Toward a neuro-fuzzy system for automatic segmentation and characterization of intravascular ultrasound images

A project of a neuro-fuzzy system for automatic segmentation and characterization of intravascular ultrasound images is presented. The segmentation is performed by the use of the Fuzzy C Means algorithm and characterization is performed by using a neuro-fuzzy system capable of learning from images manually segmented and labeled by medical doctors. The proposed system has just been developed, and, in the actual stage, preliminary tests with the Inter Class Contrast validity measure have presented encouraging results.