Room: AAPM ePoster Library
Purpose: Computational phantoms comprise a digital representation of anatomy in radiation dose calculations, for diagnostic and therapeutic procedures in nuclear medicine. Reference phantoms broadly encompassing ‘population averaged’ height, weight, and internal anatomy are used for the vast majority of internal dose calculations, but do not well represent all patients. Conversely, patient-specific phantoms uniquely match the anatomy of specific patients but generally require time-consuming manual segmentation of patient tomographic images (CT or MRI) to generate. This work demonstrates the use of automated intensity-based deformable image registration to morph existing reference phantoms to conform to patient computed tomography (CT) scans, in order to provide personalized dose estimation without requiring manual organ segmentation.
Methods: A personalized anatomical atlas was created for a sample patient, derived from their CT image. First, a linear attenuation coefficient phantom was generated by resampling the PSRK-Man reference phantom onto a voxel grid and defining organ regions with corresponding Hounsfield unit (HU) reference values. The HU phantom was then co-registered with a patient CT scan using the Plastimatch B-spline deformable registration algorithm to create a patient-specific atlas-based HU image and associated atlas-based organ map. In parallel, major organs were manually contoured to generate a ‘ground truth’ patient-specific phantom. The methods were for compared for their differences in mapping and resultant dosimetry implications.
Results: Application of the derived B-spline transform to the polygon vertices comprising the PSRK-Man yielded a deformed variant more closely matching the patient’s body contour and most organ volumes as-evaluated by Hausdorff distance and Dice metrics. S-values computed for several radionuclides for the deformed phantom using the Particle and Heavy Ion Transport code System via PARaDIM agreed well with those derived from the patient-specific analog.
Conclusion: Intensity-based deformable registration could improve the accuracy of anatomical representation with mesh-type phantoms, facilitating personalized dosimetry calculations via Monte Carlo simulation.