Automated Data Extraction and Quality Assurance Tests for High Dose Rate Brachytherapy Treatment Plans
Abstract
Purpose
High Dose Rate (HDR) brachytherapy requires a rapid turnaround from treatment planning to delivery. Medical physicists perform dose planning and quality assurance (QA) checks prior to treatment, but these tasks are time-consuming and involve repetitive manual steps that can slow workflow and introduce transcription errors. This project automates extraction of patient and plan data from the treatment planning system (TPS) to support quality assurance and retrospective analysis of HDR brachytherapy plans.
Methods
A data parsing pipeline was developed using anonymized HDR brachytherapy cases from BC Cancer Kelowna, including 45 prostate and 30 gynecological plans. Patient and plan information was extracted from TPS-generated documents. Validation was performed using 40 prostate and 15 gynecological HDR cases. Two approaches were evaluated: (1) PDF-based text extraction from TPS summary documents, and (2) DICOM-based data extraction to enable additional QA checks.
Results
A monitoring (“watcher”) script continuously observes a designated folder for new files. Once new documents are detected during the planning process, the system automatically performs a series of QA checks and generates a standardized PDF report. These reports summarize key patient information for physicist review and include QA calculations such as a total dwell time check, an independent secondary dose estimate at two reference points, and patient identification verification. The PDF-based extraction approach integrates seamlessly into the existing workflow with no additional user steps, but is sensitive to document formatting changes. The DICOM-based approach offers more robust and consistent extraction with access to a broader parameter set, at the cost of additional steps for data retrieval.
Conclusion
The core QA automation workflow is ready for clinical deployment. Future work will focus on expanding DICOM-based data extraction to support a wider range of QA tests and enabling direct querying of the DICOM server to further automate data retrieval and analysis.