Synthetic CT for single-fraction neoadjuvant partial breast irradiation on an MRI-linac.

A synthetic computed tomography (sCT) is required for daily plan optimization on an MRI-linac. Yet, only limited information is available on the accuracy of dose calculations on sCT for breast radiotherapy. This work aimed to (1) evaluate dosimetric accuracy of treatment plans for single-fraction neoadjuvant partial breast irradiation (PBI) on a 1.5 T MRI-linac calculated on a) bulk-density sCT mimicking the current MRI-linac workflow and b) deep learning-generated sCT, and (2) investigate the number of bulk-density levels required. For ten breast cancer patients we created three bulk-density sCTs of increasing complexity from the planning-CT, using bulk-density for: (1) body, lungs, and GTV (sCTBD1); (2) volumes for sCTBD1plus chest wall and ipsilateral breast (sCTBD2); (3) volumes for sCTBD2plus ribs (sCTBD3); and a deep learning-generated sCT (sCTDL) from a 1.5 T MRI in supine position. Single-fraction neoadjuvant PBI treatment plans for a 1.5 T MRI-linac were optimized on each sCT and recalculated on the planning-CT. Image evaluation was performed by assessing mean absolute error (MAE) and mean error (ME) in Hounsfield Units (HU) between the sCTs and the planning-CT. Dosimetric evaluation was performed by assessing dose differences, gamma pass rates, and dose-volume histogram (DVH) differences. The following results were obtained (median across patients for sCTBD1/sCTBD2/sCTBD3/sCTDLrespectively): MAE inside the body contour was 106/104/104/75 HU and ME was 8/9/6/28 HU, mean dose difference in the PTVGTVwas 0.15/0.00/0.00/-0.07 Gy, median gamma pass rate (2%/2 mm, 10% dose threshold) was 98.9/98.9/98.7/99.4%, and differences in DVH parameters were well below 2% for all structures except for the skin in the sCTDL. Accurate dose calculations for single-fraction neoadjuvant PBI on an MRI-linac could be performed on both bulk-density and deep learning sCT, facilitating further implementation of MRI-guided radiotherapy for breast cancer. Balancing simplicity and accuracy, sCTBD2showed the optimal number of bulk-density levels for a bulk-density approach.

DeepDose: a robust deep learning-based dose engine for abdominal tumours in a 1.5 T MRI radiotherapy system.

We present a robust deep learning-based framework for dose calculations of abdominal tumours in a 1.5 T MRI radiotherapy system. For a set of patient plans, a convolutional neural network is trained on the dose of individual multi-leaf-collimator segments following the DeepDose framework. It can then be used to predict the dose distribution per segment for a set of patient anatomies. The network was trained using data from three anatomical sites of the abdomen: prostate, rectal and oligometastatic tumours. A total of 216 patient fractions were used, previously treated in our clinic with fixed-beam IMRT using the Elekta MR-linac. For the purpose of training, 176 fractions were used with random gantry angles assigned to each segment, while 20 fractions were used for the validation of the network. The ground truth data were calculated with a Monte Carlo dose engine at 1% statistical uncertainty per segment. For a total of 20 independent abdominal test fractions with the clinical angles, the network was able to accurately predict the dose distributions, achieving 99.4% ± 0.6% for the whole plan prediction at the 3%/3 mm gamma test. The average dose difference and standard deviation per segment was 0.3% ± 0.7%. Additional dose prediction on one cervical and one pancreatic case yielded high dose agreement of 99.9% and 99.8% respectively for the 3%/3 mm criterion. Overall, we show that our deep learning-based dose engine calculates highly accurate dose distributions for a variety of abdominal tumour sites treated on the MR-linac, in terms of performance and generality.

Proof-of-concept delivery of intensity modulated arc therapy on the Elekta Unity 1.5 T MR-linac.

In this work we present the first delivery of intensity modulated arc therapy on the Elekta Unity 1.5 T MR-linac. The machine's current intensity modulated radiation therapy based control system was modified suitably to enable dynamic delivery of radiation, for the purpose of exploring MRI-guided radiation therapy adaptation modes in a research setting. The proof-of-concept feasibility was demonstrated by planning and delivering two types of plans, each investigating the performance of different parts of a dynamic treatment. A series of fixed-speed arc plans was used to show the high-speed capabilities of the gantry during radiation, while several fully modulated prostate plans-optimised following the volumetric modulated arc therapy approach-were delivered in order to establish the performance of its multi-leaf collimator and diaphragms. These plans were delivered to Delta4 Phantom+ MR and film phantoms passing the clinical quality assurance criteria used in our clinic. In addition, we also performed some initial MR imaging experiments during dynamic therapy, demonstrating that the impact of radiation and moving gantry/collimator components on the image quality is negligible. These results show that arc therapy is feasible on the Elekta Unity system. The machine's high performance components enable dynamic delivery during fast gantry rotation and can be controlled in a stable fashion to deliver fully modulated plans.

Deep learning prediction of proton and photon dose distributions for paediatric abdominal tumours.

OBJECTIVE: Dose prediction using deep learning networks prior to radiotherapy might lead tomore efficient modality selections. The study goal was to predict proton and photon dose distributions based on the patient-specific anatomy and to assess their clinical usage for paediatric abdominal tumours. MATERIAL AND METHODS: Data from 80 patients with neuroblastoma or Wilms' tumour was included. Pencil beam scanning (PBS) (5 mm/ 3%) and volumetric-modulated arc therapy (VMAT) plans (5 mm) were robustly optimized on the internal target volume (ITV). Separate 3-dimensional patch-based U-net networks were trained to predict PBS and VMAT dose distributions. Doses, planning-computed tomography images and relevant optimization masks (ITV, vertebra and organs-at-risk) of 60 patients were used for training with a 5-fold cross validation. The networks' performance was evaluated by computing the relative error between planned and predicted dose-volume histogram (DVH) parameters for 20 inference patients. In addition, the organs-at-risk mean dose difference between modalities was calculated using planned and predicted dose distributions (ΔDmean = DVMAT-DPBS). Two radiation oncologists performed a blind PBS/VMAT modality selection based on either planned or predicted ΔDmean. RESULTS: Average DVH differences between planned and predicted dose distributions were ≤ |6%| for both modalities. The networks classified the organs-at-risk Dmean difference as a gain (ΔDmean > 0) with 98% precision. An identical modality selection based on planned compared to predicted ΔDmean was made for 18/20 patients. CONCLUSION: Deep learning networks for accurate prediction of proton and photon dose distributions for abdominal paediatric tumours were established. These networks allowing fast dose visualisation might aid in identifying the optimal radiotherapy technique when experience and/or resources are unavailable.

DeepDose: Towards a fast dose calculation engine for radiation therapy using deep learning.

We present DeepDose, a deep learning framework for fast dose calculations in radiation therapy. Given a patient anatomy and linear-accelerator IMRT multi-leaf-collimator shape or segment, a novel set of physics-based inputs is calculated that encode the linac machine parameters into the underlying anatomy. These inputs are then used to train a deep convolutional network to derive the dose distribution of individual MLC shapes on a given patient anatomy. In this work we demonstrate the proof-of-concept application of DeepDose on 101 prostate patients treated in our clinic with fixed-beam IMRT. The ground-truth data used for training, validation and testing of the prediction were calculated with a state-of-the-art Monte Carlo dose engine at 1% statistical uncertainty per segment. A deep convolution network was trained using the data of 80 patients at the clinically used 3 mm3 grid spacing while 10 patients were used for validation. For another 11 independent test patients, the network was able to accurately estimate the segment doses from the clinical plans of each patient passing the clinical QA when compared with the Monte Carlo calculations, yielding on average 99.9%±0.3% for the forward calculated patient plans at 3%/3 mm gamma tests. Dose prediction using the trained network was very fast at approximately 0.9 seconds for the input generation and 0.6 seconds for single GPU inference per segment and 1 minute per patient in total. The overall performance of this dose calculation framework in terms of both accuracy and inference speed, makes it compelling for online adaptive workflows where fast segment dose calculations are needed.

Dosimetric impact of soft-tissue based intrafraction motion from 3D cine-MR in prostate SBRT.

To investigate the dosimetric impact of intrafraction translation and rotation motion of the prostate, as extracted from daily acquired post-treatment 3D cine-MR based on soft-tissue contrast, in extremely hypofractionated (SBRT) prostate patients. Accurate dose reconstruction is performed by using a prostate intrafraction motion trace which is obtained with a soft-tissue based rigid registration method on 3D cine-MR dynamics with a temporal resolution of 11 s. The recorded motion of each time-point was applied to the planning CT, resulting in the respective dynamic volume used for dose calculation. For each treatment fraction, the treatment delivery record was generated by proportionally splitting the plan into 11 s intervals based on the delivered monitor units. For each fraction the doses of all partial plan/dynamic volume combinations were calculated and were summed to lead to the motion-affected fraction dose. Finally, for each patient the five fraction doses were summed, yielding the total treatment dose. Both daily and total doses were compared to the original reference dose of the respective patient to assess the impact of the intrafraction motion. Depending on the underlying motion of the prostate, different types of motion-affected dose distributions were observed. The planning target volumes (PTVs) ensured CTV_30 (seminal vesicles) D99% coverage for all patients, CTV_35 (prostate corpus) coverage for 97% of the patients and GTV_50 (local boost) for 83% of the patients when compared against the strict planning target D99% value. The dosimetric impact due to prostate intrafraction motion in extremely hypofractionated treatments was determined. The presented study is an essential step towards establishing the actual delivered dose to the patient during radiotherapy fractions.

Towards fast online intrafraction replanning for free-breathing stereotactic body radiation therapy with the MR-linac.

The hybrid MRI-radiotherapy machines, like the MR-linac (Elekta AB, Stockholm, Sweden) installed at the UMC Utrecht (Utrecht, The Netherlands), will be able to provide real-time patient imaging during treatment. In order to take advantage of the system's capabilities and enable online adaptive treatments, a new generation of software should be developed, ranging from motion estimation to treatment plan adaptation. In this work we present a proof of principle adaptive pipeline designed for high precision stereotactic body radiation therapy (SBRT) suitable for sites affected by respiratory motion, like renal cell carcinoma (RCC). We utilized our research MRL treatment planning system (MRLTP) to simulate a single fraction 25 Gy free-breathing SBRT treatment for RCC by performing inter-beam replanning for two patients and one volunteer. The simulated pipeline included a combination of (pre-beam) 4D-MRI and (online) 2D cine-MR acquisitions. The 4DMRI was used to generate the mid-position reference volume, while the cine-MRI, via an in-house motion model, provided three-dimensional (3D) deformable vector fields (DVFs) describing the anatomical changes during treatment. During the treatment fraction, at an inter-beam interval, the mid-position volume of the patient was updated and the delivered dose was accurately reconstructed on the underlying motion calculated by the model. Fast online replanning, targeting the latest anatomy and incorporating the previously delivered dose was then simulated with MRLTP. The adaptive treatment was compared to a conventional mid-position SBRT plan with a 3 mm planning target volume margin reconstructed on the same motion trace. We demonstrate that our system produced tighter dose distributions and thus spared the healthy tissue, while delivering more dose to the target. The pipeline was able to account for baseline variations/drifts that occurred during treatment ensuring target coverage at the end of the treatment fraction.