Matching Accuracy Comparison Between Five Different Deformable Image Registration Algorithms
M Horita1*, N Hayashi2, I Kato 3, D Yamanaka1, K Yasui2, A Takemura4, (1)Graduate School of Health Sciences, Fujita Health University, Tayoake, Aichi, JP,(2)School of Medical Sciences, Fujita Health University, Tayoake, Aichi, JP, (3)Department of Medical Technology, Ogaki Municipal Hospital, Ogaki, Gifu, JP, (4) Faculty of Health Sciences, Institute of Medical, Pharmaceutical and Health Sciences, Kanazawa University, Kanazawa, Ishikawa, JP,
Presentations
PO-GePV-M-172 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: The deformable image registration (DIR) technique is useful for adaptive radiotherapy. However, there are few reports of detailed comparative analysis of DIR algorithm use in clinical patient models. The purpose of this study was to compare and analyze the impact of matching accuracy among four open-to-public and one commercially available DIR algorithm.
Methods: Four DIR algorithms, including Horn-Schnuck Optical Flow (HS), Combined HS & LK Optical Flow (HSLK), Iterative Optical Flow (IOF), and Free Form Deformation (FFD) in DIRART software and a DIR algorithm in MiradaRTx software (Mirada) were compared by performing DIR in open-source clinical patient models, which included six head and neck and five each of pelvis, abdomen, and chest models. Whole-image DIR was performed between the planning CT as the target image and the follow-up CT a few days later as the deformed image in all clinical model data. The Dice similarity coefficient (DSC) and Hausdorff distance (HD) were evaluated.
Results: There was no significant statistical difference among the five different algorithms in each specific organ, including the spinal cord, bladder, rectum, right and left kidneys, and left and right lungs. DSC values of Mirada were higher than those of the other four algorithms in all patient models; however, its HD values were lower, except in patients 4 and 5. For the IOF algorithm, even though the DSC values were poorer than those of the other four algorithms, the HD values were similar.
Conclusion: Matching accuracy among the five different DIR algorithms was quantitatively analyzed in this study. The DIR algorithm in Mirada indicated good matching accuracy compared to other algorithms, but the tendency was not always demonstrated in several organs on pelvis lesions.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email