Daniela Raicu
DePaul University, IL·Computer Science
Computers can now help doctors read medical images and flag possible disease, but most of these tools are rarely used in hospitals. A major reason is that many work as 'black boxes': they give an answer without showing why, so clinicians struggle to trust them.
- Fields
Computer Science
Biomedical Engineering
Public Health & Epidemiology
- Topics
computer-aided diagnosis
explainable AI
deep learning
- Grants
- 2 known grants, through May 2028
What they work on
- Develop a Semantic Deep-learning Neural Network (SDNN), a deep-learning model enhanced with knowledge of what visual features mean, that generates visual patterns and example images resembling the concepts radiologists use to read lung nodules.
- Work with an expert panel to determine which SDNN-generated visual patterns and example images are most informative for diagnosis, so they can serve as explanations for computer-aided lung nodule diagnosis.
- Link Electronic Medical Records with social determinants of health data and build models that predict Emergency Department readmissions for diverse patient groups at the Sinai Chicago safety-net hospital system.
- Design and evaluate an analytics dashboard for community health workers, using human-computer interaction methods to refine prediction models and training based on their feedback.
More about the research
More broadly, data tools often fail to reach the people who need them most. These two projects ask how to make data-driven health technology understandable and useful to the people who rely on it. One focuses on radiologists diagnosing lung nodules, and the other on community health workers serving high-risk patients.
Why it matters
Radiologists could gain diagnostic aids that explain their reasoning, which may build trust, reduce diagnostic errors and improve detection of lung cancer. Separately, safety-net hospitals and community health workers on Chicago's South and Southwest Sides could use better data tools to find high-risk patients, target resources and reduce Emergency Department readmissions. This could make care more sustainable in under-resourced communities.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Nabil Kahouadji
Northeastern Illinois University, IL·Mathematics & Statistics
Test the triple discriminant scoring methodology, a classification method built on many simulations across subsets of variables, for predicting colorectal cancer outcomes from Chicago electronic medical records and socio-environmental data.
Jeanne Shen
Stanford University, CA·Medicine & Clinical Research
Build a publicly available, expert-annotated multimodal colorectal cancer dataset. It will combine pathology whole-slide images, preoperative CT and MRI scans, and structured electronic medical record data.
Joann Elmore
University of California-Los Angeles, CA·Medicine & Clinical Research
Run a two-year pragmatic randomized trial of more than 154,000 3D screening mammograms assigned to interpretation with or without an FDA-cleared AI tool, measuring cancer detection and recall rates across 16 breast imaging facilities.
The school
DePaul University: 4 labs on PI Pages, $11M a year on research, 47 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.