Application of Stepwise Binary Decision Classification for Reduction of False Positives in Tuberculosis Detection from Smeared Slides

Corina Pangilinan, Ajay Divekar, Gerrit Coetzee, Dave A. Clark, Bernard Fourie, Fleming Y.M. Lure, and Sean Kennedy


Computer-Aided-Detection, Stepwise Classification, Fluorescence Microscope, Tuberculosis Acid-fast Bacilli, Binary Decision Diagram


This paper presents an approach using a stepwise binary decision classification to enable a significant reduction of non-AFB false positive (FP) objects while maintaining a similar true positive (TP) detection to improve performance. The stepwise classification (SWC) performs post-processing to remove the FP generated from the computer aided detection (CAD) system. FP objects are first analyzed and classified into several different categories such as small bright objects, beaded, dim elongated objects, etc. A SWC is developed to remove each type of FP, one at a time. For each individual classification, a binary decision diagram is created through minimization from a Boolean binary decision tree to classify different types of FP based on the extracted feature vectors associated with AFB and non-AFB objects. Each classification algorithm is developed and applied in a sequence to reduce a different type of FP such that the most dominant category of FP will be removed first. Consequently, significant amounts of FPs have been reduced while maintaining consistent true positive detection. The receiver operating characteristics (ROC) method with the area under the curve (Az) is utilized to analyze the performance of this SWC algorithm on 102 confirmed negative and 74 positive cases that had never been used in the development of the SWC. It shows the superior detection performance on the high-concentration cases (Az=0.913) and cases mixed with high-concentration and scanty AFB cases (Az=0.878). The SWC algorithm is easily extensible to newly identified FP and is backward compatible.

Important Links:

Go Back