Patient-specific dosimetry for intracavitary super(32)P-chromic phosphate colloid therapy of cystic brain tumours

Purpose: super(32)P-chromic phosphate colloid treatments of astrocytoma and craniopharyngioma cystic brain tumours in paediatric patients are conventionally based on a sphere model under the assumption of uniform uptake. The aims of this study were to determine the distribution of the absorbed dose delivered by super(32)P on a patient-specific basis and to evaluate the accuracy with which this can be predicted from a pretherapy administration of super(99m)Tc-Sn colloid. Methods: Three patients were treated with super(32)P-chromic phosphate colloid following super(99m)Tc-Sn colloid administrations. Convolution dosimetry was performed using pretherapy and posttherapy sequential SPECT imaging, and verified with EGSnrc Monte Carlo radiation transport simulations. Mean absorbed doses to the cyst wall and dose-volume histograms were also calculated and compared with those obtained by the sphere model approach. Results: Highly nonuniform uptake distributions of both the super(99m)Tc and super(32)P colloids were observed and characterized by dose-volume histograms to the cyst wall. Mean absorbed doses delivered to the cyst wall, obtained with the convolution method, were on average 21 % (SD 18 %) and 50 % (SD 30 %) lower than those predicted by the super(99m)Tc distribution and the uniform assumption of the sphere model, respectively. Conclusion: Absorbed doses delivered to the cyst wall by super(32)P are more accurately predicted from image-based patient-specific convolution dosimetry than from simple sphere models. These results indicate the necessity to perform personalized treatment planning and verification for intracavitary irradiation of cystic brain tumours treated with radiocolloids. Patient-specific dosimetry can be used to guide the frequency and levels of repeated administrations and would facilitate data collection and comparison to support the multicentre trials necessary to progress this therapy.