Tuberculosis (TB) is one of the leading causes of death from a single infectious agent. In many high-burden regions around the world, which often lack specialized healthcare professionals, Chest X-Ray (CXR) exams continue to be of vital importance in the diagnosis and follow-up of the various presentations of the disease. In this context, we investigate the benefits of automatic Pulmonary Tuberculosis (PTB) detection methods from these images.

Semantical segmentation of eye fundus structures, and disease detection from retinography, play a key role in mass screening using this tecnology. Despite the incredible progress in these fields, the lack of annotated images (due to cost), and rigor in the comparison of trained models has led to the conclusion larger and more dense network models provide more accurate results for such tasks. We present our findings on different architectures and databases in this context.

Reproducible research not only leads to proper scientific conduct but also provides other researchers the access to build upon previous work. Most importantly, the person setting up a reproducible research project will quickly realize the immediate personal benefits: an organized and structured way of working. The person that most often has to reproduce your own analysis is your future self!

Early and accurate prediction of decompensation (functional deterioration) in patients in domestic settings may help prevent deaths. Based on values that can be measured from portable devices such as heart rate, blood oxygen saturation, systolic blood pressure, temperature, and age, we study the prediction capability of machine learning algorithms to determine patient decompensation (death) in the next 24 hours.

We address the problem of reproducible research in remote photo-plethysmography (rPPG). Most of the work published in this domain is assessed on privately-owned databases, making it difficult to evaluate proposed algorithms in a standard and principled manner. As a consequence, we present a new, publicly available database containing a relatively large number of subjects recorded under two different lighting conditions. Also, three state-of-the-art rPPG algorithms from the literature were selected, implemented and released as open source free software. After a thorough, unbiased experimental evaluation in various settings, it is shown that none of the selected algorithms is precise enough to be used in a real-world scenario.

I have actively worked in computer vision and deep learning (mostly) associated to biometric recognition, with potential application to various other tasks. Contributions range from the collection of datasets, the exploration of different methods to address and assess biometric recognition vulnerabilities, domain adaptation, and remote photoplethysmography.