8. Validation of a "Demons" intensity-based deformable image registration algorithm using numerical phantoms and CT-patients. (December 2016)

• Record Type: Journal Article
• Authors: Badey, A.; Belkacemi, R.; Bodez, V.; Khamphan, C.; Jaegle, E.; Alayrach, M.E.; Martinez, P.; Alfonsi, M.; Garcia, R.
• Is Part Of: Physica medica. Volume 32(2016)Supplement 4
• Page Start: 345
• Publication Date: 2016-12
• Subjects: Adaptive radiotherapy -- Deformable image registration -- Numerical phantoms Head and neck -- Cone-beam CT; Medical physics -- Periodicals; Biophysics -- Periodicals; Biophysics -- Periodicals; Imagerie médicale -- Périodiques; Radiothérapie -- Périodiques; Rayons X -- Sécurité -- Mesures -- Périodiques; Physique -- Périodiques; Médecine -- Périodiques; 610.153
• DOI: 10.1016/j.ejmp.2016.11.059
• Languages: English
• ISSNs: 1120-1797

Abstract

Abstract : Introduction: Adaptive Radiotherapy (ART) is a technique which minimizes the dosimetric impact of anatomical changes that may occur during treatment. One of the milestones of ART is deformable image registration (DIR) between original planning images and images acquired during treatment. The aim of this study is to propose a method to evaluate the performances and validate the clinical use of a "Demons" intensity-based DIR algorithm implemented in SmartAdapt® v.13.5 (Varian Medical Systems, USA). Methods: Two types of dataset were used to perform the evaluation: numerical phantoms and CT-CBCT images of 5 patients with head and neck (H&N) cancer. IMSIMQA™ software (OSL, UK) provides a numerical phantoms library and the opportunity to generate deformation. Deformation scenarios on simple volumes were created and two sets of images were systematically obtained for each induced deformation: initialCT and deformedCT. The contour comparison was made between distorted contours obtained by SmartAdapt® and expert contours on deformedCT. Regarding patients images, the purpose was to assess DIR relevance with anatomical changes. For each H&N case, three images sets (CTD0, CBCTD16, CTre-CT) were used, and contours deformation was evaluated qualitatively and quantitatively. Analytical tools such as the Dice similarity index were used. A qualitative evaluation of the contours was performed by a senior radiation oncologist using a specific scale. Contours were corrected if necessary. An estimate of potential time-saving was also done. Results: For numerical phantoms, average Dice values are 0.84 (±2 σ = 0.13) for tested deformation scenarios. For CT-patients, average Dice values are 0.75 (min = 0.64; max = 0.92) and 0.73 (min = 0.65; max = 0.9) for registrations CTD0-CTre-CT and CTD0-CBCTD16 respectively. The dispersion of the values is mainly due to organs shape and algorithm choice. No significant variation is obtained between DIR for single-modality (CT–CT) and multi-modality (CT–CBCT). Moreover, contours propagation reduces significantly physicians required time for contouring patient images. Conclusions: Given the results obtained with numerical phantoms and CT-patients, a clinical use appears possible. SmartAdapt® enables significant time-saving compared to manual re-contouring for re-planning strategies. However, a careful review by a specialist for DIR-generated contours is essential for clinical cases.