In medical imaging, model observers are used to define a new method of task based image quality assessment. In this thesis a novel search algorithm is presented that detects possible lesions in a...Show moreIn medical imaging, model observers are used to define a new method of task based image quality assessment. In this thesis a novel search algorithm is presented that detects possible lesions in a digital anthropomorphic 2D and 3D lung phantoms and defines the detectability of the candidate lesions using a non-prewhitening matched filter with an eye filter (NPWE) model observer. Sets of phantom images were simulated for a range of noise levels and two types of noise (Gaussian white noise and CT-like noise). The candidate lesions were classified as true positives and false positives. A proof of concept study showed promising results in the detectability trends the search algorithm described. The trends showed that with increasing noise levels the detectability of true positives decreased. When comparing the detectability indexes of the true positives and false positives, the differences between them became smaller for increasing noise levels. In the future, the algorithm can be applied to the analysis of real CT scans of a lung phantom containing lesions, and used to obtain Free Response Operating Characterisic (FROC) curves.Show less