Search

Volume delineation quality assurance (QA) is particularly important in clinical trial settings where consistent protocol implementation is required, as outcomes will affect future as well current patients. Currently, where feasible, this is conducted manually, which is time consuming and resource intensive. Although previous studies mostly focused on automating delineation QA on CT, magnetic resonance imaging (MRI) is being increasingly used in radiotherapy treatment. In this work, we propose to perform automatic delineation QA on prostate MRI for both the clinical target volume (CTV) and organs-at-risk (OARs) by using delineations generated by 3D Unet variants as benchmarks for QA. These networks were trained on a small gold standard atlas set and applied on a multicentre radiotherapy clinical trial dataset to generate benchmark delineations. Then, a QA stage was designed to recommend 'pass', 'minor correction' and 'major correction' for each manual delineation in the trial set by thresholding its Dice similarity coefficient to the network generated delineation. Among all 3D Unet variants explored, the Unet with anatomical gates in an AtlasNet architecture performed the best in delineation QA, achieving an area under the receiver operating characteristics curve of 0.97, 0.92, 0.89 and 0.97 for identifying unacceptable (major correction) delineations with a sensitivity of 0.93, 0.73, 0.74 and 0.90 at a specificity of 0.93, 0.86, 0.86 and 0.95 for bladder, prostate CTV, rectum and gel spacer respectively. To the best of our knowledge, this is the first study to propose automated delineation QA for a multicentre radiotherapy clinical trial with treatment planning MRI. The methods proposed in this work can potentially improve the accuracy and consistency of CTV and OAR delineation in radiotherapy treatment planning.

Purpose. The purpose of this work was to report on the geometric uncertainty for patients treated with multi-leaf collimator (MLC) tracking for lung SABR to verify the accuracy of the system. Methods. Seventeen patients were treated as part of the MLC tracking for lung SABR clinical trial using electromagnetic beacons implanted around the tumor acting as a surrogate for target motion. Sources of uncertainties evaluated in the study included the surrogate-target positional uncertainty, the beam-surrogate tracking uncertainty, the surrogate localization uncertainty, and the target delineation uncertainty. Probability density functions (PDFs) for each source of uncertainty were constructed for the cohort and each patient. The total PDFs was computed using a convolution approach. The 95% confidence interval (CI) was used to quantify these uncertainties. Results. For the cohort, the surrogate-target positional uncertainty 95% CIs were ±2.5 mm (−2.0/3.0 mm) in left-right (LR), ±3.0 mm (−1.6/4.5 mm) in superior–inferior (SI) and ±2.0 mm (−1.8/2.1 mm) in anterior–posterior (AP). The beam-surrogate tracking uncertainty 95% CIs were ±2.1 mm (−2.1/2.1 mm) in LR, ±2.8 mm (−2.8/2.7 mm) in SI and ±2.1 mm (−2.1/2.0 mm) in AP directions. The surrogate localization uncertainty minimally impacted the total PDF with a width of ±0.6 mm. The target delineation uncertainty distribution 95% CIs were ±5.4 mm. For the total PDF, the 95% CIs were ±5.9 mm (−5.8/6.0 mm) in LR, ±6.7 mm (−5.8/7.5 mm) in SI and ±6.0 mm (−5.5/6.5 mm) in AP. Conclusion. This work reports the geometric uncertainty of MLC tracking for lung SABR by accounting for the main sources of uncertainties that occurred during treatment. The overall geometric uncertainty is within ±6.0 mm in LR and AP directions and ±6.7 mm in SI. The dominant uncertainty was the target delineation uncertainty. This geometric analysis helps put into context the range of uncertainties that may be expected during MLC tracking for lung SABR (ClinicalTrials.gov registration number: NCT02514512).

Active shape models (ASMs) have proved successful in automatic segmentation by using shape and appearance priors in a number of areas such as prostate segmentation, where accurate contouring is important in treatment planning for prostate cancer. The ASM approach however, is heavily reliant on a good initialisation for achieving high segmentation quality. This initialisation often requires algorithms with high computational complexity, such as three dimensional (3D) image registration. In this work, we present a fast, self-initialised ASM approach that simultaneously fits multiple objects hierarchically controlled by spatially weighted shape learning. Prominent objects are targeted initially and spatial weights are progressively adjusted so that the next (more difficult, less visible) object is simultaneously initialised using a series of weighted shape models. The scheme was validated and compared to a multi-atlas approach on 3D magnetic resonance (MR) images of 38 cancer patients and had the same (mean, median, inter-rater) Dice's similarity coefficients of (0.79, 0.81, 0.85), while having no registration error and a computational time of 12-15 min, nearly an order of magnitude faster than the multi-atlas approach.

Dosimetry of ionising radiation is a well-established and mature branch of physical sciences with many applications in medicine and biology. In particular radiotherapy relies on dosimetry for optimisation of cancer treatment and avoidance of severe toxicity for patients. Several novel developments in radiotherapy have introduced new challenges for dosimetry with small and dynamically changing radiation fields being central to many of these applications such as stereotactic ablative body radiotherapy and intensity modulated radiation therapy. There is also an increasing awareness of low doses given to structures not in the target region and the associated risk of secondary cancer induction. Here accurate dosimetry is important not only for treatment optimisation but also for the generation of data that can inform radiation protection approaches in the future. The article introduces some of the challenges and highlights the interdependence of dosimetric calculations and measurements. Dosimetric concepts are explored in the context of six application fields: reference dosimetry, small fields, low dose out of field, in vivo dosimetry, brachytherapy and auditing of radiotherapy practice. Recent developments of dosimeters that can be used for these purposes are discussed using spatial resolution and number of dimensions for measurement as sorting criteria. While dosimetry is ever evolving to address the needs of advancing applications of radiation in medicine two fundamental issues remain: the accuracy of the measurement from a scientific perspective and the importance to link the measurement to a clinically relevant question. This review aims to provide an update on both of these.

In-room cine-MRI guidance can provide non-invasive target localization during radiotherapy treatment. However, in order to cope with finite imaging frequency and system latencies between target localization and dose delivery, tumour motion prediction is required. This work proposes a framework for motion prediction dedicated to cine-MRI guidance, aiming at quantifying the geometric uncertainties introduced by this process for both tumour tracking and beam gating. The tumour position, identified through scale invariant features detected in cine-MRI slices, is estimated at high-frequency (25 Hz) using three independent predictors, one for each anatomical coordinate. Linear extrapolation, auto-regressive and support vector machine algorithms are compared against systems that use no prediction or surrogate-based motion estimation. Geometric uncertainties are reported as a function of image acquisition period and system latency. Average results show that the tracking error RMS can be decreased down to a [0.2; 1.2] mm range, for acquisition periods between 250 and 750 ms and system latencies between 50 and 300 ms. Except for the linear extrapolator, tracking and gating prediction errors were, on average, lower than those measured for surrogate-based motion estimation. This finding suggests that cine-MRI guidance, combined with appropriate prediction algorithms, could relevantly decrease geometric uncertainties in motion compensated treatments.

Many similarity metrics exist for inter-observer contouring variation studies, however no correlation between metric choice and prostate cancer radiotherapy dosimetry has been explored. These correlations were investigated in this study. Two separate trials were undertaken, the first a thirty-five patient cohort with three observers, the second a five patient dataset with ten observers. Clinical and planning target volumes (CTV and PTV), rectum, and bladder were independently contoured by all observers in each trial. Structures were contoured on T2-weighted MRI and transferred onto CT following rigid registration for treatment planning in the first trial. Structures were contoured directly on CT in the second trial. STAPLE and majority voting volumes were generated as reference gold standard volumes for each structure for the two trials respectively. VMAT treatment plans (78 Gy to PTV) were simulated for observer and gold standard volumes, and dosimetry assessed using multiple radiobiological metrics. Correlations between contouring similarity metrics and dosimetry were calculated using Spearman's rank correlation coefficient. No correlations were observed between contouring similarity metrics and dosimetry for CTV within either trial. Volume similarity correlated most strongly with radiobiological metrics for PTV in both trials, including TCPPoisson (ρ = 0.57, 0.65), TCPLogit (ρ = 0.39, 0.62), and EUD (ρ = 0.43, 0.61) for each respective trial. Rectum and bladder metric correlations displayed no consistency for the two trials. PTV volume similarity was found to significantly correlate with rectum normal tissue complication probability (ρ = 0.33, 0.48). Minimal to no correlations with dosimetry were observed for overlap or boundary contouring metrics. Future inter-observer contouring variation studies for prostate cancer should incorporate volume similarity to provide additional insights into dosimetry during analysis.

To clinically implement MRI simulation or MRI-alone treatment planning requires comprehensive end-to-end testing to ensure an accurate process. The purpose of this study was to design and build a geometric phantom simulating a human male pelvis that is suitable for both CT and MRI scanning and use it to test geometric and dosimetric aspects of MRI simulation including treatment planning and digitally reconstructed radiograph (DRR) generation.A liquid filled pelvic shaped phantom with simulated pelvic organs was scanned in a 3T MRI simulator with dedicated radiotherapy couch-top, laser bridge and pelvic coil mounts. A second phantom with the same external shape but with an internal distortion grid was used to quantify the distortion of the MR image. Both phantoms were also CT scanned as the gold-standard for both geometry and dosimetry. Deformable image registration was used to quantify the MR distortion. Dose comparison was made using a seven-field IMRT plan developed on the CT scan with the fluences copied to the MR image and recalculated using bulk electron densities. Without correction the maximum distortion of the MR compared with the CT scan was 7.5 mm across the pelvis, while this was reduced to 2.6 and 1.7 mm by the vendor's 2D and 3D correction algorithms, respectively. Within the locations of the internal organs of interest, the distortion was

Collected real-life clinical target volume (CTV) displacement data show that some patients undergoing external beam radiotherapy (EBRT) demonstrate significantly more fraction-to-fraction variability in their displacement ('random error') than others. This contrasts with the common assumption made by historical recipes for margin estimation for EBRT, that the random error is constant across patients. In this work we present statistical models of CTV displacements in which random errors are characterised by an inverse gamma (IG) distribution in order to assess the impact of random error variability on CTV-to-PTV margin widths, for eight real world patient cohorts from four institutions, and for different sites of malignancy. We considered a variety of clinical treatment requirements and penumbral widths. The eight cohorts consisted of a total of 874 patients and 27 391 treatment sessions. Compared to a traditional margin recipe that assumes constant random errors across patients, for a typical 4 mm penumbral width, the IG based margin model mandates that in order to satisfy the common clinical requirement that 90% of patients receive at least 95% of prescribed RT dose to the entire CTV, margins be increased by a median of 10% (range over the eight cohorts -19% to +35%). This substantially reduces the proportion of patients for whom margins are too small to satisfy clinical requirements.

The routine use of electronic portal imaging devices (EPIDs) as dosimeters for radiotherapy quality assurance is complicated by the non-water equivalence of the EPID's dose response. A commercial EPID modified to a direct-detection configuration was previously demonstrated to provide water-equivalent dose response with d(max) solid water build-up and 10 cm solid water backscatter. Clinical implementation of the direct EPID (dEPID) requires a design that maintains the water-equivalent dose response, can be incorporated onto existing EPID support arms and maintains sufficient image quality for clinical imaging. This study investigated the dEPID dose response with different configurations of build-up and backscatter using varying thickness of solid water and copper. Field size output factors and beam profiles measured with the dEPID were compared with ionization chamber measurements of dose in water for both 6 MV and 18 MV. The dEPID configured with d(max) solid water build-up and no backscatter (except for the support arm) was within 1.5% of dose in water data for both energies. The dEPID was maintained in this configuration for clinical dosimetry and image quality studies. Close agreement between the dEPID and treatment planning system was obtained for an IMRT field with 98.4% of pixels within the field meeting a gamma criterion of 3% and 3 mm. The reduced sensitivity of the dEPID resulted in a poorer image quality based on quantitative (contrast-to-noise ratio) and qualitative (anthropomorphic phantom) studies. However, clinically useful images were obtained with the dEPID using typical treatment field doses. The dEPID is a water-equivalent dosimeter that can be implemented with minimal modifications to the standard commercial EPID design. The proposed dEPID design greatly simplifies the verification of IMRT dose delivery.

The fluence exiting a patient during beam delivery can be used as treatment delivery quality assurance, either by direct comparison with expected exit fluences or by backprojection to reconstruct the patient dose. Multiple possible sources of measured exit fluence deviations exist, including changes in the beam delivery and changes in the patient anatomy. The purpose of this work is to compare the deviations caused by these sources. Machine delivery-related variability is measured by acquiring multiple dosimetric portal images (DPIs) of several test fields without a patient/phantom in the field over a time period of 2 months. Patient anatomy-related sources of fluence variability are simulated by computing transmission DPIs for a prostate patient using the same incident fluence for 11 different computed tomography (CT) images of the patient anatomy. The standard deviation (SD) and maximum deviation of the exit fluence, averaged over 5 mm x 5 mm square areas, is calculated for each test set. Machine delivery fluence SDs as large as 1% are observed for a sample patient field and as large as 2.5% for a picket-fence dMLC test field. Simulations indicate that day-to-day patient anatomy variations induce exit fluence SDs as large as 3.5%. The largest observed machine delivery deviations are 4% for the sample patient field and 7% for the picket-fence field, while the largest difference for the patient anatomy-related source is 8.5%. Since daily changes in patient anatomy can result in substantial exit fluence deviations, care should be taken when applying fluence back-projection to ensure that such deviations are properly attributed to their source.

The availability of cone beam computerized tomography (CBCT) images at the time of treatment has opened possibilities for dose calculations representing the delivered dose for adaptive radiation therapy. A significant component in the accuracy of dose calculation is the calibration of the Hounsfield unit (HU) number to electron density (ED). The aim of this work is to assess the impact of HU to ED calibration phantom insert composition and phantom volume on dose calculation accuracy for CBCT. CBCT HU to ED calibration curves for different commercial phantoms were measured and compared. The effect of the scattering volume of the phantom on the HU to ED calibration was examined as a function of phantom length and radial diameter. The resulting calibration curves were used at the treatment planning system to calculate doses for geometrically simple phantoms and a pelvic anatomical phantom to compare against measured doses. Three-dimensional dose distributions for the pelvis phantom were calculated using the HU to ED curves and compared using Chi comparisons. The HU to ED calibration curves for the commercial phantoms diverge at densities greater than that of water, depending on the elemental composition of the phantom insert. The effect of adding scatter material longitudinally, increasing the phantom length from 5 cm to 26 cm, was found to be up to 260 HU numbers for the high-density insert. The change in the HU value, by increasing the diameter of the phantom from 18 to 40 cm, was found to be up to 1200 HU for the high-density insert. The effect of phantom diameter on the HU to ED curve can lead to dose differences for 6 MV and 18 MV x-rays under bone inhomogeneities of up to 20% in extreme cases. These results show significant dosimetric differences when using a calibration phantom with materials which are not tissue equivalent. More importantly, the amount of scattering material used with the HU to ED calibration phantom has a significant effect on the dosimetric accuracy, particularly in the radial direction.

This work assessed the dosimetric performance of an amorphous silicon electronic portal imaging device (EPID) for measurement and quality assurance of enhanced dynamic wedge (EDW) profile and wedge factor. EPID measurements of EDW profiles were corrected for pixel sensitivity variation and spectral over-response relative to ion chamber and compared to ion-chamber and diode-array measurements. The dependence of EPID measurements on wedge direction and source to EPID distance was assessed. The long-term stability was investigated by weekly measurement of EDW profiles and wedge factors over a seven month interval. An empirical correction method was developed to improve EPID profile agreement with diode-array measurements. The EPID profiles differed from conventional measurements by up to 5%. Backscatter from the EPID housing was also found to affect measurements by up to 4%, resulting in changes in EPID measured profile with the direction of the moving jaw and source to EPID distance. EPID profile measurements varied by a maximum of 0.3% (1 SD) within the umbra, and wedge factors varied by 0.3% (1 SD) over the seven month interval. The correction function improved agreement between EPID and diode array to within 2% for all wedge angles and energies. Due to the ease of use and reproducibility of the EPID-measured EDW profiles the device is highly suited to regular measurement of EDW.

In this study we investigate the characteristics of a rounded leaf end multileaf collimator (MLC) that is used for delivering intensity-modulated radiotherapy (IMRT) with a Varian linear accelerator. The rounded leaf end MLC design results in an offset between the radiation field edge (the physical leaf position) and the light field (the geometric leaf position). We call this the radiation field offset (RFO). The leaf position is calibrated to the leaf tip at the mid-leaf plane. There is an additional offset between the geometric leaf position and the projected leaf tip position that varies as a function of distance from the collimator central axis due to the MLC geometry. We call this the leaf position offset (LPO). There is a lack of consistency in the interpretation and implementation of the RFO and the LPO in the literature. We investigated the RFO and the LPO on Varian's 600 C/D and 21 EX linear accelerators. We used a combination of film and ion chamber measurements of static, segmental MLC (SMLC) and dynamic MLC (DMLC) fields to quantify the leaf offsets across the range of leaf positions. We were able to improve the dosimetry at large off-axis positions with minor adjustments to the vendor's LPO file. The RFO was determined to within 0.1 mm accuracy at the collimator central axis. The measured RFO value depends on whether the method is based on the radiation field edge position or on an integral dose measurement. The integral dose method results in an RFO that is approximately 0.2 mm greater than the radiation field edge method. The difference is due to the MLC penumbra shape. We propose a methodology for measuring and implementing MLC leaf offsets that is suitable for both SMLC and DMLC IMRT. In addition, we propose some definitions that more clearly describe the MLC leaf position for accurate IMRT dosimetry.

A new tool has been developed to verify the trajectory of dynamic multileaf collimators (MLCs) used in advanced radiotherapy techniques using only the information provided by the electronic portal imaging devices (EPID) measured image frames. The prescribed leaf positions are resampled to a higher resolution in a pre-processing stage to improve the verification precision. Measured MLC positions are extracted from the EPID frames using a template matching method. A cosine similarity metric is then applied to synchronise measured and planned leaf positions for comparison. Three additional comparison functions were incorporated to ensure robust synchronisation. The MLC leaf trajectory error detection was simulated for both intensity modulated radiation therapy (IMRT) (prostate) and volumetric modulated arc therapy (VMAT) (head-and-neck) deliveries with anthropomorphic phantoms in the beam. The overall accuracy for MLC positions automatically extracted from EPID image frames was approximately 0.5 mm. The MLC leaf trajectory verification system can detect leaf position errors during IMRT and VMAT with a tolerance of 3.5 mm within 1 s.

In this study an amorphous silicon electronic portal imaging device (a-Si EPID) converted to direct detection configuration was investigated as a transit dosimeter for intensity modulated radiation therapy (IMRT). After calibration to dose and correction for a background offset signal, the EPID-measured absolute IMRT transit doses for 29 fields were compared to a MatriXX two-dimensional array of ionization chambers (as reference) using Gamma evaluation (3%, 3 mm). The MatriXX was first evaluated as reference for transit dosimetry. The accuracy of EPID measurements was also investigated by comparison of point dose measurements by an ionization chamber on the central axis with slab and anthropomorphic phantoms in a range of simple to complex fields. The uncertainty in ionization chamber measurements in IMRT fields was also investigated by its displacement from the central axis and comparison with the central axis measurements. Comparison of the absolute doses measured by the EPID and MatriXX with slab phantoms in IMRT fields showed that on average 96.4% and 97.5% of points had a Gamma index

In modern radiotherapy, it is vitally important to monitor the performance of all linac components including the collimation system. In this study, a simple measurement method and accurate algorithm are introduced for investigation of the secondary and tertiary collimator sag during radiotherapy arc treatments. The method is based on cine EPID images of a ball bearing marker fixed to the gantry head and determines the jaw and MLC sag in all directions relative to the reference at zero gantry angle. Analysis was performed using different field sizes and collimator angles, different linacs and different gantry rotation directions. The accuracy of the method was tested and was less than 0.02 mm. The repeatability and reproducibility of the method was 0.005 and 0.09 mm, respectively. The setup is easy and quick and the algorithm is fast and fully automatic with sub-pixel accuracy. This method is suitable to be included in the routine quality assurance of linacs to monitor the collimator system performance.

Early approaches to characterizing errors in target displacement during a fractionated course of radiotherapy assumed that the underlying fraction-to-fraction variability in target displacement, known as the 'treatment error' or 'random error', could be regarded as constant across patients. More recent approaches have modelled target displacement allowing for differences in random error between patients. However, until recently it has not been feasible to compare the goodness of fit of alternate models of random error rigorously. This is because the large volumes of real patient data necessary to distinguish between alternative models have only very recently become available. This work uses real-world displacement data collected from 365 patients undergoing radical radiotherapy for prostate cancer to compare five candidate models for target displacement. The simplest model assumes constant random errors across patients, while other models allow for random errors that vary according to one of several candidate distributions. Bayesian statistics and Markov Chain Monte Carlo simulation of the model parameters are used to compare model goodness of fit. We conclude that modelling the random error as inverse gamma distributed provides a clearly superior fit over all alternatives considered. This finding can facilitate more accurate margin recipes and correction strategies.

Backscatter from the metallic components in the support arm is one of the sources of inaccuracy in dosimetry with Varian amorphous silicon electronic portal imaging devices (a-Si EPIDs). In this study, the non-uniform arm backscatter is blocked by adding lead sheets between the EPID and an E-type support arm. By comparing the EPID responses on and off the arm, with and without lead and considering the extra weight on the imager, 2 mm of lead was determined as the optimum thickness for both 6 and 18 MV beam energies. The arm backscatter at the central axis with the 2 mm lead in place decreased to 0.1% and 0.2% for the largest field size of 30 × 30 cm(2) using 6 and 18 MV beams, from 2.3% and 1.3% without lead. Changes in the source-to-detector distance (SDD) did not affect the backscatter component more than 1%. The symmetry of the in-plane profiles improved for all field sizes for both beam energies. The addition of lead decreased the contrast-to-noise ratio and resolution by 1.3% and 0.84% for images taken in 6 MV and by 0.5% and 0.38% for those in 18 MV beams. The displacement of the EPID central pixel was measured during a 360° gantry rotation with and without lead which was 1 pixel different. While the backscatter reduces with increasing lead thickness, a 2 mm lead sheet seems sufficient for acceptable dosimetry results without any major degradation to the routine performance of the imager. No increase in patient skin dose was detected.