Mayara Martins Perroni

Heres the translation:Undergraduate student in Biomedical Informatics at the University of São Paulo (USP), with a focus on Machine Learning applied to healthcare. In 2022, completed a FAPESP Technical Training II, where she worked on the segmentation of computed tomography images for the development of augmented reality software aimed at surgical procedures. From 2023 to 2024, she participated in the PRECARE-ML project (Prevention of Major Adverse Cardiovascular Events) as a PUB scientific initiation fellow, applying the OMOP-CDM model from OHDSI for organizing clinical cohorts and developing machine learning models. During the project, she contributed to the standardization of medical vocabularies and developed solutions to avoid local overfitting, helping to generalize the models to different populations.In her undergraduate thesis (TCC), she applied machine learning techniques to structure free-text fields in electronic health records. The project involved the use of natural language processing (NLP) models for extracting and standardizing relevant clinical information, such as risk factors (smoking, alcohol consumption, and obesity) and symptoms described in an unstructured format in the records. Approaches such as Large Language Models (LLMs) were tested to capture complex nuances and contexts from the texts, in addition to classic techniques like TF-IDF for tokenizing and vectorizing the texts. This combination allowed for the creation of an automated pipeline to process these fields and incorporate them into machine learning predictive models, increasing the accuracy of MACE predictions across different clinical contexts. (Source: Lattes Curriculum)