Presentations
MO-B-Therapy Room-0 (Monday, 4/19/2021) 12:30 PM - 2:30 PM [Eastern Time (GMT-4)]
Automation is the key to unlocking major progress in radiation therapy. Consistently delivering a first treatment within 1-2 days of initial consult, hypofractionated treatments for chronic metastatic disease, adaptive radiotherapy: These approaches would likely result in better outcomes but are simply not feasible without automation.
While contouring, treatment planning, plan evaluation and reporting, analytics, quality assurance, etc. can all be automated to one degree or another, no solution currently exists that does not require clinical expertise at key points to drive the process. Therefore, the goal of automation is to empower clinicians to provide their input in a way that achieves the highest quality results in the least amount of time and with the smallest probability of error. This requires a holistic view into how the automation fits into the broader workflow.
Clinical physicists are uniquely positioned to guide their clinics to identify the best candidates for automation and to implement intelligent automation into the clinic's workflow, for example, AI-driven automatic contouring. However, when evaluating commercially-available tools, it is critical for clinicians to understand the hurdles involved in the development and implementation of automation.
In addition to commercially-available automation options, many institutions have been developing in-house automated radiation treatment planning tools as a way to improve clinical efficiency. Clinical implementation of in-house tools requires a team effort and assurance of safety is crucial for patient care.
Breast treatment planning using tangential beams with fluence maps has been a tedious manual process in our clinic. A machine learning-based automated treatment planning (MLAP) tool was developed to improve planning efficiency for breast treatment planning. After a promising validation study performed both in the research setting and clinical setting, the MLAP tool was launched for actual patient treatment. Performance evaluation in a clinical setting as well as in a prospective fashion is critical to identify strengths and weaknesses of MLAP to enhance MLAP and its future clinical translation. A unified quality assurance guideline also was established to assure safe translation of the in-house automation tool. This guideline covers acceptance, commissioning, and maintenance testing as well as quality assurance test for each upgrade.
This session will provide an overview of how to identify appropriate tasks for automation, as well as challenges that may present themselves while trying to implement automation. Additionally, this session will present two methods of clinical implementation of AI: a commercially available AI-driven automatic contouring tool and in-house treatment planning automation tools for breast.
Learning Objectives:
1. Learn what to consider when identifying tasks that are good candidates for automation and how to implement a solution in a way that fits your clinic.
2. Understand the key obstacles in developing and implementing automation into clinical workflows.
3. To share experience in clinical implementation of automation tools.
Keywords
Taxonomy
TH- Dataset Analysis/Biomathematics: Machine learning techniques
Contact Email