Click here to


Are you sure ?

Yes, do it No, cancel

Using Knowledge-Based Models to Train Human Planners with Lung and Mediastinum IMRT Planning

M Mistro1*, Y Sheng1,2 , Y Ge3 , J Palta4 , J Salama2 , C Kelsey2 , Q Wu1,2 , F Yin1,2 , Q Wu1,2 , (1) Duke University, Durham, NC, (2) Duke University Medical Center, Durham, NC, (3) University of North Carolina at Charlotte, Charlotte, NC, (4) Virginia Commonwealth University, Richmond, VA


(Wednesday, 7/17/2019) 1:45 PM - 3:45 PM

Room: 301

Purpose: To develop an e-learning system by incorporating knowledge-based treatment planning models to serve as informative, efficient bases to train individuals to develop optimal IMRT plans while creating confidence in utilizing these models in clinical settings.

Methods: A beam angle selection model and a DVH prediction model for lung/mediastinum IMRT planning are used as knowledge bases within a directed e-learning system guided by scoring criteria and communicated with the trainees via a user interface ran from Eclipse. The scoring serves both to illustrate relative quality of plans and to guide directed changes within the plan. One mediastinum case serves as a benchmark to show skill change from the e-learning system and is completed without intervention. Five additional lung/mediastinum cases follow in the subsequent training pipeline where the models, GUI and trainer work with trainee’s directives and guide meaningful beam selection and tradeoffs within IMRT optimization. Five trainees with minimal treatment planning background were evaluated by both the scoring criteria and a physician to look for improved planning quality and relative effectiveness compared to the clinically delivered plan.

Results: Trainees scored an average of 22.7% of total points within the scoring criteria for their benchmark yet improved to an average of 51.9%; the clinically delivered plan achieved 54.1% of total potential points. Two of the five trainee final plans were rated as comparable to the clinically delivered by a physician; all five trainees rated as noticeably improved. For the five plans within the system, trainees scored, on average, 24.5% higher than the respective clinically delivered plan.

Conclusion: This first attempt at creating a dynamic interface communicating the prior experience inherent to these models to an end-user took approximately 10 hours and rapidly improved planning quality. It brings unexperienced planners to a level comparable of experienced dosimetrists for a specific treatment site.

Funding Support, Disclosures, and Conflict of Interest: This work was partially supported by NIH/NCI 1R01CA201212


Not Applicable / None Entered.


TH- Dataset analysis/biomathematics: Informatics

Contact Email