Thesis Defense Schedule


Winter 2017


Wednesday, March 1

Maitham Naeemi

Chair: Dr. Sohini Roy Chowdhury
Candidate: Master of Science in Electrical Engineering

9:00 AM; DISC 464
Investigation of Data Modeling Strategies for Quantification of CT Image Quality

CT imaging is an increasingly vital diagnostic tool, and its growing use raises the risk of patient exposure to harmful radiation. This risk has prompted technicians and radiologists to balance image quality against radiation exposure levels, in order to maintain exposure risks As Low As Reasonably Achievable (ALARA) while ensuring diagnostic image quality. In this work, we present an automatic image quality assessment system that selects regions of interest (ROI) within abdominal CT images to estimate additive noise and, thus, quantify image quality.  The Windowed Fourier-domain Distance Metric (WFDM) is also presented, as a novel method of selecting spatially uniform ROI. This WFDM model, a Fixed Size ROI model, and a Convolutional Neural Network, are all used to classify phantom and patient CT images acquired at different CTDIvol (radiation dose index) values, and their performance is comparatively analyzed. While all the models accurately detect changes in quality due to CTDIvol for phantom images, the WFDM model outperforms the others when applied to patient images. The limiting conditions, exclusion criteria, and optimal parameters for the WFDM model and image quality metric are also identified.

Final Examination Archives

Autumn 2016