In my master's Thesis, I developed a computational system that can
autonomously detect microaneurysms, inspecting color photographs of the
retina of patients with diabetes.
As a fundamental algorithm of my system, I developed and trained a deep neural
network architecture, which was trained to make the semantic segmentation of
the pixels that belonged to a microaneurysm, thus managing to detect them.
After the experimental work, I achieved precision values, in the detection of
microaneurysms, of 64.15% and sensitivity of 60.3%, thus obtaining a competent
system with the state of the art in the matter of automatic detection of
microaneurysms.