Processamento de amostras fecais e desenvolvimento da tecnica de analises de imagens por computador, para o diagnostico das enteroparasitoses / Processing of samples feces and development of the technique of analyses of images for computer, for the diagnosis of the enteroparasitoses

AUTOR(ES)
DATA DE PUBLICAÇÃO

2008

RESUMO

In the present study, we investigated on fecal sample processings, development and evaluation of a computer system for parasite image analysis. We observed that the fecal processing step was critical, but, essential for providing clean parasite structures, with reduced amount of fecal microdebris, to the system of computer analysis. Several conventional parasitologic techniques, using or not commercial kit, were studied, and TF-Test (Three Fecal Test) showed lower rate of microdebris at the end of fecal processing. However, this microdebris rate was still great, being unsuitable for computer analysis. The TF-Test technique was modified by staining the fecal sediment components, with a Lugol-based stain, followed by an alkaline degradation of fecal microdebris, using a clarifyer solution. So, the modified TF-Test technique became capable to supply parasite structure with little fecal microdebris. The prototype design of the computer system for the image analysis incorporated: the modified TF-Test technique; a optical microscope coupled to a peristaltic suction pump for leading an aliquot of processed fecal suspension to the tubular slide or chamber, where the parasite structures or microdebris appeared in tridimensional images; a digital camera; a video monitor to calibrate and capture images; and a computer. An image database was formed, after collecting fecal samples from prevalent regions (Campinas, Botucatu, Avaré and Piraju) of the State of São Paulo for enteroparasitosis. The total fecal sample loss was low, being less than 8%, since biosecurity and laboratory quality control protocols were frequently checked. A total of 16 parasite species were identified, consisting of helminths and protozoans, which provided a collection of 1.126 parasite images to the computer. Moreover, the image database was formed by information acquired from 5.626 parasite and nonparasite components, in addition to data on their shape characteristics, texture and color. The computational analysis was based on a pipeline of image processing techniques, including the use of a technique known as Image-Foresting Transform (IFT). The pipeline consisted of: a technique for image segmentation, in order to separate parasites and microdebris from background image; a technique for feature extraction to encode shape, color and texture characteristics of parasites; and techniques for pattern recognition and delineation, permitting to distinguish parasites from microdebris, according to their own features. The technique of computer image analysis (CIA) was evaluated in comparison with the optical microscope technique, named modified TF-Test, demonstrating a high sensitivity of 95,3%, specificity of 96,4% and efficiency of 96.2%. The agreement between two techniques was ranked as Almost Perfect, since the kappa (k) index has been high as much as 0.88. This technique proved to be reproducible, in a study, in which the assay was repeated 10 times, in different occasions. Our findings present good perspectives for the industrial production of the here developed prototype, causing impact on the Public Health area, since, in the fecal examination of the population, there is a strong demand for an automated system of enteroparasite detections.

ASSUNTO(S)

parasites feces fezes parasitos image analysis analise de imagem

ACESSO AO ARTIGO

Documentos Relacionados