DVH Distortion in Deformably Propagated Dose Grids - a Practical Challenge for Adaptive Radiotherapy
J Kipritidis1*, A Quinn1, T Morgas2, S Kuckertz3, N Papenberg3, S Heldmann3, T Coradi2, F Franco2, J Kieselmann2, C Huang2, J Booth1, (1) Northern Sydney Cancer Centre, Sydney, NSW, AU, (2) Varian Medical Systems, Palo Alto, CA, (3) Fraunhofer MEVIS, Lubeck, DE
Presentations
PO-GePV-M-119 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: Dose accumulation driven by deformable image registration (DIR) is a clinical reality. However a fundamental limitation is that “plan of the day” DVHs may suffer distortion when dose is deformably propagated to another reference frame. This work demonstrates the magnitude of DVH distortions for three “classic” adaptive use cases and investigates correlations with established DIR quality metrics from AAPM TG-132.
Methods: Three radiotherapy patients (Bladder 55Gy/20Fx, Rt Tonsil 63Gy/30Fx, Rt Lung 60Gy/30Fx), underwent a simulation CT and mid-treatment re-simulation CT for dose evaluation. In each case, dose propagation between the resimulation / simulation CT was emulated using a prototype deformable registration system and the displacement fields analysed. Dice Similarity Coefficients (DSC) and negative Jacobian percentages were computed for each registered structure pair and correlated against DVH distortion, quantified as the root mean square difference (“RMS%ΔDVH”) between DVHs calculated using the resimulation dose with resimulation CT structures, versus DVHs calculated using propagated dose with simulation CT structures.
Results: The number of registered target/OAR structure pairs was N=9 for the Bladder case, 16 for the Tonsil case, and 12 for the Lung case. TG-132 metrics indicated moderate DIR performance, with average DSC / negative Jacobian values of 0.86/0% for the Bladder case, 0.79/2.6% for the Tonsil case, and 0.92/0% for the Lung case. The respective RMS%ΔDVH distributions had Mean(Min, Max) values 3.6(0.7, 11.3)%, 3.3(0.6, 12.1)% and 2.1(0.3, 6.5)%. For each patient, the maximum DVH distortion affected key OARs adjacent the target (Sigmoid, Parotid, and Carina). Application of unpaired t-tests showed predominantly non-significant (p>0.05) correlations of RMS%ΔDVH versus DSC / Jacobian metrics on a per-patient basis.
Conclusion: Deformable dose propagation can produce DVH distortions that are potentially uncorrelated with established DIR quality metrics. This issue warrants further investigation, as it could lead to mistaken adaptation decisions or perpetuate uncertainties in dose-response models.
Funding Support, Disclosures, and Conflict of Interest: Northern Sydney Cancer Center has a funded collaborative research agreement with Varian Medical Systems.
Keywords
Deformation, Registration, Dose Volume Histograms
Taxonomy
TH- External Beam- Photons: Dose reconstruction over deforming anatomies
Contact Email