A System for Enhancing Human-level Performance in COVID-19 Antibody Detection
The world currently suffers from the global COVID-19 pandemic. Billions of people have been impacted, and millions of casualties have already occurred. Therefore, it is of extreme importance to identify individuals contaminated by SARS-CoV-2, allowing governments to plan actions to reduce further impacts. In this context, this work employed machine learning to improve the detection of SARS-CoV-2 antibodies in blood exams. Models have been developed in a real-world scenario with 500 thousand exams and were deployed in a remote laboratory for experiments. Results indicate that the models averaged sensitivity and specificity of 95%, and thus, they could aid COVID-19 antibody detection and the decision-making process of biomedical specialists.