Volumetric Quantification of Metastatic Burden from SPECT/CT Images

Sammanfattning: Bone scintigraphy images are used to investigate the presence of metastases in patients with prostate cancer. An analysis of these images indicates the proportion of cancer in the bones; the Bone Scan Index (BSI). This value is an important prognostic factor used to predict the future for the patients. It is common to extend the examination with a combination of SPECT and CT images to be able to study possible metastases in more detail. The aim of this thesis is to integrate the information from SPECT and CT images to get a more accurate calculation of the BSI. The strategy has been to use the analysis of the bone scintigraphy images as an initialization of a segmentation of metastases in the SPECT images. The CT images are used to produce a simple segmentation of the bones. The problem is divided into two main parts; image registration and segmentation of metastases. The image registration is needed to align the coordinate system of the bone scintigraphy images and the SPECT images. The Morphon method has been chosen and the results are good; twelve of fifteen tested registrations are classified as successful. A combination of a more robust method to find start guesses and a more generous transformation model would probably improve the results even further. Seeded region growing has been chosen as the segmentation algorithm. An implementation with an automatic termination criterion has been created and the results seem promising. The conclusions are that the Morphon method works fine for registration of the images and that the strategy of initializing a segmentation gives good results. In this work, the size of the segmentation was not used to update the BSI but the results can still be useful. Another goal was to facilitate navigation between different types of images by aligning them to one coordinate system. A user interface has been created to reduce the amount of time spent by doctors and biomedical scientists to navigate through the images searching for possible metastases.