An Exploration of Spatial Radiomic Features in Pulmonary Sarcoidosis

Sarcoidosis is a rare, multi-systemic, inflammatory disease, primarily affecting the lungs. High-resolution computed tomography (CT) scans are used to clinically characterize pulmonary sarcoidosis. In the medical imaging field, there is growing recognition to switch from visual examination of CT images to more rapid, objective assessments of the abnormalities. In this work, we explore the usefulness of various objective measures of spatial heterogeneity---fractal dimension, Moran's I, and Geary's C ---for distinguishing between abnormal sarcoidosis and normal lung parenchyma. CT data for N=58 sarcoidosis subjects enrolled at National Jewish Health were obtained from the GRADS study. CT data for N=101 control patients were obtained from the COPDGene study. Radiomic measures were computed for each two-dimensional slice of a given scan, in the axial, coronal, and sagittal planes. Functional regression was applied to identify lung regions where CT nodules tend to proliferate. Moran's I, Geary's C and fractal dimension significantly differentiate between subjects with and without sarcoidosis throughout the majority of the lung, with disease abnormalities most apparent in the top axial, middle coronal, and outer sagittal regions. A trend appeared across Scadding stages, with CT scans from patients with Scadding stages I and III appearing the healthiest, and Scadding stage IV appearing the least healthy. The radiomic measures and techniques presented herein successfully characterize CT images in sarcoidosis by objectively and efficiently approximating what we know about the pathology of sarcoidosis.