Nanodosimetric quantity-weighted dose optimization for carbon-ion treatment planning.

Dose verification of treatment plans is an essential step in radiotherapy workflows. In this work, we propose a novel method of treatment planning based on nanodosimetric quantity-weighted dose (NQWD), which could realize biological representation using pure physical quantities for biological-oriented carbon ion-beam treatment plans and their direct verification. The relationship between nanodosimetric quantities and relative biological effectiveness (RBE) was studied with the linear least-squares method for carbon-ion radiation fields. Next, under the framework of the matRad treatment planning platform, NQWD was optimized using the existing RBE-weighted dose (RWD) optimization algorithm. The schemes of NQWD-based treatment planning were compared with the RWD treatment plans in term of the microdosimetric kinetic model (MKM). The results showed that the nanodosimetric quantity F3 - 10 had a good correlation with the radiobiological effect reflected by the relationship between RBE and F3 - 10. Moreover, the NQWD-based treatment plans reproduced the RWD plans generally. Therefore, F3 - 10 could be adopted as a radiation quality descriptor for carbon-ion treatment planning. The novel method proposed herein not only might be helpful for rapid physical verification of biological-oriented ion-beam treatment plans with the development of experimental nanodosimetry, but also makes the direct comparison of ion-beam treatment plans in different institutions possible. Thus, our proposed method might be potentially developed to be a new strategy for carbon-ion treatment planning and improve patient safety for carbon-ion radiotherapy.

DR-only Carbon-ion radiotherapy treatment planning via deep learning.

PURPOSE: To evaluate the feasibility of patient-specific digital radiography (DR)-only treatment planning for carbon ion radiotherapy in anthropomorphic thorax-and-abdomen phantom and head-and-neck patients. METHODS: The study was conducted on the anthropomorphic phantom and head-and-neck patients. We collected computed tomography (CT) and DR images of the phantom and cone beam CT (CBCT) and DR images of the patients, respectively. Two different deep neural networks were established to correlate the relationships between DR and digitally reconstructed radiograph (DRR) images, as well as DRR and CT images. The similarity between CT and predicted CT images was evaluated by computing the mean absolute error (MAE), root mean square error (RMSE), peak signal-to-noise ratio (PSNR) and structural similarity (SSIM), respectively. Dose calculations on the predicted CT images were compared against the true CT-based dose distributions for carbon-ion radiotherapy treatment planning with intensity-modulated pencil-beam spot scanning. Relative dose differences in the target volumes and organ-at-risks were computed and three-dimensional gamma analyses (3 mm, 3%) were performed. RESULTS: The average MAE, RMSE, PSNR and SSIM of the framework were 0.007, 0.144, 37.496 and 0.973, respectively. The average relative dose differences between the predicted CT- and CT-based dose distributions at the same carbon-ion irradiation settings for the phantom and the patients were <2% and ≤4%, respectively. The average gamma pass-rates were >98% for the predicted CT-based versus CT-based carbon ion plans of the phantom and the patients. CONCLUSION: We have demonstrated the feasibility of a patient-specific DR-only treatment planning workflow for heavy ion radiotherapy by using deep learning approach.

Validation and testing of a novel pencil-beam model derived from Monte Carlo simulations in carbon-ion treatment planning for different scenarios.

PURPOSE: The calculation ability of the newly-proposed accurate beam model, the double Gaussian-logistic (DG-L) model, was validated in both homogeneous and heterogeneous phantoms to provide helpful information for its future application in clinical carbon-ion treatment planning system (TPS). METHODS: MatRad was used as the new algorithm test platform. Based on Monte Carlo (MC) method, the basic database in matRad was generated, then comparative dosimetric analyses between the single Gaussian (SG), double Gaussian (DG) and DG-L models against the MC recalculations were performed on the treatment plans of a cubic water phantom, a TG119 phantom and a liver patient scenario. Absolute dose differences, dose-volume histograms (DVHs) and global γ-index analyses derived from the treatment plans were evaluated. RESULTS: Calculated with the DG-L model, the deviations of the target dose coverage (D95) for the cubic water phantom, the TG119 phantom and the liver patient case against the MC recalculations could be reduced from -2.5%, -4.6% and -6.4% to -0.3%, -2.0% and -4.5% respectively compared to the SG model, while the γ pass rates (3%/3mm) could be enhanced from 98.0%, 90.6% and 90.1% to 99.8%, 95.7% and 91.6%, respectively. The novel beam model also shows improved performance compared with the DG model, without substantially increasing the computation time. CONCLUSIONS: The DG-L model could effectively improve the dose calculation accuracy and mitigate the delivered dose deficiency in target volumes compared to the SG and DG models. The lateral heterogeneities should be considered for its future implementation in a clinical TPS.

Artificial Intelligence Radiotherapy Planning: Automatic Segmentation of Human Organs in CT Images Based on a Modified Convolutional Neural Network.

Objective: Precise segmentation of human organs and anatomic structures (especially organs at risk, OARs) is the basis and prerequisite for the treatment planning of radiation therapy. In order to ensure rapid and accurate design of radiotherapy treatment planning, an automatic organ segmentation technique was investigated based on deep learning convolutional neural network. Method: A deep learning convolutional neural network (CNN) algorithm called BCDU-Net has been modified and developed further by us. Twenty two thousand CT images and the corresponding organ contours of 17 types delineated manually by experienced physicians from 329 patients were used to train and validate the algorithm. The CT images randomly selected were employed to test the modified BCDU-Net algorithm. The weight parameters of the algorithm model were acquired from the training of the convolutional neural network. Result: The average Dice similarity coefficient (DSC) of the automatic segmentation and manual segmentation of the human organs of 17 types reached 0.8376, and the best coefficient reached up to 0.9676. It took 1.5-2 s and about 1 h to automatically segment the contours of an organ in an image of the CT dataset for a patient and the 17 organs for the CT dataset with the method developed by us, respectively. Conclusion: The modified deep neural network algorithm could be used to automatically segment human organs of 17 types quickly and accurately. The accuracy and speed of the method meet the requirements of its application in radiotherapy.

Nanodosimetric understanding to the dependence of the relationship between dose-averaged lineal energy on nanoscale and LET on ion species.

With the extension of ion species in ion-beam radiotherapy, the sole dependence of relative biological effectiveness (RBE) on linear energy transfer (LET) is insufficient when comparing RBE for ion beams with the same LET value. The aim of the present study was to provide a systematic study of the nanodosimetry for ion beams with the same LET value. Based on the calculated LET profiles of ion beams with range about 130 mm, lineal energy spectra and dose-averaged lineal energy [Formula: see text] on 4 nm site for various clinical ion beams were obtained. Then, the lineal energy spectra and [Formula: see text] values were compared for ion beams with the same LET values. The results showed that the relationships between [Formula: see text] and LET for various ion beams present an dependence on ion species. For ion beams with the same LET value, the ion beams with smaller nucleon number yielded greater [Formula: see text] values. The probability of the small-nucleon-number ion beams to generate large energy deposition events on nanoscale was higher than that of the large-nucleon-number ion beams. The dependence of the relationship between RBE and LET on ion species might be attributed to the fluctuation of energy depositions on nanometer scale.

Nanodosimetric quantities and RBE of a clinically relevant carbon-ion beam.

PURPOSE: Although carbon-ion therapy is becoming increasingly attractive to the treatment of tumors, details about the ionization pattern formed by therapeutic carbon-ion beam in tissue have not been fully investigated. In this work, systematic calculations for the nanodosimetric quantities and relative biological effectiveness (RBE) of a clinically relevant carbon-ion beam were studied for the first time. METHODS: The method combining both track structure and condensed history Monte Carlo (MC) simulations was adopted to calculate the nanodosimetric quantities. Fragments and energy spectra at different positions of the radiation field of a clinically relevant carbon-ion pencil beam were generated by means of MC simulations in water. Nanodosimetric quantities such as mean ionization cluster size ( M 1 ), the first moment of conditional cluster size ( M 1 C 2 ), cumulative probability ( F 2 ), and conditional cumulative probability ( F 3 C 2 ) at these positions were then acquired based on the spectra and the pre-calculated nanodosimetric database created by track structure MC simulations. What's more, a novel approach to calculate RBE based on the said nanodosimetric quantities was introduced. The RBE calculations were then conducted for the carbon-ion beam at different water-equivalent depths. RESULTS: Lateral distributions at various water-equivalent depths of both the nanodosimetric quantities and RBE values were obtained. The values of M 1 , M 1 C 2 , F 2 , and F 3 C 2 were 1.49, 2.67, 0.30, and 0.38 at the plateau at the beam central axis and maximized at 2.79, 5.69, 0.47, and 0.68 at the depths around the Bragg peak, respectively. At a given depth, M 1 and F 2 decreased laterally with increasing the distance to the beam central axis while M 1 C 2 and F 3 C 2 remained nearly unchanged at first and then decreased except for M 1 C 2 at the rising edge of the Bragg peak. The calculated RBE values were 1.07 at the plateau and 3.13 around the Bragg peak. Good agreement between the calculated RBE values and experimental data was obtained. CONCLUSIONS: Different nanodosimetric quantities feature the track structure of therapeutic carbon-ion beam in different manners. Detailed ionization patterns generated by carbon-ion beam could be characterized by nanodosimetric quantities. Moreover the combined method adopted in this work to calculate nanodosimetric quantities is not only valid but also convenient. Nanodosimetric quantities are significantly helpful for the RBE calculations in carbon-ion therapy.

Technical Note: Effect of magnetic fields on the microdosimetry of carbon-ion beams.

PURPOSE: To investigate the influence of magnetic fields on the microdosimetry of carbon-ion beams and the scaling effect of tissue equivalent proportional counter (TEPC) defined as the change of energy deposition in a TEPC from that in a microscopic scale region of interest due to the presence of a magnetic field in combination with the TEPC larger physical dimensions. METHODS: Geant4-based Monte Carlo simulations were conducted to calculate the microdosimetric quantities for carbon-ion beams with different initial energies (10-290 MeV/u) under magnetic fields of various strengths (0.5-3 T). The calculations were performed for a 1 µm spherical volume made of tissue, and for spherical TEPCs of 1 and 10 mm in diameter. Then, values of dose-averaged lineal energy (yD ) were acquired for the different scenarios to analyze the effect of magnetic fields on the microdosimetry of carbon-ion beams and the scaling effect of TEPC. RESULTS: The yD values and lineal energy spectra in the 1 µm spherical tissue volume for the scenarios without magnetic field and with magnetic fields of different strengths and directions remained nearly the same for the various energy carbon-ion beams. However, compared with those of the 1 µm spherical tissue volume, an increase of of yD values and an obvious shift of the lineal energy spectra for the TEPCs of 1 and 10 mm in diameter under magnetic fields were found. CONCLUSIONS: The application of magnetic fields under 3 T has no significant influence on the microdosimetric results of carbon-ion beams. However, there is definitely a scaling effect when using TEPC for microdosimetric study, which alters the reading of TEPC in the presence of magnetic fields. Novel methods to correct the reading of TEPC or scaling effect-resistant microdosimetric measurement detectors are urgently needed to perform experimental microdosimetric studies under magnetic fields.

A novel pencil beam model for carbon-ion dose calculation derived from Monte Carlo simulations.

An accurate kernel model is of vital importance for pencil-beam dose algorithm in charged particle therapy using precise spot-scanning beam delivery, in which an accurate depiction of the low dose envelope is especially crucial. Based on the Monte Carlo method, we investigated the dose contribution of secondary particles to the total dose and proposed a novel beam model to depict the lateral dose distribution of carbon-ion pencil beam in water. We demonstrated that the low dose envelope in single-spot profiles in water could be adequately modelled with the addition of a logistic distribution to a double Gaussian one, which was verified in both single carbon-ion pencil beam and superposed fields of different sizes with multiple pencil beams. Its superiority was mainly manifested at medium depths especially for high-energy beams with small fields compared with single, double and triple Gaussian models, where the secondary particles influenced the total dose considerably. The double Gaussian-logistic model could reduce the deviations from 4.1%, 1.7% to 0.3% in the plateau and peak regions, and from 19.2%, 4.9% to 1.2% in the tail region compared for the field size factor (FSF) calculations of 344 MeV/u carbon-ion pencil beam with the single and double Gaussian models. Compared with the triple Gaussian one, our newly-proposed model was on a par with it, even better than it in the plateau and peak regions. Thus our work will be helpful for improving the dose calculation accuracy for carbon-ion therapy.

Fragmentation level determines mitochondrial damage response and subsequently the fate of cancer cells exposed to carbon ions.

OBJECTIVES: Although mitochondria are known to play an important role in radiation-induced cellular damage response, the mechanisms of how radiation elicits mitochondrial responses are largely unknown. MATERIALS AND METHODS: Human cervical cancer cell line HeLa and human breast cancer cell lines MCF-7 and MDA-MB-231 were irradiated with high LET carbon ions at low (0.5 Gy) and high (3 Gy) doses. Mitochondrial functions, dynamics, mitophagy, intrinsic apoptosis and total apoptosis, and survival fraction were investigated after irradiation. RESULTS: We found that carbon ions irradiation induced two different mitochondrial morphological changes and corresponding responses in cancer cells. Cells exposed to carbon ions of 0.5 Gy exhibited only modestly truncated mitochondria, and subsequently damaged mitochondria could be eliminated through mitophagy. In contrast, mitochondria within cells insulted by 3 Gy radiation split into punctate and clustered ones, which were associated with apoptotic cell death afterward. Inhibition of mitochondrial fission by Drp1 or FIS1 knockdown or with the Drp1 inhibitor mdivi-1 suppressed mitophagy and potentiated apoptosis after irradiation at 0.5 Gy. However, inhibiting fission led to mitophagy and increased cell survival when cells were irradiated with carbon ions at 3 Gy. CONCLUSION: We proposed a stress response model to provide a mechanistic explanation for the mitochondrial damage response to high-LET carbon ions.

Evaluation of mesh- and binary-based contour propagation methods in 4D thoracic radiotherapy treatments using patient 4D CT images.

Based on four dimensional (4D) computed tomography (CT) images, mesh- and binary-based contour propagation algorithms for 4D thoracic radiotherapy treatments were evaluated. Gross tumor volumes (GTVs), lungs, hearts and spinal cords on the CT images at the end-exhale and end-inhale phases for six patients were delineated by the physician. All volumes of interest (VOIs) were automatically propagated from the end-exhale phase to the end-inhale phase using two propagation methods. The propagated VOIs were quantitatively compared with the VOIs contoured at the end-inhale phase by the physician using Dice Similarity Coefficient (DSC), Mean Slicewise Hausdorff Distance (MSHD), Center Of Mass (COM) displacement and volume difference. A two-sided Student's t test was implemented to examine the significance of the differences between the results obtained from the two algorithms. For GTVs, statistically significant differences between the two algorithms were not observed. For all the other VOIs, the mesh-based method showed higher mean DSCs for the heart, left lung, right lung and spinal cord, lower mean MSHD for the spinal cord, lower mean COM displacement for the heart, and lower mean volume differences for the left lung, right lung and spinal cord with statistically significant differences than the binary-based method. The running time for propagation was approximately 3s and 3min for the mesh- and binary-based methods, respectively. Collectively, the mesh-based algorithm provides superiorities in running time and reliability for contour propagation in 4D radiotherapy.

Respiratory motion management using audio-visual biofeedback for respiratory-gated radiotherapy of synchrotron-based pulsed heavy-ion beam delivery.

PURPOSE: To efficiently deliver respiratory-gated radiation during synchrotron-based pulsed heavy-ion radiotherapy, a novel respiratory guidance method combining a personalized audio-visual biofeedback (BFB) system, breath hold (BH), and synchrotron-based gating was designed to help patients synchronize their respiratory patterns with synchrotron pulses and to overcome typical limitations such as low efficiency, residual motion, and discomfort. METHODS: In-house software was developed to acquire body surface marker positions and display BFB, gating signals, and real-time beam profiles on a LED screen. Patients were prompted to perform short BHs or short deep breath holds (SDBH) with the aid of BFB following a personalized standard BH/SDBH (stBH/stSDBH) guiding curve or their own representative BH/SDBH (reBH/reSDBH) guiding curve. A practical simulation was performed for a group of 15 volunteers to evaluate the feasibility and effectiveness of this method. Effective dose rates (EDRs), mean absolute errors between the guiding curves and the measured curves, and mean absolute deviations of the measured curves were obtained within 10%-50% duty cycles (DCs) that were synchronized with the synchrotron's flat-top phase. RESULTS: All maneuvers for an individual volunteer took approximately half an hour, and no one experienced discomfort during the maneuvers. Using the respiratory guidance methods, the magnitude of residual motion was almost ten times less than during nongated irradiation, and increases in the average effective dose rate by factors of 2.39-4.65, 2.39-4.59, 1.73-3.50, and 1.73-3.55 for the stBH, reBH, stSDBH, and reSDBH guiding maneuvers, respectively, were observed in contrast with conventional free breathing-based gated irradiation, depending on the respiratory-gated duty cycle settings. CONCLUSIONS: The proposed respiratory guidance method with personalized BFB was confirmed to be feasible in a group of volunteers. Increased effective dose rate and improved overall treatment precision were observed compared to conventional free breathing-based, respiratory-gated irradiation. Because breathing guidance curves could be established based on the respective average respiratory period and amplitude for each patient, it may be easier for patients to cooperate using this technique.