Latent space manipulation for high-resolution medical image synthesis via the StyleGAN.

INTRODUCTION: This paper explores the potential of the StyleGAN model as an high-resolution image generator for synthetic medical images. The possibility to generate sample patient images of different modalities can be helpful for training deep learning algorithms as e.g. a data augmentation technique. METHODS: The StyleGAN model was trained on Computed Tomography (CT) and T2- weighted Magnetic Resonance (MR) images from 100 patients with pelvic malignancies. The resulting model was investigated with regards to three features: Image Modality, Sex, and Longitudinal Slice Position. Further, the style transfer feature of the StyleGAN was used to move images between the modalities. The root-mean-squard error (RMSE) and the Mean Absolute Error (MAE) were used to quantify errors for MR and CT, respectively. RESULTS: We demonstrate how these features can be transformed by manipulating the latent style vectors, and attempt to quantify how the errors change as we move through the latent style space. The best results were achieved by using the style transfer feature of the StyleGAN (58.7 HU MAE for MR to CT and 0.339 RMSE for CT to MR). Slices below and above an initial central slice can be predicted with an error below 75 HU MAE and 0.3 RMSE within 4cm for CT and MR, respectively. DISCUSSION: The StyleGAN is a promising model to use for generating synthetic medical images for MR and CT modalities as well as for 3D volumes.

Investigating conditional GAN performance with different generator architectures, an ensemble model, and different MR scanners for MR-sCT conversion.

Recent developments in magnetic resonance (MR) to synthetic computed tomography (sCT) conversion have shown that treatment planning is possible without an initial planning CT. Promising conversion results have been demonstrated recently using conditional generative adversarial networks (cGANs). However, the performance is generally only tested on images from one MR scanner, which neglects the potential of neural networks to find general high-level abstract features. In this study, we explored the generalizability of the generator models, trained on a single field strength scanner, to data acquired with higher field strengths. T2-weighted 0.35T MRIs and CTs from 51 patients treated for prostate (40) and cervical cancer (11) were included. 25 of them were used to train four different generators (SE-ResNet, DenseNet, U-Net, and Embedded Net). Further, an ensemble model was created from the four network outputs. The models were validated on 16 patients from a 0.35T MR scanner. Further, the trained models were tested on the Gold Atlas dataset, containing T2-weighted MR scans of different field strengths; 1.5T(7) and 3T(12), and 10 patients from the 0.35T scanner. The sCTs were dosimetrically compared using clinical VMAT plans for all test patients. For the same scanner (0.35T), the results from the different models were comparable on the test set, with only minor differences in the mean absolute error (MAE) (35-51HU body). Similar results were obtained for conversions of 3T GE Signa and the 3T GE Discovery images (40-62HU MAE) for three of the models. However, larger differences were observed for the 1.5T images (48-65HU MAE). The overall best model was found to be the ensemble model. All dose differences were below 1%. This study shows that it is possible to generalize models trained on images of one scanner to other scanners and different field strengths. The best metric results were achieved by the combination of all networks.

Characterization of EBT3 radiochromic films for dosimetry of proton beams in the presence of magnetic fields.

PURPOSE: Radiochromic film dosimetry is extensively used for quality assurance in photon and proton beam therapy. So far, GafchromicTM EBT3 film appears as a strong candidate to be used in future magnetic resonance (MR) based therapy systems. The response of Gafchromic EBT3 films in the presence of magnetic fields has already been addressed for different MR-linacs systems. However, a detailed evaluation of the influence of external magnetic fields on the film response and calibration curves for proton therapy has not yet been reported. This study aims to determine the dose responses of EBT3 films for clinical proton beams exposed to magnetic field strengths up to 1 T in order to investigate the feasibility of EBT3 film as an accurate dosimetric tool for a future MR particle therapy system (MRPT). METHODS: The dosimetric characteristics of EBT3 films were studied for a proton beam passing through magnetic field strengths of B = 0, 0.5, and 1 T. Absorbed dose calibration and measurements were performed using clinical proton beams in the nominal energy range of 62.4-252.6 MeV. Irradiations were done using an in-house developed PMMA slab phantom placed in the center of a dipole research magnet. Monte Carlo (MC) simulations using the GATE/Geant4 toolkit were performed to predict the effect of magnetic fields on the energy deposited by proton beams in the phantom. Planned and measured doses from 3D box cube irradiations were compared to assess the accuracy of the dosimetric method using EBT3 films with/without the external magnetic field. RESULTS: Neither for the mean pixel value nor for the net optical density, any significant deviations were observed due to the presence of an external magnetic field (B ≤ 1T) for doses up to 10 Gy. Dose-response curves for the red channel were fitted by a three-parameter function for the field-free case and for B = 1T, showing for both cases an R-square coefficient of unity and almost identical fitting parameters. Independently of the magnetic field, EBT3 films showed an under-response as high as 8% in the Bragg peak region, similarly to previously reported effects for particle therapy. No noticeable influence of the magnetic field strength was observed on the quenching effect of the EBT3 films. CONCLUSIONS: For the first time detailed absorbed dose calibrations of EBT3 films for proton beams in magnetic field regions were performed. Results showed that EBT3 films represent an attractive solution for the dosimetry of a future MRPT system. As film response functions for protons are not affected by the magnetic field strenght, they can be used for further investigations to evaluate the dosimetric effects induced due to particle beams bending in magnetic fields regions.

(R)-[18F]NEBIFQUINIDE: A promising new PET tracer for TSPO imaging.

Positron emission tomography (PET) imaging of the 18 kDa translocator protein (TSPO), has a high diagnostic potential in neurodegenerative disorders and cancer. However, TSPO is considered a challenge for molecular imaging due to the poor availability of suitable radiotracers with adequate pharmacokinetic properties. Here, we describe the development of a radiofluorinated pyridinyl isoquinoline analogue of the established TSPO PET tracer (R)-[11C]PK11195 with improved binding properties in all known human TSPO phenotypes. We conducted a complete preclinical evaluation using in vitro, in vivo and ex vivo methods to assess the performance of this novel radiotracer and observed high specific binding of the radiotracer to TSPO, as well as high metabolic stability. Therefore, we propose this radiolabeled compound for further evaluation in animal models as well as in clinical trials.

SNAPshots of the MCHR1: a Comparison Between the PET-Tracers [18F]FE@SNAP and [11C]SNAP-7941.

PURPOSE: The melanin-concentrating hormone receptor 1 (MCHR1) has become an important pharmacological target, since it may be involved in various diseases, such as diabetes, insulin resistance, and obesity. Hence, a suitable positron emission tomography radiotracer for the in vivo assessment of the MCHR1 pharmacology is imperative. The current paper contrasts the extensive in vitro, in vivo, and ex vivo assessments of the radiotracers [18F]FE@SNAP and [11C]SNAP-7941 and provides comprehensive information about their biological and physicochemical properties. Furthermore, it examines their suitability for first-in-man imaging studies. PROCEDURES: Kinetic real-time cell-binding studies with [18F]FE@SNAP and [11C]SNAP-7941 were conducted on adherent Chines hamster ovary (CHO-K1) cells stably expressing the human MCHR1 and MCHR2. Small animal imaging studies on mice and rats were performed under displacement and baseline conditions, as well as after pretreatment with the P-glycoprotein/breast cancer resistant protein inhibitor tariquidar. After the imaging studies, detailed analyses of the ex vivo biodistribution were performed. Ex vivo metabolism was determined in rat blood and brain and analyzed at various time points using a quantitative radio-HPLC assay. RESULTS: [11C]SNAP-7941 demonstrates high uptake on CHO-K1-hMCHR1 cells, whereas no uptake was detected for the CHO-K1-hMCHR2 cells. In contrast, [18F]FE@SNAP evinced binding to CHO-K1-hMCHR1 and CHO-K1-hMCHR2 cells. Imaging studies with [18F]FE@SNAP and [11C]SNAP-7941 showed an increased brain uptake after tariquidar pretreatment in mice, as well as in rats, and exhibited a significant difference between the time-activity curves of the baseline and blocking groups. Biodistribution of both tracers demonstrated a decreased uptake after displacement. [11C]SNAP-7941 revealed a high metabolic stability in rats, whereas [18F]FE@SNAP was rapidly metabolized. CONCLUSIONS: Both radiotracers demonstrate appropriate imaging properties for the MCHR1. However, the pronounced metabolic stability as well as superior selectivity and affinity of [11C]SNAP-7941 underlines the decisive superiority over [18F]FE@SNAP.

Retina dose as a predictor for visual acuity loss in 106Ru eye plaque brachytherapy of uveal melanomas.

BACKGROUND AND PURPOSE: To evaluate the retina dose as a risk factor associated with loss of visual acuity (VA) in 106Ru plaque brachytherapy. MATERIAL/METHODS: 45 patients receiving 106Ru plaques brachytherapy (median follow-up 29.5 months) were included in this study. An in-house developed treatment planning system with Monte Carlo based dose calculation was used to perform treatment planning and dose calculation. Risk factors associated with loss of VA were evaluated using the Cox proportional hazards models, Kaplan-Meier estimates and Pearson correlation coefficients. RESULTS: A significant correlation was found between VA loss and mean (r = 0.49, p = 0.001) and near maximum (r = 0.47, p = 0.001) retina dose D2% and tumor basal diameter (r = 0.50, p < 0.001). The Kaplan-Meier and Cox proportional hazards model yielded a significantly higher risk for VA loss (>0.3Snellen) for patients receiving a maximum dose of >500 Gy (p = 0.002). A Cox multivariate analysis including the macula dose (p = 0.237) and basal diameter (p = 0.791) showed that a high maximum retinal dose is the best risk factor (p = 0.013) for VA loss. CONCLUSION: The study showed that retina dose (D2% and Dmean) is a suitable predictor for VA loss.

Treatment plan optimization and robustness of 106Ru eye plaque brachytherapy using a novel software tool.

BACKGROUND AND PURPOSE: To analyze treatment plan robustness and plan optimization strategies of 106Ru eye plaque brachytherapy using a novel software tool. MATERIALS AND METHODS: A treatment planning software was developed that allows to calculate dose-volume metrics. Plaque misplacements were simulated and evaluated with respect to the effect on tumor coverage and dose changes in critical structures. Two treatment plan optimization approaches were analyzed: (a) reducing plaque size and (b) shifting the plaque away from organs-at-risk (OAR). RESULTS: Maximum tumor sizes were identified which can be covered by the prescribed dose for different robustness levels (0-2mm). For an apex height of 5mm a 1mm uncertainty yielded changes in D2% to the lens of up to ±13Gy in anterior and ±20Gy to the optic nerve in posterior tumors. By reducing the plaque size Dmean and D2% to lens, optic nerve and macula were decreased by >60% for most simulated cases. Similarly, by shifting the plaque away from the lens dose reductions of 15%/mm in anterior and even 30%/mm in central tumors were achieved. CONCLUSION: Critical structures in the treatment of uveal melanomas with 106Ru plaques can benefit from the proposed, computational treatment plan optimization.