Differentiation of digital TB images using multi-fractal analysis

Microscopic examination of stained sputum smears has remained the cornerstone of pulmonary TB (tuberculosis) diagnosis throughout the world. In this work, an attempt has been made to differentiate such digital TB positive and negative sputum smear images using multi-fractal analysis. The digital TB images (N=50) used for this analysis were captured for this analysis were captured using a fluorescence microscope from auramine stained slides. The multi-fractal analysis of the captured original images and cropped ones were subjected to multi-fractal analysis. Distinct variations were observed from the cropped images in terms of the multi-fractal signatures, maximum and minimum values of local and global information (α, f(α)). The spectral width was found to vary significant for positive and negative images. The results seem to be clinically useful for mass screening of TB images.