E5517. The Efficacy of Artificial Intelligence in Radiation Treatment Planning
Authors
Fawad Alam-Siddiqui;
University of Pikesville
Objective:
Artificial intelligence (AI) is a rapidly evolving technology that is poised to change the future of many specialties within medicine. Of these specialties, technology focused specialties such as radiology and radiation oncology are particularly susceptible to being affected by this revolutionary technology. AI and its subsets machine learning (ML) and deep learning (DL), can be used to improve efficiency and streamline workflow. AI can be employed to extend the capabilities of a dosimetrist, while decreasing the time needed per treatment plan. Image segmentation, also known as contouring, is a labor- and time-intensive process that has a high degree of variability between observers and users. Contouring can be highly variable, and there are current image models that can be applied to quicken the process; however, some of them are flawed. AI can be used to improve dose distributions and optimize protections for organs at risk (OAR).
Materials and Methods:
Electronic databases, including Google Scholar, EBSCO, National Center for Biotechnology Information, PubMed, and Cumulative Index to Nursing and Allied Health Literature, were used to conduct a literature search. Articles from 2016 or later were considered, with one exception allowed for an article from 2009. Sources lacking pertinence to the specific topic of treatment planning were discounted, and those not corroborated by at least two other sources also were excluded.
Results:
The results of this literature review indicate that AI could transform the medical landscape, including radiation treatment planning, with AI demonstrating the ability to streamline quality assurance (QA) and workflow, image segmentation, and dose optimization.
However, there are barriers that must be overcome; these barriers include initial cost, clinical application limits, and ethical dilemmas.
Conclusion:
AI has the potential to revolutionize radiation oncology. The literature has demonstrated AI’s potential in QA, image segmentation, and dose optimization. AI and its subdivisions of ML and DL are evolving continuously, and great strides are being made by computer scientists. However, AI also has limitations in terms of cost, clinical limitations, and ethics. Aspects of AI and its related technologies are being used for model-based image segmentation and dose optimization, but fully automated clinical plans are not yet widely used or accepted. However, given the data, wider clinical testing and implementation might eventually allow for acceptance of fully automated clinical plans.
