Ranking of Visual Attention Spread Improves Trypanosoma Cruzi Detection in Microscopy Videos

Visual attention is an essential ability for detecting para-sites in microscopy videos, as it involves scenarios with dynamic elements, occlusions, and low-contrast features. Par-ticularly, the parasite Trypanosoma cruzi (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T$</tex> cruzi) emerges as conspicuous due to the nature of its movement. However, microscopy videos typically exhibit other motion patterns that become intricate in the perception of these parasites. Here, we showed that irrelevant motions in microscopy videos are linked to a broad distribution of attention, whereas T. cruzi's motion results in a more constrained focus of attention. Using saliency maps, we ranked the distribution of visual attention within the scene, boosting constrained saliency maps for the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T$</tex>. cruzi detection by computational approaches. Experiments with a state-of-the-art saliency-based model and videos fea-turing 1–3 parasites, conducted under real artifact conditions and without any contrast-enhancing treatment, revealed that reducing the influence of broad saliency maps improves the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$F_{1}$</tex>-score by 9% in challenging microscopy environments.