Feasibility of Monte Carlo dropout-based uncertainty maps to evaluate deep learning-based synthetic CTs for adaptive proton therapy.

BACKGROUND: Deep learning has shown promising results to generate MRI-based synthetic CTs and to enable accurate proton dose calculations on MRIs. For clinical implementation of synthetic CTs, quality assurance tools that verify their quality and reliability are required but still lacking. PURPOSE: This study aims to evaluate the predictive value of uncertainty maps generated with Monte Carlo dropout (MCD) for verifying proton dose calculations on deep-learning-based synthetic CTs (sCTs) derived from MRIs in online adaptive proton therapy. METHODS: Two deep-learning models (DCNN and cycleGAN) were trained for CT image synthesis using 101 paired CT-MR images. sCT images were generated using MCD for each model by performing 10 inferences with activated dropout layers. The final sCT was obtained by averaging the inferred sCTs, while the uncertainty map was obtained from the HU variance corresponding to each voxel of 10 sCTs. The resulting uncertainty maps were compared to the observed HU-, range-, WET-, and dose-error maps between the sCT and planning CT. For range and WET errors, the generated uncertainty maps were projected along the 90-degree angle. To evaluate the dose distribution, a mask based on the 5%-isodose curve was applied to only include voxels along the beam paths. Pearson's correlation coefficients were calculated to determine the correlation between the uncertainty maps and HUs, range, WET, and dose errors. To evaluate the dosimetric accuracy of synthetic CTs, clinical proton treatment plans were recalculated and compared to the pCTs RESULTS: Evaluation of the correlation showed an average of r = 0.92 ± 0.03 and r = 0.92 ± 0.03 for errors between uncertainty-HU, r = 0.66 ± 0.09 and r = 0.62 ± 0.06 between uncertainty-range, r = 0.64 ± 0.06 and r = 0.58 ± 0.07 between uncertainty-WET, and r = 0.65 ± 0.09 and r = 0.67 ± 0.07 between uncertainty and dose difference for DCNN and cycleGAN model, respectively. Dosimetric comparison for target volumes showed an average 3%/3 mm gamma pass rate of 99.76 ± 0.43 (DCNN) and 99.10 ± 1.27 (cycleGAN). CONCLUSION: The observed correlations between uncertainty maps and the various metrics (HU, range, WET, and dose errors) demonstrated the potential of MCD-based uncertainty maps as a reliable QA tool to evaluate the accuracy of deep learning-based sCTs.

Assessment of residual geometrical errors of clinical target volumes and their impact on dose accumulation for head and neck radiotherapy.

PURPOSE: To assess the residual geometrical errors (dr) and their impact on the clinical target volumes (CTV) dose coverage for head and neck cancer (HNC) proton therapy patients. METHODS: We analysed 28 HNC patients treated with 70 Gy (RBE) and 54.25 Gy (RBE) to the therapeutic CTV70 and prophylactic CTV54.25, respectively. Daily cone beam CTs were converted to high quality synthetic CTs (sCTs). The CTVs from the nominal CT were propagated to the corresponding sCTs using a hybrid deformable image registration (propagated CTVs) in RayStation 11B. For 11 patients, all propagated CTVs were reviewed by our HNC radiation oncologist (physician corrected CTVs). The residual geometrical error dr was quantified as a function of the daily CTVs volume overlap with the nominal plan CTV. The errors dr(propagated CTVs) and dr(physician corrected CTVs) and the difference in dice similarity coefficients (ΔDSC) were determined. Using clinical plans, dose coverage and the tumor control probability (TCP) for the nominal, accumulated and voxel-wise minimum scenarios were determined. RESULTS: The difference in the residual geometrical error dr (propagated CTVs - physician corrected CTVs) and mean DSC (|ΔDSC|mean) were minor: Δdr(CTV70) = 0.16 mm, Δdr(CTV54.25) = 0.26 mm, |ΔDSC|mean < 0.9%. For all 28 patients, dr(CTV70) = 1.91 mm and dr(CTV54.25) = 1.90 mm. However, CTV54.25 above and below the cricoid cartilage differed substantially (1.00 mm c.f. 3.93 mm). The CTV54.25 coverage below the cricoid was then almost always lower, although the TCP of the accumulated dose was higher than the TCP of the voxel-wise minimum dose. CONCLUSIONS: Setup uncertainty setting of 2 mm is possible. The feasibility of using propagated CTVs for error determination is demonstrated.

Spot scanning proton arc therapy reduces toxicity in oropharyngeal cancer patients.

BACKGROUND: Proton arc technology has recently shown dosimetric gains for various treatment indications. The increased number of beams and energy layers (ELs) in proton arc plans, increases the degrees of freedom in plan optimization and thereby flexibility to spare dose in organs at risk (OARs). A relationship exists between dosimetric plan quality, delivery efficiency, the number of ELs -and beams in a proton arc plan. PURPOSE: This work aims to investigate the effect of the number of beams and ELs in a proton arc plan, on toxicity and delivery time for oropharyngeal cancer patients (OPC) selected for intensity modulated proton therapy (IMPT) based on the Dutch model-based approach. METHODS: The EL reduction algorithm iteratively selects ELs from beams equidistantly spaced over a 360° arc. The beams in the final plan may contain multiple ELs, making them suited for static delivery on the studied treatment machine. The produced plans can therefore be called "step and shoot" proton arc plans. The number of beams and ELs were varied to determine the relationship with the planning cost function value, normal tissue complication probability (NTCP) and delivery time. Proton arc plans with robust target coverage and optimal energy layer reduction (ELR) settings to reduce NTCP, were generated for 10 OPC patients. Proton arc plans were compared to clinical volumetric modulated arc therapy (VMAT) and IMPT plans in terms of integral dose, OAR dose, NTCP for xerostomia and dysphagia and delivery time. Furthermore, dose-weighted average linear energy transfer (LETd ) distributions were compared between the IMPT and proton arc plans. A dry run delivery of a plan containing 20 beams and 360 ELs was performed to evaluate delivery time and accuracy. RESULTS: We found 360 ELs distributed over 30 beams generated proton arc plans with near minimal expected plan toxicity. Relative to corresponding IMPT and VMAT plans, an average reduction of 21 ± 3% and 58 ± 10% in integral dose was observed. D m e a n $_{mean}$ was reduced most in the pharyngeal constrictor muscle (PCM) medius structure, with on average 9.0 ± 4.2 Gy(RBE) (p = 0.0002) compared to the clinical IMPT plans. The average NTCP for grade≥2 and grade≥3 xerostomia at 6 months after treatment significantly decreased with 4.7 ± 1.8% (p = 0.002) and 1.7 ± 0.8% (p = 0.002), respectively, while the average NTCP for grade≥2 and grade≥3 dysphagia decreased with 4.4 ± 2.9% (p = 0.002) and 0.9 ± 0.4% (p = 0.002), respectively, increasing the benefit of protons relative to VMAT. For a "step and shoot" proton arc delivery with auto beam sequencing the estimated delivery time is 11 min, similar to the delivery time of a 6-field IMPT treatment. Gamma analysis between the planned and delivered dose distribution resulted in a 99.99% pass rate using 1mm/1% dose difference/distance to agreement criteria. CONCLUSIONS: "Step and shoot" proton arc demonstrates potential to further reduce toxicity compared to IMPT and VMAT in OPC treatment. By employing 360 ELs and 30 beams in the proposed ELR method, delivery time can reach clinically acceptable levels without compromising plan toxicity when automatic beam sequencing is available.

Does the uncertainty in relative biological effectiveness affect patient treatment in proton therapy?

Clinical treatment with protons uses the concept of relative biological effectiveness (RBE) to convert the absorbed dose into an RBE-weighted dose that equals the dose for radiotherapy with photons causing the same biological effect. Currently, in proton therapy a constant RBE of 1.1 is generically used. However, empirical data indicate that the RBE is not constant, but increases at the distal edge of the proton beam. This increase in RBE is of concern, as the clinical impact is still unresolved, and clinical studies demonstrating a clinical effect of an increased RBE are emerging. Within the European Particle Therapy Network (EPTN) work package 6 on radiobiology and RBE, a workshop was held in February 2020 in Manchester with one day of discussion dedicated to the impact of proton RBE in a clinical context. Current data on RBE effects, patient outcome and modelling from experimental as well as clinical studies were presented and discussed. Furthermore, representatives from European clinical proton therapy centres, who were involved in patient treatment, laid out their current clinical practice on how to consider the risk of a variable RBE in their centres. In line with the workshop, this work considers the actual impact of RBE issues on patient care in proton therapy by reviewing preclinical data on the relation between linear energy transfer (LET) and RBE, current clinical data sets on RBE effects in patients, and applied clinical strategies to manage RBE uncertainties. A better understanding of the variability in RBE would allow development of proton treatments which are safer and more effective.

Can the mean linear energy transfer of organs be directly related to patient toxicities for current head and neck cancer intensity-modulated proton therapy practice?

BACKGROUND AND PURPOSE: The relative biological effectiveness (RBE) of proton therapy is predicted to vary with the dose-weighted average linear energy transfer (LETd). However, RBE values may substantially vary for different clinical endpoints. Therefore, the aim of this study was to assess the feasibility of relating mean D⋅LETd parameters to patient toxicity for HNC patients treated with proton therapy. MATERIALS AND METHODS: The delivered physical dose (D) and the voxel-wise product of D and LETd (D⋅LETd) distributions were calculated for 100 head and neck cancer (HNC) proton therapy patients using our TPS (Raystation v6R). The means and covariance matrix of the accumulated D and D⋅LETd of all relevant organs-at-risk (OARs) were used to simulate 2.500 data sets of different sizes. For each dataset, an attempt was made to add mean D⋅LETd parameters to a multivariable NTCP model based on mean D parameters of the same OAR for xerostomia, tube feeding and dysphagia. The likelihood of creating an NTCP model with statistically significant parameters (i.e. power) was calculated as a function of the simulated sample size for various RBE models. RESULTS: The sample size required to have a power of at least 80% to show an independent effect of mean D⋅LETd parameters on toxicity is over 15,000 patients for all toxicities. CONCLUSION: For current clinical practice, it is not feasible to directly model NTCP with both mean D and mean D⋅LETd of OARs. These findings should not be interpreted as a contradiction of previous evidence for the relationship between RBE and LETd.

Head and neck IMPT probabilistic dose accumulation: Feasibility of a 2 mm setup uncertainty setting.

OBJECTIVE: To establish optimal robust optimization uncertainty settings for clinical head and neck cancer (HNC) patients undergoing 3D image-guided pencil beam scanning (PBS) proton therapy. METHODS: We analyzed ten consecutive HNC patients treated with 70 and 54.25 GyRBE to the primary and prophylactic clinical target volumes (CTV) respectively using intensity-modulated proton therapy (IMPT). Clinical plans were generated using robust optimization with 5 mm/3% setup/range uncertainties (RayStation v6.1). Additional plans were created for 4, 3, 2 and 1 mm setup and 3% range uncertainty and for 3 mm setup and 3%, 2% and 1% range uncertainty. Systematic and random error distributions were determined for setup and range uncertainties based on our quality assurance program. From these, 25 treatment scenarios were sampled for each plan, each consisting of a systematic setup and range error and daily random setup errors. Fraction doses were calculated on the weekly verification CT closest to the date of treatment as this was considered representative of the daily patient anatomy. RESULTS: Plans with a 2 mm/3% setup/range uncertainty setting adequately covered the primary and prophylactic CTV (V95 ≥ 99% in 98.8% and 90.8% of the treatment scenarios respectively). The average organ-at-risk dose decreased with 1.1 GyRBE/mm setup uncertainty reduction and 0.5 GyRBE/1% range uncertainty reduction. Normal tissue complication probabilities decreased by 2.0%/mm setup uncertainty reduction and by 0.9%/1% range uncertainty reduction. CONCLUSION: The results of this study indicate that margin reduction below 3 mm/3% is possible but requires a larger cohort to substantiate clinical introduction.

Validation of linear energy transfer computed in a Monte Carlo dose engine of a commercial treatment planning system.

The relative biological effectiveness (RBE) of protons is highly variable and difficult to quantify. However, RBE is related to the local ionization density, which can be related to the physical measurable dose weighted linear energy transfer (LETD). The aim of this study was to validate the LETD calculations for proton therapy beams implemented in a commercially available treatment planning system (TPS) using microdosimetry measurements and independent LETD calculations (Open-MCsquare (MCS)). The TPS (RayStation v6R) was used to generate treatment plans on the CIRS-731-HN anthropomorphic phantom for three anatomical sites (brain, nasopharynx, neck) for a spherical target (Ø = 5 cm) with uniform target dose to calculate the LETD distribution. Measurements were performed at the University Medical Center Groningen proton therapy center (Proteus Plus, IBA) using a µ +-probe utilizing silicon on insulator microdosimeters capable of detecting lineal energies as low as 0.15 keV µm-1 in tissue. Dose averaged mean lineal energy [Formula: see text] depth-profiles were measured for 70 and 130 MeV spots in water and for the three treatment plans in water and an anthropomorphic phantom. The [Formula: see text] measurements were compared to the LETD calculated in the TPS and MCS independent dose calculation engine. D · [Formula: see text] was compared to D · LETD in terms of a gamma-index with a distance-to-agreement criteria of 2 mm and increasing dose difference criteria to determine the criteria for which a 90% pass rate was accomplished. Measurements of D · [Formula: see text] were in good agreement with the D · LETD calculated in the TPS and MCS. The 90% passing rate threshold was reached at different D · LETD difference criteria for single spots (TPS: 1% MCS: 1%), treatment plans in water (TPS: 3% MCS: 6%) and treatment plans in an anthropomorphic phantom (TPS: 6% MCS: 1%). We conclude that D · LETD calculations accuracy in the RayStation TPS and open MCSquare are within 6%, and sufficient for clinical D · LETD evaluation and optimization. These findings remove an important obstacle in the road towards clinical implementation of D · LETD evaluation and optimization of proton therapy treatment plans. Novelty and significance The dose weighed linear energy transfer (LETD) distribution can be calculated for proton therapy treatment plans by Monte Carlo dose engines. The relative biological effectiveness (RBE) of protons is known to vary with the LETD distribution. Therefore, there exists a need for accurate calculation of clinical LETD distributions. Previous LETD validations have focused on general purpose Monte Carlo dose engines which are typically not used clinically. We present the first validation of mean lineal energy [Formula: see text] measurements of the LETD against calculations by the Monte Carlo dose engines of the Raystation treatment planning system and open MCSquare.

Composite minimax robust optimization of VMAT improves target coverage and reduces non-target dose in head and neck cancer patients.

BACKGROUND AND PURPOSE: To assess the potential of composite minimax robust optimization (CMRO) compared to planning target volume (PTV)-based optimization for head and neck cancer (HNC) patients treated with volumetric modulated arc therapy (VMAT). MATERIALS AND METHODS: Ten HNC patients previously treated with a PTV-based VMAT plan were studied. In addition to the PTV-plan a VMAT plan was created with CMRO. For both plans an adapted planning strategy was also investigated, including a plan adaptation during the third week of treatment. The PTV-plans and CMRO-plans (adapted and non-adapted) were evaluated by means of the estimated actually given dose (EAGD). Therefore, the dose was calculated on daily acquired CBCTs, mapped onto the planning CT and accumulated. The plans were compared by dosimetric parameters and normal tissue complication probabilities (NTCPs) for tube feeding dependence, grade 2-4 dysphagia and xerostomia. The accuracy of CBCT-based dose accumulation was further quantified by comparisons of dose accumulation on weekly verification CTs. RESULTS: On average, CMRO significantly increased (1.5 Gy) the D98% of the EAGD to the clinical target volume and significantly decreased the mean dose of the ipsilateral parotid (2.8 Gy), inferior pharynx constrictor muscle (0.7 Gy) and the oral cavity (0.8 Gy). This translated into significantly reduced NTCP of tube feeding dependence (0.9%) and xerostomia (2.8%). The differences in EAGD derived from evaluation CTs or CBCTs were minimal. CONCLUSION: Minimax robust optimization led to improved target coverage and dose reduction in organs at risk in HNC patients treated with VMAT.

Quantitative comparison of commercial and non-commercial metal artifact reduction techniques in computed tomography.

OBJECTIVES: Typical streak artifacts known as metal artifacts occur in the presence of strongly attenuating materials in computed tomography (CT). Recently, vendors have started offering metal artifact reduction (MAR) techniques. In addition, a MAR technique called the metal deletion technique (MDT) is freely available and able to reduce metal artifacts using reconstructed images. Although a comparison of the MDT to other MAR techniques exists, a comparison of commercially available MAR techniques is lacking. The aim of this study was therefore to quantify the difference in effectiveness of the currently available MAR techniques of different scanners and the MDT technique. MATERIALS AND METHODS: Three vendors were asked to use their preferential CT scanner for applying their MAR techniques. The scans were performed on a Philips Brilliance ICT 256 (S1), a GE Discovery CT 750 HD (S2) and a Siemens Somatom Definition AS Open (S3). The scans were made using an anthropomorphic head and neck phantom (Kyoto Kagaku, Japan). Three amalgam dental implants were constructed and inserted between the phantom's teeth. The average absolute error (AAE) was calculated for all reconstructions in the proximity of the amalgam implants. RESULTS: The commercial techniques reduced the AAE by 22.0±1.6%, 16.2±2.6% and 3.3±0.7% for S1 to S3 respectively. After applying the MDT to uncorrected scans of each scanner the AAE was reduced by 26.1±2.3%, 27.9±1.0% and 28.8±0.5% respectively. The difference in efficiency between the commercial techniques and the MDT was statistically significant for S2 (p=0.004) and S3 (p<0.001), but not for S1 (p=0.63). CONCLUSIONS: The effectiveness of MAR differs between vendors. S1 performed slightly better than S2 and both performed better than S3. Furthermore, for our phantom and outcome measure the MDT was more effective than the commercial MAR technique on all scanners.