MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Knowledge-Based Prediction of Plan Quality Metrics in Intracranial Stereotactic Treatment Planning

S Yu*, H Zhen , X Zhang , U Langner , M Dyer , M Truong , Boston University School of Medicine, Boston, MA

Presentations

(Tuesday, 7/16/2019) 1:15 PM - 1:45 PM

Room: Exhibit Hall | Forum 7

Purpose: To develop a knowledge-based model to accurately predict dosimetric indices in CyberKnife stereotactic treatment planning.

Methods: We retrospectively investigated 40 consecutive patients treated with CyberKnife stereotactic radiosurgery or radiotherapy (SRS/SRT). Prescription dose (Rx) range was 13 Gy to 30 Gy in 1-5 fractions. A linear model was built correlating the equivalent radius of the planning target volumes (req_PTV) and the equivalent radius of volume receiving various percentage of the prescription dose (req_Vi, where Vi = V10%, V20% … V120 %). This model can be used to automate the planning process by predicting the equivalent radius of volume receiving certain doses, and consequently predicting common SRS/SRT plan metrics such as gradient measure (GM) and brain V50% (For most SRS plans, 50% of Rx was 10 Gy, so V50% = V10Gy). Four-fold cross validation was performed to evaluate the predictability of the model. The accuracy of the predictions was quantified by the mean and the standard deviation of the difference between actual clinical and predicted values, i.e. ΔGM=GMclin-GMpred, ΔV50%=V50%clin-V50%pred, and a coefficient of determination, R².

Results: Mean planning target volume was 6.78 cc (range 0.02 - 74.22 cc). Mean GM was 4.46 ± 2.04 mm, and mean brain V50% was 9.50 ± 5.72 cc. For the 40 plans, the R2 for the linear fit between req_PTV and req_vi was 0.993 ± 0.010 for isodose volumes ranging from V10% to V120%; for both the V50% and V100% isodose volumes, R² was 0.998. For the four-fold cross-validation, the average prediction error for ΔGM was 0.36 ± 0.06 mm, and for ΔV50% was 0.25 ± 0.03 cc.

Conclusion: We demonstrated a simple yet highly accurate knowledge-based model that predicts CyberKnife SRS/SRT planning dosimetric indices that may be used to automate treatment planning based on PTV volumes.

Keywords

Stereotactic Radiosurgery, Treatment Planning

Taxonomy

Not Applicable / None Entered.

Contact Email