Microcomputer system for automatic identification of the Cryptococcus neoformans and its clinical application.

Abstract

In this study, microcomputer image processing and pattern recognition technology, and the knowledge of morphology and optical characteristics of Cryptococcus neoformans were used for identification of Cryptococcus neoformans. Four groups of mice were lethally infected with standard strain, Wuhan strain, American B-2643 strain and Var. Shanghainesis of the Cryptococcus neoformans. The samples collected included mice brain, lung, kidney, liver, small intestine tissue and were observed under a light microscope. More than 600 images of the fungus were input into a microcomputer. A system of computer for automatic identification of the Cryptococcus neoformans was developed. The technique involved image preprocessing, image segmenting, coding of line-length on the edge, curve fitting, extracting of image feature, building of image library and feature data bank etc.. And then, 768 images of the clinical samples and other fungus samples whose morphological features tend to be confused with Cryptococcus neoformans were input into microcomputer and subjected to automatic identification. The Cryptococcus neoformans was accurately identified within 15 min, and the consistency rate with results of routine culture was 98%.