Artificial Intelligence for QA
K Hasse1*, A Witztum2*, S Luk3*, C Guthier4*, (1) University of California, San Francisco, San Francisco, CA, (2) University of California, San Francisco, Oxford, GB, (3) University of Vermont Medical Center, Burlington, VT, (4) Brigham and Women's Hospital, Brookline, MA
Presentations
WE-A-BRA-0 (Wednesday, 7/13/2022) 7:30 AM - 8:30 AM [Eastern Time (GMT-4)]
Ballroom A
Artificial intelligence (AI) has shown great potential to revolutionize therapeutic medical physics. In recent years, there have been many advances in the application of AI to the clinical medical physics workflow, including but not limited to contouring, planning, and quality assurance. In this session we will delve deeper into the use of AI for quality assurance (QA) purposes.
Pre-treatment measurement-based QA has been shown to be of questionable utility and yet this method remains by far the most popular form of patient-specific QA (PSQA). AI methods can be used to predict the results of pre-treatment measurement-based QA using plan complexity features. This information can be used as a clinical decision support system to improve the efficiency of the PSQA process by reducing the number of measurements required. In the first talk of the session, we will discuss whether such a clinical decision support system could also be extended to assess online adaptive plans, where a pre-treatment measurement cannot be taken.
Physics plan and chart review is a key component of quality control processes for detecting high severity incidents in radiation oncology. However, the actual performance of this process has shown to be lower than expected. Recent studies have reported the successful use of rules-based algorithms to improve the efficiency and effectiveness of plan and chart review process, but they are limited in adaptability and ability to perform nuanced reasoning. AI could potentially overcome these limits and complement with the rules to provide a comprehensive assistance on physics plan and chart review. In the second talk of this session, we will discuss the recent developments of AI models that assist plan and chart reviews and the hurdles and potential solutions towards clinical implementation of these models.
There are unique challenges associated with the clinical development and implementation of AI platforms in radiation oncology. In the final talk of this session, we will review current guidelines and what we can learn from other technologies for a safe implementation of AI with respect to initial commissioning and ongoing QA. Once the system is fully understood, AI can be used either as extensions for contouring and planning tools but also as standalone applications for quality control. The latter is of particular interest since it can provide additional independent checks in real-time.
Specific learning objectives include:
1. Understand how AI can be used as a clinical decision support system to reduce the number of PSQA measurements required.
2. Understand how AI can assist physics plan and chart review, and recognize the hurdles and potential solutions towards implementing AI tools for plan review clinically.
3. Understand how to successfully commission and QA AI systems to be used clinically, and recognize how AI can be used as a quality control tool for various tasks.
Funding Support, Disclosures, and Conflict of Interest: Ownership stake in Foretell Med LLC which develops machine learning models for PSQA
Handouts
- 175-63529-16291659-186916-2004552809.pdf (Samuel Ming Ho Luk)
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email