Image analysis methods for brain tumor treatment follow-up

Official URL: http://risc01.sabanciuniv.edu/record=b1534419 (Table of Contents)

Assessment of the progression of the tumors in current clinical practice is based on maximum diameter measurements, which are related to the volumetric changes. With the advent of the spatially localized radiotherapy techniques (i.e. Cyberknife, IMRT, Gammaknife, Tomotherapy) not only the volumes of the tumors but also the geometric changes need to be considered to measure the effectiveness and to improve the applied therapy. In this thesis, image analysis techniques are developed for assessment of the changes of the tumor geometry between MRI volumes acquired after and before the therapy. Three main parts of the thesis are: Segmentation of brain tumors on MRI; change quantification in temporal MRI series of brain tumors; and deformable registration of brain MRI volumes with tumors. The results obtained by the developed semi-automatic brain tumor segmentation method, Tumor-cut, are comparable with those of state-of-the-art techniques in the field. The quantification of tumor evolution using the invariants of the Lagrange strain tensor provide measures that are more correlated with the clinical outcome than the volumetric measures. The deformable registration of longitudinal data provides a novel framework to study brain deformations, in vivo, and more accurate assessment of the changes.