Towards real-time EPID-based 3D in vivo dosimetry for IMRT with Deep Neural Networks: A feasibility study.

We investigate the potential of the Deep Dose Estimate (DDE) neural network to predict 3D dose distributions inside patients with Monte Carlo (MC) accuracy, based on transmitted EPID signals and patient CTs. The network was trained using as input patient CTs and first-order dose approximations (FOD). Accurate dose distributions (ADD) simulated with MC were given as training targets. 83 pelvic CTs were used to simulate ADDs and respective EPID signals for subfields of prostate IMRT plans (gantry at 0∘). FODs were produced as backprojections from the EPID signals. 581 ADD-FOD sets were produced and divided into training and test sets. An additional dataset simulated with gantry at 90∘ (lateral set) was used for evaluating the performance of the DDE at different beam directions. The quality of the FODs and DDE-predicted dose distributions (DDEP) with respect to ADDs, from the test and lateral sets, was evaluated with gamma analysis (3%,2 mm). The passing rates between FODs and ADDs were as low as 46%, while for DDEPs the passing rates were above 97% for the test set. Meaningful improvements were also observed for the lateral set. The high passing rates for DDEPs indicate that the DDE is able to convert FODs into ADDs. Moreover, the trained DDE predicts the dose inside a patient CT within 0.6 s/subfield (GPU), in contrast to 14 h needed for MC (CPU-cluster). 3D in vivo dose distributions due to clinical patient irradiation can be obtained within seconds, with MC-like accuracy, potentially paving the way towards real-time EPID-based in vivo dosimetry.

Optimization of Phase Space files from clinical linear accelerators.

This work proposes a methodology to produce an optimized phase-space (PhSp) for the Elekta Synergy linac by tuning the energy and direction of particles inside the 6-MV Elekta Precise PhSp, provided by the International Atomic Energy Agency (IAEA), for Monte Carlo (MC) simulations. First, the energies of the particles emerging from the original PhSp were increased by different factors, producing new PhSps. Percentage depth dose (PDD) profiles were simulated and compared to measured data from a Synergy linac for 6-MV photon beam. This process was repeated until a minimum difference was reached. Particles' directions were then manipulated following identified correlations to lateral profiles, resulting in two distinct perturbation factors based on inline and crossline profiles. Both factors were merged into one single optimal factor. For energy optimization, an increase of 0.32 MeV applied to all particles inside the original PhSp, but to 0.511 MeV annihilation photons, provided the best results. The direction optimization factor was the combination of the individual factors for inline (0.605%) and crossline (0.051%). The agreement between measured and simulated profiles, when using the optimized PhSp, improved considerably in comparison to simulations performed with the original IAEA PhSp. For all fields and depths analyzed, the discrepancies for PDD, inline and crossline profiles dropped from 11.2%, 15.7% and 27.5% to under 1.4%, 4.7% and 13.2%, respectively. The optimized PhSp should not replace the full linac modelling, however it offers an alternative for MC dose calculations when neither geometric details nor validated IAEA PhSp are available to the user.

A novel approach to EPID-based 3D volumetric dosimetry for IMRT and VMAT QA.

Intensity modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT) are relatively complex treatment delivery techniques and require quality assurance (QA) procedures. Pre-treatment dosimetric verification represents a fundamental QA procedure in daily clinical routine in radiation therapy. The purpose of this study is to develop an EPID-based approach to reconstruct a 3D dose distribution as imparted to a virtual cylindrical water phantom to be used for plan-specific pre-treatment dosimetric verification for IMRT and VMAT plans. For each depth, the planar 2D dose distributions acquired in air were back-projected and convolved by depth-specific scatter and attenuation kernels. The kernels were obtained by making use of scatter and attenuation models to iteratively estimate the parameters from a set of reference measurements. The derived parameters served as a look-up table for reconstruction of arbitrary measurements. The summation of the reconstructed 3D dose distributions resulted in the integrated 3D dose distribution of the treatment delivery. The accuracy of the proposed approach was validated in clinical IMRT and VMAT plans by means of gamma evaluation, comparing the reconstructed 3D dose distributions with Octavius measurement. The comparison was carried out using (3%, 3 mm) criteria scoring 99% and 96% passing rates for IMRT and VMAT, respectively. An accuracy comparable to the one of the commercial device for 3D volumetric dosimetry was demonstrated. In addition, five IMRT and five VMAT were validated against the 3D dose calculation performed by the TPS in a water phantom using the same passing rate criteria. The median passing rates within the ten treatment plans was 97.3%, whereas the lowest was 95%. Besides, the reconstructed 3D distribution is obtained without predictions relying on forward dose calculation and without external phantom or dosimetric devices. Thus, the approach provides a fully automated, fast and easy QA procedure for plan-specific pre-treatment dosimetric verification.