Encrypted login | home

Program Information

A Statistical-Based Treatment Plan Prediction Method

no image available
J Lu

J Lu*, W Hu , J Wang , J Fan , G Qing , L HUANG , H Ying , Fudan University Shanghai Cancer center, Shanghai, Shanghai


SU-I-GPD-T-365 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: To develop an automated IMRT treatment plan dose distribution and DVH prediction workflow.

Methods: Binary data from pinnacle(n=20) was used to train the dataset and build a kernel density model (KDM) according to the relationship between doses and related spatial parameters. To predict a treatment plan for the new patient, the contours approved by the physicians were used in the KDM. The KDM will analysis the distances between the target and organs at risks and predict the dose distribution and DVHs. The dosimetric comparisons between manual and predicted DVH were performed on 5 patients. For the test purpose, we evaluated our workflow in the rectal cancers.

Results: The mean dose of left femoral for manually optimized plans was 19.9 Gy (SD 5.8), and was 21.5 Gy (SD 5.6) for predicted plans. The V(30 Gy) point-wise value was 0.15 (SD 0.18) for manual plans, and was 0.17 (SD 0.20) for predicted ones.

Conclusion: The estimated plan was highly close to the manually made ones and is possible to be used as a reference for optimizing treatment plans. Reduction of plan variance and improvement of time efficiency can be achieved. Further work needs to be done to improve the accuracy and generalize our method to other tumor types.

Contact Email: