Daniel La Russa

The use of treatment plan characteristics to predict patient-specific quality assurance (QA) measurement results has recently been reported as a strategy to help facilitate automated pre-treatment verification workflows or to provide a virtual assessment of delivery quality. The goal of this work is to investigate the potential of using treatment plan characteristics and linac performance metrics (i.e. quality control test results) in combination with machine learning techniques to predict the results of VMAT patient-specific QA measurements. Using features that describe treatment plan complexity and linac performance metrics, we trained a linear support vector classifier (SVC) to classify the results of VMAT patient-specific QA measurements. The ‘targets’ in this model were simple classes representing median dose difference between measured and expected dose distributions—‘hot’ if the median dose deviation was >1%, ‘cold’ if it was <−1%, and ‘normal’ if it was within ±1%. A total of 1620 unique patient-specific QA measurements were available for model development and testing. 75% of the data were used to develop and cross-validate the model, and the remaining 25% were used for an independent assessment of model performance. For the model development phase, a recursive feature elimination (RFE) cross-validation technique was used to eliminate unimportant features. Model performance was assessed using receiver operator characteristic (ROC) curve metrics. Of the ten features found to be most predictive of patient-specific QA measurement results, half were derived from treatment plan characteristics and half from quality control (QC) metrics characterizing linac performance. The model achieved a micro-averaged area under the ROC curve of 0.93, and a macro-averaged area under the ROC curve of 0.88. This work demonstrates the potential of using both treatment plan characteristics and routine linac QC results in the development of machine learning models for VMAT patient-specific QA measurements.

Recent technological innovations have created new opportunities for the increased adoption of virtual reality (VR) and augmented reality (AR) applications in medicine. While medical applications of VR have historically seen greater adoption from patient-as-user applications, the new era of VR/AR technology has created the conditions for wider adoption of clinician-as-user applications. Historically, adoption to clinical use has been limited in part by the ability of the technology to achieve a sufficient quality of experience. This article reviews the definitions of virtual and augmented reality and briefly covers the history of their development. Currently available options for consumer-level virtual and augmented reality systems are presented, along with a discussion of technical considerations for their adoption in the clinical environment. Finally, a brief review of the literature of medical VR/AR applications is presented prior to introducing a comprehensive conceptual framework for the viewing and manipulation of medical images in virtual and augmented reality. Using this framework, we outline considerations for placing these methods directly into a radiology-based workflow and show how it can be applied to a variety of clinical scenarios.

A consortium of radiotherapy centers in North America is in the process of evaluating a novel new ⁶⁰Co teletherapy device, called the GammaPod™ (Xcision Medical Systems, Columbia Maryland), designed specifically for breast SBRT. The GammaPod consists of 36 collimated ⁶⁰Co sources with a total activity of 4320 Ci. The sources are housed in a hemispherical source carrier that rotates during treatment to produce a cylindrically symmetric cone of primary beam spanning 16° – 54° degrees from the horizontal. This unique beam geometry presents challenges when designing or evaluating room shielding for the purposes of meeting regulatory requirements, and for ensuring the safety of staff and the public in surrounding areas. Conventional methods for calculating radiation barrier thicknesses have been adapted so that barrier transmission factors for the GammaPod can be determined from a few relevant distances and characteristics of the primary beam. Simple formalisms have been determined for estimating shielding requirements for primary radiation (with a rotating and non‐rotating source carrier), patient‐scattered radiation, and leakage radiation. When making worst case assumptions, it was found that conventional barrier thicknesses associated with linac treatment suites are sufficient for shielding all sources of radiation from the GammaPod.

Purpose: The purpose of this project is to develop a robust method of parameter estimation for a Poisson‐based TCP model using Bayesian inference.Methods: Bayesian inference was performed using the PyMC3 probabilistic programming framework written in Python. A Poisson‐based TCP regression model that accounts for clonogen proliferation was fit to observed rates of local relapse as a function of equivalent dose in 2 Gy fractions for a population of 623 stage‐I non‐small‐cell lung cancer patients. The Slice Markov Chain Monte Carlo sampling algorithm was used to sample the posterior distributions, and was initiated using the maximum of the posterior distributions found by optimization. The calculation of TCP with each sample step required integration over the free parameter α, which was performed using an adaptive 24‐point Gauss‐Legendre quadrature. Convergence was verified via inspection of the trace plot and posterior distribution for each of the fit parameters, as well as with comparisons of the most probable parameter values with their respective maximum likelihood estimates.Results: Posterior distributions for α, the standard deviation of α (σ), the average tumour cell‐doubling time (Td), and the repopulation delay time (Tk), were generated assuming α/β = 10 Gy, and a fixed clonogen density of 10⁷ cm−³. Posterior predictive plots generated from samples from these posterior distributions are in excellent agreement with the observed rates of local relapse used in the Bayesian inference. The most probable values of the model parameters also agree well with maximum likelihood estimates.Conclusion: A robust method of performing Bayesian inference of TCP data using a complex TCP model has been established.

Purpose:: In this report the authors present the validation of a Monte Carlo dose calculation algorithm (XiO EMC from Elekta Software) for electron beams.Methods:: Calculated and measured dose distributions were compared for homogeneous water phantoms and for a 3D heterogeneous phantom meant to approximate the geometry of a trachea and spine. Comparisons of measurements and calculated data were performed using 2D and 3D gamma index dose comparison metrics.Results:: Measured outputs agree with calculated values within estimated uncertainties for standard and extended SSDs for open applicators, and for cutouts, with the exception of the 17 MeV electron beam at extended SSD for cutout sizes smaller than 5 × 5 cm². Good agreement was obtained between calculated and experimental depth dose curves and dose profiles (minimum number of measurements that pass a 2%/2 mm agreement 2D gamma index criteria for any applicator or energy was 97%). Dose calculations in a heterogeneous phantom agree with radiochromic film measurements (>98% of pixels pass a 3 dimensional 3%/2 mm γ‐criteria) provided that the steep dose gradient in the depth direction is considered.Conclusions:: Clinically acceptable agreement (at the 2%/2 mm level) between the measurements and calculated data for measurements in water are obtained for this dose calculation algorithm. Radiochromic film is a useful tool to evaluate the accuracy of electron MC treatment planning systems in heterogeneous media.

The treatment planning software SharePlan is designed to convert dose distributions generated by the TomoTherapy planning station into step‐and‐shoot IMRT plans deliverable on a c‐arm linear accelerator. Five anal canal patients who were planned for TomoTherapy treatments were exported into a SharePlan system and plans were generated for delivery on an Elekta Synergy unit. A total of 80 plans were generated for those five patients, with either seven, nine, eleven or twenty‐one gantry angles and different priorities between focusing on matching either the target doses or healthy tissue sparing of the TomoTherapy plan. The plans generated by SharePlan, while often not matching target coverage at prescription, matched well the TomoTherapy coverage at 95% and 105% of the prescription dose. Organ at risk dose, when heavily emphazied in the SharePlan calculations matched or bettered the TomoTherapy dose due to the placement of the beams and the sharper sup‐inf fall off of the dose distribution on a linac. For one of the patients, it was possible to produce a better DVH with SharePlan than the original TomoTherapy plan for those reasons. The TomoTherapy plans boasted significantly shorter delivery times than the plans generated with SharePlan.

The aim of this work was to apply failure modes and effect analysis (FMEA) to assess risk in two radiation planning and treatment processes; our on-call (out-of-clinical hours) process and our tomotherapy process. The motivation was provided by analysis of 2506 adverse incidents reported over a 5 year period, the on-call process for giving rise to a higher than expected number of incidents and our tomotherapy process for the reverse. For the on-call scenario, three separate processes were analysed: our current process, our current process incorporating a software upgrade eliminating several planning steps and a fully integrated process in which the patient is imaged, planned and treated on a single platform (TomoTherapy Hi Art, Accuray Incorporated, Sunnyvale, CA). After construction of a detailed process map for each case, a multidisciplinary group identified potential failure modes for each process step, the effects of each failure and existing controls. Risk probability numbers were determined from severity, frequency of occurrence and detectability scores assigned to each failure mode according to a standard scale. The results were analysed to identify and prioritise feasible and effective process improvements. For the on-call process, our current workflow was identified as incurring the highest risk of the three processes analysed, demonstrating quantitatively the value of the software upgrade and providing a clear rationale for the associated expense. In summary, we have found FMEA to be a feasible tool for assessing relative risk in a clinical process. However, operational and resource issues must be considered separately.