Fast and accurate dose predictions for novel radiotherapy treatments in heterogeneous phantoms using conditional 3D-UNet generative adversarial networks
Florian Mentzel, Kevin Kröninger, Michael Lerch, Olaf Nackenhorst, Jason Paino, Anatoly Rosenfeld, Ayu Saraswati, Ah Chung Tsoi, Jens Weingarten, Markus Hagenbu… 展开作者
Florian Mentzel, Kevin Kröninger, Michael Lerch, Olaf Nackenhorst, Jason Paino, Anatoly Rosenfeld, Ayu Saraswati, Ah Chung Tsoi, Jens Weingarten, Markus Hagenbuchner, Susanna Guatelli
Novel radiotherapy techniques like synchrotron X-ray microbeam radiation therapy (MRT), require fast dose distribution predictions that are accurate at the sub-mm level, especially close to tissue/bone/air interfaces. Monte Carlo physics simulations are recognised to be one of the most accurate tools to predict the dose delivered in a target tissue but can be very time consuming and therefore prohibitive for treatment planning. Faster dose prediction algorithms are usually developed for clinically deployed treatments only. In this work, we explore a new approach for fast and accurate dose estimations suitable for novel treatments using digital phantoms used in pre-clinical development and modern machine learning techniques. We develop a generative adversarial network (GAN) model, which is able to emulate the equivalent Geant4 Monte Carlo simulation with adequate accuracy, and use it to predict the radiation dose delivered by a broad synchrotron beam to various phantoms. The energy depositions used for the training of the GAN are obtained using full Geant4 Monte Carlo simulations of a synchrotron radiation broad beam passing through the phantoms. The energy deposition is scored and predicted in voxel matrices of size 140x18x18 with a voxel edge length of 1 mm. The GAN model consists of two competing 3D convolutional neural networks, which are conditioned on the photon beam and phantom properties. The energy deposition predictions inside all phantom geometries under investigation show deviations of less than 3% of the maximum deposited energy from the simulation for roughly 99% of the voxels in the field of the beam. The computing time for a single prediction is reduced from several hundred hours using Geant4 simulation to less than a second using the GAN model.
A new variant of the pencil-beam (PB) algorithm for dose distribution calculation for radiotherapy with protons and heavier ions, the grid-dose spreading (GDS) algorithm, is proposed. The GDS algorithm is intrinsically faster than conventional PB algorithms due to approximations in convolution integral, where physical calculations are decoupled from simple grid-to-grid energy transfer. It was effortlessly implemented to a carbon-ion radiotherapy treatment planning system to enable realistic beam blurring in the field, which was absent with the broad-beam (BB) algorithm. For a typical prostate …
In radiotherapy, a trade-off exists between computational workload/speed and dose calculation accuracy. Calculation methods like pencil-beam convolution can be much faster than Monte-Carlo methods, but less accurate. The dose difference, mostly caused by inhomogeneities and electronic disequilibrium, is highly correlated with the dose distribution and the underlying anatomical tissue density. We hypothesize that a conversion scheme can be established to boost low-accuracy doses to high-accuracy, using intensity information obtained from computed tomography (CT) images. A deep learning-driven f…
Cloud computing allows for vast computational resources to be leveraged quickly and easily in bursts as and when required. Using the Amazon Elastic Compute Cloud and the Amazon Simple Storage Solution, we describe a technique that allows for Monte Carlo radiotherapy dose calculations to be performed using GEANT4 and executed in the cloud. Simulation cost and completion time was evaluated as a function of instance count using compute instances acquired via biding on the Elastic Compute Cloud spot market. Bidding for instances on the instance spot market was found to be 35-60% of the cost of on-…
Dose painting has been developed to modulate the required dose in the target area without increasing the toxicity in healthy areas. Apart from determining the accurate location and size of tumors, quantitative functional imaging can be used to implement the dose painting. Functional imaging, such as multi-parameter MRI and PET CT, allows us to achieve biological dose escalation by increasing the dose in certain areas or voxels that are therapy-resistant in the gross tumor volume while reducing the dose in the less aggressive area or voxels. Functional imaging can serve as an indicator of thera…
Voxel-Based Conversion of Hypofractionated Radiotherapy Dose Distributions to 2 Gy-Equivalent OAR Constraints: Proof-of-Concept Demonstrating the Radiobiological Benefits of Hypofractionation in a Prostate Radiotherapy Case
In carbon-ion radiotherapy, single-beam delivery each day in alternate directions has been commonly practiced for operational efficiency, taking advantage of the Bragg peak and the relative biological effectiveness (RBE) for uniform dose conformation to a tumor. The treatment plans are usually evaluated with total RBE-weighted dose, which is however deficient in relevance to the biological effect in the linear-quadratic model due to its quadratic-dose term, or the dose-fractionation effect. In this study, we reformulate the extrapolated response dose (ERD), or synonymously BED, which normalize…
Purpose: Various dose calculation algorithms are available for radiation therapy for cancer patients. However, these algorithms are faced with the tradeoff between efficiency and accuracy. The fast algorithms are generally less accurate, while the accurate dose engines are often time consuming. In this work, we try to resolve this dilemma by exploring deep learning (DL) for dose calculation. Methods: We developed a new radiotherapy dose calculation engine based on a modified Hierarchically Densely Connected U-net (HD U-net) model and tested its feasibility with prostate intensity-modulated rad…
Dose distribution (depthwise and laterally) to organs outside the radiotherapy treatment field can be significant and therefore is of clinical interest from the radiation protection point of view. In the present work, measurements were performed in a locally fabricated polystyrene phantom using TLD chips (LiF-100) for different teletherapy units (Cobalt-60 gamma ray, 120 kVp X-ray and 250 kVp X-ray) to estimate the dose distribution at distances up to 40 cm from the field edge along the central axes of the field size. Finally, the dose distribution for Cobalt-60 beam energy is parameterized as…
This work addresses computing techniques for dose calculations in treatment planning with proton and ion beams, based on an efficient kernel-convolution method referred to as grid-dose spreading (GDS) and accurate heterogeneity-correction method referred to as Gaussian beam splitting. The original GDS algorithm suffered from distortion of dose distribution for beams tilted with respect to the dose-grid axes. Use of intermediate grids normal to the beam field has solved the beam-tilting distortion. Interplay of arrangement between beams and grids was found as another intrinsic source of artifac…
Due to the increasing complexity of radiotherapy delivery, accurate dose verification has become an essential part of the clinical treatment process. The purpose of this work was to develop an electronic portal image (EPI) based pre-treatment verification technique capable of quickly reconstructing 3D dose distributions from both coplanar and non-coplanar treatments. The dose reconstruction is performed in a spherical water phantom by modulating, based on EPID measurements, pre-calculated Monte Carlo (MC) doselets defined on a spherical coordinate system. This is called the spherical doselet m…
To enable magnetic resonance (MR)-only radiotherapy and facilitate modelling of radiation attenuation in humans, synthetic-CT (sCT) images need to be generated. Considering the application of MR-guided radiotherapy and online adaptive replanning, sCT generation should occur within minutes. This work aims at assessing whether an existing deep learning network can rapidly generate sCT images to be used for accurate MR-based dose calculations in the entire pelvis. A study was conducted on data of 91 patients with prostate, rectal and cervical cancer who underwent external beam radiotherapy acquir…
The RefleXion X1 is the first biology-guided radiotherapy (BgRT) system. Its dual 90-degree PET detector collects fewer pair production events compared to a full-ring diagnostic PET system. In the proposed BgRT workflow, a short scan is acquired before treatment delivery to ensure image quality and consistency. The shorter scan time, a quarter of the simulation scan time, also leads to fewer coincidence events and hence reduced image quality. In this study, we proposed a patient-specific mean teacher UNet (MT-UNet) to enhance PET image quality and low-dose PET reconstruction on RefleXion X1. P…
Targeting at the development of an accurate and efficient dose calculation engine for online adaptive radiotherapy, we have implemented a finite size pencil beam (FSPB) algorithm with a 3D-density correction method on GPU. This new GPU-based dose engine is built on our previously published ultrafast FSPB computational framework [Gu et al. Phys. Med. Biol. 54 6287-97, 2009]. Dosimetric evaluations against Monte Carlo dose calculations are conducted on 10 IMRT treatment plans (5 head-and-neck cases and 5 lung cases). For all cases, there is improvement with the 3D-density correction over the con…
Ultra high dose rate electron beams also known as FLASH radiotherapy is becoming of importance in several preclinial cancer treatment studies. However, due to the dose rate used during the irradiation sessions, no real time dose monitoring device exists to date. In this work, we present the development of a beam current transformer (BCT), from the choice of the ferromagnetic component and the realisation of the shielding to the design of a front-end electronics based on a trans-impedance circuit in order to perform a low noise optimization of the detector. The BCT prototype is able to monitor …
Recent works in automated radiotherapy treatment planning have used machine learning based on historical treatment plans to infer the spatial dose distribution for a novel patient directly from the planning image. We present an atlas-based approach which learns a dose prediction model for each patient (atlas) in a training database, and then learns to match novel patients to the most relevant atlases. The method creates a spatial dose objective, which specifies the desired dose-per-voxel, and therefore replaces any requirement for specifying dose-volume objectives for conveying the goals of tr…
In multi-stage radiotherapy, a patient is treated in several stages separated by weeks or months. This regimen has been motivated mostly by radiobiological considerations, but also provides an approach to reduce normal tissue dose by exploiting tumor shrinkage. The paper considers the optimal design of multi-stage treatments, motivated by the clinical management of large liver tumors for which normal liver dose constraints prohibit the administration of an ablative radiation dose in a single treatment. We introduce a dynamic tumor model that incorporates three factors: radiation induced cell k…
An essential component in proton radiotherapy is the algorithm to calculate the radiation dose to be delivered to the patient. The most common dose algorithms are fast but they are approximate analytical approaches. However their level of accuracy is not always satisfactory, especially for heterogeneous anatomic areas, like the thorax. Monte Carlo techniques provide superior accuracy, however, they often require large computation resources, which render them impractical for routine clinical use. Track-repeating algorithms, for example the Fast Dose Calculator, have shown promise for achieving …
Treatment with high energy ionizing radiation is one of the main methods in modern cancer therapy that is in clinical use. During the last decades, two main approaches to dose calculation were used, Monte Carlo simulations and semi-empirical models based on Fermi-Eyges theory. A third way to dose calculation has only recently attracted attention in the medical physics community. This approach is based on the deterministic kinetic equations of radiative transfer. Starting from these, we derive a macroscopic partial differential equation model for electron transport in tissue. This model involve…
Purpose: Variations in proton Relative Biological Effectiveness (RBE) with Linear Energy Transfer (LET) remain one of the largest sources of uncertainty in proton radiotherapy. This work seeks to identify physics-based metrics which can be applied to reduce this biological uncertainty. Materials and Methods: Three different physical metrics - dose, dose $\times$ LET and a complexity-weighted dose (CWD, Dose $\times$ (1+$κLET_D$) ) were compared with in vitro experimental studies of proton RBE and clinical treatment plans analysed using RBE models. The biological effects of protons in each syst…
It has long been known that radiation biology plays an important role and it is necessary for radiotherapy treatments. The radiation effects on normal and malignant tissues after exposure range from a femtosecond to months and years thereafter [1,2]. Therefore, to optimize treatment, it is crucial to explain and understand these mechanisms [3-5]. Providing a conceptual basis for radiotherapy and identifying the mechanisms and processes that underlie the tumor and normal tissue responses to irradiation can help to explain the observed phenomena [6]. Examples include understanding hypoxia, reoxy…
Automating the radiotherapy treatment planning process is a technically challenging problem. The majority of automated approaches have focused on customizing and inferring dose volume objectives to used in plan optimization. In this work we outline a multi-patient atlas-based dose prediction approach that learns to predict the dose-per-voxel for a novel patient directly from the computed tomography (CT) planning scan without the requirement of specifying any objectives. Our method learns to automatically select the most effective atlases for a novel patient, and then map the dose from those at…
The limits of TDF (time, dose, and fractionation) and linear quadratic models have been known for a long time. Medical physicists and physicians are required to provide fast and reliable interpretations regarding the delivered doses or any future prescriptions relating to treatment changes. We therefore propose a calculation interface under the GNU license to be used for equivalent doses, biological doses, and normal tumor complication probability (Lyman model). The methodology used draws from several sources: the linear-quadratic-linear model of Astrahan, the repopulation effects of Dale, and…
The next great leap toward improving treatment of cancer with radiation will require the combined use of online adaptive and magnetic resonance guided radiation therapy techniques with automatic X-ray beam orientation selection. Unfortunately, by uniting these advancements, we are met with a substantial expansion in the required dose information and consequential increase to the overall computational time imposed during radiation treatment planning, which cannot be handled by existing techniques for accelerating Monte Carlo dose calculation. We propose a deep convolutional neural network appro…
The aim of this study is to investigate the dose-deposition enhancement by alpha-particle irradiation in a cellular model using carbon and oxygen chemical compositions.A simulation study was performed to study dose enhancement due to carbon and oxygen for a human cell where Geant4 code used for the alpha-particle irradiation to the cellular phantom. The characteristic of dose enhancement in the nucleus and cytoplasm by the alpha-particle radiation was investigated based on concentrations of the carbon and oxygen compositions and was compared with those by gold and gadolinium.The results show t…
Dose painting of hypoxic tumour sub-volumes using positron-emission tomography (PET) has been shown to improve tumour control in silico in several sites. Pancreatic cancer presents a more stringent challenge, given its proximity to critical organs-at-risk (OARs) and anatomic motion. A radiobiological model was developed to estimate clonogen survival fraction (SF), using 18F-fluoroazomycin arabinoside PET (FAZA PET) images from ten patients with pancreatic cancer to quantify oxygen enhancement effects. For each patient, four simulated five-fraction stereotactic body radiotherapy (SBRT) plans we…
Background and purpose: To investigate the possible contribution of indirect damage and damage saturation to tumour control obtained with SBRT/SRS treatments for early-stage NSCLC and brain metastases. Methods and Materials: We have constructed a dataset of early-stage NSCLC and brain metastases dose-response. These data were fitted to models based on the linear-quadratic (LQ), the linear-quadratic-linear (LQL), and phenomenological modifications of the LQ-model to account for indirect cell damage. We use the Akaike-Information-Criterion formalism to compare performance, and studied the stabil…
This paper describes the dose delivery system used at the Centro Nazionale di Adroterapia Oncologica (CNAO) for ion beam modulated scanning radiotherapy. CNAO Foundation, INFN and University of Torino have developed and commissioned a Dose Delivery System (DDS) to monitor and guide ion beams accelerated by a synchrotron and to distribute the dose with a 3D scanning technique. The target volume, segmented in several layers orthogonally to the beam direction, is irradiated by thousands of pencil beams which must be steered and held to the prescribed positions until the prescribed number of parti…
Monte Carlo (MC) simulation is commonly considered to be the most accurate dose calculation method in radiotherapy. However, its efficiency still requires improvement for many routine clinical applications. In this paper, we present our recent progress towards the development a GPU-based MC dose calculation package, gDPM v2.0. It utilizes the parallel computation ability of a GPU to achieve high efficiency, while maintaining the same particle transport physics as in the original DPM code and hence the same level of simulation accuracy. In GPU computing, divergence of execution paths between th…
Carbon ion is the mostly common used particle in heavy ion radiotherapy. In this paper, carbon ion dose in tumor treatment for human eye was simulated with FLUKA code, 80 MeV/u carbon beam was irradiated into the human eye from two directions, The first is from the lateral-forward direction which was a typical therapeutic condition, maximum dose was deposited in the tumor volume. The second one was that beam irradiated into eyes from the forward direction which may cause certain medical accident. The calculated results are compared with other reports. The agreement indicates that this method c…
The Radiotherapy treatment planning optimization process based on a quasi-Newton algorithm with an object function containing dose-volume constraints is not guaranteed to converge when the dose value in the dose-volume constraint is a critical value of the dose distribution. This is caused by finite differentiability of the dose-volume histogram at such values. A closer look near such values reveals that convergence is most likely not at stake, but it might be slowed down.
\item[Purpose] A recent study revealed that polyethylene (PE) would cause extra carbon-ion attenuation per range shift by 0.45\%/cm due to compositional differences in nuclear interactions. The present study aims to assess the influence of PE range compensators on tumor dose in carbon-ion radiotherapy. \item[Methods] Carbon-ion radiation was modeled to be composed of primary carbon ions and secondary particles, for each of which the dose and the relative biological effectiveness (RBE) were estimated at a tumor depth in the middle of spread-out Bragg peak. Assuming exponential behavior for atte…
Toward CT-Equivalent Image Quality in Low-Dose Radiotherapy Planning: Conditional Diffusion-Based CBCT-to-CT Synthesis and the Impact of CBCT Input Representation
Jordan University of Science and Technology（约旦科技大学）；University of Dundee（邓迪大学）；NHS Ninewells Dundee（NHS邓迪奈恩韦尔斯医院）；University of Glasgow（格拉斯哥大学）
Monte Carlo simulation is the most accurate method for absorbed dose calculations in radiotherapy. Its efficiency still requires improvement for routine clinical applications, especially for online adaptive radiotherapy. In this paper, we report our recent development on a GPU-based Monte Carlo dose calculation code for coupled electron-photon transport. We have implemented the Dose Planning Method (DPM) Monte Carlo dose calculation package (Sempau et al, Phys. Med. Biol., 45(2000)2263-2291) on GPU architecture under CUDA platform. The implementation has been tested with respect to the origina…
Cancer is a primary cause of morbidity and mortality worldwide. The radiotherapy plays a more and more important role in cancer treatment. In the radiotherapy, the dose distribution maps in patient need to be calculated and evaluated for the purpose of killing tumor and protecting healthy tissue. Monte Carlo (MC) radiation transport calculation is able to account for all aspects of radiological physics within 3D heterogeneous media such as the human body and generate the dose distribution maps accurately. However, an MC calculation for doses in radiotherapy usually takes a great mass of time t…
Purpose: The purpose of the present article is to propose a modified version of the D-shaped collimator used at the Univ. Hosp. of Essen for retinoblastoma treatment that reduces even further the irradiation field with the scope to reduce the risk of radio-induced secondary malignancies. The new dedicated D-shaped collimator is easier to build and produces dose distributions that only differ on the field size with respect to the dose distributions obtained by the current collimator in use. Methods: The Monte Carlo code PENELOPE was used to study the effect that the structural elements of the c…
This research considers potential dose increase in target due to cisplatin (Pt) concentration and radiation type. Cisplatin concentrations from 0.003 to 120 mM were used. Monte-Carlo simulation of Linear accelerator (Elekta Synergy) and X-ray tube (Xstrahl300) was carried out using Geant4 and PClab. As the first step of this research, we performed simulation of energy spectrum from radiotherapy units (spectrum model). The next step was the modeling of linear accelerator head and X-ray tube, and the distribution of dose in the water phantom (PDD model). At the second stage, dose changes were in…
Purpose: To investigate the validity of two Monte Carlo simulation absolute dosimetry approaches in the case of a small field dedicated `D'-shaped collimator used for the retinoblastoma treatment with external photon beam radiotherapy. Methods: The Monte Carlo code {\sc penelope} is used to simulate the linac, the dedicated collimator and a water phantom. The absolute doses (in Gy per monitor unit) for the field sizes considered are obtained within the approach of Popescu {\it et al.} in which the tallied backscattered dose in the monitor chamber is accounted for. The results are compared to e…
With the continued evolution of modern radiation therapy towards high precision delivery of high therapeutic doses to the tumour while optimally sparing surrounding healthy tissue, imaging becomes a crucial component to identify the intended target, properly position it at the treatment site and, in more advanced research applications, visualize the treatment delivery. This contribution reviews the main role of imaging in modern external beam radiation therapy, with special emphasis on emerging ion beam therapy techniques, which aim at exploiting the favourable properties of ion interaction in…
A novel phase-space source implementation has been designed for GPU-based Monte Carlo dose calculation engines. Due to the parallelized nature of GPU hardware, it is essential to simultaneously transport particles of the same type and similar energies but separated spatially to yield a high efficiency. We present three methods for phase-space implementation that have been integrated into the most recent version of the GPU-based Monte Carlo radiotherapy dose calculation package gDPM v3.0. The first method is to sequentially read particles from a patient-dependent phase-space and sort them on-th…
FLASH is a new treatment modality that requires optimization of dose, dose rate, and LET. Here we validate these three quantities under FLASH conditions, which includes the quantum uncertainty in the time-dependent instantaneous dose rate (IDR) curves and LET spectra that underlie the newly proposed integrated optimization framework. Measurements of dose, IDR, and LET were performed at the Emory Proton Therapy Center using a FLASH proton beam with a nominal energy of 250 MeV and a 3D printed ridge filter. Because 3D printing resin is made from a proprietary chemical formula, we developed a met…
