Big data analytics tools provide clinical decision support

Researchers at the University of São Paulo are developing computer systems to process and extract information from large datasets provided by public hospitals. Their goal is to create a database that can be queried by physicians and clinical specialists to help diagnose and treat patients.

By Heitor Shimizu in Paris  |  Agência FAPESP – Sophisticated computer systems capable of storing, indexing, analyzing and making sense of large datasets that cannot be processed by traditional software could become essential tools to support decision making in the medical field.

Research on such systems has been conducted by the Database and Imagebase Group (GBDI) in the University of São Paulo’s Mathematics and Computer Sciences Institute (ICMC-USP) at São Carlos, São Paulo State, Brazil. 

Professor Agma Traina discussed the topic in a lecture delivered to FAPESP Week France.

“One of the biggest challenges in computer science is integrating, organizing and mining large amounts of multimodal data from a wide array of platforms to support decision making. In other words, we need to know how to help health workers make use of a variety of data sources, from lab tests to patient monitoring and treatment records, so that they can garner information on similar problems and build a better understanding of a specific case,” Traina told Agência FAPESP.

The research conducted at the GBDI’s laboratory involves masses of complex data from public hospitals in São Paulo State. The group mainly work with images and videos that provide clinicians with information on similar cases treated in the past.

“In analyzing a patient’s chest X-ray, for example, a physician may recall seeing similar results in the past but is unlikely to remember exactly when and where, let alone the names of patients in previous cases,” Traina said. “If you can search a database and instantly find similar past cases, tests, results and treatments, you can make decisions with less effort and more confidence.”

Part of the research is supported by FAPESP via a Thematic Project for which Traina is principal investigator. The project, she explained, involves database organization, metric access methods (to speed up the assessment of similar cases), and image processing and viewing, all of which can give physicians and clinical specialists the tools, algorithms and other means to assemble and analyze highly valuable information on past and present cases.

“To do this we have to bring together professionals in machine learning, database management, data lineage [the origins and lifecycle of data], and image visualization and processing. Our group includes computer scientists, physicians, mathematicians and other researchers who work in an integrated manner to solve the problems posed,” said Taina, who is a member of FAPESP’s Area Panel for Computer Science and Engineering.

The size and complexity of the databases that contain electronic patient records are a major processing challenge, in terms both of the development and application of analytics and knowledge extraction techniques, and of support for the development of practical tools for clinical use, according to Traina.

“Yet they also offer endless opportunities to create algorithms and methods for displaying relevant information on a particular patient or group of patients. The massive amount of data would normally make this kind of information inaccessible,” she said.

“In addition, efficient big data management makes electronic patient records more useful to health workers in dealing with fast-track medical applications, as well as to support strategic governmental decisions on health-related matters.”

FAPESP Week France took place on November 21-27, 2019, thanks to a partnership between FAPESP and the Universities of Lyon and Paris. For more news on the event, visit www.fapesp.br/week2019/france.