Artificial Intelligence >
Advancing Biomedical Discovery and Transforming Health.
Biomedical artificial intelligence (AI) encompasses both the development of new computational methods and their application to problems in biology, medicine and public health. Methodological work includes causal inference methods to identify mechanisms beyond mere correlation; probabilistic and Bayesian models that quantify uncertainty in high-stakes decisions; large language models adapted to the vocabulary and reasoning patterns of clinical medicine; and multimodal approaches that integrate images, text, genomics and time-series data into unified representations. Application work spans three core areas. Biomedical discovery uses AI to unlock insights from large, heterogeneous, and combined experimental and observational datasets to predict disease onset, identify drug targets and analyze single-cell and multi-omics data. Clinical decision support uses AI to extract knowledge from clinical text and deliver person-specific information to clinicians and patients, intelligently filtered and timed to enhance real-time decision-making. Population health applies these methods to population-level disease prevention, biosurveillance and epidemic modeling.
Education
Our graduate training program, offering PhD, MS, and certificate options, prepares the next generation of leaders in biomedical informatics and AI research. Our students range from computer scientists and data scientists to biologists and clinicians. Coursework, mentored research and hands-on projects are tightly integrated from the first semester, covering algorithms, data structures and core AI concepts, as well as their application to biomedical problems. Students also develop structured competency in using generative AI across the research lifecycle, from literature review and study design to data analysis, writing, and scientific communication.
We are also pioneering AI education for researchers who come from noncomputational backgrounds. A new introductory course, designed specifically for PhD and MS students across the School of Medicine and other health sciences schools, brings rigorous AI training to students with expertise is in biology, clinical medicine and the health sciences.
At the undergraduate level, our Artificial Intelligence Biomedical Informatics and Data Science (AIBIDS) program offers a full-time, 10-week research internship for students interested in AI and biomedical informatics. At the high school level, our Computer Science, Biology, and Biomedical Informatics (CoSBBI) program is a summer experience that combines a crash course in biomedical informatics and AI, a programming boot camp, and mentored collaboration on an active research project.
Research
Our faculty bring expertise across a broad range of AI methods, including symbolic AI, probabilistic AI, machine learning, causal inference, generative AI, deep learning and human-computer interaction. Their research spans clinical informatics, clinical research informatics, population health informatics, translational bioinformatics and imaging informatics, applied to clinical domains such as anesthesiology, neurology, pediatrics, critical care, oncology and infectious diseases.
Our research is grounded in the understanding that effective biomedical AI must encompass the full arc of the field, from developing new AI methods to studying how those tools can be deployed safely and effectively. To that end, our faculty and trainees draw on methods from evaluation science, implementation science, data science and human-computer interaction.
Example projects are listed below.
| AI method | Informatics subfield | Example project |
|---|---|---|
| Symbolic AI | Clinical research informatics | Graph topology and semantic relationships for biological discovery across heterogeneous multi-omics datasets (Jonathan Silverstein) |
| Symbolic AI | Population health informatics | Pharmacovigilance using electronic health records (Shyam Visweswaran) |
| Probabilistic AI | Clinical informatics | Predicting the survival of breast cancer patients (Xia Jiang) |
| Probabilistic AI | Translational bioinformatics | Integrating multiomic data (genomics, proteomics, transcriptomics, epigenomics) to predict disease mechanisms (Lujia Chen) |
| Machine learning | Clinical informatics | Developing machine learning-based clinical decision support (Shyam Visweswaran) |
| Machine learning | Imaging informatics | Predicting executive dysfunction from brain imaging (Rafael Ceschin) |
| Causal inference | Clinical informatics | Identifying root causes of disease from data (Eric Strobl) |
| Causal inference | Translational bioinformatics | Developing methods for tumor-specific causal inference (Xinghua Lu) |
| Generative AI | Clinical informatics | Mitigating health inequity risks in large language models (Yanshan Wang) |
| Generative AI | Clinical research informatics | Analysis of medical notes |
| Deep learning | Translational bioinformatics | Predicting protein phosphorylation (Hatice Osmanbeyoglu) |
| Human-computer interaction | Clinical informatics | Impact of explanations on risk models for clinical users (Harry Hochheiser) |
Commercialization
Our faculty members are active in commercializing biomedical AI through startups. PredxBio, co-founded by Michael Becich, MD, PhD, uses AI to analyze whole-slide images of tumor microenvironments and predict patient response to immunotherapy. DeepRx, co-founded by Xinghua Lu, MD, PhD, Gregory F. Cooper, MD, PhD, and Lujia Chen, PhD, develops AI-based clinical decision support for precision oncology, with a lead product targeting drug selection in colorectal cancer. READE.ai, co-founded by Shyam Visweswaran, MD, PhD, applies AI to real-time detection of brain ischemia during surgery. Dr. Visweswaran also serves as chief medical officer of ThetaRho, a startup offering a natural language AI interface to streamline access to patient data at the point of care within electronic health records.