New prospects in CT image processing via mathematical morphology.

New prospects are foreseen in CT image processing via mathematical morphology (M.M.). M.M. is a set theory based upon the concept of "structuring elements" and allows texture analysis and pattern recognition by performing sequences of neighborhood transformations. This results in a wide range of quantifications. Mathematical morphology applied to CT densitometry allows obtaining the mean density of an organ after its automatic extraction from the background. This operation eliminates the approximations due to the manual selection of the Region of Interest (ROI). Precision, reliability and reproducibility are therefore improved. In this study, we performed the automatic isolation of a sole vertebral body in thoracic or abdominal CT slices to quantify numerous parameters, among them, density of the vertebral trabecular and cortical bones. On scans of the second and third metacarpals, morphomathematical analysis allowed quantification of the parameters of bone evolution: mean cortical density, directional cortico-diaphyseal index, and medullar area. In a different connection, M.M. has allowed automatic segmentation of lung parenchyma, the measurement of its mean density and the determination of the relative importance of the vascular network.