With the creation of an algorithm, the project aims to analyze textual reports and predict the absence of coronary diseases in emergency patients more quickly, at lower costs and with greater assertiveness
An initiative between Einstein and Philips foresees the development of a solution to transform the identification of coronary heart disease in patients. The project uses Large Language Models (Large Language Models, LLMs) to analyze textual exam reports and, based on the indicators, identify mainly the greatest probability of absence of coronary disease in patients treated in emergency departments. Einstein is responsible for creating the algorithm that will serve as the basis for this solution.
Coronary heart disease, or coronary artery disease, is a condition that affects the coronary arteries, often leading to reduced blood flow and an increased risk of heart attack. In Brazil, it is one of the leading causes of death, affecting approximately 400,000 people annually, according to data from the Ministry of Health. The high incidence and severity of these conditions highlight the importance of advancing early detection and accurate diagnosis of these conditions.
“Philips has innovation in its DNA, and is once again participating in the development of solutions that aim to solve prevalent challenges in national and global healthcare institutions. The algorithm to be developed in this project aims to support the clinical decision-making of non-cardiologist physicians in emergency care and reduce requests for referrals and exams (especially invasive ones) for non-cardiopath patients. Philips’ Clinical Informatics plant in Blumenau exports Tasy, the electronic medical record and healthcare management system that has always been developed to meet healthcare and operational needs, to the world. It continues to expand its modules considering the needs of healthcare managers and professionals,” says Felipe Basso, president of Philips Latin America.
The goal of this project is to develop a tool capable of extracting 50 different variables in various formats to build robust predictive models. The use of LLMs is essential for the project, as more than 50,000 reports will be processed retrospectively, potentially reaching 100,000 documents. The major challenge (and the project's distinguishing feature) is that these reports often do not follow a uniform pattern. Therefore, LLM is a powerful tool for organizing data for optimal model training.
“These language models not only facilitate the extraction of anonymized data, but also allow this task to be performed at scale with excellent quality,” says Rodrigo Demarch, Einstein’s Executive Director of Innovation. “In the first experiments, the LLM technology applied in the project demonstrated an accuracy of 99% compared to variables extracted manually by experts, ensuring highly reliable results,” he adds.
The project has a multidisciplinary team of four data scientists and seven medical specialists dedicated to developing and validating the tool. The partnership between these professionals demonstrates the importance of multidisciplinarity in creating safe, high-quality products. The data used to build the algorithm comes from Einstein itself. The current focus is on analyzing the extraction of variables to ensure good data structuring and a satisfactory level of accuracy for doctors, aiming to identify the highest probability of absence of coronary artery disease. The next step will be to assess the accuracy of the prediction and diagnosis.
“This collaboration between Einstein and Philips not only represents a significant advance in the use of artificial intelligence in medicine, but also reaffirms the commitment of both organizations to innovation and continuous improvement in healthcare,” explains Marcos Queiróz, director of diagnostic medicine at Einstein.
“The use of validated clinical algorithms helps to improve accuracy in the diagnostic process, reduce the time taken to produce reports and, consequently, allows patients to start appropriate treatment more quickly,” explains Patricia Frossard, country manager of Philips Brazil.
EN 
PT 
