Computer-Assisted Screening of Ziehl-Neelsen–Stained Tissue for Mycobacteria

• DOI: 10.1309/ajcpmr3blvbh8thv
• Publication Date: 2010-05-14T19:42:55Z
• Authors: Paul J. Tadrous
• Keywords: Ziehl–Neelsen stain

Screening Ziehl-Neelsen (ZN)-stained sections for acid-alcohol–fast bacilli (AAFB) is laborious, and sparse are easily missed. This article presents an automatic screening algorithm using digital image analysis designed to assist human diagnosis of tissue sections. The uses multiderivative source potentiators suppressors feeding into interconnected product nodes that result in a probability value each (the likelihood it contains AAFB) spatial map showing the position any bacillus. For study, 3,000 images from ZN-stained tissues were captured, 1,000 used train algorithm, 2,000 test it. successfully ranked AAFB-containing as highest data sets, despite only single being present (occupying 0.0024% image) staining artifacts. These results suggest this automated assistance method has potential save time money, which especially important resource-poor health services.