Gross failure rates and failure modes for a commercial AI-based auto-segmentation algorithm in head and neck cancer patients.

PURPOSE: Artificial intelligence (AI) based commercial software can be used to automatically delineate organs at risk (OAR), with potential for efficiency savings in the radiotherapy treatment planning pathway, and reduction of inter- and intra-observer variability. There has been little research investigating gross failure rates and failure modes of such systems. METHOD: 50 head and neck (H&N) patient data sets with "gold standard" contours were compared to AI-generated contours to produce expected mean and standard deviation values for the Dice Similarity Coefficient (DSC), for four common H&N OARs (brainstem, mandible, left and right parotid). An AI-based commercial system was applied to 500 H&N patients. AI-generated contours were compared to manual contours, outlined by an expert human, and a gross failure was set at three standard deviations below the expected mean DSC. Failures were inspected to assess reason for failure of the AI-based system with failures relating to suboptimal manual contouring censored. True failures were classified into 4 sub-types (setup position, anatomy, image artefacts and unknown). RESULTS: There were 24 true failures of the AI-based commercial software, a gross failure rate of 1.2%. Fifteen failures were due to patient anatomy, four were due to dental image artefacts, three were due to patient position and two were unknown. True failure rates by OAR were 0.4% (brainstem), 2.2% (mandible), 1.4% (left parotid) and 0.8% (right parotid). CONCLUSION: True failures of the AI-based system were predominantly associated with a non-standard element within the CT scan. It is likely that these non-standard elements were the reason for the gross failure, and suggests that patient datasets used to train the AI model did not contain sufficient heterogeneity of data. Regardless of the reasons for failure, the true failure rate for the AI-based system in the H&N region for the OARs investigated was low (∼1%).

4DCT and VMAT for lung patients with irregular breathing.

PURPOSE: Irregular breathing in lung cancer patients is a common contra-indication to 4D computerized tomography (4DCT), which may then limit radiotherapy treatment options. For irregular breathers, we investigated whether 3DCT or 4DCT (1) better represents tumor motion, (2) better represents average tumor densities, and (3) better allows for volumetric modulated arc threarpy (VMAT) plans delivered with acceptable dosimetric accuracy. METHODS: Ten clinical breathing traces were identified with irregularities in phase and amplitude, and fed to a programmable moving platform incorporating an anthropomorphic lung tumor phantom. 3DCT and 4DCT data resorted by phase (4DCT-P) and amplitude (4DCT-A) were acquired for each trace. Tumors were delineated by Hounsfield unit (HU) thresholding and apparent motion range assessed. HU profiles were extracted from each image and agreement with calculated expected profiles quantified using area-under-curve (AUC) scoring. Clinically representative VMAT plans were created for each image, delivered to the irregularly moving phantom, and measured with a small-volume ion chamber at the tumor center. RESULTS: Median difference from expected tumor motion range for 3DCT, 4DCT-P, and 4DCT-A was 2.5 [1.6-3.6] cm, 1.1 [0.1-1.9] cm, and 1.3 [0.4-1.9] cm, respectively (p = 0.005, 4DCT-P vs. 3DCT). Median AUC scores (ideal = 0) for 3DCT, 4DCT-P, and 4DCT-A were 0.25 [0.14-0.49], 0.12 [0.05-0.42], and 0.13 [0.09-0.44], respectively (p = 0.005, 4DCT-P vs. 3DCT). Nine of ten 4DCT-P plans and all 4DCT-A plans measured within 2.5% of expected dose in the treatment planning system (TPS), compared with seven 3DCT plans. CONCLUSION: For the cases studied tumor motion range and average density was better represented with 4DCT compared with 3DCT, even in the presence of irregular breathing. 4DCT images allowed for delivery of VMAT plans with acceptable dosimetric accuracy. No significant differences were detected between phase and amplitude resorting. In combination with 4D cone beam imaging at treatment, our findings have given us confidence to introduce 4DCT and VMAT for lung radiotherapy patients with irregular breathing.

Evaluation of failure modes and effect analysis for routine risk assessment of lung radiotherapy at a UK center.

PURPOSE: Explore the feasibility of adopting failure modes and effects analysis (FMEA) for risk assessment of a high volume clinical service at a UK radiotherapy center. Compare hypothetical failure modes to locally reported incidents. METHOD: An FMEA for a lung radiotherapy service was conducted at a hospital that treats ~ 350 lung cancer patients annually with radical radiotherapy. A multidisciplinary team of seven people was identified including a nominated facilitator. A process map was agreed and failure modes identified and scored independently, final failure modes and scores were then agreed at a face-to-face meeting. Risk stratification methods were explored and staff effort recorded. Radiation incidents related to lung radiotherapy reported locally in a 2-year period were analyzed to determine their relation to the identified failure modes. The final FMEA was therefore a combination of prospective evaluation and retrospective analysis from an incident learning system. RESULTS: Thirty-six failure modes were identified for the pre-existing clinical service. The top failure modes varied according to the ranking method chosen. The process required 30 h of combined staff time. Over the 2-year period chosen, 38 voluntarily reported incidents were identified as relating to lung radiotherapy. Of these, 13 were not predicted by the identified failure modes, with six relating to delays in the process, three issues with appointment times, one communication error, two instances of a failure to image, and one technical fault deemed unpredictable by the manufacturer. Four additional failure modes were added to the FMEA following the incident analysis. CONCLUSION: FMEA can be effectively applied to an established high volume service as a risk assessment method. Facilitation by an individual familiar with the FMEA process can reduce resource requirement. Prospective evaluation of risks should be combined with an incident reporting and learning system to produce a more comprehensive analysis of risk.

Chemoradiotherapy of locally-advanced non-small cell lung cancer: Analysis of radiation dose-response, chemotherapy and survival-limiting toxicity effects indicates a low α/ß ratio.

PURPOSE: To analyse changes in 2-year overall survival (OS2yr) with radiotherapy (RT) dose, dose-per-fraction, treatment duration and chemotherapy use, in data compiled from prospective trials of RT and chemo-RT (CRT) for locally-advanced non-small cell lung cancer (LA-NSCLC). MATERIAL AND METHODS: OS2yr data was analysed for 6957 patients treated on 68 trial arms (21 RT-only, 27 sequential CRT, 20 concurrent CRT) delivering doses-per-fraction ≤4.0 Gy. An initial model considering dose, dose-per-fraction and RT duration was fitted using maximum-likelihood techniques. Model extensions describing chemotherapy effects and survival-limiting toxicity at high doses were assessed using likelihood-ratio testing, the Akaike Information Criterion (AIC) and cross-validation. RESULTS: A model including chemotherapy effects and survival-limiting toxicity described the data significantly better than simpler models (p < 10-14), and had better AIC and cross-validation scores. The fitted α/ß ratio for LA-NSCLC was 4.0 Gy (95%CI: 2.8-6.0 Gy), repopulation negated 0.38 (95%CI: 0.31-0.47) Gy EQD2/day beyond day 12 of RT, and concurrent CRT increased the effective tumour EQD2 by 23% (95%CI: 16-31%). For schedules delivered in 2 Gy fractions over 40 days, maximum modelled OS2yr for RT was 52% and 38% for stages IIIA and IIIB NSCLC respectively, rising to 59% and 42% for CRT. These survival rates required 80 and 87 Gy (RT or sequential CRT) and 67 and 73 Gy (concurrent CRT). Modelled OS2yr rates fell at higher doses. CONCLUSIONS: Fitted dose-response curves indicate that gains of ~10% in OS2yr can be made by escalating RT and sequential CRT beyond 64 Gy, with smaller gains for concurrent CRT. Schedule acceleration achieved via hypofractionation potentially offers an additional 5-10% improvement in OS2yr. Further 10-20% OS2yr gains might be made, according to the model fit, if critical normal structures in which survival-limiting toxicities arise can be identified and selectively spared.

Reply to comment on 'origins of the changing detector response in small megavoltage photon radiation fields'.

Andreo and Benmakhlouf (2017 Phys. Med. Biol. 62 1518-32) have disputed a finding of Scott et al (2012 Phys. Med. Biol. 57 4461-76) that the variation with field-size of the response of small ion chambers and solid-state dosimeters in small megavoltage photon radiation fields can largely be attributed to density. Further evidence for this finding was provided by Fenwick et al (2018 Phys. Med. Biol. 63 125003), but Andreo and Benmakhlouf (2018 Phys. Med. Biol. 63 125003) have now challenged the methodology used in that study. Specifically, Andreo and Benmakhlouf suggest that mass stopping-powers of fictitious materials used in Monte Carlo radiation transport calculations should be adjusted with material density according to the polarization effect, as if the materials were real and created by compressing other real materials. In this reply, we observe that fictitious materials are not real, and therefore their densities, mass stopping-powers and microscopic radiation interaction cross-sections can be freely and independently chosen to provide the clearest answers to the questions being studied. And we note that the key role played by density in small field detector response was further confirmed by our group back in 2013, using fictitious materials in which mass stopping-powers were adjusted with density, as preferred by Andreo and Benmakhlouf, as well as being held fixed, with very similar results being obtained in both circumstances (Underwood et al 2013a Med. Phys. 40 082102).

Origins of the changing detector response in small megavoltage photon radiation fields.

Differences in detector response between measured small fields, f clin, and wider reference fields, f msr , can be overcome by using correction factors [Formula: see text] or by designing detectors with field-size invariant responses. The changing response in small fields is caused by perturbations of the electron fluence within the detector sensitive volume. For solid-state detectors, it has recently been suggested that these perturbations might be caused by the non-water-equivalent effective atomic numbers Z of detector materials, rather than by their non-water-like densities. Using the EGSnrc Monte Carlo code we have analyzed the response of a PTW 60017 diode detector in a 6 MV beam, calculating the [Formula: see text] correction factor from computed doses absorbed by water and by the detector sensitive volume in 0.5 × 0.5 and 4 × 4 cm2 fields. In addition to the 'real' detector, fully modelled according to the manufacturer's blue-prints, we calculated doses and [Formula: see text] factors for a 'Z → water' detector variant in which mass stopping-powers and microscopic interaction coefficients were set to those of water while preserving real material densities, and for a 'density → 1' variant in which densities were set to 1 g cm-3, leaving mass stopping-powers and interaction coefficients at real levels. [Formula: see text] equalled 0.910 ± 0.005 (2 standard deviations) for the real detector, was insignificantly different at 0.912 ± 0.005 for the 'Z → H2O' variant, but equalled 1.012 ± 0.006 for the 'density → 1' variant. For the 60017 diode in a 6 MV beam, then, [Formula: see text] was determined primarily by the detector's density rather than its atomic composition. Further calculations showed this remained the case in a 15 MV beam. Interestingly, the sensitive volume electron fluence was perturbed more by detector atomic composition than by density; however, the density-dependent perturbation varied with field-size, whereas the Z-dependent perturbation was relatively constant, little affecting [Formula: see text].

The suitability of common metrics for assessing parotid and larynx autosegmentation accuracy.

Contouring structures in the head and neck is time-consuming, and automatic seg-mentation is an important part of an adaptive radiotherapy workflow. Geometric accuracy of automatic segmentation algorithms has been widely reported, but there is no consensus as to which metrics provide clinically meaningful results. This study investigated whether geometric accuracy (as quantified by several commonly used metrics) was associated with dosimetric differences for the parotid and larynx, comparing automatically generated contours against manually drawn ground truth contours. This enabled the suitability of different commonly used metrics to be assessed for measuring automatic segmentation accuracy of the parotid and larynx. Parotid and larynx structures for 10 head and neck patients were outlined by five clinicians to create ground truth structures. An automatic segmentation algorithm was used to create automatically generated normal structures, which were then used to create volumetric-modulated arc therapy plans. The mean doses to the automatically generated structures were compared with those of the corresponding ground truth structures, and the relative difference in mean dose was calculated for each structure. It was found that this difference did not correlate with the geometric accuracy provided by several metrics, notably the Dice similarity coefficient, which is a commonly used measure of spatial overlap. Surface-based metrics provided stronger correlation and are, therefore, more suitable for assessing automatic seg-mentation of the parotid and larynx.

Relative plan robustness of step-and-shoot vs rotational intensity-modulated radiotherapy on repeat computed tomographic simulation for weight loss in head and neck cancer.

INTRODUCTION: Interfractional anatomical alterations may have a differential effect on the dose delivered by step-and-shoot intensity-modulated radiotherapy (IMRT) and volumetric-modulated arc therapy (VMAT). The increased degrees of freedom afforded by rotational delivery may increase plan robustness (measured by change in target volume coverage and doses to organs at risk [OARs]). However, this has not been evaluated for head and neck cancer. MATERIALS AND METHODS: A total of 10 patients who required repeat computed tomography (CT) simulation and replanning during head and neck IMRT were included. Step-and-shoot IMRT and VMAT plans were generated from the original planning scan. The initial and second CT simulation scans were fused and targets/OAR contours transferred, reviewed, and modified. The plans were applied to the second CT scan and doses recalculated without repeat optimization. Differences between step-and-shoot IMRT and VMAT for change in target volume coverage and doses to OARs between first and second CT scans were compared by Wilcoxon signed rank test. RESULTS: There were clinically relevant dosimetric changes between the first and the second CT scans for both the techniques (reduction in mean D95% for PTV2 and PTV3, Dmin for CTV2 and CTV3, and increased mean doses to the parotid glands). However, there were no significant differences between step-and-shoot IMRT and VMAT for change in any target coverage parameter (including D95% for PTV2 and PTV3 and Dmin for CTV2 and CTV3) or dose to any OARs (including parotid glands) between the first and the second CT scans. CONCLUSIONS: For patients with head and neck cancer who required replanning mainly due to weight loss, there were no significant differences in plan robustness between step-and-shoot IMRT and VMAT. This information is useful with increased clinical adoption of VMAT.

Parametrized rectal dose and associations with late toxicity in prostate cancer radiotherapy.

OBJECTIVE: We investigated possible associations between planned dose-volume parameters and rectal late toxicity in 170 patients having radical prostate cancer radiotherapy. METHODS: For each patient, the rectum was outlined from anorectal junction to sigmoid colon, and rectal dose was parametrized using dose-volume (DVH), dose-surface (DSH) and dose-line (DLH) histograms. Generation of DLHs differed from previous studies in that the rectal dose was parametrized without first unwrapping onto 2-dimensional dose-surface maps. Patient-reported outcomes were collected using a validated Later Effects in Normal Tissues Subjective, Objective, Management and Analytic questionnaire. Associations between dose and toxicity were assessed using a one-sided Mann-Whitney U test. RESULTS: Associations (p < 0.05) were found between equieffective dose (EQD23) and late toxicity as follows: overall toxicity with DVH and DSH at 13-24 Gy; proctitis with DVH and DSH at 25-36 Gy and with DVH, DSH and DLH at 61-67 Gy; bowel urgency with DVH and DSH at 10-20 Gy. None of these associations met statistical significance following the application of a Bonferroni correction. CONCLUSION: Independently confirmed associations between rectal dose and late toxicity remain elusive. Future work to increase the accuracy of the knowledge of the rectal dose, either by accounting for interfraction and intrafraction rectal motion or via stabilization of the rectum during treatment, may be necessary to allow for improved dose-toxicity comparisons. ADVANCES IN KNOWLEDGE: This study is the first to use parametrized DLHs to study associations with patient-reported toxicity for prostate radiotherapy showing that it is feasible to model rectal dose mapping in three dimensions.

AutoLock: a semiautomated system for radiotherapy treatment plan quality control.

A semiautomated system for radiotherapy treatment plan quality control (QC), named AutoLock, is presented. AutoLock is designed to augment treatment plan QC by automatically checking aspects of treatment plans that are well suited to computational evaluation, whilst summarizing more subjective aspects in the form of a checklist. The treatment plan must pass all automated checks and all checklist items must be acknowledged by the planner as correct before the plan is finalized. Thus AutoLock uniquely integrates automated treatment plan QC, an electronic checklist, and plan finalization. In addition to reducing the potential for the propagation of errors, the integration of AutoLock into the plan finalization workflow has improved efficiency at our center. Detailed audit data are presented, demonstrating that the treatment plan QC rejection rate fell by around a third following the clinical introduction of AutoLock.

Dose intensified hypofractionated intensity-modulated radiotherapy with synchronous cetuximab for intermediate stage head and neck squamous cell carcinoma.

BACKGROUND: For stage II and III head and neck squamous cell carcinoma (HNSCC) treated with radiotherapy alone, loco-regional recurrence is the main cause of treatment failure. Strategies to improve loco-regional control should not be at the expense of increased late normal tissue toxicity. We investigated dose-intensified hypofractionated intensity-modulated radiotherapy (IMRT) with synchronous cetuximab. MATERIAL AND METHODS: In a phase I/II trial, 27 patients with stage III or high risk stage II HNSCC were recruited. They received three dose level simultaneous integrated boost IMRT, 62.5 Gy in 25 daily fractions to planning target volume one over five weeks with synchronous cetuximab. The primary endpoint was acute toxicity. Secondary endpoints included: late toxicity and quality of life; loco-regional control, cause-specific and overall survival. RESULTS: Radiotherapy was completed by 26/27 patients; for one (4%) the final fraction was omitted due to skin toxicity. All cycles of cetuximab were received by 23/27 patients. Grade 3 acute toxicities included: pain (81%), oral mucositis (78%) and dysphagia (41%). There were few grade 3 physician-recorded late toxicities, including: pain (11%), problems with teeth (8%) and weight loss (4%). At 12 months, only one (4%) patient required a feeding tube, inserted prior to treatment due to dysphagia. The maximal/peak rates of patient-reported late toxicities included: severe pain (11%), any dry mouth (89%) and swallowing dysfunction that required a soft/liquid diet (23%). At 12 months, all quality of life and most symptoms mean scores had resolved to baseline or were only a little worse; dry mouth, sticky saliva and dentition scores remained very much worse. At a median follow-up of 47 months, there were five (18.5%) loco-regional recurrences and the overall cause-specific survival was 79% (95% CI 53-92). CONCLUSIONS: This regimen is safe with acceptable acute toxicity, low rates of late toxicity and impact on quality of life at 12 months following treatment. Further evaluation is recommended.

Simulation of realistic linac motion improves the accuracy of a Monte Carlo based VMAT plan QA system.

PURPOSE: To investigate the use of a software-based pre-treatment QA system for VMAT, which incorporates realistic linac motion during delivery. METHODS: A beam model was produced using the GATE platform for GEANT4 Monte Carlo dose calculations. Initially validated against static measurements, the model was then integrated with a VMAT delivery emulator, which reads plan files and generates a set of dynamic delivery instructions analogous to the linac control system. Monte Carlo simulations were compared to measurements on dosimetric phantoms for prostate and head and neck VMAT plans. Comparisons were made between calculations using fixed control points, and simulations of continuous motion utilising the emulator. For routine use, the model was incorporated into an automated pre-treatment QA system. RESULTS: The model showed better agreement with measurements when incorporating linac motion: mean gamma pass (Γ<1) over 5 prostate plans was 100.0% at 3%/3mm and 97.4% at 2%/2mm when compared to measurement. For the head and neck plans, delivered to the anatomical phantom, gamma passes were 99.4% at 4%/4mm and 94.94% at 3%/3mm. For example simulations within patient CT data, gamma passes were observed which are within our centre's tolerance for pre-treatment QA. CONCLUSIONS: Through comparison to phantom measurements, it was found that the incorporation of a realistic linac motion improves the accuracy of the model compared to the simulation of fixed control points. The ability to accurately calculate dose as a second check of the planning system, and determine realistic delivery characteristics, may allow for the reduction of machine-based pre-treatment plan QA for VMAT.

Evaluation of the impact of dental artefacts on intensity-modulated radiotherapy planning for the head and neck.

BACKGROUND AND PURPOSE: High density materials create severe artefacts in the computed tomography (CT) scans used for radiotherapy dose calculations. Increased use of intensity-modulated radiotherapy (IMRT) to treat oropharyngeal cancers raises concerns over the accuracy of the resulting dose calculation. This work quantifies their impact and evaluates a simple corrective technique. MATERIALS AND METHODS: Fifteen oropharyngeal patients with severe artefacts were retrospectively planned with IMRT using two different CT/density look-up tables. Each plan was recalculated using a corrected CT dataset to evaluate the dose distribution delivered to the patient. Plan quality in the absence of dental artefacts was similarly assessed. A range of dosimetric and radiobiological parameters were compared pre- and post-correction. RESULTS: Plans using a standard CT/density look-up table (density 1.8 g/cm(3)) revealed inconsistent inter-patient errors, mostly within clinical acceptance, although potentially significantly reducing target coverage for individual patients. Using an extended CT/density look-up table (density 10.0 g/cm(3)) greatly reduced the errors for 13/15 patients. In 2/15 patients with residual errors the CTV extended into the severely affected region and could be corrected by applying a simple manual correction. CONCLUSIONS: Use of an extended CT/density look-up table together with a simple manual bulk density correction reduces the impact of dental artefacts on head and neck IMRT planning to acceptable levels.

Accuracy and precision of an IGRT solution.

Image-guided radiotherapy (IGRT) can potentially improve the accuracy of delivery of radiotherapy treatments by providing high-quality images of patient anatomy in the treatment position that can be incorporated into the treatment setup. The achievable accuracy and precision of delivery of highly complex head-and-neck intensity modulated radiotherapy (IMRT) plans with an IGRT technique using an Elekta Synergy linear accelerator and the Pinnacle Treatment Planning System (TPS) was investigated. Four head-and-neck IMRT plans were delivered to a semi-anthropomorphic head-and-neck phantom and the dose distribution was measured simultaneously by up to 20 microMOSFET (metal oxide semiconductor field-effect transmitter) detectors. A volumetric kilovoltage (kV) x-ray image was then acquired in the treatment position, fused with the phantom scan within the TPS using Syntegra software, and used to recalculate the dose with the precise delivery isocenter at the actual position of each detector within the phantom. Three repeat measurements were made over a period of 2 months to reduce the effect of random errors in measurement or delivery. To ensure that the noise remained below 1.5% (1 SD), minimum doses of 85 cGy were delivered to each detector. The average measured dose was systematically 1.4% lower than predicted and was consistent between repeats. Over the 4 delivered plans, 10/76 measurements showed a systematic error > 3% (3/76 > 5%), for which several potential sources of error were investigated. The error was ultimately attributable to measurements made in beam penumbrae, where submillimeter positional errors result in large discrepancies in dose. The implementation of an image-guided technique improves the accuracy of dose verification, particularly within high-dose gradients. The achievable accuracy of complex IMRT dose delivery incorporating image-guidance is within +/- 3% in dose over the range of sample points. For some points in high-dose gradients, submillimeter errors in position can lead to errors > 3%. The precision of the delivery system was demonstrated to be within the experimental noise of the detector system of 1.5% (1 SD).

Evaluation of larynx-sparing techniques with IMRT when treating the head and neck.

PURPOSE: Concern exists that widespread implementation of whole-field intensity-modulated radiotherapy (IMRT) for the treatment of head-and-neck cancer has resulted in increased levels of dysphagia relative to those seen with conventional planning. Other investigators have suggested an alternative junctioned-IMRT (J-IMRT) method, which matches an IMRT plan to a centrally blocked neck field to restrict the laryngeal dose and reduce dysphagia. The effect on target coverage and sparing of organs at risk, including laryngeal sparing, in the optimization was evaluated and compared with that achieved using a J-IMRT technique. METHODS AND MATERIALS: A total of 13 oropharyngeal cancer whole-field IMRT plans were planned with and without including laryngeal sparing in the optimization. A comparison of the target coverage and sparing of organs at risk was made using the resulting dose-volume histograms and dose distribution. The nine plans with disease located superior to the level of the larynx were replanned using a series of J-IMRT techniques to compare the two laryngeal-sparing techniques. RESULTS: An average mean larynx dose of 29.1 Gy was achieved if disease did not extend to the level of the larynx, with 38.8 Gy for disease extending inferiorly and close to the larynx (reduced from 46.2 and 47.7 Gy, respectively, without laryngeal sparing). Additional laryngeal sparing could be achieved with J-IMRT (mean dose 24.4 Gy), although often at the expense of significantly reduced coverage of the target volume and with no improvement to other areas of the IMRT plan. CONCLUSION: The benefits of J-IMRT can be achieved with whole-field IMRT if laryngeal sparing is incorporated into the class solution. Inclusion of laryngeal sparing had no effect on other parameters in the plan.

Design and implementation of a head-and-neck phantom for system audit and verification of intensity-modulated radiation therapy.

The head and neck is a challenging anatomic site for intensity-modulated radiation therapy (IMRT), requiring thorough testing of planning and treatment delivery systems. Ideally, the phantoms used should be anatomically realistic, have radiologic properties identical to those of the tissues concerned, and allow for the use of a variety of devices to verify dose and dose distribution in any target or normaltissue structure. A phantom that approaches the foregoing characteristics has been designed and built; its specific purpose is verification for IMRT treatments in the head-andneck region. This semi-anatomic phantom, HANK, is constructed of Perspex (Imperial Chemical Industries, London, U.K.) and provides for the insertion of heterogeneities simulating air cavities in a range of fixed positions. Chamber inserts are manufactured to incorporate either a standard thimble ionization chamber (0.125 cm3: PTW, Freiburg, Germany) or a smaller PinPoint chamber (0.015 cm3: PTW), and measurements can be made with either chamber in a range of positions throughout the phantom. Coronal films can also be acquired within the phantom, and additional solid blocks of Perspex allow for transverse films to be acquired within the head region. Initial studies using simple conventional head-and-neck plans established the reproducibility of the phantom and the measurement devices to within the setup uncertainty of +/- 0.5 mm. Subsequent verification of 9 clinical head-and-neck IMRT plans demonstrated the efficacy of the phantom in making a range of patient-specific dose measurements in regions of dosimetric and clinical interest. Agreement between measured values and those predicted by the Pinnacle3 treatment planning system (Philips Medical Systems, Andover, MA) was found to be generally good, with a mean error on the calculated dose to each point of +0.2% (range: -4.3% to +2.2%; n = 9) for the primary planning target volume (PTV), -0.1% (range: -1.5% to +2.0%; n = 8) for the nodal PTV, and +0.0% (range: -1.8% to +4.3%, n = 9) for the spinal cord. The suitability of the phantom for measuring combined dose distributions using radiographic film was also evaluated. The phantom has proved to be a valuable tool in the development and implementation of clinical head-and-neck IMRT, allowing for accurate verification of absolute dose and dose distributions in regions of clinical and dosimetric interest.

Development of an optimum photon beam model for head and-neck intensity-modulated radiotherapy.

Intensity-modulated radiotherapy (IMRT) for complex sites such as tumors of the head and neck requires a level of accuracy in dose calculation beyond that currently used for conformal treatment planning. Recent advances in treatment planning systems have aimed to improve the dose calculation accuracy by improving the modeling of machine characteristics such as interleaf leakage, tongue and groove, and rounded multileaf collimator (MLC) leaf ends. What is uncertain is the extent to which these model parameters improve the agreement between dose calculation and measurements for IMRT treatments. We used Pinnacle version 7.4f (Philips Medical Systems, Andover, MA) to carry out optimization of additional photon-beam model parameters for both an Elekta Precise (Elekta, Stockholm, Sweden) and a Varian (Varian Medical Systems, Palo Alto, CA) linear accelerator (LINAC). One additional parameter was added to the beam models in turn, and associated models were commissioned to investigate the dosimetric impact of each model parameter on 5 clinical head-and-neck IMRT plans. The magnitude and location of differences between the models was determined from gamma analysis of the calculated planar dose maps. A final model that incorporated all of the changes was then commissioned. For the Elekta Precise, the impact of all the changes was determined using a gamma analysis as compared with measured films. Cumulative differences of up to more than 3%/3 mm were observed when dose distributions with and without all of the model changes were compared. Individually, for both LINACs, the addition of modeling for the rounded MLC leaf ends caused the most dramatic change to the calculation of the dose distribution, generating a difference of 3%/3 mm in up to 5% of pixels for the 5 patient plans sampled. The effect of tongue-and-groove modeling was more significant for the Varian LINAC (at 1%/1 mm, mean of 25% of pixels as compared with 5% of pixels with the Elekta Precise LINAC). The combined changes to the Elekta model were found to improve agreement with measurement. Current commercially available treatment planning systems offer accuracy sufficient for clinical implementation of head-and-neck IMRT. For this treatment site, the ability to accurately model the rounded MLC leaf ends has the greatest affect on the similarity of the calculated dose distribution to measurements. In addition, for the Varian LINAC, modeling of the tongue-and-groove effect was also advantageous.

Effect of contrast enhanced CT scans on heterogeneity corrected dose computations in the lung.

The aim of this study was to investigate and, if possible, compensate for the effect of intravenous contrast-enhanced CT scans on the treatment planning dose distributions for lung patients. The contrast and noncontrast CT scans of 3 patients were registered, and the effect of contrast on the Hounsfield units (HU) was assessed. The effect of contrast was then simulated in the CT scans of 18 patients receiving radiotherapy of the lung by modification of the CT numbers for relevant sections of noncontrast-enhanced CT scans. All treatment planning was performed on the Pinnacle3 planning system. The dose distributions computed from simulated contrast CT scans were compared to the original dose distributions by comparison of the monitor units (MUs) for each beam in the treatment plan required to deliver the prescribed dose to the isocenter as well as a comparison of the total MUs for each patient, a percentage change in required MUs being equivalent to a percentage change in the dose. A correction strategy to enable the use of contrast-enhanced CT scans in treatment planning was developed, and the feasibility of applying the strategy was investigated by calculating dose distributions for both the original and simulated contrast CT scans. A mean increase in the overall patient MUs of 1.0 +/- 0.8% was found, with a maximum increase of 3.3% when contrast was simulated on the original CT scans. The simulated contrast scans confirmed that the use of contrast-enhanced CT scans for routine treatment planning would result in a systematic change in the dose delivered to the isocenter. The devised correction strategy had no clinically relevant effect on the dose distribution for the original CT scans. The application of the correction strategy to the simulated contrast CT scans led to a reduction of the mean difference in the overall MUs to 0.1 +/- 0.2% compared to the original scan, demonstrating that the effect of contrast was eliminated with the correction strategy. This work has highlighted the problems associated with using contrast-enhanced CT scans in heterogeneity corrected dose computation. Contrast visible in the CT scan is transient and should not be accounted for in the treatment plan. A correction strategy has been developed that minimizes the effect of intravenous contrast while having no clinical effect on noncontrast CT scans. The correction strategy allows the use of contrast without detriment to the treatment plan.

A geometric model for evaluating the effects of inter-fraction rectal motion during prostate radiotherapy.

A geometric model is presented which allows calculation of the dosimetric consequences of rectal motion in prostate radiotherapy. Variations in the position of the rectum are measured by repeat CT scanning during the courses of treatment of five patients. Dose distributions are calculated by applying the same conformal treatment plan to each imaged fraction and rectal dose-surface histograms produced. The 2D model allows isotropic expansion and contraction in the plane of each CT slice. By summing the dose to specific volume elements tracked by the model, composite dose distributions are produced that explicitly include measured inter-fraction motion for each patient. These are then used to estimate effective dose-surface histograms (DSHs) for the entire treatment. Results are presented showing the magnitudes of the measured target and rectal motion and showing the effects of this motion on the integral dose to the rectum. The possibility of using such information to calculate normal tissue complication probabilities (NTCP) is demonstrated and discussed.

Use of artificial neural networks to predict biological outcomes for patients receiving radical radiotherapy of the prostate.

BACKGROUND AND PURPOSE: This paper discusses the application of artificial neural networks (ANN) in predicting biological outcomes following prostate radiotherapy. A number of model-based methods have been developed to correlate the dose distributions calculated for a patient receiving radiotherapy and the radiobiological effect this will produce. Most widely used are the normal tissue complication probability and tumour control probability models. An alternative method for predicting specific examples of tumour control and normal tissue complications is to use an ANN. One of the advantages of this method is that there is no need for a priori information regarding the relationship between the data being correlated. PATIENTS AND METHODS: A set of retrospective clinical data from patients who received radical prostate radiotherapy was used to train ANNs to predict specific biological outcomes by learning the relationship between the treatment plan prescription, dose distribution and the corresponding biological effect. The dose and volume were included as a differential dose-volume histogram in order to provide a holistic description of the available data. RESULTS: It was shown that the ANNs were able to predict biochemical control and specific bladder and rectum complications with sensitivity and specificity of above 55% when the outcomes were dichotomised. It was also possible to analyse information from the ANNs to investigate the effect of individual treatment parameters on the outcome. CONCLUSION: ANNs have been shown to learn something of the complex relationship between treatment parameters and outcome which, if developed further, may prove to be a useful tool in predicting biological outcomes.