|
Abstract:
|
Computed tomography images of the lung play an increasingly important role in diagnostics and clinical care decisions in interstitial lung diseases, such as sarcoidosis and idiopathic pulmonary fibrosis. To date, most lung image analysis has been done visually by trained radiologists. Although incredibly useful, it is time consuming and has poor inter-rater reliability. As such, quantitative imaging has emerged as an important field to augment visual scoring. Radiomics is one branch of quantitative imaging where large numbers of first and second order features are computed on each image as potential biomarkers. The radiomic biomarkers are then assessed for their discriminative ability. In this poster we first define radiomics statistically and clinically and then work through an application of radiomics in sarcoidosis. We show how radiomic biomarkers relate to visual lung patterns and predict lung function better than either Scadding stage (the gold standard for staging) or visual scoring. We end by showing how the radiomic profiles of a cohort of sarcoidosis patients cluster into four new disease phenotypes that appear to better represent disease progression.
|
