Lung Cancer Risk Assessment Using Quantitative Imaging
Patients at high-risk of developing lung cancer undergo an increasing amount of medical imaging procedures to detect and assess tumors. The resolution of these clinical scans generates rich dataset which are visually read by radiologists, using only the human-perceptible information to identify, diagnosis, and evaluate disease states. We propose to use the rich image information more completely by extracting objective, quantitative characteristics of tumor texture, shape, and size. The image characteristics will then be ranked by their effectiveness at describing a disease state, specifically cancerous versus noncancerous distinctions. From the ranking, a computer-aided-diagnosis tool will be created using artificial intelligence techniques to predict a tumor's cancer risk. This tool can be integrated into the clinical pipeline providing a risk summary for radiologists, which would improve treatment planning and reduce data waste.