SU-E-T-510: Static and Dynamic Polynomial Correlations of Internal/external Marker Positions in Radiotherapy

Abstract

Purpose: Image guided radiotherapy offers the potential to determine the precise localization of mobile tumors in real time. This study investigates the use of polynomial curves to correlate the position of external skin surface markers with the position of the internally implanted fiducial markers. Methods: The simultaneous tracked 3D internal fiducial marker motion and 1D skin surface marker motion of 8 patients and 120 treatment fractions has been performed in this study. Three different correlation algorithms have been analyzed. The first method, the static correlation approach, only calculates the correlation once based on motion data of the first three breathing cycles and is used to calculate the internal position for the rest of the treatment fraction. The second method, the dynamic correlation approach, computes and updates the correlation in an online fashion at each end‐of‐exhale state based the motion of the most recent three breathing cycles is used to derive the correlation in an online fashion. The third method, the piecewise dynamic correlation approach, identifies the separate the breathing phases (including Exhale, End‐of‐Exhale, and Inhale) and the algorithm correlation approach is then applied for each breathing phase based on the motion signal of the corresponding phases from the most recent three breathing cycles. Results: The resulting RMS errors are 1.10mm, 1.01mm, 0.78mm, and 0.69mm for normalization only, static correlation, dynamic correlation, and piecewise dynamic correlation, respectively. Additionally, the accuracy of the correlation algorithm remains the roughly the same (<5% variation) when reducing the imaging rate from 30Hz to 15Hz and 10Hz. In most cases, the correlation was even slightly more accurate when applied to the motion data acquired at smaller imagining frequencies. Conclusions: In general, internal/external correlation based on polynomial functions results in smaller RMS errors. The dynamic correlation approaches, especially with the piecewise dynamic correlation, outperformed the static correlation.