SU‐FF‐J‐98: A Feature Matching Approach for the Automatic Correlation of Internal and External Motion in Lung Tumors

Purpose: To develop an automatic procedure to correlate the motion of internal features in CT with external RPM© (Real‐time Position Management) signal. Method and Materials: A cross‐correlation based algorithm was implemented to perform template matching for detecting anatomical structures in cine‐mode CT data acquired for 4D‐CT. The method was used to track the spatial movement of pulmonary bifurcations in 4 patients over short durations in time. For each subject, 12 points, equally distributed from top to bottom in both lungs, were selected. A graphical interface was developed to allow users to navigate through the un‐binned CTimages, acquired in cine‐mode with a 4‐slice scanner. Interactive methods were developed for temporal image browsing, template selection, definition of couch positions to be searched, and verification of results. The cross‐correlation matrix was computed for each slice in the search volume. To increase the resolution along the superior‐inferior (SI) direction, a 2nd order spline was computed to interpolate the Z position in the 4 slices of each cine‐mode chunk. The detected internal movement of features in 3D was then retrospectively synchronized with the RPM signal, and the correlation index R2 was computed. Results: Peak‐to‐peak values of feature motion, along the SI direction, ranged from 0.83 mm (upper lung) to 25.33 mm (lower lung). Some patient exhibited relevant motion also in the latero‐lateral (10.60 mm) and anterior‐posterior (12.22 mm) directions. The median±quartile of R2 in SI direction was 0.89±0.09. No statistical difference was found between upper and lower lung (0.89±0.10 vs 0.89±0.08). Conclusion: The developed automatic procedure allowed a fast analysis for external/internal correlation of lunganatomy. Such a study is particularly relevant in IGRT(Image Guided Radiation Therapy) since most techniques rely on external fiducial monitoring to assess motion.