CBCT-DRRs superior to CT-DRRs for target-tracking applications for pancreatic SBRT.

Objective.In current radiograph-based intra-fraction markerless target-tracking, digitally reconstructed radiographs (DRRs) from planning CTs (CT-DRRs) are often used to train deep learning models that extract information from the intra-fraction radiographs acquired during treatment. Traditional DRR algorithms were designed for patient alignment (i.e.bone matching) and may not replicate the radiographic image quality of intra-fraction radiographs at treatment. Hypothetically, generating DRRs from pre-treatment Cone-Beam CTs (CBCT-DRRs) with DRR algorithms incorporating physical modelling of on-board-imagers (OBIs) could improve the similarity between intra-fraction radiographs and DRRs by eliminating inter-fraction variation and reducing image-quality mismatches between radiographs and DRRs. In this study, we test the two hypotheses that intra-fraction radiographs are more similar to CBCT-DRRs than CT-DRRs, and that intra-fraction radiographs are more similar to DRRs from algorithms incorporating physical models of OBI components than DRRs from algorithms omitting these models.Approach.DRRs were generated from CBCT and CT image sets collected from 20 patients undergoing pancreas stereotactic body radiotherapy. CBCT-DRRs and CT-DRRs were generated replicating the treatment position of patients and the OBI geometry during intra-fraction radiograph acquisition. To investigate whether the modelling of physical OBI components influenced radiograph-DRR similarity, four DRR algorithms were applied for the generation of CBCT-DRRs and CT-DRRs, incorporating and omitting different combinations of OBI component models. The four DRR algorithms were: a traditional DRR algorithm, a DRR algorithm with source-spectrum modelling, a DRR algorithm with source-spectrum and detector modelling, and a DRR algorithm with source-spectrum, detector and patient material modelling. Similarity between radiographs and matched DRRs was quantified using Pearson's correlation and Czekanowski's index, calculated on a per-image basis. Distributions of correlations and indexes were compared to test each of the hypotheses. Distribution differences were determined to be statistically significant when Wilcoxon's signed rank test and the Kolmogorov-Smirnov two sample test returnedp≤ 0.05 for both tests.Main results.Intra-fraction radiographs were more similar to CBCT-DRRs than CT-DRRs for both metrics across all algorithms, with allp≤ 0.007. Source-spectrum modelling improved radiograph-DRR similarity for both metrics, with allp< 10-6. OBI detector modelling and patient material modelling did not influence radiograph-DRR similarity for either metric.Significance.Generating DRRs from pre-treatment CBCT-DRRs is feasible, and incorporating CBCT-DRRs into markerless target-tracking methods may promote improved target-tracking accuracies. Incorporating source-spectrum modelling into a treatment planning system's DRR algorithms may reinforce the safe treatment of cancer patients by aiding in patient alignment.

Evaluation of deep learning based implanted fiducial markers tracking in pancreatic cancer patients.

Real-time target position verification during pancreas stereotactic body radiation therapy (SBRT) is important for the detection of unplanned tumour motions. Fast and accurate fiducial marker segmentation is a Requirement of real-time marker-based verification. Deep learning (DL) segmentation techniques are ideal because they don't require additional learning imaging or prior marker information (e.g., shape, orientation). In this study, we evaluated three DL frameworks for marker tracking applied to pancreatic cancer patient data. The DL frameworks evaluated were (1) a convolutional neural network (CNN) classifier with sliding window, (2) a pretrained you-only-look-once (YOLO) version-4 architecture, and (3) a hybrid CNN-YOLO. Intrafraction kV images collected during pancreas SBRT treatments were used as training data (44 fractions, 2017 frames). All patients had 1-4 implanted fiducial markers. Each model was evaluated on unseen kV images (42 fractions, 2517 frames). The ground truth was calculated from manual segmentation and triangulation of markers in orthogonal paired kV/MV images. The sensitivity, specificity, and area under the precision-recall curve (AUC) were calculated. In addition, the mean-absolute-error (MAE), root-mean-square-error (RMSE) and standard-error-of-mean (SEM) were calculated for the centroid of the markers predicted by the models, relative to the ground truth. The sensitivity and specificity of the CNN model were 99.41% and 99.69%, respectively. The AUC was 0.9998. The average precision of the YOLO model for different values of recall was 96.49%. The MAE of the three models in the left-right, superior-inferior, and anterior-posterior directions were under 0.88 ± 0.11 mm, and the RMSE were under 1.09 ± 0.12 mm. The detection times per frame on a GPU were 48.3, 22.9, and 17.1 milliseconds for the CNN, YOLO, and CNN-YOLO, respectively. The results demonstrate submillimeter accuracy of marker position predicted by DL models compared to the ground truth. The marker detection time was fast enough to meet the requirements for real-time application.

In-line MRI-LINAC depth dose measurements using an in-house plastic scintillation dosimeter.

Plastic scintillation dosimeters (PSDs) have many properties that make them desirable for relative dosimetry with MRI-LINACs. An in-house PSD, Farmer ionisation chamber and Gafchromic EBT3 film were used to measure central axis percentage depth dose distributions (PDDs) at the Australian MRI-LINAC Mean errors were calculated between each detector's responses, where the in-house PSD was on average within 0.7% of the Farmer chamber and 1.4% of film, while the Farmer chamber and film were on average within 1.1% of each other. However, the PSD systematically over-estimated the dose as depth increased, approaching a maximum overestimation of the order of 3.5% for the smallest field size measured. This trend was statistically insignificant for all other field sizes measured; further investigation is required to determine the source of this effect. The calculated values of mean absolute error are comparable to the those of trusted dosimeters reported in the literature. These mean absolute errors, and the ubiquity of desirable dosimetric qualities inherent to PSDs suggest that PSDs in general are accurate for relative dosimetry with the MRI-LINAC. Further investigation is required into the source of the reported systematic trends dependent on field-size and depth of measurement.

Temporal modelling of beryllium oxide ceramics' real-time OSL for dosimetry with a superficial 140 kVp X-ray beam.

A new analysis method for the rtOSL of BeO ceramics is presented, using temporal curve fitting of an expected rtOSL signal to measured rtOSL signals. The presented technique does not require heavy signal averaging to determine the OSL bleaching correction associated with the ΔrtOSL method, reducing uncertainties in the post-correction rtOSL. The corrected rtOSL signal was demonstrated to be linear with dose, and dose-rate independent. The presented technique is expected to be applicable for many other dosimeters capable of the rtOSL technique. The presented technique achieved relative uncertainties in the corrected rtOSL between 3.4% and 6.5%. The initial measurements are promising, but uncertainties are required to be further improved upon before the technique can be used clinically.

MRI-LINAC beam profile measurements using a plastic scintillation dosimeter.

Plastic scintillation dosimeters (PSDs) possess many desirable qualities for dosimetry with LINACs. These qualities are expected to make PSDs effective for MRI-LINAC dosimetry, however little research has been conducted investigating their dosimetric performance with MRI-LINACs. In this work, an in-house PSD was used to measure 8 beam profiles with an in-line MRI-LINAC, compared with film measurements. One dimensional global gamma indices (γ) and corresponding γ pass rates were calculated to compare PSD and film profiles for the 1%/1 mm, 2%/2 mm and 3%/3 mm criterion. The mean global pass rates were 85.8%, 97.5% and 99.4% for the 1%/1 mm, 2%/2 mm and 3%/3 mm criteria, respectively. The majority of the γ failures occurred in the penumbral regions. Penumbra widths were measured to be slightly narrower with the PSD compared to film, however, the uncertainties in the measured penumbra widths brought the PSD and film penumbra widths into agreement. Differences in dose were calculated between the PSD and film, and remained within 2.2% global agreement for the central regions and 1.5% global agreement for out of field regions. These values for range of agreement were similar to the those reported in the literature for other dosimeters which are trusted for relative MRI-LINAC dosimetry.

First measurements with a plastic scintillation dosimeter at the Australian MRI-LINAC.

MRI-LINACs combine MRI and LINAC technologies with the potential for image guided radiation therapy with optimal soft-tissue contrast. In this work, we present the advantages and limitations of plastic scintillation dosimeters (PSDs) for relative dosimetry with MRI-LINACs. PSDs possess many desirable qualities, including magnetic field insensitivity and irradiation angle independence, which are expected to make them suitable for dosimetry with MRI-LINACs. An in-house PSD was used to measure field size output factors as well as a percent depth dose distribution and the beam quality index TPR20/10 at a [Formula: see text] cm2 field size. Measurements were repeated with a Scanditronix/Wellhofer FC65-G ionisation chamber and PTW 60019 microDiamond detector for comparison. Relative differences were calculated between the three detectors, where the mean difference in dose was 1.2% between the PSD and ionisation chamber, 1.9% between the PSD and microDiamond detector and 1.3% between the microDiamond detector and the ionisation chamber. The closeness between the three mean differences in doses suggests that PSDs are feasible for relative dosimetry with MRI-LINACs.

An algorithmic approach to single-probe Cherenkov removal in pulsed x-ray beams.

PURPOSE: The removal of Cherenkov light in an optical dosimetry system is an important process to ensure accurate dosimetry without compromising spatial resolution. Many solutions have been presented in the literature, each with advantages and disadvantages. We present a methodology to remove Cherenkov light from a scintillator fiber optic dosimeter in a pulsed megavoltage x-ray beam using the temporal waveform across the pulse. METHODS: A sample waveform of Cherenkov light can be measured by exposing only the fiber to the beam. By assuming that the Cherenkov waveform closely matches the intensity of incident radiation, this waveform can be convoluted with the instantaneous scintillation response function to generate an expected scintillation signal. By finding the least-squares fit between these two functions and the experimental data, the estimated Cherenkov contribution can be subtracted off the net signal. This can be applied for arbitrarily complex Cherenkov waveforms (within the 2 ns timing resolution of the data acquisition), and in fact, the results suggest more fluctuations in the waveforms provide a better fit to data. RESULTS: Four beam profiles for different field sizes and energies were found with this method. They closely matched references data measured with ionization chamber with average differences across the beam no more than 4%. Noisy waveforms are assumed to be the primary cause of differences between the analyzed scintillator and IC results. We propose methods for improving the results and optimizing the data acquisition and analysis processes. CONCLUSIONS: These results demonstrate that it is possible and effective with a single probe to use function fitting of expected data to experimental to remove a complicated Cherenkov signal from the net light signal in pulsed-beam optical dosimetry.

Temporal separation of Cerenkov radiation and scintillation using a clinical LINAC and artificial intelligence.

Convolutional neural network (CNN) type artificial intelligences were trained to estimate the Cerenkov radiation present in the temporal response of a LINAC irradiated scintillator-fiber optic dosimeter. The CNN estimate of Cerenkov radiation is subtracted from the combined scintillation and Cerenkov radiation temporal response of the irradiated scintillator-fiber optic dosimeter, giving the sole scintillation signal, which is proportional to the scintillator dose. The CNN measured scintillator dose was compared to the background subtraction measured scintillator dose and ionisation chamber measured dose. The dose discrepancy of the CNN measured dose was on average 1.4% with respect to the ionisation chamber measured dose, matching the 1.4% average dose discrepancy of the background subtraction measured dose with respect to the ionisation chamber measured dose. The developed CNNs had an average time of 3 ms to calculate scintillator dose, permitting the CNNs presented to be applicable for dosimetry in real time.

Temporal separation of Cerenkov radiation and scintillation using artificial neural networks in Clinical LINACs.

The irradiation of scintillator-fiber optic dosimeters by clinical LINACs results in the measurement of scintillation and Cerenkov radiation. In scintillator-fiber optic dosimetry, the scintillation and Cerenkov radiation responses are separated to determine the dose deposited in the scintillator volume. Artificial neural networks (ANNs) were trained and applied in a novel single probe method for the temporal separation of scintillation and Cerenkov radiation. Six dose profiles were measured using the ANN, with the dose profiles compared to those measured using background subtraction and an ionisation chamber. The average dose discrepancy of the ANN measured dose was 2.2% with respect to the ionisation chamber dose and 1.2% with respect to the background subtraction measured dose, while the average dose discrepancy of the background subtraction dose was 1.6% with respect to the ionisation chamber dose. The ANNs performance was degraded when compared with background subtraction, arising from an inaccurate model used to synthesise ANN training data.

A comparison of temporal Cherenkov separation techniques in pulsed signal scintillator dosimetry.

Cherenkov radiation is the primary source of unwanted light in a scintillator dosimetry system. In this work we compare two techniques for temporally separating Cherenkov radiation from a slow scintillator signal. These techniques are applicable to a pulsed radiation beam. We found that by analysing the rising edge of the light pulse to identify the fast Cherenkov light only removed 74% of the Cherenkov light. By integrating the tail of the signal where only scintillation light is present a more accurate result is achieved. The average of the results of the two methods provides up to a 90% improvement in the accuracy of the relative dose when compared to ionisation chamber, in certain measurements. This work demonstrates an alternative methodology for the removal of Cherenkov light using signal analysis, while preserving all the scintillation light signal and minimising the bulk of the experimental equipment.

Temporally separating Cherenkov radiation in a scintillator probe exposed to a pulsed X-ray beam.

Cherenkov radiation is generated in optical systems exposed to ionising radiation. In water or plastic devices, if the incident radiation has components with high enough energy (for example, electrons or positrons with energy greater than 175keV), Cherenkov radiation will be generated. A scintillator dosimeter that collects optical light, guided by optical fibre, will have Cherenkov radiation generated throughout the length of fibre exposed to the radiation field and compromise the signal. We present a novel algorithm to separate Cherenkov radiation signal that requires only a single probe, provided the radiation source is pulsed, such as a linear accelerator in external beam radiation therapy. We use a slow scintillator (BC-444) that, in a constant beam of radiation, reaches peak light output after 1 microsecond, while the Cherenkov signal is detected nearly instantly. This allows our algorithm to separate the scintillator signal from the Cherenkov signal. The relative beam profile and depth dose of a linear accelerator 6MV X-ray field were reconstructed using the algorithm. The optimisation method improved the fit to the ionisation chamber data and improved the reliability of the measurements. The algorithm was able to remove 74% of the Cherenkov light, at the expense of only 1.5% scintillation light. Further characterisation of the Cherenkov radiation signal has the potential to improve the results and allow this method to be used as a simpler optical fibre dosimeter for quality assurance in external beam therapy.