Aspectos clínicos laborais e o uso da modelagem computacional no diagnóstico da leishmaniose visceral

Abstract

Leishmaniasis is caused by parasites of the genus Leishmania. There are three main forms of leishmaniasis: visceral, also known as Calazar, which is a more serious form of the disease, cutaneous, a more common and mucocutaneous. According to data from the world health organization, it is estimated that 700,000 to 1 million new cases and about 26,000 to 65,000 deaths occur. Different methods can be used for the diagnosis of Visceral Leishmaniasis (VL). Currently, there are a variety of techniques, but none have 100% sensitivity and specificity. The basic, safe and determining requirement (gold standard) in the laboratory diagnosis of VL is the application of amastigote forms in the material applied by bone marrow puncture. However, manual identification processes are work, demonstrated and prone to human errors. Detecting Leishmania automatically is a challenging computational task. The current scenario of efficient image analysis tools that assist in this detection. A potential new way to get around a technical manual is a machine learning technique from which to save a neural convolutional. This is a retrospective and observational study conducted with the clinical and laboratory data of twenty patients diagnosed with Visceral Leishmaniasis (VL) from January 2018 to May 2019 and aims to develop resources to aid in the diagnosis of VL. In this work, we present a construction of a deep convolutional neural network to automate the detection process of Leishmania with an accuracy of 83% and also describe characteristics of the parasitological examination and clinical data of patients that can also corroborate whith diagnosis increasing the sensivity of the method.