CAREER: AI-Driven Multimodal Feature Engineering for Personalized Biomedical System Design

This CAREER project will develop artificial intelligence (AI) tools to better understand and monitor neurological diseases. The research will combine brain images, genetic information, and clinical information. These are complex data sets with uncertainties, which makes it difficult to analyze them separately. The AI tools developed in this project will integrate and interpret the combined data. The resulting analysis will identify patient specific disease mechanisms and predict how diseases progress. The outcomes of the project will advance personalized healthcare, disease monitoring, and pattern detection. The project will also support education and hands-on training in biomedical AI, statistical inference, and modeling. Outreach activities include summer coding camps and data science workshops for high school students, and tutorials at national conferences. Software and models will be released as open-source tools to promote reproducibility and wide adoption. These activities will increase participation in biomedical engineering, train a skilled workforce, and improve public understanding of AI-enabled healthcare technologies. A unified, AI-driven framework for multimodal feature engineering will be used to model complex neurological diseases. The framework will combine machine learning (ML) with statistical causal inference to unify multimodal data across biological domains and timescales, producing interpretable representations that identify patient-specific mechanisms, predict individualized disease trajectories, and support precise diagnosis, treatment stratification, and real-time disease monitoring. The project will introduce three core system-level innovations: (1) domain-aware causal inference that integrates probabilistic and generative modeling to uncover latent features linking imaging, genetic, and clinical data; (2) customized transformer-based architectures that fuse structured and unstructured biomedical data, such as imaging embeddings, genetic variants, and clinical narratives, using cross-modal attention and contrastive learning strategies; and (3) domain-specialized large language models (LLMs) trained on multimodal features and biomedical text to translate complex outputs into interpretable clinical summaries. Reusable, open-source tools and models will be developed for clinicians and researchers, enabling analysis of large, heterogeneous biomedical datasets. Education and training activities will include interdisciplinary coursework in AI and causal inference for biomedical systems, hands-on research experiences, and outreach programs for high school learners. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2543636 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Rongjie Liu | Institution: University of Georgia Research Foundation Inc, ATHENS, GA | Award Amount: $546,167 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2543636 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2543636.html