Intra-thoracic tumor motion estimation from CT imaging using 3D optical flow method

Purpose: To develop and validate an automated method for intra-thoracic tumor motion estimation from respiratory gated CT imaging using 3D optical flow method (OFM). Materials and Methods: A deformable image registration algorithm using 3D OFM is modified to track user delineated tumor surfaces and volumes. The location of a tumor delineated on a reference CT volume is estimated on the target CT volume from the 3D OFM calculated displacement vectors. A thoracic phantom CT image volume with generated 1.2 and 2.4 cm displacements is used for validation of motion estimates. One esophagus and one lung cancer cases with inspiration and expiration breath-hold CT image volumes are input to a 3D thin-plate spline (TPS) algorithm to generate a ficticious known displacement for validation. The 3D OFM is applied to the exhalation and inhalation CT image sets. The resulting displacement vectors are applied to the tumor surface and content providing an estimate of tumor surface and volume motion. The root mean square error (rms) is calculated for the known displacements. Results: The displacements of the phantom tumor were 1.20 and 2.40 cm with full-width at tenth maximum (FWTM) of 0.008 and 0.006 cm respectively. The maximum error of any single voxel's motion estimate is 1.1 mm or approximately one third of the z-dimension voxel size. Comparison of the 3D TPS generated and the 3D OFM calculated displacements revealed a rms error less than 0.25 mm. For the patient cases, the respiratory displacement of the esophageal tumors was non-uniform and up to 1.4 cm. The esophagus tumor motion was greatest inferiorly. A lung tumor displacement of 2.4 cm was found in the case evaluated. The lung tumor motion was not uniform and deformation was noted. Conclusion: 3D OFM was shown to provide an accurate estimation of intra-thoracic tumor motion, with estimated errors less than the voxel dimension. Surprisingly esophageal tumor motion was large.