Segmentação de imagens microscópicas para quantificação da infecção de Leishmaniose em macrófagos com visão computacional

Abstract

Leishmanisis is one of the twenty neglected diseases according to the World Health Organization (WHO) classification. Considered a major public health problem, it represents a complex of diseases with an important clinical spectrum and epidemiological diversity. The Oswaldo Cruz Foundation (Fiocruz) in Brazil conducts the process of identifying the level of the disease, through the manual count of amastigotes found in macrophages, being a process susceptible to errors and exhausting for specialists in the area. This work proposes the automation of the process of counting leishmania in macrophages through semantic segmentation using Neural Networks. The dataset was built with images captured by smartphones, with samples generated in the laboratory to perform the experiment. The results were compared with statistics obtained by experts in manual counting. The proposed approach was able to segment and count macrophages and leishmania, reaching a Coefficient of determination of 0.8. Therefore, the automation of leishmaniasis counting using neural networks proves to be a promising methodology to assist the work of specialist physicians in calculating the rate of leishmaniasis infection.