Radiation Physics

PV QA 3 - Poster Viewing Q&A 3

TU_12_3232 - Adoption of Knowledge-Based Treatment Planning Models

Tuesday, October 23
1:00 PM - 2:30 PM
Location: Innovation Hub, Exhibit Hall 3

Adoption of Knowledge-Based Treatment Planning Models
J. Baker1, A. Sharma1, Y. Cao1, J. Antone1, J. Rogers2, B. Hamilton2, and L. Potters1; 1Zucker School of Medicine at Hofstra/Northwell, Hempstead, NY, 2Alyzen Medical Physics, Jonesboro, AZ

Purpose/Objective(s): Knowledge-Base Planning (KBP) is valuable in achieving quality and efficiency in treatment planning. Modeling protocols requires a steep learning curve and early adoption of KBP is minimal. This study outlines our experience in developing KBP models and adopting them real-time in the clinic.

Materials/Methods: Our large, multi-site department has been using treatment directives for the past 8 years, which has created a library of consistently contoured cases. Coupled with peer review rounds, our contours are reliable for each directive. This case library was utilized for KBP model development. Model development took place on commercially available treatment planning software. Our initial trial was a generalized head and neck (GH&N) model. Additional models developed include prostate with or without seminal vesicles (PSV), prostate with nodes (P+N), prostate SBRT (PSBRT), and gynecologic with pelvic nodes (GP). Initial model training includes all plans for a model site in the last 30 months. KBP objectives utilize the best case scenario, so a less than ideal plan is ignored. The resulting models consist of 206, 45, 74, 28 and 81 plans respectively. Validation consisted of statistical analysis and removing outlying data points. Additional patients were used to adjust generalized optimization objectives.

Results: We found that the models generate plans in three categories. Less complicated geometries produce plans with minimal intervention. This includes the PSV, P+N, PSBRT and GP models. Although, a substantial change in planning methodology may result in a less effective model, which is corrected through retraining and validation. Another scenario results with ineffective optimization objectives, but DVH estimations help ensure quality. Lastly, we find that some models cannot be implemented. This is primarily stereotactic cases where objectives are in absolute volume, as there is currently no method to implement absolute volumes. This may be more the implementation of the optimization rather than the KBP. On average, use of KBP adds about 3-6 minutes for each plan, but it can save significant time in adjusting optimization objectives. One advantage is the consistency in quality that is achieved.

Conclusion: This is one of the largest series of model development using KBP. Initial model development was a long process, but subsequent models were easier as we streamlined the implementation process and developed less complicated models. KBP is implemented in the clinic and to-date find the prostate models are utilized, and the GP models less so. The GH&N model fails to obtain PTV coverage and is not used for planning. Nevertheless, all models guide towards proper OAR sparing and we believe that implementation and use will increase. A Key to model development is a consistent library of past plans. Overall, model development, testing and implementation currently results in mixed enthusiasm due to the amount of intervention required during planning.

Author Disclosure: J. Baker: None. A. Sharma: None. Y. Cao: None. B. Hamilton: None. L. Potters: None.

Jameson Baker, PhD


Send Email for Jameson Baker


TU_12_3232 - Adoption of Knowledge-Based Treatment Planning Models

Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Adoption of Knowledge-Based Treatment Planning Models