Explore multimodal AI research for medical imaging.

The Imaging Informatics Lab focuses on extracting knowledge from medical imaging data. Medical imaging data contains vast amounts of anatomical and physiological information to diagnose, classify, and understand the progression of disease. Our group specializes in combining the deep insights gleaned from medical imaging data, e.g., a tumor volume, with medical record data, e.g., patient demographics and genetic history, through standards-based data models to enable this multimodal research.
Our team includes imaging scientists who focus on deep learning methods for medical image segmentation, clinical imaging specialists in Radiology and Ophthalmology, and informaticists. Our group includes Paul Nagy, PhD, a Fellow of the Society of Imaging Informatics in Medicine (SIIM), past society Chair, and past Chair of the College of SIIM Fellows; and Teri Sippel Schmidt, MS, a Fellow of SIIM and a recipient of the SIIM Gold Medal. Our projects focus on integrating the DICOM imaging data with EHR data using the OHDSI Common Data Model (CDM) to generate open, reproducible, computational evidence.
Sample Projects:
- Medical research should be transparent and reproducible. Our lab is focusing on replicating published Alzheimer’s research papers, at scale, using AI for image segmentation, and combining this information with clinical information from the EHR using the OHDSI OMOP Imaging Extensions. Recently, we implemented the OHDSI Imaging Extensions to include DICOM image attributes. Combining EHR and DICOM criteria, we were able to computationally identify a cohort with inclusion and exclusion criteria. We used the OpenMapT1 segmentation algorithm to quantify hippocampus volumes and stored that quantified imaging data into the OMOP CDM, comparing the hippocampus volumes to the clinical cognitive scores from the EHR. This is novel because cohort, segmentation, correlations to clinical data, and analysis were automated, significantly reducing time and effort, replicating papers on public data sets, and creating transparent research.
- We are currently incorporating all of the CT and MR neuro imaging data which is available at Johns Hopkins, for patients over the age of 55, and adding the DICOM metadata to the existing OMOP Common Data Model (CDM) database, thereby combining demographic and clinical data with imaging. Once this project is complete, researchers will be able to quickly and easily create patient cohorts which include imaging studies and measurements. Sharing aggregated research results, which include imaging, will be enabled with global research partners.
These processes began with Alzheimer’s and neuro, these methods will be applicable to every clinical issue, including the Johns Hopkins Cancer Informatics Lab.
Recent Grants
Recent Publications
- Coming soon: Jen Park, et all
- Development of Medical Imaging Data Standardization for Imaging-Based Observational Research: OMOP Common Data Model Extension. J Digit Imaging. Inform. med. 37, 899–908 (2024).
- Advancing Medical Imaging Research Through Standardization: The Path to Rapid Development, Rigorous Validation, and Robust Reproducibility. Investigative Radiology ():10.1097
Faculty Involved
Blake Dewey, PhD
Paul Nagy, PhD, FSIIM
Teri Sippel Schmidt , MS, FSIIM
Brad Genereaux, FSIIM
Recommended Courses
In addition to the Core Courses, recommended Imaging Informatics Lab courses include:
- ME.250.782 Observational Health Research Methods on Medical Records (OMOP)
- ME.250.783 Imaging Informatics
- ME.250.XXX Deep Learning of Medical Imaging