SU-GG-J-13: A New Self-Consistent Inverse Deformation Field Generator and Its Applications
Changshun Yan, Hangyu Zhong, M Murphy, Elisabeth Weiss, Jeffrey V. Siebers · Medical Physics · 2008
Purpose: To provide a tool to map image intensity, dose, and contours back-and-forth consistently between two images. Method and Materials: A new concept, a self-consistent Inverse Displacement Vector Field (IDVF) is defined and an algorithm to generate the IDVF is implemented. A deformable image registration (DIR) algorithm was chosen to generate a DVFA-B between reference image A and study image B. The same DIR algorithm is also used to generate DVFB-A and our IDVF generator is used to create IDVFB-A, which is consistent with DVFA-B. To test whether DVFB-A or IDVFB-A better maps information from A-B back to A for a lung 4D CT dataset, an arbitrary set of points SA is chosen and are warped to SB by using DVFA-B. DVFB-A is used to warp SB back to A to get . Similarly, IDVFB-A is used to warp SB back to A to get . For each point in (or ), we know its original point in SA. Hence, one can calculate the Euclidean distance between each point in (or ) and its original point in SA to quantify the error in the back-and-forth image warping. Similarly, dose on A is warped to B and back to A. Results: The IDVF warped points from study back to reference with maximum error about 10 times smaller than using DVFB-A. Similarly, the IDVFB-A dose maximum warping error is about 2 times smaller than using DVFB-A. Conclusion: This work shows that it is possible to generate self-consistent IDVF with small mapping error. This IDVF is useful to warp contours and dose for IGART. (Work supported by NIH P01CA116602).