Room: AAPM ePoster Library
Purpose: We investigate the feasibility of using simultaneously acquired megavoltage (MV) and kilovoltage (kV) images to determine spine target position on a standard LINAC.
Methods: Experiment was performed with an anthropomorphic body phantom on an Elekta LINAC. Small couch shifts (~4 mm) were introduced along lateral, longitudinal and vertical directions. Pre- and post-shift CBCTs were obtained and registered to the reference CT to determine the ground-truth shifts. Post-shift MV and kV images were acquired using the electronic portal imaging device (EPID, 6 MV, 10×10 cm field size, 4 MUs) and the gantry-mounted kV imaging system (120 kV and 6.4 mAs). Each paired orthogonal MV/kV images were registered to their corresponding kV projections of pre-shift CBCT to determine 2D shifts. The following image processing steps were applied before image registration: 1) center corrections for both MV and kV images to compensate source/detector shift related to gantry sag; 2) image filtering to MV image to reduce noise and enhance contrast; 3) histogram matching to improve image similarity between MV and reference kV projection. The 2D shifts for each MV/kV pair were then used to triangulate 3D shifts. Residual registration errors, defined by the differences between the ground-truth and the calculated shifts, were used to assess the accuracy of this approach.
Results: For a total of 24 MV/kV image pairs (8 gantry angles and 3 spine sites, i.e., cervical, thoracic, and lumbar), the mean and standard deviation of residual errors were 0.23 ± 0.16 mm for lateral (range 0.01–0.54 mm), 0.55 ± 0.10 for longitudinal (0.39–0.71) and 0.31 ± 0.26 for the vertical direction (0.02–0.99).
Conclusion: Our results indicate that it is possible to achieve sub-millimeter accuracy using combined MV/kV imaging to determine spine position. It could be a simple alternative to mid-treatment CBCT for intra-fraction target verification during spine SBRT delivery.
Not Applicable / None Entered.
Not Applicable / None Entered.