Quantitative Analysis of Longitudinal Molecular Images to Monitor Lung Infection

• Activity Number: 588
• Type: Invited
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Bruno Michel Jedynak*+
• Companies: The Johns Hopkins University
• Keywords: imaging ; registration ; segmentation ; molecular imaging ; lddmm ; deformation

Abstract

Longitudinal molecular imaging offers the possibility to monitor a condition in real time, to assess response to therapy. However, most of the time the analysis require the manual delineation of a region of interest, which is not only time-consuming but also subject to error and bias. We present a computational method used to segment a target organ and register all serial images onto a common template organ. Contrarily to usual registration techniques, the proposed method does not require the user to tune registration parameters by hand. Instead, the user highlights parts of the organ of interest and few elements of the image background. Once registration is obtained, a statistical model is built from controlled images. Subject images are compared to the statistical model to detect regions of higher activity. We will demonstrate this technique on the analysis of combined FDG PET/CT images to monitor the inflammatory response to tuberculosis (TB) treatment in a mouse model. This is a joint work with the Center for Imaging Science and the Center for Tuberculosis Research of the Johns Hopkins University.