Fuzzy clustering to detect tuberculous meningitis-associated hyperdensity in CT images

Abstract

Hyperdensity in head CT images has been shown to be a specific feature for diagnosing tuberculous meningitis (TBM) in children. We describe the extraction of hyperdense regions using fuzzy c-means clustering and fuzzy maximum likelihood estimation, thus providing a tool for the enhancement of an often subtle radiological feature. We calculate an asymmetry measure and confirm that normal and TBM images have different patterns of hyperdensity. Our results may be used in computer-assisted diagnosis of TBM.