Correlation between changes in morphology and dose distribution for adaptive radiation therapy in the head&neck region

Purpose/Objective(s) Morphologic changes during a treatment course cause changes in the dose distribution compared to the initial treatment planning. A way to study these changes is to transfer the initial treatment plan to repeated CT-datasets using elastic image registration. The morphology of the treated volume is subject to a number of influences: shrinking or swelling of tissues, torsions and distortions due to imperfect patient immobilization. The choice of an adequate corrective action in adaptive radiation therapy (ART) requires knowledge about the type and extent of these errors. Aim of this study is a feasibility study to answer the two questions, (1) to what extent the dose distributions for target volumes and organs at risk are changing and (2) whether it is possible to judge the type of deformation and to compile prospective predications about dosimetric changes using statistical parameters of the transformation matrices. Materials/Methods We developed software for elastic image registration. Therewith it is possible to calculate transformation vectors, representing translations and deformations of tissues between different datasets. Fifteen patients received an additional CT-scan during the 4th week of their IMRT-course. The two CT-scans were elastically registered. We used the resulting transformation matrices to adapt delineations. We parameterized the matrices for statistical evaluation. By transferring the initial treatment plan to the consecutive CT-scan the correlations between morphologic and dosimetric changes were evaluated. For the PTV and parotid glands mean dose, D2% and D98% values were analyzed, for spinal cord and brain stem maximum dose and D2%. To evaluate morphologic changes our software allows to analyze the voxel data of the calculated transformation matrices, optionally using band pass-filters, for delineated volumes. Mean and standard deviations of vector lengths were evaluated. Results Our patients had received 36Gy on average at the time of the repeated CT-scans. Mean volumetric changes at this point in time were -5% (PTV) and -17% (parotid glands), maximum -25% and -35%, respectively. Mean and D98% for the PTVs changed by +1.2% and -4.3%, respectively. A significant increase was found for Dmax to the spinal cord (+14%). The mean dose to parotid glands changed by 6%. Morphologic changes, represented by the mean vector length in the filtered transformation matrices, correlated well with the observed dosimetric changes. Conclusions The aim of ART is re-optimization of the dose plan according to changes in form, shape and location of the treated volume. In order to perform a well-directed correction depending on the type of morphologic change the proposed method constitutes a promising approach.