Room: AAPM ePoster Library
Purpose: To derive a quantitative relationship between random and systematic seed placement errors and post-implant CTV dosimetry in permanent breast seed implant brachytherapy.
Methods: The CTV was modeled as a 9.2cm³ sphere, representing an average volume, expanded 10mm to create the PTV. Eight clinically-acceptable treatment plans were created. 1000 simulations were performed to model the impact of implant errors on resultant CTV coverage (V90%) for each treatment plan. Random and systematic seed placement errors were simulated by shifting needles from planned positions. Random errors were sampled from a Gaussian distribution with mean of zero. Distributions with a range of standard deviations from 1-9mm were sampled in three orthogonal directions to generate random errors of total magnitude 3-12mm for each needle. Systematic errors were simulated by shifting all needles by a magnitude 0-12mm in the same direction. Least squares regression was performed to derive second order polynomial surface equations relating magnitude of systematic and random error, to median CTV V90% coverage. Equations were compared over the range of errors at 0.1mm granularity to determine differences.
Results: A six-term equation, given by V90%(med)=95+1.9r+0.97s-0.18r²-0.064s²-0.1rs resulted in the highest R2 value (R²=0.98). In the equation, r and s are the magnitude of random and systematic errors respectively. A four-term equation was also fit (R²=0.92). The equation was given by V90%(med)=102-0.059r²-0.011s²-0.024rs. All residuals were within 4.0% and 5.6% for the six-term and four-term respectively. The average V90% difference between the fits was 0.9% (standard deviation=1.8%, 11,011 evaluations) with four-term equation overestimating CTV V90%. Absolute differences between the equations are <2% for clinically observed random and systematic errors, though differences up to 10% exist for 12mm random and systematic errors.
Conclusion: In clinical range, differences between the two fits are small. The four-term polynomial equation can be used to estimate CTV V90% coverage for most cases.
Funding Support, Disclosures, and Conflict of Interest: Michael Roumeliotis and Tyler Meyer are founders of Okolo Health, which commercialized the inverse optimization software used in this study.