CLHSS Launches Innovative Methods & Data Science Program

This past November, CLHSS launched a new Innovative Methods & Data Science (IMDS) program. Co-directed by Julian Wolfson, PhD, and Rui Zhang, PhD, the goal of the program is to help researchers develop new artificial intelligence (AI) and data analysis tools and improve existing ones.

The IMDS program is dedicated to creating novel and improving existing tools researchers have available to study multimodal biomedical big data. This allows researchers to study multimodal data representing patients with 360 angles, giving them better answers to questions on health outcomes.

Dr. Wolfson and Dr. Zhang started the program to create a coordinated effort to assist researchers working with multimodal data. The IMDS program involves collaboration between units from different areas like biostatistics and computational health sciences working on different aspects of computational and statistical data analysis. The co-directors hope to expand this collaboration to include academic institutions and industrial partners outside of the University of Minnesota.

One of the main goals of the IMDS program is to make accurate, fair and interpretable methods for researchers to easily access new types of data . New, multi-modal forms of data like clinical texts, medical images, log file, sensing data, and longitudinal patient data offer researchers a much more complete picture of patient health. This allows them to better study the real-world impact of changes to clinical practice, and examine additional differences between patients beyond demographics. However, much of this data is private and structured in a way that is difficult to directly use for clinical and translational research. The IMDS team is working on developing multimodal AI methods to organize and catalog this data so that researchers can more easily use it.

The IMDS team is also educating researchers on the value new AI and statistical methods can bring to their work. They are hoping to eventually host workshops for University of Minnesota researchers on methods and data analysis tools.

While the IMDS program is still getting off the ground, members have already been brainstorming potential projects. Some of these involve developing AI tools for understanding patient data, like a natural language processing model that automatically extracts social determinants of health from electronic health records, federative learning for medical imaging analysis and many others. The IMDS team is also interested in examining risk prediction model accuracy and fairness to ensure that models that guide physician decision making are equitable and perform equally well across different groups of people.

Learn more about the IMDS program: med.umn.edu/clhss/imds