Predicting Radiotherapy Patient Outcomes with Real-Time Clinical Data Using Mathematical Modelling.

Longitudinal tumour volume data from head-and-neck cancer patients show that tumours of comparable pre-treatment size and stage may respond very differently to the same radiotherapy fractionation protocol. Mathematical models are often proposed to predict treatment outcome in this context, and have the potential to guide clinical decision-making and inform personalised fractionation protocols. Hindering effective use of models in this context is the sparsity of clinical measurements juxtaposed with the model complexity required to produce the full range of possible patient responses. In this work, we present a compartment model of tumour volume and tumour composition, which, despite relative simplicity, is capable of producing a wide range of patient responses. We then develop novel statistical methodology and leverage a cohort of existing clinical data to produce a predictive model of both tumour volume progression and the associated level of uncertainty that evolves throughout a patient's course of treatment. To capture inter-patient variability, all model parameters are patient specific, with a bootstrap particle filter-like Bayesian approach developed to model a set of training data as prior knowledge. We validate our approach against a subset of unseen data, and demonstrate both the predictive ability of our trained model and its limitations.

Essentially unedited deep-learning-based OARs are suitable for rigorous oropharyngeal and laryngeal cancer treatment planning.

Quality of organ at risk (OAR) autosegmentation is often judged by concordance metrics against the human-generated gold standard. However, the ultimate goal is the ability to use unedited autosegmented OARs in treatment planning, while maintaining the plan quality. We tested this approach with head and neck (HN) OARs generated by a prototype deep-learning (DL) model on patients previously treated for oropharyngeal and laryngeal cancer. Forty patients were selected, with all structures delineated by an experienced physician. For each patient, a set of 13 OARs were generated by the DL model. Each patient was re-planned based on original targets and unedited DL-produced OARs. The new dose distributions were then applied back to the manually delineated structures. The target coverage was evaluated with inhomogeneity index (II) and the relative volume of regret. For the OARs, Dice similarity coefficient (DSC) of areas under the DVH curves, individual DVH objectives, and composite continuous plan quality metric (PQM) were compared. The nearly identical primary target coverage for the original and re-generated plans was achieved, with the same II and relative volume of regret values. The average DSC of the areas under the corresponding pairs of DVH curves was 0.97 ± 0.06. The number of critical DVH points which met the clinical objectives with the dose optimized on autosegmented structures but failed when evaluated on the manual ones was 5 of 896 (0.6%). The average OAR PQM score with the re-planned dose distributions was essentially the same when evaluated either on the autosegmented or manual OARs. Thus, rigorous HN treatment planning is possible with OARs segmented by a prototype DL algorithm with minimal, if any, manual editing.

Technical feasibility of novel immunostimulatory low-dose radiation for polymetastatic disease with CBCT-based online adaptive and conventional approaches.

PURPOSE: A workflow/planning strategy delivering low-dose radiation therapy (LDRT) (1 Gy) to all polymetastatic diseases using conventional planning/delivery (Raystation/Halcyon = "conventional") and the AI-based Ethos online adaptive RT (oART) platform is developed/evaluated. METHODS: Using retrospective data for ten polymetastatic non-small cell lung cancer patients (5-52 lesions each) with PET/CTs, gross tumor volumes (GTVs) were delineated using PET standardized-uptake-value (SUV) thresholding. A 1 cm uniform expansion of GTVs to account for setup/contour uncertainty and organ motion-generated planning target volumes (PTVs). Dose optimization/calculation used the diagnostic CT from PET/CT. Dosimetric objectives were: Dmin,0.03cc ≥ 95% (acceptable variation (Δ) ≥ 90%), V100% ≥ 95% (Δ ≥ 90%), and D0.03cc ≤ 120% (Δ ≤ 125%). Additionally, online adaptation was simulated. When available, subsequent diagnostic CT was used to represent on-treatment CBCT. Otherwise, the CT from PET/CT used for initial planning was deformed to simulate clinically representative changes. RESULTS: All initial plans generated, both for Raystation and Ethos, achieved clinical goals within acceptable variation. For all patients, Dmin,0.03cc ≥ 95%, V100% ≥ 95%, and D0.03cc ≤ 120% goals were achieved for 84.8%/99.5%, 97.7%/98.7%, 97.4%/92.3%, in conventional/Ethos plans, respectively. The ratio of 50% isodose volume to PTV volume (R50%), maximum dose at 2 cm from PTV (D2cm), and the ratio of the 100% isodose volume to PTV volume (conformity index) in Raystation/Ethos plans were 7.9/5.9; 102.3%/88.44%; and 0.99/1.01, respectively. In Ethos, online adapted plans maintained PTV coverage whereas scheduled plans often resulted in geographic misses due to changes in tumor size, patient position, and body habitus. The average total duration of the oART workflow was 26:15 (min:sec) ranging from 6:43 to 57:30. The duration of each oART workflow step as a function of a number of targets showed a low correlation coefficient for influencer generation and editing (R2 = 0.04 and 0.02, respectively) and high correlation coefficient for target generation, target editing and plan generation (R2 = 0.68, 0.63 and 0.69, respectively). CONCLUSIONS: This study demonstrates feasibility of conventional planning/treatment with Raystation/Halcyon and highlights efficiency gains when utilizing semi-automated planning/online-adaptive treatment with Ethos for immunostimulatory LDRT conformally delivered to all sites of polymetastatic disease.

A head and neck treatment planning strategy for a CBCT-guided ring-gantry online adaptive radiotherapy system.

PURPOSE: A planning strategy was developed and the utility of online-adaptation with the Ethos CBCT-guided ring-gantry adaptive radiotherapy (ART) system was evaluated using retrospective data from Head-and-neck (H&N) patients that required clinical offline adaptation during treatment. METHODS: Clinical data were used to re-plan 20 H&N patients (10 sequential boost (SEQ) with separate base and boost plans plus 10 simultaneous integrated boost (SIB)). An optimal approach, robust to online adaptation, for Ethos-initial plans using clinical goal prioritization was developed. Anatomically-derived isodose-shaping helper structures, air-density override, goals for controlling hotspot location(s), and plan normalization were investigated. Online adaptation was simulated using clinical offline adaptive simulation-CTs to represent an on-treatment CBCT. Dosimetric comparisons were based on institutional guidelines for Clinical-initial versus Ethos-initial plans and Ethos-scheduled versus Ethos-adapted plans. Timing for five components of the online adaptive workflow was analyzed. RESULTS: The Ethos H&N planning approach generated Ethos-initial SEQ plans with clinically comparable PTV coverage (average PTVHigh V100%  = 98.3%, Dmin,0.03cc  = 97.9% and D0.03cc  = 105.5%) and OAR sparing. However, Ethos-initial SIB plans were clinically inferior (average PTVHigh V100%  = 96.4%, Dmin,0.03cc  = 93.7%, D0.03cc  = 110.6%). Fixed-field IMRT was superior to VMAT for 93.3% of plans. Online adaptation succeeded in achieving conformal coverage to the new anatomy in both SEQ and SIB plans that was even superior to that achieved in the initial plans (which was due to the changes in anatomy that simplified the optimization). The average adaptive workflow duration for SIB, SEQ base and SEQ boost was 30:14, 22.56, and 14:03 (min: sec), respectively. CONCLUSIONS: With an optimal planning approach, Ethos efficiently auto-generated dosimetrically comparable and clinically acceptable initial SEQ plans for H&N patients. Initial SIB plans were inferior and clinically unacceptable, but adapted SIB plans became clinically acceptable. Online adapted plans optimized dose to new anatomy and maintained target coverage/homogeneity with improved OAR sparing in a time-efficient manner.

Maintaining dosimetric quality when switching to a Monte Carlo dose engine for head and neck volumetric-modulated arc therapy planning.

Head and neck cancers present challenges in radiation treatment planning due to the large number of critical structures near the target(s) and highly heterogeneous tissue composition. While Monte Carlo (MC) dose calculations currently offer the most accurate approximation of dose deposition in tissue, the switch to MC presents challenges in preserving the parameters of care. The differences in dose-to-tissue were widely discussed in the literature, but mostly in the context of recalculating the existing plans rather than reoptimizing with the MC dose engine. Also, the target dose homogeneity received less attention. We adhere to strict dose homogeneity objectives in clinical practice. In this study, we started with 21 clinical volumetric-modulated arc therapy (VMAT) plans previously developed in Pinnacle treatment planning system. Those plans were recalculated "as is" with RayStation (RS) MC algorithm and then reoptimized in RS with both collapsed cone (CC) and MC algorithms. MC statistical uncertainty (0.3%) was selected carefully to balance the dose computation time (1-2 min) with the planning target volume (PTV) dose-volume histogram (DVH) shape approaching that of a "noise-free" calculation. When the hot spot in head and neck MC-based treatment planning is defined as dose to 0.03 cc, it is exceedingly difficult to limit it to 105% of the prescription dose, as we were used to with the CC algorithm. The average hot spot after optimization and calculation with RS MC was statistically significantly higher compared to Pinnacle and RS CC algorithms by 1.2 and 1.0 %, respectively. The 95% confidence interval (CI) observed in this study suggests that in most cases a hot spot of ≤107% is achievable. Compared to the 95% CI for the previous clinical plans recalculated with RS MC "as is" (upper limit 108%), in real terms this result is at least as good or better than the historic plans.

Unlocking a closed system: dosimetric commissioning of a ring gantry linear accelerator in a multivendor environment.

The Halcyon™ platform is self-contained, combining a treatment planning (Eclipse) system TPS) with information management and radiation delivery components. The standard TPS beam model is configured and locked down by the vendor. A portal dosimetry-based system for patient-specific QA (PSQA) is also included. While ensuring consistency across the user base, this closed model may not be optimal for every department. We set out to commission independent TPS (RayStation 9B, RaySearch Laboratories) and PSQA (PerFraction, Sun Nuclear Corp.) systems for use with the Halcyon linac. The output factors and PDDs for very small fields (0.5 × 0.5 cm2 ) were collected to augment the standard Varian dataset. The MLC leaf-end parameters were estimated based on the various static and dynamic tests with simple model fields and honed by minimizing the mean and standard deviation of dose difference between the ion chamber measurements and RayStation Monte Carlo calculations for 15 VMAT and IMRT test plans. Two chamber measurements were taken per plan, in the high (isocenter) and lower dose regions. The ratio of low to high doses ranged from 0.4 to 0.8. All percent dose differences were expressed relative to the local dose. The mean error was 0.0 ± 1.1% (TG119-style confidence limit ± 2%). Gamma analysis with the helical diode array using the standard 3%Global/2mm criteria resulted in the average passing rate of 99.3 ± 0.5% (confidence limit 98.3%-100%). The average local dose error for all detectors across all plans was 0.2% ± 5.3%. The ion chamber results compared favorably with our recalculation with Eclipse and PerFraction, as well as with several published Eclipse reports. Dose distribution gamma analysis comparisons between RayStation and PerFraction with 2%Local/2mm criteria yielded an average passing rate of 98.5% ± 0.8% (confidence limit 96.9%-100%). It is feasible to use the Halcyon accelerator with independent planning and verification systems without sacrificing dosimetric accuracy.

Pretreatment CT and PET radiomics predicting rectal cancer patients in response to neoadjuvant chemoradiotherapy.

BACKGROUND: The purpose of this study was to characterize pre-treatment non-contrast computed tomography (CT) and 18F-fluorodeoxyglucose positron emission tomography (PET) based radiomics signatures predictive of pathological response and clinical outcomes in rectal cancer patients treated with neoadjuvant chemoradiotherapy (NACR T). MATERIALS AND METHODS: An exploratory analysis was performed using pre-treatment non-contrast CT and PET imaging dataset. The association of tumor regression grade (TRG) and neoadjuvant rectal (NAR) score with pre-treatment CT and PET features was assessed using machine learning algorithms. Three separate predictive models were built for composite features from CT + PET. RESULTS: The patterns of pathological response were TRG 0 (n = 13; 19.7%), 1 (n = 34; 51.5%), 2 (n = 16; 24.2%), and 3 (n = 3; 4.5%). There were 20 (30.3%) patients with low, 22 (33.3%) with intermediate and 24 (36.4%) with high NAR scores. Three separate predictive models were built for composite features from CT + PET and analyzed separately for clinical endpoints. Composite features with α = 0.2 resulted in the best predictive power using logistic regression. For pathological response prediction, the signature resulted in 88.1% accuracy in predicting TRG 0 vs. TRG 1-3; 91% accuracy in predicting TRG 0-1 vs. TRG 2-3. For the surrogate of DFS and OS, it resulted in 67.7% accuracy in predicting low vs. intermediate vs. high NAR scores. CONCLUSION: The pre-treatment composite radiomics signatures were highly predictive of pathological response in rectal cancer treated with NACR T. A larger cohort is warranted for further validation.

The 2019 mathematical oncology roadmap.

Whether the nom de guerre is Mathematical Oncology, Computational or Systems Biology, Theoretical Biology, Evolutionary Oncology, Bioinformatics, or simply Basic Science, there is no denying that mathematics continues to play an increasingly prominent role in cancer research. Mathematical Oncology-defined here simply as the use of mathematics in cancer research-complements and overlaps with a number of other fields that rely on mathematics as a core methodology. As a result, Mathematical Oncology has a broad scope, ranging from theoretical studies to clinical trials designed with mathematical models. This Roadmap differentiates Mathematical Oncology from related fields and demonstrates specific areas of focus within this unique field of research. The dominant theme of this Roadmap is the personalization of medicine through mathematics, modelling, and simulation. This is achieved through the use of patient-specific clinical data to: develop individualized screening strategies to detect cancer earlier; make predictions of response to therapy; design adaptive, patient-specific treatment plans to overcome therapy resistance; and establish domain-specific standards to share model predictions and to make models and simulations reproducible. The cover art for this Roadmap was chosen as an apt metaphor for the beautiful, strange, and evolving relationship between mathematics and cancer.

Comprehensive evaluation of the high-resolution diode array for SRS dosimetry.

A high-resolution diode array has been comprehensively evaluated. It consists of 1013 point diode detectors arranged on the two 7.7 × 7.7 cm2 printed circuit boards (PCBs). The PCBs are aligned face to face in such a way that the active volumes of all diodes are in the same plane. All individual correction factors required for accurate dosimetry have been validated for conventional and flattening filter free (FFF) 6MV beams. That included diode response equalization, linearity, repetition rate dependence, field size dependence, angular dependence at the central axis and off-axis in the transverse, sagittal, and multiple arbitrary planes. In the end-to-end tests the array and radiochromic film dose distributions for SRS-type multiple-target plans were compared. In the equalization test (180° rotation), the average percent dose error between the normal and rotated positions for all diodes was 0.01% ± 0.1% (range -0.3 to 0.4%) and -0.01% ± 0.2% (range -0.9 to 0.9%) for 6 MV and 6MV FFF beams, respectively. For the axial angular response, corrected dose stayed within 2% from the ion chamber for all gantry angles, until the beam direction approached the detector plane. In azimuthal direction, the device agreed with the scintillator within 1% for both energies. For multiple combinations of couch and gantry angles, the average percent errors were -0.00% ± 0.6% (range: -2.1% to 1.6%) and -0.1% ± 0.5% (range -1.6% to 2.1%) for the 6MV and 6MV FFF beams, respectively. The measured output factors were largely within 2% of the scintillator, except for the 5 mm 6MV beam showing a 3.2% deviation. The 2%/1 mm gamma analysis of composite SRS measurements produced the 97.2 ± 1.3% (range 95.8-98.5%) average passing rate against film. Submillimeter (≤0.5 mm) dose profile alignment with film was demonstrated in all cases.

A Monte Carlo Method for Determining the Response Relationship between Two Commonly Used Detectors to Indirectly Measure Alpha Particle Radiation Activity.

Using targeted ligands to deliver alpha-emitting radionuclides directly to tumor cells has become a promising therapeutic strategy. To calculate the radiation dose to patients, activities of parent and daughter radionuclides must be measured. Scintillation detectors can be used to quantify these activities; however, activities found in pre-clinical and clinical studies can exceed their optimal performance range. Therefore, a method of correcting scintillation detector measurements at higher activities was developed using Monte Carlo modeling. Because there are currently no National Institute of Standards and Technology traceable Actinium-225 (225Ac) standards available, a well-type ionization chamber was used to measure 70.3 ± 7.0, 144.3 ± 14.4, 222.0 ± 22.2, 299.7 ± 30.0, 370.0 ± 37.0, and 447.7 ± 44.7 kBq samples of 225Ac obtained from Oak Ridge National Lab. Samples were then placed in a well-type NaI(Tl) scintillation detector and spectra were obtained. Alpha particle activity for each species was calculated using gamma abundance per alpha decay. MCNP6 Monte Carlo software was used to simulate the 4π-geometry of the NaI(Tl) detector. Using the ionization chamber reading as activity input to the Monte Carlo model, spectra were obtained and compared to NaI(Tl) spectra. Successive simulations of different activities were run until a spectrum minimizing the mean percent difference between the two was identified. This was repeated for each sample activity. Ionization chamber calibration measurements showed increase in error from 3% to 10% as activities decreased, resulting from decreasing detection efficiency. Measurements of 225Ac using both detector types agreed within 7% of Oak Ridge stated activities. Simulated Monte Carlo spectra of 225Ac were successfully generated. Activities obtained from these spectra differed with ionization chamber readings up to 156% at 147.7 kBq. Simulated spectra were then adjusted to correct NaI(Tl) measurements to be within 1%. These were compared to ionization chamber readings and a response relationship was determined between the two instruments. Measurements of 225Ac and daughter activity were conducted using a NaI(Tl) scintillation detector calibrated for energy and efficiency and an ionization chamber calibrated for efficiency using a surrogate calibration reference. Corrections provided by Monte Carlo modeling improve the accuracy of activity quantification for alpha-particle emitting radiopharmaceuticals in pre-clinical and clinical studies.

Development of Targeted Alpha Particle Therapy for Solid Tumors.

Targeted alpha-particle therapy (TAT) aims to selectively deliver radionuclides emitting α-particles (cytotoxic payload) to tumors by chelation to monoclonal antibodies, peptides or small molecules that recognize tumor-associated antigens or cell-surface receptors. Because of the high linear energy transfer (LET) and short range of alpha (α) particles in tissue, cancer cells can be significantly damaged while causing minimal toxicity to surrounding healthy cells. Recent clinical studies have demonstrated the remarkable efficacy of TAT in the treatment of metastatic, castration-resistant prostate cancer. In this comprehensive review, we discuss the current consensus regarding the properties of the α-particle-emitting radionuclides that are potentially relevant for use in the clinic; the TAT-mediated mechanisms responsible for cell death; the different classes of targeting moieties and radiometal chelators available for TAT development; current approaches to calculating radiation dosimetry for TATs; and lead optimization via medicinal chemistry to improve the TAT radiopharmaceutical properties. We have also summarized the use of TATs in pre-clinical and clinical studies to date.

The Evolution of Tumour Composition During Fractionated Radiotherapy: Implications for Outcome.

Current protocols for delivering radiotherapy are based primarily on tumour stage and nodal and metastases status, even though it is well known that tumours and their microenvironments are highly heterogeneous. It is well established that the local oxygen tension plays an important role in radiation-induced cell death, with hypoxic tumour regions responding poorly to irradiation. Therefore, to improve radiation response, it is important to understand more fully the spatiotemporal distribution of oxygen within a growing tumour before and during fractionated radiation. To this end, we have extended a spatially resolved mathematical model of tumour growth, first proposed by Greenspan (Stud Appl Math 51:317-340, 1972), to investigate the effects of oxygen heterogeneity on radiation-induced cell death. In more detail, cell death due to radiation at each location in the tumour, as determined by the well-known linear-quadratic model, is assumed also to depend on the local oxygen concentration. The oxygen concentration is governed by a reaction-diffusion equation that is coupled to an integro-differential equation that determines the size of the assumed spherically symmetric tumour. We combine numerical and analytical techniques to investigate radiation response of tumours with different intratumoral oxygen distribution profiles. Model simulations reveal a rapid transient increase in hypoxia upon regrowth of the tumour spheroid post-irradiation. We investigate the response to different radiation fractionation schedules and identify a tumour-specific relationship between inter-fraction time and dose per fraction to achieve cure. The rich dynamics exhibited by the model suggest that spatial heterogeneity may be important for predicting tumour response to radiotherapy for clinical applications.

Predicting Patient-Specific Radiotherapy Protocols Based on Mathematical Model Choice for Proliferation Saturation Index.

Radiation is commonly used in cancer treatment. Over 50% of all cancer patients will undergo radiotherapy (RT) as part of cancer care. Scientific advances in RT have primarily focused on the physical characteristics of treatment including beam quality and delivery. Only recently have inroads been made into utilizing tumor biology and radiobiology to design more appropriate RT protocols. Tumors are composites of proliferating and growth-arrested cells, and overall response depends on their respective proportions at irradiation. Prokopiou et al. (Radiat Oncol 10:159, 2015) developed the concept of the proliferation saturation index (PSI) to augment the clinical decision process associated with RT. This framework is based on the application of the logistic equation to pre-treatment imaging data in order to estimate a patient-specific tumor carrying capacity, which is then used to recommend a specific RT protocol. It is unclear, however, how dependent clinical recommendations are on the underlying tumor growth law. We discuss a PSI framework with a generalized logistic equation that can capture kinetics of different well-known growth laws including logistic and Gompertzian growth. Estimation of model parameters on the basis of clinical data revealed that the generalized logistic model can describe data equally well for a wide range of the generalized logistic exponent value. Clinical recommendations based on the calculated PSI, however, are strongly dependent on the specific growth law assumed. Our analysis suggests that the PSI framework may best be utilized in clinical practice when the underlying tumor growth law is known, or when sufficiently many tumor growth models suggest similar fractionation protocols.

A hybrid volumetric dose verification method for single-isocenter multiple-target cranial SRS.

A commercial semi-empirical volumetric dose verification system (PerFraction [PF], Sun Nuclear Corp.) extracts multi-leaf collimator positions from the electronic portal imaging device movies collected during a pre-treatment run, while the rest of the delivered control point information is harvested from the accelerator log files. This combination is used to reconstruct dose on a patient CT dataset with a fast superposition/convolution algorithm. The method was validated for single-isocenter multi-target SRS VMAT treatments against absolute radiochromic film measurements in a cylindrical phantom. The targets ranged in size from 0.8 to 3.6 cm and in number from 3 to 10 per plan. A total of 17 films rotated at different angles around the cylinder axis were analyzed. Each of 27 total targets was intercepted by at least one film, and 2-4 different films were analyzed per plan. Film dose was always scaled to the ion chamber measurement in a high-dose, low-gradient area deliberately created at the isocenter. The planar dose agreement between PF and film using 3%(Global dose-difference normalization)/1 mm gamma analysis was on average 99.2 ± 1.1%. The point dose difference in the low-gradient area in the middle of every target was below 3%, while PF-reconstructed and film dose centroids for individual targets showed submillimeter agreement when measured on a well aligned accelerator. Volumetrically, all voxels in all plans agreed between PF and the primary treatment planning system at the 3%/1 mm level. With proper understanding of its advantages and shortcomings, the tool can be applied to patient-specific QA in routine radiosurgical clinical practice.

Practical quantification of image registration accuracy following the AAPM TG-132 report framework.

The AAPM TG 132 Report enumerates important steps for validation of the medical image registration process. While the Report outlines the general goals and criteria for the tests, specific implementation may be obscure to the wider clinical audience. We endeavored to provide a detailed step-by-step description of the quantitative tests' execution, applied as an example to a commercial software package (Mirada Medical, Oxford, UK), while striving for simplicity and utilization of readily available software. We demonstrated how the rigid registration data could be easily extracted from the DICOM registration object and used, following some simple matrix math, to quantify accuracy of rigid translations and rotations. The options for validating deformable image registration (DIR) were enumerated, and it was shown that the most practically viable ones are comparison of propagated internal landmark points on the published datasets, or of segmented contours that can be generated locally. The multimodal rigid registration in our example did not always result in the desired registration error below ½ voxel size, but was considered acceptable with the maximum errors under 1.3 mm and 1°. The DIR target registration errors in the thorax based on internal landmarks were far in excess of the Report recommendations of 2 mm average and 5 mm maximum. On the other hand, evaluation of the DIR major organs' contours propagation demonstrated good agreement for lung and abdomen (Dice Similarity Coefficients, DSC, averaged over all cases and structures of 0.92 ± 0.05 and 0.91 ± 0.06, respectively), and fair agreement for Head and Neck (average DSC = 0.73 ± 0.14). The average for head and neck is reduced by small volume structures such as pharyngeal constrictor muscles. Even these relatively simple tests show that commercial registration algorithms cannot be automatically assumed sufficiently accurate for all applications. Formalized task-specific accuracy quantification should be expected from the vendors.

The future of personalised radiotherapy for head and neck cancer.

Radiotherapy has long been the mainstay of treatment for patients with head and neck cancer and has traditionally involved a stage-dependent strategy whereby all patients with the same TNM stage receive the same therapy. We believe there is a substantial opportunity to improve radiotherapy delivery beyond just technological and anatomical precision. In this Series paper, we explore several new ideas that could improve understanding of the phenotypic and genotypic differences that exist between patients and their tumours. We discuss how exploiting these differences and taking advantage of precision medicine tools-such as genomics, radiomics, and mathematical modelling-could open new doors to personalised radiotherapy adaptation and treatment. We propose a new treatment shift that moves away from an era of empirical dosing and fractionation to an era focused on the development of evidence to guide personalisation and biological adaptation of radiotherapy. We believe these approaches offer the potential to improve outcomes and reduce toxicity.

Superficial and peripheral dose in compensator-based FFF beam IMRT.

Flattening filter-free (FFF) beams produce higher dose rates. Combined with compensator-based intensity modulated radiotherapy (IMRT) techniques, the dose delivery for each beam can be much shorter compared to the flattened beam MLC-based or flattened beam compensator-based IMRT. This 'snap shot' IMRT delivery is beneficial to patients for tumor motion management. Due to softer energy, superficial doses in FFF beam treatment are usually higher than those from flattened beams. Due to no flattening filter, thus less photon scattering, peripheral doses are usually lower in FFF beam treatment. However, in compensator-based IMRT using FFF beams, the compensator is in the beam pathway. Does it introduce beam hardening effects and scattering such that the superficial dose is lower and peripheral dose is higher compared to FFF beam MLC-based IMRT? This study applied Monte Carlo techniques to investigate the superficial and peripheral doses in compensator-based IMRT using FFF beams and compared it to the MLC-based IMRT using FFF beams and flattened beams. Besides varying thicknesses of brass slabs to simulate varying thicknesses of compensators, a simple cone-shaped compensator was simulated to mimic a clinical application. The dose distribution in water phantom by the cone-shaped compensator was then simulated by multiple MLC-defined FFF and flattened beams with varying apertures. After normalization to the maximum dose, Dmax, the superficial and peripheral doses were compared between the FFF beam compensator-based IMRT and FFF/flattened beam MLC-based IMRT. The superficial dose at the central 0.5 mm depth was about 1% (of Dmax) lower in the compensator-based 6 MV FFF (6FFF) IMRT compared to the MLC-based 6FFF IMRT, and about 8% higher than the flattened 6 MV MLC-based IMRT dose. At 8 cm off-axis at depth of central maximum dose, dmax, the peripheral dose between the 6FFF and flattened 6 MV MLC demonstrated similar doses, while the compensator dose was about 1% (of Dmax) higher. Compensators reduce the superficial doses slightly compared to open FFF beams, but increases the peripheral doses due to scatter in the compensator.

Validation of a GPU-Based 3D dose calculator for modulated beams.

A superposition/convolution GPU-accelerated dose computation algorithm (the Calculator) has been recently incorporated into commercial software. The algorithm requires validation prior to clinical use. Three photon energies were examined: conventional 6 MV and 15 MV, and 10 MV flattening filter free (10 MVFFF). For a set of IMRT and VMAT plans based on four of the five AAPM Practice Guideline 5a downloadable datasets, ion chamber (IC) measurements were performed on the water-equivalent phantoms. The average difference between the Calculator and IC was -0.3 ± 0.8% (1SD). The same plans were projected on a phantom containing a biplanar diode array. We used the forthcoming criteria for routine gamma analysis, 3% dose-error (global (G) normalization, 2 mm distance to agreement, and 10% low dose cutoff). The γ (3%G/2 mm) average passing rate was 98.9 ± 2.1%. Measurement-guided three-dimensional dose reconstruction on the patient CT dataset (excluding the Lung) resulted in a similar average agreement rate with the Calculator: 98.2 ± 2.0%. The mean γ (3%G/2 mm) passing rate comparing the Calculator to the TPS (again excluding the Lung) was 99.0 ± 1.0%. Because of the significant inhomogeneity, the Lung case was investigated separately. The calculator has an alternate heterogeneity correction mode that can change the results in the thorax for higher-energy beams (15 MV). As this correction is nonphysical and was optimized for simple slab geometries, its application leads to mixed results when compared to the TPS and independent Monte Carlo calculations, depending on the CT dataset and the plan. The Calculator vs. TPS 15 MV Guideline 5a IMRT and VMAT plans demonstrate 96.3% and 93.4% γ (3%G/2 mm) passing rates respectively. For the lower energies, which should be predominantly used in the thoracic region, the passing rates for the same plans and criteria range from 98.6 to 100%. Overall, the Calculator accuracy is sufficient for the intended use.

Reproducibility of F18-FDG PET radiomic features for different cervical tumor segmentation methods, gray-level discretization, and reconstruction algorithms.

Site-specific investigations of the role of radiomics in cancer diagnosis and therapy are emerging. We evaluated the reproducibility of radiomic features extracted from 18 Flourine-fluorodeoxyglucose (18 F-FDG) PET images for three parameters: manual versus computer-aided segmentation methods, gray-level discretization, and PET image reconstruction algorithms. Our cohort consisted of pretreatment PET/CT scans from 88 cervical cancer patients. Two board-certified radiation oncologists manually segmented the metabolic tumor volume (MTV1 and MTV2 ) for each patient. For comparison, we used a graphical-based method to generate semiautomated segmented volumes (GBSV). To address any perturbations in radiomic feature values, we down-sampled the tumor volumes into three gray-levels: 32, 64, and 128 from the original gray-level of 256. Finally, we analyzed the effect on radiomic features on PET images of eight patients due to four PET 3D-reconstruction algorithms: maximum likelihood-ordered subset expectation maximization (OSEM) iterative reconstruction (IR) method, fourier rebinning-ML-OSEM (FOREIR), FORE-filtered back projection (FOREFBP), and 3D-Reprojection (3DRP) analytical method. We extracted 79 features from all segmentation method, gray-levels of down-sampled volumes, and PET reconstruction algorithms. The features were extracted using gray-level co-occurrence matrices (GLCM), gray-level size zone matrices (GLSZM), gray-level run-length matrices (GLRLM), neighborhood gray-tone difference matrices (NGTDM), shape-based features (SF), and intensity histogram features (IHF). We computed the Dice coefficient between each MTV and GBSV to measure segmentation accuracy. Coefficient values close to one indicate high agreement, and values close to zero indicate low agreement. We evaluated the effect on radiomic features by calculating the mean percentage differences (d¯) between feature values measured from each pair of parameter elements (i.e. segmentation methods: MTV1 -MTV2 , MTV1 -GBSV, MTV2 -GBSV; gray-levels: 64-32, 64-128, and 64-256; reconstruction algorithms: OSEM-FORE-OSEM, OSEM-FOREFBP, and OSEM-3DRP). We used |d¯| as a measure of radiomic feature reproducibility level, where any feature scored |d¯| ±SD ≤ |25|% ± 35% was considered reproducible. We used Bland-Altman analysis to evaluate the mean, standard deviation (SD), and upper/lower reproducibility limits (U/LRL) for radiomic features in response to variation in each testing parameter. Furthermore, we proposed U/LRL as a method to classify the level of reproducibility: High- ±1% ≤ U/LRL ≤ ±30%; Intermediate- ±30% < U/LRL ≤ ±45%; Low- ±45 < U/LRL ≤ ±50%. We considered any feature below the low level as nonreproducible (NR). Finally, we calculated the interclass correlation coefficient (ICC) to evaluate the reliability of radiomic feature measurements for each parameter. The segmented volumes of 65 patients (81.3%) scored Dice coefficient >0.75 for all three volumes. The result outcomes revealed a tendency of higher radiomic feature reproducibility among segmentation pair MTV1 -GBSV than MTV2 -GBSV, gray-level pairs of 64-32 and 64-128 than 64-256, and reconstruction algorithm pairs of OSEM-FOREIR and OSEM-FOREFBP than OSEM-3DRP. Although the choice of cervical tumor segmentation method, gray-level value, and reconstruction algorithm may affect radiomic features, some features were characterized by high reproducibility through all testing parameters. The number of radiomic features that showed insensitivity to variations in segmentation methods, gray-level discretization, and reconstruction algorithms was 10 (13%), 4 (5%), and 1 (1%), respectively. These results suggest that a careful analysis of the effects of these parameters is essential prior to any radiomics clinical application.