Author: Jie Liu 👨🔬
Affiliation: Inova Health System 🌍
Purpose:
The respiratory signal commonly acquired by using external surrogates or detecting the diaphragm position only represents the phase information and has poor correlation with the tumor motion amplitude. Direct tumor motion measurement is limited in clinical practice. This study proposes a method to estimate the primary tumor motion along the Superior-Inferior (SI) direction directly from daily CBCT projections.
Methods:
A first-order approximation formula was derived to estimate the real-time position of a representative marker of tumor along the SI direction based on the corresponding detected coordinates on CBCT projections. The validity of the method was tested in silico using the tumor motion data of 6 patients previously treated in the clinic, with the data extracted from tumor tracking by electromagnetic transponders implanted in the liver. The transponder traces from each treatment session were divided into segments each with 30-60 seconds length. The CBCT projections for each segment were simulated with a sampling rate of 15Hz. The simulation was repeated in various scenarios where the marker centroid relative to the isocenter is at different positions along the axial, radial or diagonal planes, with the extreme off-isocenter location at (20,20,20) cm.
Results:
With a total of 573 transponder segments and ~400,000 sampling points used in the simulations, the 5%-95% percentile estimation errors for the SI motion were within the range of (-0.2, 0.3) cm for all simulated scenarios. The estimation error mainly increases with the centroid location along the axial direction and is almost independent of the radial direction. For centroid location within 10cm along the axial direction from isocenter, the estimation error is in sub-millimeter.
Conclusion:
The proposed method shows promising accuracy in estimating the tumor motion with respiration along the SI direction based on CBCT projections. Further validation may require phantom measurements with benchmark data in clinical settings.