Room: Exhibit Hall | Forum 3
Purpose: Stereotactic body radiotherapy (SBRT) delivers high biological effective dose and is successful at controlling certain diseases. The technique can require more resources than traditional treatments, specifically with regards to treatment planning time. This study aims to evaluate the performance of automated, iterative, patient-specific spine SBRT planning routine.
Methods: Spine SBRT plans were extracted from the clinical database (78 Raystation and 142 Eclipse). All plan processing and optimization was performed using the Raystation Python API. Dosimetric quality metrics (QM) optimized with this routine were dose to 1%, 5% and 10% of the spinal cord and the two closest organs-at-risk (OARs) to the treated target. Due to the varying disease presentation and physician intent, the original target coverage and overall maximum dose constraints were used in the planning process. Cord dose limits were set using values from high-quality plans. The doses were made conformal with constraints using the specific target volume and doses using a â€˜first principlesâ€™ approach. Doses to the two nearest OARs were then further reduced based on the planning system dose distribution. Optimizations were run after each set of dose limits were applied. No optimization structures were used in this presented model and the technique is generalizable to any planning system. Performance was indicated by the difference between the QMs: dQM=QM(clinical)-QM(autoplan).
Results: The dQM using a testing set of plans showed dose reduction for all investigated metrics: dCord_1%=66.7Â±307.9, dCord_5%=180.4Â±337.9, dCord_10%=244.2Â±367.3, dOAR1_1%=87.4Â±252.6, dOAR1_5%=116.5Â±250.4, dOAR1_10%=149.1Â±248.1, dOAR2_1%=97.5Â±282.1, dOAR2_5%=83.3Â±277.8 and dOAR2_10%=75.0Â±266.2. Additionally, the planning process was completed in approximately 20 minutes.
Conclusion: These results show spine SBRT is amenable to auto-planning routines. An effective and robust model was created from the data. With the accrual of more data and using varying approaches a more sophisticated model is possible. In addition, input from physicians will be crucial to constructing clinically reliable treatment plans.