ePoster Forums
Purpose: To estimate the clinical impact of the differences between planned and delivered dose distributions due to inter-fractional patient setup and organ deformation uncertainties using normal tissue complication probability (NTCP) modeling.
Methods: 47 prostate cancer patients were treated with IMRT using CT-on-rails (CTOR) for daily IGRT. The delivered dose distributions to bladder and rectum were calculated using a hybrid DIR algorithm (intensity- and contour-based) and were compared with the respective planned doses. For the calculation of the normal tissue complication probabilities (NTCP), the Relative Seriality model was used with the following parameter values: D50 = 81.0 Gy, γ = 0.44, s = 0.0001 for bladder and D50 = 67.0 Gy, γ = 2.5, s = 1.0 for rectum.
Results: The planned mean doses to bladder and rectum were 44.9±13.6Gy and 42.8±7.3Gy, respectively. The corresponding estimated delivered doses were 46.1±13.4Gy and 41.3±8.7Gy, respectively. For bladder, the V50 metric was 42.7±23.9% for the planned doses and 44.8±24.8% for the delivered doses. For rectum, the D10cc metric was found to be 64.1±7.6Gy for the planned doses and 60.1±9.2Gy for the delivered doses. The NTCP values of the treatment plan was 22.3±8.4% for bladder and 12.6±5.9% for rectum. The values for the estimated delivered doses were 23.2±8.4% for bladder and 9.9±8.3% for rectum, respectively. Of the 25 patients that had follow-up data, 7 showed urinary complications (28%) and 3 had rectal complications (12%).
Conclusion: The average NTCP values of the estimated delivered doses to bladder and rectum were similar and lower, respectively compared to the values calculated from the respective plans. This study indicates that an automated workflow involving a DIR based dose delivery estimation may help to achieve a better OAR sparing in the cases where higher than expected doses are delivered to those organs in the early phase of the treatment.
NTCP, Prostate Therapy, Image Guidance
TH- Response Assessment: Modeling: other than machine learning