A Comparative Evaluation of Deformable Image Registration Algorithms Based On the Jacobian
E Li1,2, Z Zhong1,2, Y An1,2, S Huang1, W Zheng1,3, J Lian1,4, X Yang1*, (1) Sun Yat-sen University Cancer Center; State Key Laboratory of Oncology in South China; Collaborative Innovation Center for Cancer Medicine; Guangdong Key Laboratory of Nasopharyngeal Carcinoma Diagnosis and Therapy, Guangzhou, Guangdong province, 510060, China,(2) Guangzhou Xinhua College, Guangzhou, Guangdong, 510520, China,(3) Department of Radiation Oncology, Southern Theater Air Force Hospital of the People's Liberation Army, Guangzhou, Guangdong province, 510050, China,(4) Department of Radiation Oncology, The First Affiliated Hospital of Guangzhou University of Traditional Chinese Medicine, Guangzhou, Guangdong province, 510405, China
Presentations
PO-GePV-M-175 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: To effectively quantify the evaluation of the registration quality of various deformable image registration (DIR) algorithms based on the Jacobian.
Methods: We proposed two new indexes for DIR: Symmetric Jacobian Error (SJE), Positive/Inverse Jacobian Negative Percentages(P/IJNP). 62 lung cancer patients and 26 nasopharynx cancer patients were randomly selected for the comparative evaluation on the propose two new indexes. Based on rigorous convergence analysis, we designed a control mechanism for modulating the number of times of the current iterate, via aiming at the smooth constraints and quantitative analyzed of the DVF of the algorithm. The results of bidirectional voxel mapping provided by inverse consistent registration were also analyzed by visualization and inverse consistency error. The performance of different DIR algorithms was then evaluated based on the grid Jacobian.
Results: Different algorithms have different sensitivities to different types of cancer on the two new indexes. The SJE for lung cancer is about 3.6% smaller than the nasopharynx cancer in Optical Flow (OF) algorithm. And the comparison between OF algorithm and Demon algorithm shows that the error of change in lung cancer is about 5.54%, but in nasopharynx cancer is almost the same. The average smooth perturbation rate with and without inverse DVF changes by 62.85%. The smoothness of the OF algorithm increases after the modulation iteration and Gauss low-pass filter. The values of P/IJNP can assist the calculation of fold percentage of DVF in SJE.
Conclusion: The proposed indexes are more discriminating and have the preliminary screening effect to the suitable algorithm for different kinds of cancer. Compared by a series of baseline, we find out that OF algorithm is more suitable for lung cancer.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email