A project by researchers from USP’s Faculty of Medical Sciences of Ribeirão Preto (FMRP) to validate and improve machine learning models for predicting cardiovascular risk has won the EP PerMed video competition, promoted by the European Partnership for Personalized Medicine coordination platform, which brings together European Ministries, agencies and funding organizations. The winners were announced during a conference in Berlin, Germany, on February 11 and 12.
The FMRP project, called Precare ML – Prediction of Cardiovascular Risk, was created to validate and improve machine learning (ML) models in different hospital systems and populations, integrate them into hospital information systems and evaluate their impact on the hospital’s daily routine. Furthermore, based on the validated models, the researchers intend to address effective risk communication strategies to effect patient behavioral change.
Paulo Mazzoncini, professor in the Department of Medical Imaging, Hematology, and Clinical Oncology at FMRP, explains that ML is an area of artificial intelligence that allows computers to learn to recognize patterns in large volumes of data automatically. The professor explains that machine learning is currently being used extensively in healthcare, financial fraud detection, and natural language processing, among other areas.
Mazzoncini points out that the use of ML enables preventive actions, such as the early identification of patients at high risk of cardiovascular diseases, such as myocardial infarction, ischemic heart disease, and cerebral vascular accident. “It will be possible to inform the physician accompanying the patient and, if necessary, refer them for appropriate treatment, reducing the possibility of future major cardiovascular events.” However, the researcher warns that although many models have been developed in the last few years, validation is still rare. “We still don’t know how the models work in different clinical environments or populations; moreover, as the models use numerous and diverse predictors, it is difficult to transfer models to other healthcare systems.”
Click on the player below to watch the video of the winning project:
Future
According to Mazzoncini, the FMRP's research is still ongoing and is supported by the Medical University of Graz and the Styrian Hospitals Organization, Austria; Hasso Plattner Institute for Digital Health at the University of Potsdam, Germany; and Karolinska Institute, Sweden. In addition to the collaborators, the project also involves stages of standardization of data extracted from Electronic Health Record (EHR), such as implementation, training, and validation of machine learning models, integration of the models into the hospitals' information systems, and monitoring their effectiveness for use in real clinical environments.
Still on the subject of ML, Mazzoncini details that several well-established machine learning models are available for customization and informs that the award-winning project uses classic models trained for the specific application using data extracted from electronic health records.
The project is partially funded by the São Paulo Research Foundation (Fapesp) and has the participation of Kátia Suzuki, Systems Analyst, and Business Intelligence Manager at the FMRP’s Center for Development and Continuing Education in Biomedical Informatics, Gladys Pierri and Hilton Vicente Cesar, Fapesp technical training fellowship holders, and professors Elen Almeida Romão, José Abrão Cardeal da Costa, João Mazzoncini de Azevedo Marques, and Sandro Scarpelini.
The European Partnership for Personalized Medicine initiative aims to improve results in sustainable healthcare systems through research, development, innovation, and implementation of personalized medicine approaches for the benefit of patients. More information about the initiative can be found by clicking here.
The video about FMRP’s Precare ML – Prediction of Cardiovascular Risk project is available at this link.
.*Intern at Faculty of Medical Sciences of Ribeirão Preto under the supervision of Rose Talamone
