 keV). In such detectors, an X-ray is converted into electron-hole pairs, which travel to (pixelated) electrodes, on which they induce a current pulse. However, the cost-effective synthesis of material of sufficient quality to make this a stable and reliable detection process appears to remain an issue. Thus, the aim of this thesis is to explore the photon-counting performance, e.g., the rate capability and energy resolution, of an alternative detector concept based on a scintillator, which converts an X-ray into a light pulse, and a silicon photomultiplier (SiPM), which detects the light. Since such a detector relies on light rather than charge transport, it may enable cost-effective manufacturing of stable and reliable PCDs...
... Read more

In this research the effectiveness of analytical neural networks compared to the maximum likelihood method on the prediction of spatial and DOI positioning of a Gamma detector with a NaI(Tl) scintillator of size 590mm x 470mm x 40mm (x,y,z), with a glass lightguide of size 620mm x 500mm x 4mm and a PMT area of 620mm x 500mm x 40mm with 2-inch round PMTs with a Bialkali photocathode is presented. This is done by training neural networks with different cost function, different amounts of hidden layers and different amounts of neurons per hidden layer, trained on different amounts of training data. The resolution of the predictions of the testing data are compared with those of the maximum likelihood method. It was concluded that the neural network with best spatial resolution, had the Huber loss function as cost function, 4 hidden layers and 512 neurons per hidden layer and was trained on 29,970 datapoints. The FWHM and the FWTM were 3.83 ± 0.54 mm and 12.49 ± 1.19 mm respectively, while the FWHM and the FWTM of the maximum likelihood method were 3.31 mm and 12.13 mm respectively. The resolution of the neural network was lower than that of the maximum likelihood method. The same was done for the DOI resolution, here a neural network with mean squared error as cost function, 4 hidden layers and 64 neurons per hidden layer trained on 9,990 datapoints, gave the best the resolution with FWHM and FWTM equal to 6.00 ±0.50 mm and 11.94 ± 0.94 mm respectively. The FWHM and FWTM of the maximum likelihood method were 6.16 mm and 11.22 mm respectively. This made the DOI resolution of the neural network higher then that of the maximum likelihood method. Finally different ideas were presented to increase the resolution of the neural network. These were: training the neural network on independent data, split the neural network in a spatial part and a DOI part, create a more complex architecture and making use of a convolutional neural network.
... Read more

Introduction: Radiation is an effective treatment to increase overall mean survival of patients with metastatic brain tumours, however, damage to healthy tissue is inevitable. Radiation can cause dysfunction of the cerebrovasculature which is hypothesised to induce cognitive decline in patients after radiotherapy (RT). A new method, the cerebrovascular stress test, is able to visualise cerebrovascular reactivity (CVR) which is the ability of the vessels to dilate after a vasoactive stimulus. Research suggests a link between reduced CVR and cognitive impairment in patients, however, current studies have not yet shown if CVR is reduced in patients with metastatic brain tumours. This thesis aims to assess CVR in patients with metastatic brain tumour at baseline and after RT.
Methods: In this thesis, 13 patients with metastatic brain metastases were included and underwent a magnetic resonance imaging (MRI) scan with a vasoactive stimulus at baseline and three months after the same MRI-scan with stimulus. On the same day as the baseline MRI-scan, the patients received RT. CVR maps were calculated using the MRI-scan with the vasoactive stimulus. An additional computed tomography scan was obtained from each patient prior to their first MRI scan. All scanning data was brought into spatial correspondence with a developed image registration pipeline. After the scanning data was registered image analysis was performed using a VOI- and dose-based analysis.
Results: The performance of the image registration pipeline was close to optimal for the MRI scans, and 69% for the baseline CT scan. The image analysis found a significant increase of CVR at an increasing distance from the tumour for white matter (WM) (p = 0.050). For grey matter (GM) and WM, a significant increase of CVR was found at 14 pixels away from the tumour in comparison to 2 pixels away from the tumour (WM: p = 0.039, and GM: p = 0.046). In the dose-based analysis, a nonsignificant decrease of mean CVR was found after RT. The decrease in CVR after RT did also not depend on the received dose.
Conclusions: This thesis developed an image registration pipeline that can be used in further analysis with this specific patient group and scanning data. The image analysis showed an significant increase in CVR at a distance from the tumour for GM and WM. These results indicate that BM influences the CVR of these patients. However, no conclusions can be drawn based on the dose-based analysis. Additional research needs to be done to relate changes in CVR to cognitive decline in patients with metastatic brain tumours.
... Read more

Biomedical researchers and clinicians are interested in (ab)normal foetal development because it can aid in better understanding human anatomy. To capture this foetal development non-destructive three dimensional (3D) imaging techniques like computed tomography (CT) are used. Visualising foetuses remains a challenge however, as foetuses consist mostly of soft-tissue. Visualisation of soft-tissue with CT scans is difficult because X-rays easily pass through. This consequently results into images with low contrast. Therefore, improving contrast is artificially gained by using chemical compounds called stains. The most effective stain is considered to be Lugol’s solution. A downside of using Lugol’s solution is that the staining process causes extensive soft-tissue shrinkage which is detrimental for morphological analysis. The mechanism of Lugol-induced shrinkage is largely unknown. Some research suggest it is due to an osmotic imbalance between tissue and solution, while others point towards acidification of Lugol’s solution. The goal of this study is to develop an optimum (buffered) Lugol’s solution staining protocol for post-mortem human foetal CT imaging to diminish soft-tissue shrinkage and achieve homogeneous staining. Several variables in the protocol are taken into account such as staining solution concentration, staining time and specimen size. To develop this protocol, multiple tests and measurements (pH, osmolarity, optical density, weight and CT scans) were performed on pork liver samples and two post-mortem human foetuses to monitor acidification of the staining solution, staining progress and staining intensity, while applying two distinct methods: the AMC- and Arthurs method. The main difference between these methods is that the AMC method fixates tissue well before staining (conventional method), while Arthurs method uses a mixture of a fixative and stain simultaneously on fresh tissue. The research suggests that Arthurs method seems best. Even though, both methods led to a homogeneous staining, the AMC method resulted in an average shrinkage of 4.82%, while Arthurs method resulted in a shrinkage of only 1.08%. In addition, Arthurs method leads to a shorter staining protocol. ... Read more

With multi-pinhole collimation systems, high resolutions can be reached with both SPECT and PET. Using tracers with higher-energy gamma-photons increases the effects of pinhole edge penetration, decreasing the resolution of the system. This research aims to design a collimator using 3x3 twisted pinhole clusters to be used in the VECTor system. This new cluster design should allow high-energy tracers such as 89-Zr to be imaged. The sensitivity of this new design was compared at 511 keV and 909 keV with the sensitivity of the VECTor collimator by simulating the sensitivity using GATE.

A collimator design is characterised by its degree of multiplexing and detector coverage. These values are aimed to be the same as for the VECTor collimator to ensure a fair comparison at the simulation stage. Several design approaches have been tested. The final design had the clusters placed in 5 rows on the collimator. The three inner rows had 21 clusters, the two outer rows had 15 clusters. The inner radius of the collimator was increased to prevent pinholes from intersecting.

The pinhole diameters of the final design were modified such that the resolution of the system was the same as the VECTor resolution for both 511 keV and 909 keV photons. To prevent the pinholes from intersecting, either the inner radius of the collimator had to be further increased, or the pinholes had to be smaller. It was chosen to test both options, resulting in a total of three collimator designs to be simulated. A scan lasting 1 hour was simulated with a source the size of the VECTor CFOV. This was done with 511~keV photons using 2 MBq/mL 18-F, and with 909 keV photons using 2 MBq/mL 89-Zr with their respective designs. These simulations resulted in a 1.77 times higher sensitivity to direct photons for the 511 keV collimator, and either a 2.04 or a 2.69 times higher sensitivity to direct photons for the 909 keV photons. Additionally, the total sensitivity of the 511 keV collimator did not significantly change, and the total sensitivity of the 909 keV collimator designs increased with either 8% or 40%. It is concluded that the implementation of 3x3 twisted pinhole clusters in a collimator to be used in the VECTor system has significant benefits over the current collimator when used with high-energy gamma-photons. ... Read more

Preclinical SPECT systems such as the U-SPECT have been able to achieve sub-half-millimetre spatial resolution with the use of cylindrical pinhole collimators. Utilising this type of collimator comes at the cost of a reduction in the total number of detection events that take place. In order to compensate for this either the activity of the radio pharmaceutical must be increased or the exposure time must be extended. Ideally the dose received by a patient or subject is kept to a minimum. The goal of this study was therefore to investigate the application of deep learning to SPECT imaging, specifically to improve low count images to resemble high count images in the projection domain. The projection domain was chosen over the image domain as projections are easily generated in large quantities, while reconstructed images take large amounts of time and computation to generate. A previous BSc Thesis study constructed a neural network (a Perceptual Loss Network) to this end, and it was concluded that the training set was too small and specific for the neural network to be more generally applicable. In this study therefore the training set of the neural network was expanded on with different phantom types such as Derenzo hot rod phantoms, Jaszczak phantoms and uniform phantoms of various shapes and sizes. Phantoms were simulated and measured using the EXIRAD-3D, and projection pairs (low and high count) were generated. Several network architecture improvements were also explored, such as processing the projection images from different detectors together using width concatenation or applying down sampling. Phantom test sets were used in order to determine whether expanding the training set and making adjustments such as width concatenation or applying down sampling had a positive effect on the neural network’s ability to improve varying SPECT projections. The projections of these test sets were improved by the neural networks and then reconstructed to 3D arrays in the image domain. The reconstructions would then be compared against the low count reconstruction as well as those produced by the original neural network. It became apparent that the neural networks do not perform well on very low count projections. It is assumed that this is because the projections have too little information contained within them for the neural networks to determine how to improve them. It may be possible to improve the neural networks by expanding the training set further with very low count projections. The quality of the reconstructions was determined quantitatively using the Contrast to Noise Ratio (CNR) for Derenzo type phantoms and uniformity for uniform type phantoms. Expanding the training set showed slight improvement in reconstruction CNR but was not considered significantly better. Applying width concatenation as well as expanding the training set seemed to improve results further, but the increase in resource requirements and computing time may not be justified for the marginal increase in CNR. Expanding the training set and applying down sampling proved to be very promising, increasing the CNR from anywhere between 0.35 to upwards of 0.75 in some cases. It also showed the most potential when it came to improving physically measured SPECT projections. The uniform phantoms had varying results. Very low count uniform cylinder phantoms were able to be improved by neural networks, but did not seem to benefit much from training on the expanded training set. It also seemed that trying to improve higher count uniform cylinder phantoms was difficult for the neural networks as there seem to be artifacts in the neural network reconstructions that decrease uniformity. It is recommended to further examine and improve the down sampling technique used in this study. It may also be interesting to combine the individual improvements, expanding the training set, using width concatenation and applying down sampling simultaneously, to see whether this can offer a better neural network for improving low count SPECT projections. ... Read more

Breast cancer, being the most common cancer among females, is nowadays routinely diagnosed using X-ray mammography. Though this technique has proven its effectiveness in many cases, X-ray mammography has some disadvantages like reduced diagnostic sensitivity for dense breasts, need for strong breast compression and inability to assess tissues at the molecular level. 
Therefore, there is a need for alternative imaging modalities to improve breast cancer diagnosis. One option is breast scintigraphy, which images the distribution of radiolabelled molecules, called tracers, that concentrate in the tumours in breasts with a planar gamma detector. Different tracers react in different physiological processes with tumours. Therefore imaging a specific tracer can reveal the specific pathological process that is specific for a certain kind of breast tumour. Despite the fact that breast scintigraphy has been reported to have improved diagnostic sensitivity in dense breasts compared to X-ray mammography and does not require strong compression, it offers only 2D images and information on the third dimension is thus lost. In this research we proposed a molecular breast tomosynthesis scanner which provides 3D images of the radiotracers in the breast. In the proposed system, the patient would lie prone on a patient bed with a hole in which the breast is inserted. Subsequently, two gamma cameras equipped with multi-pinhole collimators (therefore the technique is called multi-pinhole molecular breast tomosynthesis, MP-MBT) scan the pendant breast from both sides. 
To estimate the performance of MP-MBT, the system was modelled in Monte Carlo simulations in a clinically realistic setting. The results assured us that it was worth building a prototype of MP-MBT to further investigate its imaging capability. Besides, voxelized raytracing (VRT) software developed earlier in our group to accelerate simulations and facilitate system optimisations was validated with the Monte Carlo simulation results. Subsequently, VRT was used in further studies in this project. 
The promising results of MP-MBT simulations partly relied on a gamma detector with high spatial linearity over the whole detector surface. However, conventional gamma detectors used in clinical practice have large dead edges, i.e. about 4 cm from the detector edges is unusable, and a detector with small dead edges would be very expensive, which may make MP-MBT a less competitive technology. Therefore, in order to have a gamma detector suitable for MP-MBT, we came up with a few different designs with NaI(Tl) scintillators and photomultiplier tube (PMT) array readouts and evaluated their performances with Monte Carlo simulations. From the simulation results, we eventually chose a design with a staggered layout of 15 square PMTs, among which two PMTs detected the optical photons from the scintillator through extra-long additional light-guides. This gamma detector was built in our lab, and it turned out to have only about 15 mm dead edge (mainly due to the 12 mm sealing). 
The customised gamma detector was equipped with a lead multi-pinhole collimator design based on previous research. The whole gamma camera was mounted on a robot arm to create a movable scanner. We calibrated the scanner with a point source and scanned a resolution phantom and a breast phantom to evaluate MP-MBT's performance. In the phantom study, the scanner showed the capability of detecting tumours down to 5 mm when a realistic tracer (technetium sestamibi) concentration was administered. 
However, the current prototype is still far from a device that can be used in the clinic and we have found several problems with MP-MBT, especially the noise pattern in the reconstructed images, which should be given special attention in the future research. ... Read more

Introduction: High Dose Rate (HDR) Brachytherapy is a radiotherapy modality that involves temporarily introducing a highly radioactive source into the target volume with the use of an applicator. With respect to HDR brachytherapy for prostate cancer, an 192Iridium source is driven into the target volume through catheters implanted into the prostate. The dose delivered to a point in the prostate depends on the time the source dwells at a given position. Treatment planning for brachytherapy involve the optimization of dwell times and dwell positions. The aim of the treatment plan is to deliver the prescribed dose to the target volume, the prostate, while minimizing the dose to the organs at risk (OAR), namely the urethra, bladder and rectum. In current clinical practice, the process of treatment planning involves the manual manipulation of the parameters of an optimizer until the desired dose distribution is achieved. This implies that the plan quality depends on the experience of the planner, and there is variation in plan quality between planners. The aim of this project was to develop an automated treatment planning system that would able to generate clinically acceptable plans with minimal human intervention. The brachytherapy treatment planning module is named B-iCycle and may be integrated in the future with the treatment planning software suite, called Erasmus-iCycle, developed at the Erasmus MC.
Materials and methods: At the core of the treatment planning system (TPS) is a precise and fast dose engine that is able to simulate the dose to be delivered. In this project, we employ the TG-43 dose calculation formalism as it is the most widely implemented method in dose engines for brachytherapy treatment planning systems. The dose engine is then verified against the dose engine of the clinical treatment planning system. B-iCycle uses the 2-phase ϵ-constraint (2pϵc) algorithm to optimize the dwell times and positions. The 2pϵc algorithm requires a ‘wish-list’, which encapsulates the treatment protocol as goals and constraints for each critical structure. For this project three treatment protocols were chosen, four fractions of 9.5 Gy, single fraction of 19 Gy and single fraction of 20 Gy, and wish-lists were generated for each protocol. Three patient groups with different catheter geometries were selected. Treatment plans were generated for each patient and compared against the plans that were generated, for the same patients, in the clinic. The treatment plans that were generated in B-iCycle were then exported to the clinical treatment planning system (Oncentra from Elekta) to obtain the dose characteristics. The plans were compared based on the dose characteristics and the Conformity Index (COIN). The plans were also verified by a radiation oncologist.
Results: The TG-43 dose engine was successfully verified against the clinical dose engine. The Gamma analysis showed that only 0.68% of the voxels failed the gamma analysis and these voxels were located within the catheters therefore they can be ignored as no tissue lies at these positions. With regard to plans that were generated, the physician confirmed that the clinically acceptable B-iCycle plans are very comparable to the clinical plans. The B-iCycle plans are better at minimizing the dose to the urethra. When comparing B-iCycle plans to the clinical plans using COIN, B-iCycle was found to be better than the clinical procedure. B-iCycle can generate a treatment plan in approximately 10 seconds, which is much faster than the clinical procedure, which averages at 10 minutes. It is also able to avoid the issue of treatment planner variability and is able to generate consistent, high quality treatment plans.
... Read more

The need of image (frame) acquisitions within short time intervals is of major importance for preclinical SPECT imaging. The short frame times enable higher temporal resolution which is required in bio-distribution and pharmacokinetic studies where fast dynamic imaging is performed. The present study evaluates and compares the performance of two different preclinical multipinhole SPECT systems (NanoScan, VECTor) for short frames acquisitions.
Prior to the systems comparison, the comparison and selection of the best performing fast imaging mode provided by NanoScan system (Mediso) was performed. The fast imaging modes of this system provide acquisitions with 1,2 and 3 detector position around the animal bed. This comparison was performed by using uniform phantoms (syringes) and the rods of the NEMA NU4IQ phantom (frames: 6-30s). The down-sized version of NU4IQ phantom (SPECTIQ phantom) was used in this study to compare the performance of VECTor (MILabs) and NanoScan when performing acquisitions with short frame times (18s-600s, whole body scans). The quality of the acquired images was assessed in terms of absolute quantification (recovery coefficient), noise levels and visual evaluation.
The quantification with the NanoScan was accurate (±5%) regardless of times frames duration and activity concentrations when imaging large structures. The increase in number of detector positions yielded images with lower noise levels. In the case of small structures, acquisition with 3 detector positions (Semi-3 mode) appeared to provide more accurate activity recovery compared to acquisitions with 1 (stationary mode) and 2 detector positions (Semi-2
mode). Especially in the case of the 2mm diameter rod of the NU4IQ phantom, the Semi-3 mode appears to provide significantly more accurate activity recovery (30s frame). The systems comparison showed activity recovery with up to 5% deviation from the dose calibrator measurement when imaging the uniform region of SPECTIQ phantom (d = 21mm). Both systems could recover the three largest rods (d = 1.5, 1.0, 0.75mm) for the longest frames used(180,360,600s). None of the systems could recover the two smallest rods of the phantom (d=0.5,0.35mm). As the frame time decreased, both systems could recover less number of rods. VECTor appeared to provide higher activity recovery than NanoScan for the three largest rods of the phantom. However, as the frame time decreased the differences became less significant. Furthermore, VECTor provided and 22.2% and 46.6% less spillover in airand water-filled phantom regions (after reaching convergence) than NanoScan did.
The performances of two preclinical SPECT systems (NanoScan, VECTor) for short time acquisitionswere compared. The conducted experiments showed that the systems perform equally when conducting short frames imaging. Furthermore, the fast imaging mode of NanoScan employing three detector positions showed better performance than the other two fast imaging modes provided by this system.



... Read more