Table of contents

Computed tomography perfusion (CTP) is a functional imaging that allows for providing capillary-level hemodynamics information of the desired tissue in clinics. In this paper, we aim to offer insight into CTP imaging which covers the basics and current state of CTP imaging, then summarize the technical applications in the CTP imaging as well as the future technological potential. At first, we focus on the fundamentals of CTP imaging including systematically summarized CTP image acquisition and hemodynamic parameter map estimation techniques. A short assessment is presented to outline the clinical applications with CTP imaging, and then a review of radiation dose effect of the CTP imaging on the different applications is presented. We present a categorized methodology review on known and potential solvable challenges of radiation dose reduction in CTP imaging. To evaluate the quality of CTP images, we list various standardized performance metrics. Moreover, we present a review on the determination of infarct and penumbra. Finally, we reveal the popularity and future trend of CTP imaging.

This paper reviews the ecosystem of GATE, an open-source Monte Carlo toolkit for medical physics. Based on the shoulders of Geant4, the principal modules (geometry, physics, scorers) are described with brief descriptions of some key concepts (Volume, Actors, Digitizer). The main source code repositories are detailed together with the automated compilation and tests processes (Continuous Integration). We then described how the OpenGATE collaboration managed the collaborative development of about one hundred developers during almost 20 years. The impact of GATE on medical physics and cancer research is then summarized, and examples of a few key applications are given. Finally, future development perspectives are indicated.

Objective. A 2-dimensional pre-clinical SFRT (GRID) collimator was designed for use on the ultra-high dose rate (UHDR) 10 MV ARIEL beamline at TRIUMF. TOPAS Monte Carlo simulations were used to determine optimal collimator geometry with respect to various dosimetric quantities. Approach. The GRID-averaged peak-to-valley dose ratio (PVDR) and mean dose rate of the peaks were investigated with the intent of maximizing both values in a given design. The effects of collimator thickness, focus position, septal width, and hole width on these metrics were found by testing a range of values for each parameter on a cylindrical GRID collimator. For each tested collimator geometry, photon beams with energies of 10, 5, and 1 MV were transported through the collimator and dose rates were calculated at various depths in a water phantom located 1.0 cm from the collimator exit. Main results. In our optimization, hole width proved to be the only collimator parameter which increased both PVDR and peak dose rates. From the optimization results, it was determined that our optimized design would be one which achieves the maximum dose rate for a PVDR $\geqslant 5$ at 10 MV. Ultimately, this was achieved using a collimator with a thickness of 75 mm, 0.8 mm septal and hole widths, and a focus position matched to the beam divergence. This optimized collimator maintained the PVDR of 5 in the phantom between water depths of 0–10 cm at 10 MV and had a mean peak dose rate of $3.06\pm 0.02$${\rm{Gy}}\,{{\rm{s}}}^{-1}$ at 0–1 cm depth. Significance. We have investigated the impact of various GRID-collimator design parameters on the dose rate and spatial fractionation of 10, 5, and 1 MV photon beams. The optimized collimator design for the 10 MV ultra-high dose rate photon beam could become a useful tool for radiobiology studies synergizing the effects of ultra-high dose rate (FLASH) delivery and spatial fractionation.

In this work we present an advanced random forest-based machine learning (ML) model, trained and tested on Geant4 simulations. The developed ML model is designed to improve the performance of the hybrid detector for microdosimetry (HDM), a novel hybrid detector recently introduced to augment the microdosimetric information with the track length of particles traversing the microdosimeter. The present work leads to the following improvements of HDM: (i) the detection efficiency is increased up to 100%, filling not detected particles due to scattering within the tracker or non-active regions, (ii) the track reconstruction algorithm precision. Thanks to the ML models, we were able to reconstruct the microdosimetric spectra of both protons and carbon ions at therapeutic energies, predicting the real track length for every particle detected by the microdosimeter. The ML model results have been extensively studied, focusing on non-accurate predictions of the real track lengths. Such analysis has been used to identify HDM limitations and to understand possible future improvements of both the detector and the ML models.

Objective. The accuracy of radiotherapy for patients with locally advanced cancer is compromised by independent motion of multiple targets. To date, MLC tracking approaches have used 2D geometric optimisation where the MLC aperture shape is simply translated to correspond to the target's motion, which results in sub-optimal delivered dose. To address this limitation, a dose-optimised multi-target MLC tracking method was developed and evaluated through simulated locally advanced prostate cancer treatments. Approach. A dose-optimised multi-target tracking algorithm that adapts the MLC aperture to minimise 3D dosimetric error was developed for moving prostate and static lymph node targets. A fast dose calculation algorithm accumulated the planned dose to the prostate and lymph node volumes during treatment in real time, and the MLC apertures were recalculated to minimise the difference between the delivered and planned dose with the included motion. Dose-optimised tracking was evaluated by simulating five locally advanced prostate plans and three prostate motion traces with a relative interfraction displacement. The same simulations were performed using geometric-optimised tracking and no tracking. The dose-optimised, geometric-optimised, and no tracking results were compared with the planned doses using a 2%/2 mm γ criterion. Main results. The mean dosimetric error was lowest for dose-optimised MLC tracking, with γ-failure rates of 12% ± 8.5% for the prostate and 2.2% ± 3.2% for the nodes. The γ-failure rates for geometric-optimised MLC tracking were 23% ± 12% for the prostate and 3.6% ± 2.5% for the nodes. When no tracking was used, the γ-failure rates were 37% ± 28% for the prostate and 24% ± 3.2% for the nodes. Significance. This study developed a dose-optimised multi-target MLC tracking method that minimises the difference between the planned and delivered doses in the presence of intrafraction motion. When applied to locally advanced prostate cancer, dose-optimised tracking showed smaller errors than geometric-optimised tracking and no tracking for both the prostate and nodes.

Introduction. To propose a novel deep learning-based method called RG-Net (reconstruction and generation network) for highly accelerated MR parametric mapping by undersampling k-space and reducing the acquired contrast number simultaneously. Methods. The proposed framework consists of a reconstruction module and a generative module. The reconstruction module reconstructs MR images from the acquired few undersampled k-space data with the help of a data prior. The generative module then synthesizes the remaining multi-contrast images from the reconstructed images, where the exponential model is implicitly incorporated into the image generation through the supervision of fully sampled labels. The RG-Net was trained and tested on the T_1ρ mapping data from 8 volunteers at net acceleration rates of 17, respectively. Regional T_1ρ analysis for cartilage and the brain was performed to assess the performance of RG-Net. Results. RG-Net yields a high-quality T_1ρ map at a high acceleration rate of 17. Compared with the competing methods that only undersample k-space, our framework achieves better performance in T_1ρ value analysis. Conclusion. The proposed RG-Net can achieve a high acceleration rate while maintaining good reconstruction quality by undersampling k-space and reducing the contrast number simultaneously for fast MR parametric mapping. The generative module of our framework can also be used as an insertable module in other fast MR parametric mapping methods.

Background and Objective. Skin lesion classification by using deep learning technologies is still a considerable challenge due to high similarity among classes and large intraclass differences, serious class imbalance in data, and poor classification accuracy with low robustness. Approach. To address these issues, a two-stage framework for dermoscopy lesion classification using adversarial training and a fuzzy rank-based ensemble of multilayer feature fusion convolutional neural network (CNN) models is proposed. In the first stage, dermoscopy dataset augmentation based on generative adversarial networks is proposed to obtain realistic dermoscopy lesion images, enabling significant improvement for balancing the number of lesions in each class. In the second stage, a fuzzy rank-based ensemble of multilayer feature fusion CNN models is proposed to classify skin lesions. In addition, an efficient channel integrating spatial attention module, in which a novel dilated pyramid pooling structure is designed to extract multiscale features from an enlarged receptive field and filter meaningful information of the initial features. Combining the cross-entropy loss function with the focal loss function, a novel united loss function is designed to reduce the intraclass sample distance and to focus on difficult and error-prone samples to improve the recognition accuracy of the proposed model. Main results. In this paper, the common dataset (HAM10000) is selected to conduct simulation experiments to evaluate and verify the effectiveness of the proposed method. The subjective and objective experimental results demonstrate that the proposed method is superior over the state-of-the-art methods for skin lesion classification due to its higher accuracy, specificity and robustness. Significance. The proposed method effectively improves the classification performance of the model for skin diseases, which will help doctors make accurate and efficient diagnoses, reduce the incidence rate and improve the survival rates of patients.

Objective. Combined proton–photon treatments, where most fractions are delivered with photons and only a few are delivered with protons, may represent a practical approach to optimally use limited proton resources. It has been shown that, when organs at risk (OARs) are located within or near the tumor, the optimal multi-modality treatment uses protons to hypofractionate parts of the target volume and photons to achieve near-uniform fractionation in dose-limiting healthy tissues, thus exploiting the fractionation effect. These plans may be sensitive to range and setup errors, especially misalignments between proton and photon doses. Thus, we developed a novel stochastic optimization method to directly incorporate these uncertainties into the biologically effective dose (BED)-based simultaneous optimization of proton and photon plans. Approach. The method considers the expected value $E\left(b\right)$ and standard deviation $\sigma \left(b\right)$ of the cumulative BED $b$ in every voxel of a structure. For the target, a piecewise quadratic penalty function of the form ${\left[{b}^{\min }-\left(E\left(b\right)-2\sigma \left(b\right)\right)\right]}_{+}^{2}$ is minimized, aiming for plans in which the expected BED minus two times the standard deviation exceeds the prescribed BED ${b}^{\min }.$ Analogously, ${\left[\left(E\left(b\right)+2\sigma \left(b\right)\right)-{b}^{\max }\right]}_{+}^{2}$ is considered for OARs. Main results. Using a spinal metastasis case and a liver cancer patient, it is demonstrated that the novel stochastic optimization method yields robust combined treatment plans. Tumor coverage and a good sparing of the main OARs are maintained despite range and setup errors, and especially misalignments between proton and photon doses. This is achieved without explicitly considering all combinations of proton and photon error scenarios. Significance. Concerns about range and setup errors for safe clinical implementation of optimized proton–photon radiotherapy can be addressed through an appropriate stochastic planning method.

Objective. Prediction of the response to neoadjuvant chemotherapy (NAC) in breast cancer is important for patient outcomes. In this work, we propose a deep learning based approach to NAC response prediction in ultrasound (US) imaging. Approach. We develop recurrent neural networks that can process serial US imaging data to predict chemotherapy outcomes. We present models that can process either raw radio-frequency (RF) US data or regular US images. The proposed approach is evaluated based on 204 sequences of US data from 51 breast cancers. Each sequence included US data collected before the chemotherapy and after each subsequent dose, up to the 4th course. We investigate three pre-trained convolutional neural networks (CNNs) as back-bone feature extractors for the recurrent network. The CNNs were pre-trained using raw US RF data, US b-mode images and RGB images from the ImageNet dataset. The first two networks were developed using US data collected from malignant and benign breast masses. Main results. For the pre-treatment data, the better performing network, with back-bone CNN pre-trained on US images, achieved area under the receiver operating curve (AUC) of 0.81 (±0.04). Performance of the recurrent networks improved with each course of the chemotherapy. For the 4th course, the better performing model, based on the CNN pre-trained with RGB images, achieved AUC value of 0.93 (±0.03). Statistical analysis based on the DeLong test presented that there were no significant differences in AUC values between the pre-trained networks at each stage of the chemotherapy (p-values > 0.05). Significance. Our study demonstrates the feasibility of using recurrent neural networks for the NAC response prediction in breast cancer US.

Objective. The red bone marrow (RBM) and bone endosteum (BE), which are required for effective dose calculation, are macroscopically modeled in the reference phantoms of the international commission on radiological protection (ICRP) due to their microscopic and complex histology. In the present study, the detailed bone models were developed to simplify the dose calculation process for skeletal dosimetry. Approach. The detailed bone models were developed based on the bone models developed at the University of Florida. A new method was used to update the definition of BE region by storing the BE location indices using virtual sub-voxels. The detailed bone models were then installed in the spongiosa regions of the ICRP mesh-type reference computational phantoms (MRCPs) via the parallel geometry feature of the Geant4 code. Main results. Comparing the results between the detailed-bone-installed MRCPs and the original MRCPs with the absorbed dose to spongiosa and fluence-to-dose response function (DRF)-based methods, the DRF-based method showed much smaller but still significant differences. Compared with the values given in ICRP Publications 116 and 133, the differences were very large (i.e. several orders of magnitudes), due mainly to the anatomical improvement of the skeletal system in the MRCPs; that is, spongiosa and medullary cavity are fully enclosed by cortical bone in the MRCPs but not in the ICRP-110 phantoms. Significance. The detailed bone models enable the direct calculation of the absorbed doses to the RBM and BE, simplifying the dose calculation process and potentially improving the consistency and accuracy of skeletal dosimetry.

Objective. The development of alpha-emitting radiopharmaceuticals using ²²⁵Ac (t_½ = 9.92 d) benefits from the quantitative determination of its biodistribution and is not always easy to directly measure. An element-equivalent matched-pair would allow for more accurate biodistribution and dosimetry estimates. ²²⁶Ac (t_½ = 29.4 h) is a candidate isotope for in vivo imaging of preclinical ²²⁵Ac radiopharmaceuticals, given its 158 keV and 230 keV gamma emissions making it suitable for quantitative SPECT imaging. This work aimed to conduct a performance assessment for ²²⁶Ac imaging and presents the first-ever ²²⁶Ac SPECT images. Approach. To establish imaging performance with regards to contrast and noise, image quality phantoms were scanned using a microSPECT/CT system. To assess the resolution, a hot rod phantom with cylindrical rods with diameters between 0.85 and 1.70 mm was additionally imaged. Two collimators were evaluated: a high-energy ultra-high resolution (HEUHR) collimator and an extra ultra-high sensitivity (UHS) collimator. Images were reconstructed from two distinct photopeaks at 158 keV and 230 keV. Main results. The HEUHR SPECT image measurements of high activity concentration regions were consistent with values determined independently via gamma spectroscopy, within 9% error. The lower energy 158 keV photopeak images demonstrated slightly better contrast recovery. In the resolution phantom, the UHS collimator only resolved rods ≥1.30 mm and ≥1.50 mm for the 158 keV and 230 keV photopeaks, respectively, while the HEUHR collimator clearly resolved all rods, with resolution <0.85 mm. Significance. Overall, the feasibility of preclinical imaging with ²²⁶Ac was demonstrated with quantitative SPECT imaging achieved for both its 158 keV and 230 keV photopeaks. The HEUHR collimator is recommended for imaging ²²⁶Ac activity distributions in small animals due to its resolution <0.85 mm. Future work will explore the feasibility of using ²²⁶Ac both as an element-equivalent isotope for ²²⁵Ac radiopharmaceuticals, or as a standalone therapeutic isotope.

Objective. Obtaining the intrinsic dose distributions in particle therapy is a challenging problem that needs to be addressed by imaging algorithms to take advantage of secondary particle detectors. In this work, we investigate the utility of deep learning methods for achieving direct mapping from detector data to the intrinsic dose distribution. Approach. We performed Monte Carlo simulations using GATE/Geant4 10.4 simulation toolkits to generate a dataset using human CT phantom irradiated with high-energy protons and imaged with compact in-beam PET for realistic beam delivery in a single-fraction (∼2 Gy). We developed a neural network model based on conditional generative adversarial networks to generate dose maps conditioned on coincidence distributions in the detector. The model performance is evaluated by the mean relative error, absolute dose fraction difference, and shift in Bragg peak position. Main results. The relative deviation in the dose and range of the distributions predicted by the model from the true values for mono-energetic irradiation between 50 and 122 MeV lie within 1% and 2%, respectively. This was achieved using 10⁵ coincidences acquired five minutes after irradiation. The relative deviation in the dose and range for spread-out Bragg peak distributions were within 1% and 2.6% uncertainties, respectively. Significance. An important aspect of this study is the demonstration of a method for direct mapping from detector counts to dose domain using the low count data of compact detectors suited for practical implementation in particle therapy. Including additional prior information in the future can further expand the scope of our model and also extend its application to other areas of medical imaging.

Objective. CT-mesh hybrid phantoms (or 'hybrid(s)') made from integrated patient CT data and mesh-type reference computational phantoms (MRCPs) can be beneficial for patient-specific whole-body dose evaluation, but this benefit has yet to be evaluated for second cancer risk prediction. The purpose of this study is to compare the hybrid's ability to predict risk throughout the body with a patient-scaled MRCP against ground truth whole-body CTs (WBCTs). Approach. Head and neck active scanning proton treatment plans were created for and simulated on seven hybrids and the corresponding scaled MRCPs and WBCTs. Equivalent dose throughout the body was calculated and input into five second cancer risk models for both excess absolute and excess relative risk (EAR and ERR). The hybrid phantom was evaluated by comparing equivalent dose and risk predictions against the WBCT. Main results. The hybrid most frequently provides whole-body second cancer risk predictions which are closer to the ground truth when compared to a scaled MRCP alone. The performance of the hybrid relative to the scaled MRCP was consistent across ERR, EAR, and all risk models. For all in-field organs, where the hybrid shares the WBCT anatomy, the hybrid was better than or equal to the scaled MRCP for both equivalent dose and risk prediction. For out-of-field organs across all patients, the hybrid's equivalent dose prediction was superior than the scaled MRCP in 48% of all comparisons, equivalent for 34%, and inferior for 18%. For risk assessment in the same organs, the hybrid's prediction was superior than the scaled MRCP in 51.8% of all comparisons, equivalent in 28.6%, and inferior in 19.6%. Significance. Whole-body risk predictions from the CT-mesh hybrid have shown to be more accurate than those from a reference phantom alone. These hybrids could aid in risk-optimized treatment planning and individual risk assessment to minimize second cancer incidence.

Objective. To establish an open framework for developing plan optimization models for knowledge-based planning (KBP). Approach. Our framework includes radiotherapy treatment data (i.e. reference plans) for 100 patients with head-and-neck cancer who were treated with intensity-modulated radiotherapy. That data also includes high-quality dose predictions from 19 KBP models that were developed by different research groups using out-of-sample data during the OpenKBP Grand Challenge. The dose predictions were input to four fluence-based dose mimicking models to form 76 unique KBP pipelines that generated 7600 plans (76 pipelines × 100 patients). The predictions and KBP-generated plans were compared to the reference plans via: the dose score, which is the average mean absolute voxel-by-voxel difference in dose; the deviation in dose-volume histogram (DVH) points; and the frequency of clinical planning criteria satisfaction. We also performed a theoretical investigation to justify our dose mimicking models. Main results. The range in rank order correlation of the dose score between predictions and their KBP pipelines was 0.50–0.62, which indicates that the quality of the predictions was generally positively correlated with the quality of the plans. Additionally, compared to the input predictions, the KBP-generated plans performed significantly better (P < 0.05; one-sided Wilcoxon test) on 18 of 23 DVH points. Similarly, each optimization model generated plans that satisfied a higher percentage of criteria than the reference plans, which satisfied 3.5% more criteria than the set of all dose predictions. Lastly, our theoretical investigation demonstrated that the dose mimicking models generated plans that are also optimal for an inverse planning model. Significance. This was the largest international effort to date for evaluating the combination of KBP prediction and optimization models. We found that the best performing models significantly outperformed the reference dose and dose predictions. In the interest of reproducibility, our data and code is freely available.

Objective. To develop a new model (Mayo Clinic Florida microdosimetric kinetic model, MCF MKM) capable of accurately describing the in vitro clonogenic survival at low and high linear energy transfer (LET) using single-event microdosimetric spectra in a single target. Methodology. The MCF MKM is based on the 'post-processing average' implementation of the non-Poisson microdosimetric kinetic model and includes a novel expression to compute the particle-specific quadratic-dependence of the cell survival with respect to dose (β of the linear-quadratic model). A new methodology to a priori calculate the mean radius of the MCF MKM subnuclear domains is also introduced. Lineal energy spectra were simulated with the Particle and Heavy Ion Transport code System (PHITS) for ¹H, ⁴He, ¹²C, ²⁰Ne, ⁴⁰Ar, ⁵⁶Fe, and ¹³²Xe ions and used in combination with the MCF MKM to calculate the ion-specific LET-dependence of the relative biological effectiveness (RBE) for Chinese hamster lung fibroblasts (V79 cell line) and human salivary gland tumor cells (HSG cell line). The results were compared with in vitro data from the Particle Irradiation Data Ensemble (PIDE) and in silico results of different models. The possibility of performing experiment-specific predictions to explain the scatter in the in vitro RBE data was also investigated. Finally, a sensitivity analysis on the model parameters is also included. Main results. The RBE values predicted with the MCF MKM were found to be in good agreement with the in vitro data for all tested conditions. Though all MCF MKM model parameters were determined a priori, the accuracy of the MCF MKM was found to be comparable or superior to that of other models. The model parameters determined a priori were in good agreement with the ones obtained by fitting all available in vitro data. Significance. The MCF MKM will be considered for implementation in cancer radiotherapy treatment planning with accelerated ions.

Objective. This study aims at comparing dosimetric parameters of 126 MeV antiprotons and protons using microdosimetric approach. Approach. Microdosimetric distributions of 126 MeV proton and antiproton beams at 1 μm site size are calculated using the Monte Carlo-based FLUKA code. The distributions are calculated at various depths along the central axis in water phantom as well as at different off-axis locations. The study also includes calculations of secondary radiations produced by antiprotons and protons. Mean quality factor, $\bar{Q}$ is calculated using the ICRP 60 and ICRU 40 recommendations. The Relative Biological Effectiveness (RBE) of HSG tumour cell at 10% survival level is calculated based on Microdosimetric Kinetic Model. Main results.${\bar{Q}}_{ICRP,}$${\bar{Q}}_{{ICRU}}$ and RBE for antiprotons are higher by a factor of about 3.60, 3.41 and 1.24, respectively, at Bragg-peak and higher by a factor of about 1.41, 1.76 and 1.05, respectively, at plateau region of depth-dose profile when compared to protons. At 15 cm depth along the central axis, ${\bar{Q}}_{{ICRP}},$${\bar{Q}}_{{ICRU}\,}$ and RBE for protons are higher by a factor of about 1.42, 1.66 and 1.26, respectively, when compared to antiprotons. At the off-axis distance (L_d) of 6 cm (at 11.5 cm depth in water), ${\bar{Q}}_{{ICRP}}$ and ${\bar{Q}}_{{ICRU}}$ of protons are higher than that of antiproton whereas the trend is opposite at off-axis distance of 4 cm. At L_d = 4 cm (at 11.5 cm depth in water), RBE of antiprotons is higher by about 4% than protons whereas at L_d = 6 cm, RBE of protons is higher by about 13% than antiprotons. Significance. The study shows that antiproton radiotherapy is advantageous as compared to protons considering enhancements in the absorbed dose and RBE-weighed dose values at the Bragg-peak.

Objective. Traditional radiotherapy (RT) treatment planning of non-small cell lung cancer (NSCLC) relies on population-wide estimates of organ tolerance to minimize excess toxicity. The goal of this study is to develop a personalized treatment planning based on patient-specific lung radiosensitivity, by combining machine learning and optimization. Approach. Sixty-nine non-small cell lung cancer patients with baseline and mid-treatment [18]F-fluorodeoxyglucose (FDG)-PET images were retrospectively analyzed. A probabilistic Bayesian networks (BN) model was developed to predict the risk of radiation pneumonitis (RP) at three months post-RT using pre- and mid-treatment FDG information. A patient-specific dose modifying factor (DMF), as a surrogate for lung radiosensitivity, was estimated to personalize the normal tissue toxicity probability (NTCP) model. This personalized NTCP was then integrated into a NTCP-based optimization model for RT adaptation, ensuring tumor coverage and respecting patient-specific lung radiosensitivity. The methodology was employed to adapt the treatment planning of fifteen NSCLC patients. Main results. The magnitude of the BN predicted risks corresponded with the RP severity. Average predicted risk for grade 1–4 RP were 0.18, 0.42, 0.63, and 0.76, respectively (p < 0.001). The proposed model yielded an average area under the receiver-operating characteristic curve (AUROC) of 0.84, outperforming the AUROCs of LKB-NTCP (0.77), and pre-treatment BN (0.79). Average DMF for the radio-tolerant (RP grade = 1) and radiosensitive (RP grade ≥ 2) groups were 0.8 and 1.63, p < 0.01. RT personalization resulted in five dose escalation strategies (average mean tumor dose increase = 6.47 Gy, range = [2.67–17.5]), and ten dose de-escalation (average mean lung dose reduction = 2.98 Gy [0.8–5.4]), corresponding to average NTCP reduction of 15% [4–27]. Significance. Personalized FDG-PET-based mid-treatment adaptation of NSCLC RT could significantly lower the RP risk without compromising tumor control. The proposed methodology could help the design of personalized clinical trials for NSCLC patients.

Objective. Intrafraction motion is a major concern for the safety and effectiveness of high dose stereotactic body radiotherapy (SBRT) in the upper abdomen. In this study, the impact of the intrafraction motion on the delivered dose was assessed in a patient group that underwent MR-guided radiotherapy for upper abdominal malignancies with an abdominal corset. Approach. Fast online 2D cine MRI was used to extract tumor motion during beam-on time. These tumor motion profiles were combined with linac log files to reconstruct the delivered dose in 89 fractions of MR-guided SBRT in twenty patients. Aside the measured tumor motion, motion profiles were also simulated for a wide range of respiratory amplitudes and drifts, and their subsequent dosimetric impact was calculated in every fraction. Main results. The average (SD) D_99% of the gross tumor volume (GTV), relative to the planned D_99%, was 0.98 (0.03). The average (SD) relative D_0.5cc of the duodenum, small bowel and stomach was 0.99 (0.03), 1.00 (0.03), and 0.97 (0.05), respectively. No correlation of respiratory amplitude with dosimetric impact was observed. Fractions with larger baseline drifts generally led to a larger uncertainty of dosimetric impact on the GTV and organs at risk (OAR). The simulations yielded that the delivered dose is highly dependent on the direction of on baseline drift. Especially in anatomies where the OARs are closely abutting the GTV, even modest LR or AP drifts can lead to substantial deviations from the planned dose. Significance. The vast majority of the fractions was only modestly impacted by intrafraction motion, increasing our confidence that MR-guided SBRT with abdominal compression can be safely executed for patients with abdominal tumors, without the use of gating or tracking strategies.

Objective. To propose a novel moment-based loss function for predicting 3D dose distribution for the challenging conventional lung intensity modulated radiation therapy plans. The moment-based loss function is convex and differentiable and can easily incorporate clinical dose volume histogram (DVH) domain knowledge in any deep learning (DL) framework without computational overhead. Approach. We used a large dataset of 360 (240 for training, 50 for validation and 70 for testing) conventional lung patients with 2 Gy × 30 fractions to train the DL model using clinically treated plans at our institution. We trained a UNet like convolutional neural network architecture using computed tomography, planning target volume and organ-at-risk contours as input to infer corresponding voxel-wise 3D dose distribution. We evaluated three different loss functions: (1) the popular mean absolute error (MAE) loss, (2) the recently developed MAE + DVH loss, and (3) the proposed MAE + moments loss. The quality of the predictions was compared using different DVH metrics as well as dose-score and DVH-score, recently introduced by the AAPM knowledge-based planning grand challenge. Main results. Model with (MAE + moment) loss function outperformed the model with MAE loss by significantly improving the DVH-score (11%, p < 0.01) while having similar computational cost. It also outperformed the model trained with (MAE + DVH) by significantly improving the computational cost (48%) and the DVH-score (8%, p < 0.01). Significance. DVH metrics are widely accepted evaluation criteria in the clinic. However, incorporating them into the 3D dose prediction model is challenging due to their non-convexity and non-differentiability. Moments provide a mathematically rigorous and computationally efficient way to incorporate DVH information in any DL architecture. The code, pretrained models, docker container, and Google Colab project along with a sample dataset are available on our DoseRTX GitHub (https://github.com/nadeemlab/DoseRTX)

Classification of arteries and veins in cerebral angiograms can increase the safety of neurosurgical procedures, such as StereoElectroEncephaloGraphy, and aid the diagnosis of vascular pathologies, as arterovenous malformations. We propose a new method for vessel classification using the contrast medium dynamics in rotational digital subtraction angiography (DSA). After 3D DSA and angiogram segmentation, contrast enhanced projections are processed to suppress soft tissue and bone structures attenuation effect and further enhance the CM flow. For each voxel labelled as vessel, a time intensity curve (TIC) is obtained as a linear combination of temporal basis functions whose weights are addressed by simultaneous algebraic reconstruction technique (SART 3.5D), expanded to include dynamics. Each TIC is classified by comparing the areas under the curve in the arterial and venous phases. Clustering is applied to optimize the classification thresholds. On a dataset of 60 patients, a median value of sensitivity (90%), specificity (91%), and accuracy (92%) were obtained with respect to annotated arterial and venous voxels up to branching order 4–5. Qualitative results are also presented about CM arrival time mapping and its distribution in arteries and veins respectively. In conclusion, this study shows a valuable impact, at no protocol extra-cost or invasiveness, concerning surgical planning related to the enhancement of arteries as major organs at risk. Also, it opens a new scope on the pathophysiology of cerebrovascular dynamics and its anatomical relationships.