Implicit modeling and virtual endoscopy of pulmonic cavities

Pulmonary cavities are a devastating consequence of severe pneumonia and other necrotizing pulmonary disorders. As part of a project to understand cavity formation and healing, the authors evaluated different techniques for determining the shape of the cavities. The pulmonary cavities were segmented from helical computed tomography scans of the thorax. The authors parameterized the shape of the cavities using two implicit functions, an ellipsoid and a superquadric. The original cavities and the parameterized models of the cavities were then visualized using a virtual endoscopy system. Both functions produced reasonable fits to the data. RMS errors were lower with the superquadric although the superquadric tended to deform inappropriately if a bronchial segment communicated with the cavity or if the top or bottom of the cavity was not capped. The authors found that fitting the cavity data to both ellipsoids and superquadrics provides straightforward estimation of the global shape.