Leishmaniasis is a neglected tropical zoonosis affecting humans and animals worldwide, with endemic presence in more than 90 countries, including Brazil. Visceral leishmaniasis (VL), primarily caused by Leishmania infantum, is the most severe form of the disease, with mortality rates exceeding 95% when untreated. Domestic dogs are the main reservoir hosts, particularly in socioeconomically vulnerable regions such as Northeast Brazil. Accurate diagnosis is essential for disease control; however, current diagnostic methods still face important limitations. Molecular techniques such as PCR provide high sensitivity and specificity but remain costly and inaccessible in many settings. Direct parasitological diagnosis through microscopic identification of amastigotes is more affordable but depends heavily on specialist expertise and is subject to observer variability.
To develop an artificial intelligence-assisted diagnostic support system capable of identifying Leishmania amastigote candidates in microscopy images and supporting healthcare and veterinary professionals during routine diagnostic analysis. A preliminary deep learning framework is being developed using microscopy images from the publicly available AIR Leish dataset. The proposed approach employs computer vision and deep learning techniques to learn morphological characteristics associated with Leishmania amastigotes and automatically highlight candidate regions of interest in digital microscopy images. Dataset preparation, model training, validation, and performance evaluation protocols are currently under development.
The proposed framework aims to assist healthcare and veterinary professionals in the interpretation of microscopy images by highlighting candidate Leishmania amastigote regions. As a diagnostic support tool, the system is expected to promote more standardized image analysis workflows and improve efficiency during routine diagnostic assessment of visceral leishmaniasis.