Accuracy quantification of deformable image registration on megavoltage computed tomography images for nasopharyngeal carcinoma adaptive radiotherapy applications

ความถูกต้องของการซ้อนทับภาพชนิดเปลี่ยนรูปร่างของภาพเอกซเรย์คอมพิวเตอร์ระดับเมกะโวลต์เพื่อใช้ประโยชน์ในการปรับเปลี่ยนแผนรังสีรักษาผู้ป่วยมะเร็งหลังโพรงจมูก

Name: Wannapha Nobnop

ThaSH: Nasal fossa -- Cancer

Classification :.DDC: 616.99

ThaSH: X-rays

Abstract: Helical megavoltage computed tomography (MVCT), which is a volumetric imaging modality of Helical Tomotherapy unit, is adopted with the primary purpose of more accurate target localization. Moreover, the information about inter-fractional anatomical variations have become more accessible. Deformable image registration (DIR) is the keystone for modifying structures according to anatomical changes for observing dosimetric effect. However, the ability to resolve soft-tissue contrast differences in MVCT is fundamentally limited by the number of photons used to create the image. These issues may be behind the degradation of deformable registration accuracy. Therefore, the two purposes of the study were to quantify the performance of multiple deformable image registration (DIR) methods on MVCT images assessed in a phantom and nasopharyngeal carcinoma (NPC) patients and to evaluate their impact for estimating the dose accumulation in target and normal organs for nasopharyngeal carcinoma patients. Regarding the DIR accuracy on MVCT images quantification, the MVCT images of the phantom are observed using twelve shapes of acrylic and tissue equivalent material for the rigid and the non-rigid volume changes for known offset investigation, the 1st and i the 20th fractions of the MVCT images of five NPC patients were used to evaluate the DIR application for unknown offset investigation. The eight DIR methods were employed using the DIRART software. The accuracy levels of the methods were compared using an intensity-based, volume-based, and deformation field analysis. Moreover, the DIR on kVCT images were compared to assess the accuracy. As regards the impact of DIR methods on dose accumulation, the 1st day MVCT and weekly MVCT images at the 1st, 6th, 11th, 16th, 21st, 26th, and 31st fractions of five NPC patients were used to assess the dose accumulation by the first three DIR methods. The weekly registration accuracy and the cumulative dose deviations from the initial treatment plan were analyzed, and correlations of these variables with the accuracy of DIR were explored. The accuracy of DIR on the MVCT images was concordant between known and unknown offset investigation. The Horn and Schunck optical flow mostly showed better performance for known and unknown offset investigation. The symmetric with Horn and Schunck in the forward mapping (SymHSFW) showed the best agreement for rigid volume change with the dice similarity coefficient (DSC) = 0.899 ± 0.03 for phantom and 0.792 ± 0.07 for NPC patient images. Whereas, the asymmetric with Demon in the forward mapping (AsyDMFW) showed the best performance for non-rigid volume changes with a mean of DSC = 0.812 ± 0.07 for NPC patient images. A comparison of DIR accuracy between kVCT and MVCT images revealed that they were not significantly different, based on intensity, volume, and physical characteristics of the deformation field. The best three DIR methods were the SymHSFW, AsyDMFW and the asymmetric with Horn and Schunck in the forward mapping (AsyHSFW). As regards the dose accumulation, the target dose received a slightly different result, by less than 0.5% from the initial plan at the end of the treatment. The normal organ dose differences increased as the treatment progressed to 6.8% (range: 2.2 to 10.9%), 15.2% (range: -1.7 to 36.3%), and 6.4% (range: -1.6 to 13.2%) for the right parotid, the left parotid, and the spinal cord, respectively. There was concordant between the accuracy of structure deformation and discrepancy of dose accumulation. The DIR method that j yielded the highest DSC value was considered the best method for dose accumulation with the lowest variation from the reference dose. The mean uncertainty values to estimate the accumulated doses for all the DIR methods were 0.21 ± 0.11 Gy (target dose), 1.99 ± 0.76 Gy (right parotid), 1.19 ± 0.26 Gy (left parotid), and 0.41 ± 0.04 Gy (spinal cord). The accuracy of DIR on MVCT images showed clinically useful methods for adaptive applications based on the intensity-based, volume-based, and deformation field analysis. The accuracy of the DIR methods affects the estimation of dose accumulation on both the target dose and the normal organ dose. The DIR methods on MVCT images provide an adequate dose estimation technique for observation as a result of interfractional anatomic changes and are beneficial for adaptive treatment strategies.

Chiang Mai University Library

Address: CHIANG MAI

Email: cmulibref@cmu.ac.th

Name: Taweap Sanghangthum

Role: Chairman

Name: Hudsaleark Neamin

Role: Member

Name: Imjai Chitapanarux

Role: Member

Name: Somsak Wanwilairat

Role: Member

Name: Vicharn Lorvidhaya

Role: Member

Name: Juntima Euathrongchit

Role: Member

Created: 2017

Modified: 2020-01-05

Issued: 2020-01-05

วิทยานิพนธ์/Thesis

application/pdf

CallNumber: Th 616.99 W249A

eng

DegreeName: Doctor of Philosophy

Level: Master's Degree

Descipline: Biomedical Science

Grantor: Chiang Mai University

©copyrights Chiang Mai University

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	medbio80717wnbnp_full.pdf	3.85 MB	4	2023-05-05 17:15:58

ใช้เวลา

0.025277 วินาที