Program Information
Evaluation of Modified Newton Method Used for Inverting Non-Linear Transformations
P Greer1 , J Hatton2 , B McCurdy3 , P Siciarz4*, C Tang5 , (1) Calvary Mater Newcastle, Waratah, NSW, (2) Mater Hospital, Callaghan, NSW, (3) CancerCare Manitoba, Winnipeg, MB, (4) CancerCare Manitoba, Winnipeg, MB, (5) Sir Charles Gairdner Hospital, Nedlands, WA
Presentations
PO-BPC-Exhibit Hall-16 (Saturday, March 5, 2016) Room: Exhibit Hall
Purpose:The goal of this project was to evaluate the accuracy of Modified Newton Method (MNM) used for inverting forward non-linear transformations obtained through CT-CBCT deformable image registration.
Methods:Forward transformations were acquired by CT-CBCT deformable registration applying Dense Anatomical Block Matching (DABM) algorithm that was proved to be superior for CT-CBCT registration compared to well known Bspline and Demons algorithms. To the best of author knowledge it is the first study which attempts to investigate the performance of Modified Newton Method used for inverting forward transformations of DABM algorithm.
Pre-treatment CT images of five prostate patients undergoing IMRT were selected for this work. CBCT data sets were acquired in the middle of the radiotherapy treatment. Deformable registration for each algorithm was followed by the initial rigid alignment of considered images. Backward transformations were than obtained by MNM that inverted forward transformations.
The accuracy of Modified Newton Method was assessed by the application of Inverse Consistency Error (ICE). Detailed analysis consists of: ICE Maps, Normalized Distributions of ICE over the patients volume, mean value of ICE as well as the correlation between ICE and Jacobian determinants of forward transformations.
Results:Early results clearly indicates very promising performance of MNM for challenging task of inverting non-linear transformations. The most frequently received values of ICE are below 0.005mm for all five patients with mean ICE staying in the range from 0.0034mm to 0.0058mm. Also, negative correlation was found between ICE and Jacobian determinant which means that the more the transformation is preserved the smaller ICE is expected.
Conclusion:Modified Newton Method allows to accurately calculate inverse transform for DABM algorithm in the case of CT-CBCT deformable registration for patients with the prostate cancer. Observed lack of large ICE values makes the method reliable for calculations of accumulated dose delivered to patients in adaptive radiation therapy applications.
Funding Support, Disclosures, and Conflict of Interest: Cancer Care Manitoba; University of Manitoba
Contact Email: