DESENVOLVIMENTO DE SISTEMA ESTRUTURADO COM INTELIGÊNCIA ARTIFICIAL PARA APOIO NO DIAGNÓSTICO DE PARASITOSES INTESTINAIS.

Introduction: Artificial intelligence systems are promising technologies for health care and laboratory diagnosis, which can be implemented as support methods for the diagnosis of intestinal parasitoses. This study aimed to develop an artificial intelligence software that assists the laboratory diagnosis of intestinal parasitoses with high sensitivity and specificity. Methods: The software was developed using two neural networks, Inception and MobileNet. First, images of eggs from the parasites Ascaris lumbricoides, Trichuris trichiura, Taenia sp., Hymenolepis nana, Schistosoma mansoni and Strongyloides stercoralis larvae were used to train the database. Then, 2,740 images provided by the Parasitology Laboratory of the Universidade do Oeste de Santa Catarina were tested in the software. Results: The software had a sensitivity of 82.3% (95% confidence interval [CI], 71.9% - 89.1%) and a specificity of 95.1% (95% CI, 94.3% - 97,8%) for MobileNet and a sensitivity of 72.1% (95% CI, 52.6% - 115%) and a specificity of 92.1% (95% CI, 91.7% - 97.7%) for Inception. Conclusion: The software showed promising results in the analysis of intestinal parasites, reinforcing that, in the future, the presence of diagnostic support systems for parasitoses may become faster and more efficient. [ABSTRACT FROM AUTHOR]