University of Twente Student Theses


Generation of Lung CT Images Using Semantic Layouts

Wu, Sheng Chih (2021) Generation of Lung CT Images Using Semantic Layouts.

[img] PDF
Abstract:Diagnosing lung nodules is still an underlying challenge. It is tedious and the accuracy heavily relies on the experience of radiologists. To address this challenge, a computer assisted learning (CAL) system is required for training and assessment of radiologists in diagnosis with advanced computer aided diagnosis (CAD) systems. CAL systems should not only render photorealistic lung CT images, but also allow user-interactive manipulation. Previous researches utilised unconditional image synthesis for lung CT images generation, for which the users could not control the output. We propose to use spatially-adaptive denormalization (SPADE), a state-of-the-art semantic image synthesis method for synthesizing images given an input semantic layout. Experiments demonstrate the usability of SPADE as a CAL system in lung CT image synthesis, regarding visual fidelity, manipulability and variety. Moreover, the effectiveness of SPADE as a data augmentation method is also explored.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:50 technical science in general, 54 computer science
Programme:Electrical Engineering MSc (60353)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page