Mathematical morphology for automatic detection of brain lesions at magnetic resonance imaging

An original method of image analysis has been developed, using 'mathematical morphology', in order to detect brain tumors at magnetic resonance imaging of the head following injection of gadolinium-DPTA. The main steps of the analysis are as follows: 1) Spatial reduction by automatic determination of a region of interest, 2) elimination of noise by morphologic filtering (increasing alternate closing-opening sequence), 3) localization of tumor by searching for r-h maxima, and 4) extraction of tumor contours using three-dimensional structuring elements. The preliminary results of this automatic and robust method encourage further studies of the potential of tissue analysis as an aid to tumor diagnosis.