An automated graphical probabilistic framework for the detection of lung tumors in thoracic CT scan images

• Main Author: Δαμόπουλος, Δημήτριος
• Other Authors: Νικήτα, Κωνσταντίνα
• Format: Thesis
• Language: English
• Published: 2017
• Subjects: Lung nodule; Automatic segmentation; MRF; AdaBoost; Gabor filters; 616.994 240 757 22; Πνευμονικοί όγκοι; Αυτοματοποιημένη κατάτμηση; Φίλτρα Gabor

The subject of this thesis is the development of a automated graphical probabilistic framework to segment lung tumors from thoracic CT scan images. The feature space that we use to represent the images consists mostly of the responses of the initial images to convolution kernels of a Gabor lter bank. We are then constructing an AdaBoost strong classi er over a collection of simple decision trees. The output of our strong classi er is used to construct the unary potentials of a Markov Random Field (MRF) model, that improves our results by enforcing structure.