Proposal of CT Fexture Factor Analysis with Validation of -950Hu Density Mask

• Activity Number: 117
• Type: Contributed
• Date/Time: Monday, August 9, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Consulting
• Author(s): Hyun J. Kim*+ and David Gjertson and James Sayre and Jonathan Goldin
• Companies: University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles
• Keywords: texture ; factor analysis ; density mask

Abstract

Density mask is known to be a good CT quantitative structural measurement of chronic disease emphysema. It results from threshold of attenuation present in emphysematous lung. This study proposes the texture factors, which can be more informative in subtle changes than density mask. From eight regions of interest (ROI) (2 arch,2 hemidia of Right and Left ), two donuts (axial, peripheral), and three thirds (anterior, middle, posterior) per patient, local texture information was extracted and represented by a 18-dimensional vector that contained statistical moments of the CT attenuation distribution, acquisition-length parameters, and co-occurrence descriptors. Principal component analyses (PCA) and were used to reduce four components such as mean, dispersion, correlation, and percentile components. Factor analyses (FA) were used to reduce to two factors. Multivariate analysis of variance test and Hotelling t test with Bonferroni adjustment were used to validate the two factors using categories of density masks. 140 cases of each ROI resulted as same two factors from PCA and FA. The two texture factors can be used to distinguish severity in emphysema.