Image Processing for Quantitative Scintillation-Camera Imaging. Application to Radionuclide Therapy

Abstract: Individual-based determinations of the absorbed dose in radionuclide therapy largely rely on absolute measurement of the activity distribution and its redistribution over time. Scintillation-camera imaging is the most commonly employed measuring technique, applied in planar or SPECT mode, sometimes in combination with structural images from CT. In this thesis, methods for processing and analysis of scintillation-camera images have been studied, with the intent to compensate for the physical limitations involved in the imaging process and thereby improve the quantification accuracy. In particular, methods for image registration and segmentation have been developed. Evaluation of the earlier SPECT based quantification program revealed limitations for the measurement of small volumes, their activity and contrast, in situations where the object contrast deviated from 100%. A 3D deformable segmentation method was developed and applied for SPECT images. In preliminary evaluations, this method gave an adequate boundary, also in situations with low contrast and high levels of noise. For registration of SPECT and CT images, a method was developed where the possibility to use images acquired in the Compton scatter region was explored. The registration accuracy was obtained to about 10 mm, where the largest deviations occurred for slices with heterogeneous activity distributions. To achieve automation, this method was further developed by applying mutual information as similarity measure and spatial transformations in 3D. This registration method was applied for activity quantification and absorbed dose calculations in 3D, using the combined information from CT and SPECT. For planar image activity quantification, a 2D method for registration of whole-body emission and transmission images was developed, including a tailored spatial transformation. The accuracy was obtained to below pixel level (< 3.6 mm) for simulated images, to 9 mm in comparison to point markers for patients. A quantification method based on conjugate scintillation-camera images was also developed, in which a registered CT image was used for attenuation correction, background compensation and segmentation of organs. Evaluation using Monte Carlo simulated images showed an accuracy of within 10% for organ activity quantification. Throughout this thesis, the major application has been images acquired in association with radionuclide therapy using an 131I labeled monoclonal antibody.