Predictive Classification for Correlated Objects with Application to CT Perfusion Images of Liver Metastases

• Activity Number: 338
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Yuan Wang*+ and Brian P. Hobbs and Jianhua Hu and Kim-Ahn Do
• Companies: MD Anderson Cancer Center and MD Anderson Cancer Center and MD Anderson Cancer Center and MD Anderson Cancer Center
• Keywords: Bayesian ; decision theoretic ; classification ; correlated data ; functional imaging

Abstract

Perfusion computed tomography (CTp) is an emerging functional imaging modality that uses physiological models to quantify characteristics pertaining to the passage of fluid through blood vessels. CTp measurements acquired from multiple intra-patient regions tend to be highly correlated due to common vasculature, and may provide a quantifiable basis for early detection of cancer. In this paper, we propose a Bayesian simultaneous predictive classification method for discrimination of neuroendocrine liver metastases from normal tissue using the CTp measurements, when in particular multiple regions are observed for each patient. The proposed method uses a Kronecker product structure for the covariance to accommodate correlation between neighboring regions from the same patient. The results show that the proposed method is competitive with the classical methods when there is weak or no correlation between different regions, and significantly improves the performance in the presence of strong correlation.