ESTRO meets Asia 2024 - Abstract Book

S332

Physics – Motion management and adaptive radiotherapy

ESTRO meets Asia 2024

Automating the extraction of Radixact Synchrony performance data for quality improvement

Scott B Crowe 1,2,3 , Jemma Walsh 1,4 , Liting Yu 1,3 , Philip Chan 1 , Rachel Effeney 1 , Gregory Rattray 1 , Tanya Kairn 1,5,6

1 Cancer Care Services, Royal Brisbane & Women's Hospital, Brisbane, Australia. 2 School of Electrical Engineering & Computer Science, University of Queensland, Brisbane, Australia. 3 School of Chemistry & Physics, Queensland University of Technology, Brisbane, Australia. 4 School of Clinical Sciences, Queensland University of Technology, Brisbane, Australia. 5 School of Electrical Engineering & Computer Science, Royal Brisbane & Women's Hospital, Brisbane, Australia. 6 School of Chemistry & Physics, Royal Brisbane & Women's Hospital, Brisbane, Australia

Purpose/Objective:

The Radixact Synchrony radiotherapy system enables real-time tracking and adaptation of treatments in response to motion of treatment volumes. It achieves this through frequent x-ray imaging of the target or fiducials, and where respiratory motion is involved, the monitoring of surrogate LEDs placed on the patient during treatment delivery. To maximise the benefits of the platform in our department, this study aimed to develop a system for the programmatic extraction and analysis of performance measures, particularly those generated during initial clinical use of the system. This data could then be incorporated in departmental operating procedures, compared against performance specifications and results reported by other departments, and used as baselines for evaluating the impact of changes in practice in the future.

Material/Methods:

Sources of data included MOSAIQ, DICOM data exported from the Precision planning system, Delivery Analysis (DA) cache files and delivery quality assurance (DQA) measurement results. Software for programmatic extraction and analysis of available data (summarised in Table 1) was developed in-house using the Python programming language, and ancillary packages (e.g., Pydicom). Data extracted or produced by the software is summarised in Table 1.

Table 1. Treatment data being processed automatically (or manually, where indicated)

Source

Data

MOSAIQ (manual)

Site, diagnosis, TNM and overall staging, grading, and pathology data. Patient sex, age, weight, and body-mass-index. Treatment delivery records, including notes on Synchrony tolerances and pauses. Prescription and number of fractions. Beam parameters, including pitch, field width, and gantry rotation period. Leaf position sinograms and associated complexity metrics. Planning parameters, including optimisation objectives and number of fiducials. Treatment delivery dates, times, durations, and pauses. Daily pre-treatment images and associated metadata (e.g., protocol). Target motion trajectories, potential difference scores, and rigid body differences. Displacement metrics, including histograms, percentiles, and ranges. Acquired radiographs, with timestamps.

Precision DICOM exports DA cache

DQA results

Phantom dose measurements and gamma comparison results.

Functions were developed to extract and analyse data on a delivery fragment, fraction, plan, patient, and cohort level basis. Data extracted or produced by the software was exported to CSV files, to facilitate exploratory statistical analysis in the future.

Results: