Purpose: The Linac-MR system enables real-time treatment adjustments by tracking tumor motion, making accurate dose calculation crucial for adaptive RT. Standard treatment planning system (TPS) computational times are impractical for real-time application, necessitating an ultrafast photon dose calculation algorithm. As a first step, we investigate the model generalization of iDoTA, an AI-based dose calculation algorithm, in conventional RT treatment scenarios.

 

Methods: The iDoTA model is based on a CNN and transformer architecture. The speed and accuracy of the trained model are tested using a water phantom and conventional RT plans. The model’s predictions are compared to AcurosXB (Varian Eclipse15.6). Gamma-analysis and average relative error (ARE) quantify dosimetric differences between iDoTA and AcurosXB.

 

Results: The model is evaluated for a 10x10cm² field, 100cm SAD, delivering 100cGy in one fraction for photon energies 6, 10, and 15MV. All energies are evaluated with a gamma-test criteria of 1-3%/3mm. The ARE/gamma pass-rates (1%/3mm) are 4.5%/98.6% for 6MV, 5.2%/87.5% for 10MV, and 5.7%/86.8% for 15MV. The highest relative errors occur in the build-up and penumbra regions. Predictions using patient plans show reduced error in the build-up region.

 

Conclusions: Errors in the penumbra align with those reported in iDoTA; however, build-up region errors are significantly higher in the phantom geometry. This may stem from iDoTA’s training on patient geometries alone. Further work includes re-training with a combination of patient and phantom geometries. Ultimately, we will re-train iDoTA with a parallel magnetic field, at which point we will compare to TOPAS MC predictions (with magnetic field).

Purpose Real-time plastic scintillation dosimeters (PSDs) are a promising alternative to passive dosimeters for dosimetry of ultrahigh dose rate (UHDR) beams. The aim of this study was to characterize the response of three scintillators and an optical fiber exposed to 9MeV UHDR electrons.

Methods A 4-channel PSD was designed with BCF-12, EJ-212 and proprietary Medscint scintillators and a blank fiber. Light signals from the four channels were measured independently using the Medscint Hyperscint RP200 dosimetry platform. The 9MeV UHDR electron beam generated by a Mobetron was used to investigate PSD response to instantaneous dose rates (IDR) from (4.78-11.12)E5Gy/s, average dose rates (ADR) from 21.04-181.63Gy/s, dose per pulse (DPP) of up to 3.39Gy, and accumulated dose of 2.2kGy. Recovery of light output and spectral changes after radiation damage was assessed over time (18-84 days).

 

Results: BCF-12, Medscint and EJ-212 response was found to vary by 12.5, 13.5 and 26% respectively for different IDRs but remained constant within 1.5% for all ADRs measured. PSD response was linear with DPP of up to 3.39Gy, even after 2.2kGy of accumulated dose in the probe. Loss of light output of 0.4-3.2%/kGy and spectral shift towards yellower wavelengths were reported after 2.2kGy. Light output recovered to 82.7-95.8% of its initial value after 53 days, with BCF-12 being the least damaged after recovery. Spectral recovery was nearly complete for all channels.

 

Conclusions: While PSDs are affected by radiation damage, they maintain their linear response to dose and exhibit significant recovery, enabling dose measurements if recalibrated before each use.

Purpose: Commissioning a Gamma Knife® (GK) (Elekta, Sweden) unit requires precise dosimetric reference measurements, particularly for small fields. The IAEA TRS-483/AAPM TG-155 code of practice (2017, 2021) provides correction factors for reference and relative dosimetry, but advancements in detectors and phantom materials necessitate validation and updates. This study calculates the beam correction factor k_Q, accounting for field-size, geometry, phantom material, and beam quality differences between conventional and machine-specific reference fields.

 

Methods: Monte Carlo simulations of a GK Perfexion with various phantom and ionization chamber combinations were performed using EGSnrc. Phantom materials included Solid Water, PMMA, polystyrene, ABS, and Lucy. Ionization chambers with different electrode materials, including aluminum, C-552, and steel, were sourced from Standard Imaging, IBA, and PTW. The reference field was a tabulated Cobalt-60 beam (10×10 cm², 100 cm SSD), while the machine-specific field used a 16x16 mm² GK Perfexion phase space. Simulations measured absorbed dose to water and dose to chamber air cavities for different phantom and chamber combinations.

 

Results: Correction factors were determined for various phantom/chamber combinations in parallel and perpendicular orientations. Linear trends with phantom electron density were observed across all chambers. Electrode material differences had the most pronounced effect on orientation-dependent perturbations.

 

Conclusion: This study updates TRS-483 correction factors for GK commissioning and quality assurance, emphasizing the influence of chamber orientation and electrode material. Findings confirm the linear scaling of photon fluence ratios with electron density, contributing to improved dosimetric accuracy in small field radiotherapy.

Purpose HDR brachytherapy is widely used for prostate cancer treatment but remains prone to errors, including manual procedures and source positioning inaccuracies. In-vivo dosimetry (IVD) becomes essential for precise treatment delivery. This study proposes a simple calibration method for plastic scintillation dosimeters (PSD), bypassing dose gradient and positioning issues in HDR brachytherapy.

 

Methods The PRB-0057 PSD (Medscint, Québec, Canada), consisting of a 1×1-mm scintillating fiber coupled to a 20-m Eska-GH-4001 clear optical fiber (Mitsubishi-Rayon, Tokyo, Japan), was connected to the Medscint Hyperscint-RP200 spectrometer for light-yield collection. Spectral calibration was performed at a LINAC following standard vendor guidelines, independently removing stem effects. Dose calibration was performed with a 6MV beam at dmax before brachytherapy measurements with a Sk=29447U Iridium-192 source...

 

Results Dose rates were measured at 10-Hz along source’s vertical z-axis at an x/y distance of ~1.2-cm with a 0.2-cm step. RDs remained below 2.5% at ~1.2-cm, corresponding to a positional uncertainty below 70-µm. At depths up to 8-cm, RDs increased to ~5%, corresponding to positional uncertainties reaching 3-mm due to reduced light-yield.

 

Conclusion Results confirm the effectiveness of a standard 6MV external beam calibration using the hyper-spectral technique for PSD in time-resolved HDR brachytherapy IVD. Uncertainties near the source are consistent with afterloader/IBA motorized unit reproducibility.

Purpose: To investigate the performance of a novel ultra-high dose rate (UHDR) electron beam capable of delivering dose per pulse up to 270 Gy.

 

Materials and Methods: Rapid and accurate characterization required a combination of detectors. Portable graphite and aluminum calorimeters were used to provide real-time measurements. The ratio from two similar, but independent, calorimeters was used to verify system integrity and provide redundancy. Dose profiles were measured using Ashland HD-V2 gafchromic film which enabled a volume averaging correction for the calorimeters. Absorbed dose to water was measured directly using alanine pellets directly trac...

 

Results: For the two different radial beam profiles, the calorimeters agreed to within 0.8 % which is consistent with uncertainties. The calorimeter/ICT ratio showed very good stability over one day (standard deviation < 0.2 %) but showed larger variations from day to day. Measurements using alanine for the highest dose per pulse yielded a value of (276 ± 4) Gy, consistent at the 2 % level with data supplied at installation.

 

Conclusions: Combining complementary dosimetry systems enables accurate characterization of UHDR beams, providing robust absorbed dose measurements with an uncertainty of around 1.5 %.

Purpose: To develop a novel real-time 3D dosimeter for quality assurance of linear accelerators used in external beam radiotherapy.

 

Methods: An experimental setup was designed using a 1000 cm³ scintillating cube, two plane mirrors positioned to capture three orthogonal views of the volume in a single image, and a CCD camera. Because scintillation is proportional to the deposited dose, the light emission pattern also reflects the dose distribution. A deep learning model, leveraging advantages from transformers and convolutional neural networks, was trained to reconstruct the 3D light distribution from a single image, achieving a resolution of 1 mm³ in 34.23±0.11 ms on an NVIDIA GeForce RTX 3090 GPU. The model was pre-trained using 30,000 synthetic data points generated with Python and fine-tuned using 250 experimental data points. Each data point featured an image containing three orthogonal views and the corresponding dose distribution calculated by the treatment planning system used as ground truth.

 

Results: Evaluation was performed on test data values exceeding 10% of the maximum dose. On test dataset of 25 samples, the model achieved a mean gamma success rate of 89.2±0.7% (3%/3mm criterion). The mean values for structural similarity index measure and mean squared error were 0.875±0.006 and 0.0050±0.0004, respectively. Gamma maps showed strong agreement, with only minimal discrepancies in high dose gradient regions.

 

Conclusion: This approach demonstrates the potential of combining plastic scintillators and deep learning for real-time 3D dose measurements. Future investigation will improve the model’s architecture, expand the dataset to mitigate overfitting, and introduce temporal resolution.