Detection of tuberculosis in sputum smear images using two one-class classifiers

Abstract

We present a method for the identification of <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Mycobacterium</i> <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">tuberculosis</i> in images of Ziehl-Neelsen stained sputum smears obtained using a bright field microscope. We use two stages of classification; the first is a one-class pixel classifier, after which geometric transformation invariant features are extracted. The second stage is a one-class object classifier. Different classifiers are compared; the sensitivity of all tested classifiers is above 90% for the identification of a single bacillus object using all extracted features. Our results may be used to reduce technician involvement in screening for tuberculosis, and will be particularly useful in laboratories in countries with a high burden of tuberculosis.