A feasibility study on adaptive 18F-FDG-PET-guided radiotherapy for recurrent and second primary head and neck cancer in the previously irradiated territory.

PURPOSE: To evaluate feasibility, disease control, survival, and toxicity after adaptive 18F-fluorodeoxyglucose (FDG) positron emisson tomography (PET) guided radiotherapy in patients with recurrent and second primary head and neck squamous cell carcinoma. METHODS: A prospective trial investigated the feasibility of adaptive intensity modulated radiotherapy (IMRT) ± concomitant cetuximab in 10 patients. The primary endpoint was achieving a 2-year survival free of grade >3 toxicity in ≥30% of patients. Three treatment plans based on 3 PET/CT scans were consecutively delivered in 6 weeks. The range of dose painting was 66.0-85.0 Gy in the dose-painted tumoral volumes in 30 fractions. RESULTS: Two-year locoregional and distant control rates were 38 and 76%, respectively. Overall and disease-free survival at 2 years was 20%. No grade 4 or 5 acute toxicity was observed in any of the patients, except for arterial mucosal hemorrhage in 1 patient. Three months after radiotherapy, grade 4 dysphagia and mucosal wound healing problems were observed in 1/7 and 1/6 of patients, respectively. Grade 5 toxicity (fatal bleeding) was seen in 2 patients, at 3.8 and 4.1 months of follow-up. Data on 2­year toxicity could only be assessed in 1 of the 2 surviving patients, in whom grade 4 mucosal wound healing problems were observed; no other grade >3 toxicity was observed. In this respect, a 30% 2­year survival free of grade >3 toxicity will not be achieved. CONCLUSIONS: Adaptive PET-guided reirradiation is feasible. However, due to slow accrual and treatment results that seemed inconsistent with achieving the primary endpoint, the trial was stopped early.

Helical tomotherapy in head and neck cancer: a European single-center experience.

BACKGROUND: We report on a retrospective analysis of 147 patients with early and locoregionally advanced squamous cell head and neck cancer (SCCHN) treated with helical tomotherapy (HT). PATIENTS AND METHODS: Included were patients with SCCHN of the oral cavity (OC), oropharynx (OP), hypopharynx (HP), or larynx (L) consecutively treated in one radiotherapy center in 2008 and 2009. The prescribed HT dose was 60-66 Gy in the postoperative setting (group A) and 66-70 Gy when given as primary treatment (group B). HT was given alone, concurrent with systemic therapy (ST), that is, chemotherapy, biotherapy, or both, and with or without induction therapy (IT). Acute and late toxicities are reported using standard criteria; locoregional failure/progression (LRF), distant metastases (DM), and second primary tumors (SPT) were documented, and event-free survival (EFS) and overall survival (OS) were calculated from the start of HT. RESULTS: Group A patients received HT alone in 22 cases and HT + ST in 20 cases; group B patients received HT alone in 17 cases and HT + ST in 88 cases. Severe (grade ≥ 3) acute mucosal toxicity and swallowing problems increased with more additional ST. After a median follow-up of 44 months, grade ≥2 late toxicity after HT + ST was approximately twice that of HT alone for skin, subcutis, pharynx, and larynx. Forty percent had grade ≥2 late xerostomia, and 29% had mucosal toxicity. At 3 years, LRF/DM/SPT occurred in 7%/7%/17% and 25%/13%/5% in groups A and B, respectively, leading to a 3-year EFS/OS of 64%/74% and 56%/63% in groups A and B, respectively. CONCLUSION: The use of HT alone or in combination with ST is feasible and promising and has a low late fatality rate. However, late toxicity is nearly twice as high when ST is added to HT.

Multi-institutional generalizability of a plan complexity machine learning model for predicting pre-treatment quality assurance results in radiotherapy.

Background and purpose: Treatment plans in radiotherapy are subject to measurement-based pre-treatment verifications. In this study, plan complexity metrics (PCMs) were calculated per beam and used as input features to develop a predictive model. The aim of this study was to determine the robustness against differences in machine type and institutional-specific quality assurance (QA). Material and methods: A number of 567 beams were collected, where 477 passed and 90 failed the pre-treatment QA. Treatment plans of different anatomical regions were included. One type of linear accelerator was represented. For all beams, 16 PCMs were calculated. A random forest classifier was trained to distinct between acceptable and non-acceptable beams. The model was validated on other datasets to investigate its robustness. Firstly, plans for another machine type from the same institution were evaluated. Secondly, an inter-institutional validation was conducted on three datasets from different centres with their associated QA. Results: Intra-institutionally, the PCMs beam modulation, mean MLC gap, Q1 gap, and Modulation Complexity Score were the most informative to detect failing beams. Eighty-tree percent of the failed beams (15/18) were detected correctly. The model could not detect over-modulated beams of another machine type. Inter-institutionally, the model performance reached higher accuracy for centres with comparable equipment both for treatment and QA as the local institute. Conclusions: The study demonstrates that the robustness decreases when major differences appear in the QA platform or in planning strategies, but that it is feasible to extrapolate institutional-specific trained models between centres with similar clinical practice. Predictive models should be developed for each machine type.

The potential of helical tomotherapy in the treatment of head and neck cancer.

A decade after its first introduction into the clinic, little is known about the clinical impact of helical tomotherapy (HT) on head and neck cancer (HNC) treatment. Therefore, we analyzed the basics of this technique and reviewed the literature regarding HT's potential benefit in HNC. The past two decades have been characterized by a huge technological evolution in photon beam radiotherapy (RT). In HNC, static beam intensity-modulated radiotherapy (IMRT) has shown superiority over three-dimensional conformal RT in terms of xerostomia and is considered the standard of care. However, the next-generation IMRT, the rotational IMRT, has been introduced into the clinic without any evidence of superiority over static beam IMRT other than being substantially faster. Of these rotational techniques, HT is the first system especially developed for IMRT in combination with image-guided RT. HT is particularly promising for the treatment of HNC because its sharp dose gradients maximally spare the many radiosensitive organs at risk nearby. In addition, HT's integrated computed tomography scan assures a very precise dose administration and allows for some adaptive RT. Because HT is specifically developed for IMRT in combination with (integrated) image-guidance, it allows for precise dose distribution ("dose painting"), patient setup, and dose delivery. As such, it is an excellent tool for difficult HNC irradiation. The literature on the clinical results of HT in HNC all show excellent short-term (≤2 years) results with acceptable toxicity profiles. However, properly designed trials are still warranted to further substantiate these results.

A geometry and dose-volume based performance monitoring of artificial intelligence models in radiotherapy treatment planning for prostate cancer.

Background and Purpose: Clinical Artificial Intelligence (AI) implementations lack ground-truth when applied on real-world data. This study investigated how combined geometrical and dose-volume metrics can be used as performance monitoring tools to detect clinically relevant candidates for model retraining. Materials and Methods: Fifty patients were analyzed for both AI-segmentation and planning. For AI-segmentation, geometrical (Standard Surface Dice 3 mm and Local Surface Dice 3 mm) and dose-volume based parameters were calculated for two organs (bladder and anorectum) to compare AI output against the clinically corrected structure. A Local Surface Dice was introduced to detect geometrical changes in the vicinity of the target volumes, while an Absolute Dose Difference (ADD) evaluation increased focus on dose-volume related changes. AI-planning performance was evaluated using clinical goal analysis in combination with volume and target overlap metrics. Results: The Local Surface Dice reported equal or lower values compared to the Standard Surface Dice (anorectum: (0.93 ± 0.11) vs (0.98 ± 0.04); bladder: (0.97 ± 0.06) vs (0.98 ± 0.04)). The ADD metric showed a difference of (0.9 ± 0.8)Gy for the anorectum D1cm3. The bladder D5cm3 reported a difference of (0.7 ± 1.5)Gy. Mandatory clinical goals were fulfilled in 90 % of the DLP plans. Conclusions: Combining dose-volume and geometrical metrics allowed detection of clinically relevant changes, applied to both auto-segmentation and auto-planning output and the Local Surface Dice was more sensitive to local changes compared to the Standard Surface Dice. This monitoring is able to evaluate AI behavior in clinical practice and allows candidate selection for active learning.

Machine learning-based treatment couch parameter prediction in support of surface guided radiation therapy.

Purpose: A fully independent, machine learning-based automatic treatment couch parameters prediction was developed to support surface guided radiation therapy (SGRT)-based patient positioning protocols. Additionally, this approach also acts as a quality assurance tool for patient positioning. Materials/Methods: Setup data of 183 patients, divided into four different groups based on used setup devices, was used to calculate the difference between the predicted and the acquired treatment couch value. Results: Couch parameters can be predicted with high precision µ = 0.90 , σ = 0.92 . A significant difference (p < 0.01) between the variances of Lung and Brain patients was found. Outliers were not related to the prediction accuracy, but are due to inconsistencies during initial patient setup. Conclusion: Couch parameters can be predicted with high accuracy and can be used as starting point for SGRT-based patient positioning. In case of large deviations (>1.5 cm), patient setup has to be verified to optimally use the surface scanning system.

Machine learning-based detection of aberrant deep learning segmentations of target and organs at risk for prostate radiotherapy using a secondary segmentation algorithm.

Objective.The output of a deep learning (DL) auto-segmentation application should be reviewed, corrected if needed and approved before being used clinically. This verification procedure is labour-intensive, time-consuming and user-dependent, which potentially leads to significant errors with impact on the overall treatment quality. Additionally, when the time needed to correct auto-segmentations approaches the time to delineate target and organs at risk from scratch, the usability of the DL model can be questioned. Therefore, an automated quality assurance framework was developed with the aim to detect in advance aberrant auto-segmentations.Approach. Five organs (prostate, bladder, anorectum, femoral head left and right) were auto-delineated on CT acquisitions for 48 prostate patients by an in-house trained primary DL model. An experienced radiation oncologist assessed the correctness of the model output and categorised the auto-segmentations into two classes whether minor or major adaptations were needed. Subsequently, an independent, secondary DL model was implemented to delineate the same structures as the primary model. Quantitative comparison metrics were calculated using both models' segmentations and used as input features for a machine learning classification model to predict the output quality of the primary model.Main results. For every organ, the approach of independent validation by the secondary model was able to detect primary auto-segmentations that needed major adaptation with high sensitivity (recall = 1) based on the calculated quantitative metrics. The surface DSC and APL were found to be the most indicated parameters in comparison to standard quantitative metrics for the time needed to adapt auto-segmentations.Significance. This proposed method includes a proof of concept for the use of an independent DL segmentation model in combination with a ML classifier to improve time saving during QA of auto-segmentations. The integration of such system into current automatic segmentation pipelines can increase the efficiency of the radiotherapy contouring workflow.

Probabilistic dose prediction using mixture density networks for automated radiation therapy treatment planning.

We demonstrate the application of mixture density networks (MDNs) in the context of automated radiation therapy treatment planning. It is shown that an MDN can produce good predictions of dose distributions as well as reflect uncertain decision making associated with inherently conflicting clinical tradeoffs, in contrast to deterministic methods previously investigated in the literature. A two-component Gaussian MDN is trained on a set of treatment plans for postoperative prostate patients with varying extents to which rectum dose sparing was prioritized over target coverage. Examination on a test set of patients shows that the predicted modes follow their respective ground truths well, both spatially and in terms of their dose-volume histograms. A special dose mimicking method based on the MDN output is used to produce deliverable plans and thereby showcase the usability of voxel-wise predictive densities. Thus, this type of MDN may serve to support clinicians in managing clinical tradeoffs and has the potential to improve the quality of plans produced by an automated treatment planning pipeline.

A novel procedure for determining the optimal MLC configuration parameters in treatment planning systems based on measurements with a Farmer chamber.

Modelling of the multi-leaf collimator (MLC) in treatment planning systems (TPS) is crucial for the dose calculation accuracy of intensity-modulated radiation therapy plans. However, no standardised methodology for their configuration exists to date. In this study we present a method that separates the effect of each dosimetric characteristic of the MLC, offering comprehensive equations for the determination of the configuration parameters used in the TPS model. The main advantage of the method is that it only requires prior knowledge of the nominal leaf width and is based on doses measured with a Farmer chamber, which is a very well established and robust methodology. Another significant advantage is the required time, since measuring the tests takes only about 30 minutes per energy. Firstly, we provide a theoretical general formalism in terms of the primary fluence constructed from the transmission map obtained from an MLC model for synchronous and asynchronous sweeping beams. Secondly, we apply the formalism to the RayStation TPS as a proof of concept and we derive analytical expressions that allow the determination of the configuration parameters (leaf tip width, tongue-and-groove width, x-position offset and MLC transmission) and describe how they intertwine. Finally, we apply the method to Varian's Millennium120 and HD120 MLCs in a TrueBeam linear accelerator for different energies and determine the optimal configuration parameters. The proposed procedure is much faster and streamlined than the typical trial-and-error methods and increases the accuracy of dose calculation in clinical plans. Additionally, the procedure can be useful for standardising the MLC configuration process and it exposes the limitations of the implemented MLC model, providing guidance for further improvement of these models in TPSs.

Outcome and toxicity of hypofractionated image-guided SABR for spinal oligometastases.

BACKGROUND: To investigate progression free survival (PFS), local control (LC) and overall survival (OS) outcomes for patients treated with spine hypofractionated stereotactic ablative radiotherapy (SABR) and to evaluate possible predictors of rapid progression in view of a correct patient selection for this potentially curative SABR. MATERIALS AND METHODS: A cohort of 59 patients with spinal metastases were treated with SABR. Patient selection criteria were the following: histologically proven diagnosis of a solid tumor, a World Health Organization (WHO) score ≤ 2, life expectancy > 6 months, Spinal Instability Neoplastic Score (SINS) ≤ 12 points and presenting with radically treated oligometastatic disease (≤5 lesions) or stable polymetastatic disease with an oligoprogressive lesion. RESULTS: From March 2015 to June 2019, 59 patients were treated with Linac-based SABR to 64 spinal metastases with a median follow-up of 55 months. SABR was standard delivered every other day in 3 to 10 fractions with median prescription dose of 27 Gy (range 21-49 Gy).The 1-,2- and 5-year PFS was 98%, 85% and 75% for all patients. OS at 5 years for all patients was 92%. Metachronous lesions (p < 0.01; HR = 7.1) and oligometastatic (vs. oligoprogressive) lesions (p = 0.02; HR = 0.3) were associated with higher PFS in uni- and multivariate Cox regression analysis. No significant predictors in multivariate analysis were demonstrated for rapid progressors.Vertebral compression fractures developed de novo in 6.3% (4/64) of cases. The median time to fracture was 11 months (range 7-15) after treatment. No other adverse events ≥ 3 grade were observed. CONCLUSIONS: Tumor control and toxicity after high-dose hypofractionated SABR was evaluated in patients with spinal oligometastases. High rates of efficacy and minimal toxicity were demonstrated. Oligometastatic patients with metachronous spinal metastases seem to benefit the most from SABR.

Variation in current prescription practice of stereotactic body radiotherapy for peripherally located early stage non-small cell lung cancer: Recommendations for prescribing and recording according to the ACROP guideline and ICRU report 91.

BACKGROUND AND PURPOSE: In 2017 the ACROP guideline on SBRT for peripherally located early stage NSCLC was published. Later that year ICRU-91 about prescribing, recording and reporting was published. The purpose of this study is to quantify the current variation in prescription practice in the institutions that contributed to the ACROP guideline and to establish the link between the ACROP and ICRU-91 recommendations. MATERIAL AND METHODS: From each of the eight participating centres, 15 SBRT plans for stage I NSCLC were analyzed. Plans were generated following the institutional protocol, centres prescribed 3 × 13.5 Gy, 3 × 15 Gy, 3 × 17 Gy or 3 × 18 Gy. Dose parameters of the target volumes were reported as recommended by ICRU-91 and also converted to BED10Gy. RESULTS: The intra-institutional variance in D98%, Dmean and D2% of the PTV and GTV/ITV is substantially smaller than the inter-institutional spread, indicating well protocollised planning procedures are followed. The median values per centre ranged from 56.1 Gy to 73.1 Gy (D2%), 50.4 Gy to 63.3 Gy (Dmean) and 40.5 Gy to 53.6 Gy (D98%) for the PTV and from 57.1 Gy to 73.6 Gy (D2%), 53.7 Gy to 68.7 Gy (Dmean) and 48.5 Gy to 62.3 Gy (D98%) for the GTV/ITV. Comparing the variance in PTV D98% with the variance in GTV Dmean per centre, using an F-test, shows that four centres have a larger variance in GTV Dmean, while one centre has a larger variance in PTV D98% (p values <0.01). This shows some centres focus on achieving a constant PTV coverage while others aim at a constant GTV coverage. CONCLUSION: More detailed recommendations for dose planning and reporting of lung SBRT in line with ICRU-91 were formulated, including a minimum PTV D98% of 100 Gy BED10Gy and minimum GTV/ITV mean dose of 150 Gy BED10Gy and a D2% in the range of 60-70 Gy.

Evaluation of the optimal combinations of modulation factor and pitch for Helical TomoTherapy plans made with TomoEdge using Pareto optimal fronts.

BACKGROUND: Modulation factor (MF) and pitch have an impact on Helical TomoTherapy (HT) plan quality and HT users mostly use vendor-recommended settings. This study analyses the effect of these two parameters on both plan quality and treatment time for plans made with TomoEdge planning software by using the concept of Pareto optimal fronts. METHODS: More than 450 plans with different combinations of pitch [0.10-0.50] and MF [1.2-3.0] were produced. These HT plans, with a field width (FW) of 5 cm, were created for five head and neck patients and homogeneity index, conformity index, dose-near-maximum (D2), and dose-near-minimum (D98) were analysed for the planning target volumes, as well as the mean dose and D2 for most critical organs at risk. For every dose metric the median value will be plotted against treatment time. A Pareto-like method is used in the analysis which will show how pitch and MF influence both treatment time and plan quality. RESULTS: For small pitches (≤0.20), MF does not influence treatment time. The contrary is true for larger pitches (≥0.25) as lowering MF will both decrease treatment time and plan quality until maximum gantry speed is reached. At this moment, treatment time is saturated and only plan quality will further decrease. CONCLUSION: The Pareto front analysis showed optimal combinations of pitch [0.23-0.45] and MF > 2.0 for a FW of 5 cm. Outside this range, plans will become less optimal. As the vendor-recommended settings fall within this range, the use of these settings is validated.

Fast Helical Tomotherapy in a head and neck cancer planning study: is time priceless?

BACKGROUND: The last few years, in radiotherapy there has been a growing focus on speed of treatment delivery (largely driven by economical and commercial interests). This study investigates the influence of treatment time on plan quality for helical tomotherapy (HT), using delivery times with Volumetric Modulated Arc Therapy (VMAT; Rapid Arc [RA]) as reference. METHODS: In a previous study, double arc RA (Eclipse) and standard HT plans (TomoHD™) were created for five oropharyngeal cancer patients and reported according to ICRU 83 guidelines. By modifying the beam width from 2.5 to 5.0 cm, elevating the pitch and lowering the modulation factor, "TomoFast" (TF) plans were generated with treatment times equal to RA plans. To quantify the impact of TF's craniocaudal gradient, similar plans were generated on TomoEdge(TM) (TomoEdgeFast;TEF). The homogeneity index (HI), conformity index (CI), mean dose, Dnear-max (D2) and Dnear-min (D98) of the PTVs were analyzed as well as the mean dose, specific critical doses and volumes of 26 organs at risk (OARs). Data were analyzed using repeated measures ANOVA. RESULTS: With a mean treatment time of 3.05 min (RA), 2.89 min (TF) and 2.95 min (TEF), PTVtherapeutic coverage was more homogeneous with TF (HI.07;SE.01) and TEF (HI.08;SE.01) compared to RA (HI.10;SE.01), while PTVprophylactic was most homogeneous with RA. Mean doses to parotid glands were comparable for RA, TF, TEF: 25.62, 25.34, 23.09 Gy for contralateral and 32.02, 31.96, 30.01 Gy for ipsilateral glands, respectively. OARs' mean doses varied between different approaches not favoring a particular technique. TF's higher dose to OARs at the cranial-caudal edges of the PTVs and its higher integral dose, both due to the extended cranial-caudal gradient, seems to be solved by the new TomoEdge™ software. However, all these faster techniques lose part of standard TomoHD's OAR sparing capacity CONCLUSION: It is possible to treat oropharyngeal cancer patients using HT (TF/TEF) within time-frames observed for RA maintaining comparable target coverage and sparing of OARs. This study indicates that treatment time is not technology specific, rather an operator's decision on balancing efficiency and quality.