A computer vision approach to emphysema detection and quantification

Emphysema is a pulmonary disease visible on postero-antero chest radiographs and characterized by a vascular rarefaction in the lungs. The authors have developed several tools to assist practitioners in quantifying the lung vascular density. Their approach is based on the extraction of the blood vessels in the image through the tracking of their medial axes. This basic algorithm referred to as fuzzy multiscale grey-level medial axis transform (MMAT) is related to the core technique promoted by the University of North Carolina team. Spurious vessels are detected mainly near rib edges. To reject them, ribs are localized by making another use of the MMAT at scales compatible with rib sizes. The resulting mapping of the vessel contrast is then regularized with a Markov random fields method to calculate the vascular density. A clinical validation has been started to evaluate the usefulness of the authors´ computer vision techniques for early and reliable diagnosis of emphysema