SafetyNet: Streamlining and Automating QA in radiotherapy.

Proper quality assurance (QA) of the radiotherapy process can be time-consuming and expensive. Many QA efforts, such as data export and import, are inefficient when done by humans. Additionally, humans can be unreliable, lose attention, and fail to complete critical steps that are required for smooth operations. In our group we have sought to break down the QA tasks into separate steps and to automate those steps that are better done by software running autonomously or at the instigation of a human. A team of medical physicists and software engineers worked together to identify opportunities to streamline and automate QA. Development efforts follow a formal cycle of writing software requirements, developing software, testing and commissioning. The clinical release process is separated into clinical evaluation testing, training, and finally clinical release. We have improved six processes related to QA and safety. Steps that were previously performed by humans have been automated or streamlined to increase first-time quality, reduce time spent by humans doing low-level tasks, and expedite QA tests. Much of the gains were had by automating data transfer, implementing computer-based checking and automation of systems with an event-driven framework. These coordinated efforts by software engineers and clinical physicists have resulted in speed improvements in expediting patient-sensitive QA tests.

Investigating the Incidence of Pulmonary Abnormalities as Identified by Parametric Response Mapping in Patients With Lung Cancer Before Radiation Treatment.

Purpose: Parametric response mapping (PRM) of high-resolution, paired inspiration and expiration computed tomography (CT) scans is a promising analytical imaging technique that is currently used in diagnostic applications and offers the ability to characterize and quantify certain pulmonary pathologies on a patient-specific basis. As one of the first studies to implement such a technique in the radiation oncology clinic, the goal of this work was to assess the feasibility for PRM analysis to identify pulmonary abnormalities in patients with lung cancer before radiation therapy (RT). Methods and Materials: High-resolution, paired inspiration and expiration CT scans were acquired from 23 patients with lung cancer as part of routine treatment planning CT acquisition. When applied to the paired CT scans, PRM analysis classifies lung parenchyma, on a voxel-wise basis, as normal, small airways disease (SAD), emphysema, or parenchymal disease (PD). PRM classifications were quantified as a percent of total lung volume and were evaluated globally and regionally within the lung. Results: PRM analysis of pre-RT CT scans was successfully implemented using a workflow that produced patient-specific maps and quantified specific phenotypes of pulmonary abnormalities. Through this study, a large prevalence of SAD and PD was demonstrated in this lung cancer patient population, with global averages of 10% and 17%, respectively. Moreover, PRM-classified normal and SAD in the region with primary tumor involvement were found to be significantly different from global lung values. When present, elevated levels of PD and SAD abnormalities tended to be pervasive in multiple regions of the lung, indicating a large burden of underlying disease. Conclusions: Pulmonary abnormalities, as detected by PRM, were characterized in patients with lung cancer scheduled for RT. Although further study is needed, PRM is a highly accessible CT-based imaging technique that has the potential to identify local lung abnormalities associated with chronic obstructive pulmonary disease and interstitial lung disease. Further investigation in the radiation oncology setting may provide strategies for tailoring RT planning and risk assessment based on pre-existing PRM-based pathology.

Machine learning for automated quality assurance in radiotherapy: A proof of principle using EPID data description.

PURPOSE: Developing automated methods to identify task-driven quality assurance (QA) procedures is key toward increasing safety, efficacy, and efficiency. We investigate the use of machine learning (ML) methods for possible visualization, automation, and targeting of QA, and assess its performance using multi-institutional data. METHODS: To enable automated analysis of QA data given its higher dimensional nature, we used nonlinear kernel mapping with support vector data description (SVDD) driven approaches. Instead of using labeled data as in typical support vector machine (SVM) applications, which requires exhaustive annotation, we applied a clustering extension of SVDD, which identifies the minimal enclosing hypersphere in the feature space defined by a kernel function separating normal operations from possible failures (i.e., outliers). In our case, QA test data are mapped by a Gaussian kernel to a higher dimensional feature space and then the minimal enclosing sphere was identified. This sphere, when mapped back to the input data space along the principal components, can separate the data into several components, each enclosing a separate cluster of QA points that could be used to evaluate tolerance boundaries and test reliability. We evaluated this approach for gantry sag, radiation field shift, and [multileaf collimator (MLC)] offset data acquired using electronic portal imaging devices (EPID), as representative examples. RESULTS: Data from eight LINACS and seven institutions (n = 119) were collected. A standardized EPID image of a phantom with fiducials provided deviation estimates between the radiation field and phantom center at four cardinal gantry angles. Deviation measurements in the horizontal direction (0°, 180°) were used to determine the gantry sag and deviations in the vertical direction (90°, 270°) were used to determine the field shift. These measurements were fed into the SVDD clustering algorithm with varying hypersphere radii (Gaussian widths). For gantry sag analysis, two clusters were identified one of which contained 2.5% of the outliers and also exceeded the 1 mm tolerance set by TG-142. In the case of field shifts, SVM clustering identified two distinct classes of measurements primarily driven by variations in the second principal component at 270°. Results from MLC analysis identified one outlier cluster (0.34%) along Leaf offset Constancy (LoC) axis that coincided with TG-142 limits. CONCLUSION: Machine learning methods based on SVDD clustering are promising for developing automated QA tools and providing insights into their reliability and reproducibility.