TY - JOUR
T1 - Biomechanical modeling of radiation dose-induced volumetric changes of the parotid glands for deformable image registration
AU - McCulloch, Molly M.
AU - Cazoulat, Guillaume
AU - Ford, Alexandra C.
AU - Elgohari, Baher
AU - Bahig, Houda
AU - Kim, Andrew D.
AU - Elhalawani, Hesham
AU - He, Renjie
AU - Wang, Jihong
AU - Ding, Yao
AU - Mohamed, Abdallah S.R.
AU - Polan, Daniel F.
AU - King, Jason B.
AU - Peterson, Christine B.
AU - Ohrt, Andrea N.
AU - Fuller, Clifton D.
AU - Lai, Stephen Y.
AU - Brock, Kristy K.
N1 - Publisher Copyright:
© 2020 Institute of Physics and Engineering in Medicine.
PY - 2020/8/21
Y1 - 2020/8/21
N2 - Purpose. Early animal studies suggest that parotid gland (PG) toxicity prediction could be improved by an accurate estimation of the radiation dose to sub-regions of the PG. Translation to clinical investigation requires voxel-level dose accumulation in this organ that responds volumetrically throughout treatment. To date, deformable image registration (DIR) has been evaluated for the PG using only surface alignment. We sought to develop and evaluate an advanced DIR technique capable of modeling these complex PG volume changes over the course of radiation therapy. Materials and methods. Planning and mid-treatment magnetic resonance images from 19 patients and computed tomography images from nine patients who underwent radiation therapy for head and neck cancer were retrospectively evaluated. A finite element model (FEM)-based DIR algorithm was applied between the corresponding pairs of images, based on boundary conditions on the PG surfaces only (Morfeus-spatial). To investigate an anticipated improvement in accuracy, we added a population model-based thermal expansion coefficient to simulate the dose distribution effect on the volume change inside the glands (Morfeus-spatialDose). The model accuracy was quantified using target registration error for magnetic resonance images, where corresponding anatomical landmarks could be identified. The potential clinical impact was evaluated using differences in mean dose, median dose, D98, and D50 of the PGs. Results. In the magnetic resonance images, the mean (±standard deviation) target registration error significantly reduced by 0.25 ± 0.38 mm (p = 0.01) when using Morfeus-spatialDose instead of Morfeus-spatial. In the computed tomography images, differences in the mean dose, median dose, D98, and D50 of the PGs reached 2.9 ± 0.8, 3.8, 4.1, and 3.8 Gy, respectively, between Morfeus-spatial and Morfeus-spatialDose. Conclusion. Differences between Morfeus-spatial and Morfeus-spatialDose may be impactful when considering high-dose gradients of radiation in the PGs. The proposed DIR model can allow more accurate PG alignment than the standard model and improve dose estimation and toxicity prediction modeling.
AB - Purpose. Early animal studies suggest that parotid gland (PG) toxicity prediction could be improved by an accurate estimation of the radiation dose to sub-regions of the PG. Translation to clinical investigation requires voxel-level dose accumulation in this organ that responds volumetrically throughout treatment. To date, deformable image registration (DIR) has been evaluated for the PG using only surface alignment. We sought to develop and evaluate an advanced DIR technique capable of modeling these complex PG volume changes over the course of radiation therapy. Materials and methods. Planning and mid-treatment magnetic resonance images from 19 patients and computed tomography images from nine patients who underwent radiation therapy for head and neck cancer were retrospectively evaluated. A finite element model (FEM)-based DIR algorithm was applied between the corresponding pairs of images, based on boundary conditions on the PG surfaces only (Morfeus-spatial). To investigate an anticipated improvement in accuracy, we added a population model-based thermal expansion coefficient to simulate the dose distribution effect on the volume change inside the glands (Morfeus-spatialDose). The model accuracy was quantified using target registration error for magnetic resonance images, where corresponding anatomical landmarks could be identified. The potential clinical impact was evaluated using differences in mean dose, median dose, D98, and D50 of the PGs. Results. In the magnetic resonance images, the mean (±standard deviation) target registration error significantly reduced by 0.25 ± 0.38 mm (p = 0.01) when using Morfeus-spatialDose instead of Morfeus-spatial. In the computed tomography images, differences in the mean dose, median dose, D98, and D50 of the PGs reached 2.9 ± 0.8, 3.8, 4.1, and 3.8 Gy, respectively, between Morfeus-spatial and Morfeus-spatialDose. Conclusion. Differences between Morfeus-spatial and Morfeus-spatialDose may be impactful when considering high-dose gradients of radiation in the PGs. The proposed DIR model can allow more accurate PG alignment than the standard model and improve dose estimation and toxicity prediction modeling.
KW - biomechanical modeling
KW - deformable image registration
KW - dose-response
KW - finite element modeling
KW - head and neck cancer
UR - http://www.scopus.com/inward/record.url?scp=85091083003&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85091083003&partnerID=8YFLogxK
U2 - 10.1088/1361-6560/ab8bf1
DO - 10.1088/1361-6560/ab8bf1
M3 - Article
C2 - 32320955
AN - SCOPUS:85091083003
SN - 0031-9155
VL - 65
JO - Physics in medicine and biology
JF - Physics in medicine and biology
IS - 16
M1 - 165017
ER -