Adaptable AI model capable of performing various biomedical tasks

BiomedGPT is described as a new type of artificial intelligence (AI), designed to support a wide range of medical and scientific tasks. In a recent study, published in the journal Nature Medicine, the AI model is described as, “the first open-source and lightweight vision-language foundation model, designed as a generalist capable of performing various biomedical tasks.”

The Lehigh University study, conducted in collaboration with multiple institutions, explained that the AI model does not need to be specialized for specific tasks, like analyzing X-rays or summarizing documents. The tool is considered a “generalist” model, capable of performing various tasks with the same underlying technology.

“This work combines two types of AI into a decision support tool for medical providers,” Lichao Sun, PhD, assistant professor of computer science and engineering at Lehigh University, and lead author of the study, said in a university release. “One side of the system is trained to understand biomedical images, and one is trained to understand and assess biomedical text. The combination of these allows the model to tackle a wide range of biomedical challenges, using insight gleaned from databases of biomedical imagery and from the analysis and synthesis of scientific and medical research reports.”

Sun elaborated that the AI model is based on foundation models, which are pre-trained AI systems that can be adapted to different tasks and situations with minimal training required. The authors of the study explained that the model achieved 16 overwhelmingly positive results after evaluation of 25 datasets across nine different biomedical tasks. The open-source codebase has also been made available for other researchers to utilize for further developments.

“The potential impact of such technology is significant,” said Kai Zhang, a Lehigh University PhD student advised by Sun, who served as the first author of the Nature Medicine article. “[The technology] could streamline many aspects of healthcare and research, making them faster and more accurate. Out method demonstrates that effective training with diverse data can lead to more practical biomedical AI for improving diagnosis and workflow efficiency.”

Receiving validation of the model’s effectiveness and applicability in real-world healthcare settings was a critical step in the validation process. Collaboration between Lehigh University and Massachusetts General Hospital was integral in this process. Other contributors to BiomedGPT analysis include researchers from the University of Georgia, Samsung Research America, the University of Pennsylvania, Stanford University, the University of Central Florida, UC-Santa Cruz, the University of Texas-Health, Children’s Hospital of Philadelphia and the Mayo Clinic.

“This was a true team effort,” Sun said. “Creating something that can truly help the medical community improve patient outcomes across a wide range of issues is a very complex challenge. With such complexity, collaboration is key to creating impact through the application of science and engineering.”