Predictive Modeling with High-Dimensional Colorimetric Image Data for Lung Cancer Detection

• Activity Number: 415
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Xiaofeng Wang*+ and Peter J. Mazzone
• Companies: Cleveland Clinic Lerner Research Institute and Cleveland Clinic Foundation
• Keywords: Variable selection ; Classification ; Semi-supervised learning ; Predictive modeling ; Colorimetric sensor image array ; Lung cancer

Abstract

Active lung cancer tumor cells outgas metabolic byproducts that have been shown to be detectable in breath. The colorimetric sensor image array is a novel technique for lung cancer breath test, which has hundreds of distinct receptor types and capable of discriminating a wide range of odorants. In this study, we propose a statistical modeling strategy with semi-supervised learning that utilizes both the image array data and the clinical data to identify subtypes of lung cancer and uses this knowledge to diagnose future patients. A real data analysis is presented and we show that the diagnostic procedure accurately predicts the cancer type.