A functional model for classifying metastatic lesions integrating scans and biomarkers

Perfusion computed tomography is an emerging functional imaging modality that uses physiological models to quantify characteristics pertaining to the passage of fluid through blood vessels. Perfusion characteristics provide physiological correlates for neovascularization induced by tumor angiogenesis and thus a quantitative basis for cancer detection, prognostication, and treatment monitoring. We consider a liver cancer study where patients underwent a dynamic computed tomography protocol to enable evaluation of multiple perfusion characteristics derived from interrogating the time-attenuation of the concentration of the intravenously administered contrast medium. The objective is to determine the effectiveness of using perfusion characteristics to identify and discriminate between regions of liver that contain malignant tissues from normal tissue. Each patient contributes multiple regions of interest which are spatially correlated due to the shared vasculature. We propose a multivariate functional data model to disclose the correlation over time and space as well as the correlation among multiple perfusion characteristics. We further propose a simultaneous classification approach that utilizes all the correlation information to predict class assignments for collections of regions. The proposed method outperforms conventional classification approaches in the presence of strong spatial correlation. The method offers maximal relative improvement in the presence of temporal sparsity wherein measurements are obtainable at only a few time points.