Convolutional Neural Network U-Net for Trypanosoma cruzi Segmentation

Chagas disease is a mortal silent illness caused by the parasite Trypanosoma cruzi that affects many people worldwide. A blood test is one of the preferred methods to get an accurate diagnosis of the disease but takes a long time and requires too much effort from the experts to analyze blood samples in the search of the parasites presence. Therefore, it is very useful to have an automatic system to detect the parasite in blood sample microscopic images. In this paper we present a deep learning method to segment the T. cruzi parasite on blood sample images. We implemented a convolutional neural network based on the U-Net model and we trained it with different loss functions to get accurate results. We report an F2 value of 0.8013, a recall value of 0.8702, a precision value of 0.6304 and a Dice score value of 0.6825.