Toward Automation of the Diagnosis of Enteroparasitosis via Computational Image Analysis

Human intestinal parasites constitute a serious public health problem in most tropical countries. In this work, we present methods and preliminary results toward the fully automation of the human enteroparasitosis diagnosis. This system consists of a sensitive parasitological technique, called TF-Test (R); a peristaltic pump; a digital camera coupled with the microscope; and a computer for image analysis. The images were manually acquired and a database was prepared with 1,126 images with samples from 16 most commonly found species of parasites in Brazil. The automatic image analysis has shown sensitivity 95.3%, specificity 96.4%, and efficiency 96.2%, with kappa coefficient 0.88 (almost perfect concordance). These results indicate the viability of the automatic diagnosis of enteroparasitosis through computational image analysis.