Medical diagnostics includes a wide range of complex tasks such as detecting and localizing disease, estimating risk, predicting outcomes, and determining efficacy of treatments. Data-driven methods have increasingly been applied to diverse medical decisions, although the promise of AI in healthcare is largely unrealized despite the hype. This talk presents a survey of our lab’s recent efforts to apply data driven solutions for: a) image generation from tomographic systems; b) rib fracture detection and localization in pediatric radiographs; and c) overcoming multimodal learning challenges in cancer detection.
This work relies on merging computer vision methods with conventional approaches reliant on model-based objective functions and optimization techniques.
Adam Alessio is a professor in the departments of Computational Mathematics, Science, and Engineering (CMSE), Biomedical Engineering (BME), and Radiology and is the Interim Chair of BME at Michigan State University. He received his Ph.D. in Electrical Engineering at the University of Notre Dame and post-doctoral training at the University of Washington, where he spent 15 years on faculty in Radiology. As an imaging scientist and nuclear medicine physicist, he has authored of over 90 peer-reviewed publications, holds 6 patents, and has grant funding from the NIH to advance non-invasive cardiac and cancer imaging. Current efforts center on translational medical research projects for topics including machine learning for quantitative diagnostics and prognostics, interpretability and generalizability of machine learning, cardiac perfusion estimation, and quantitative PET/CT imaging. Details on Dr. Alessio’s background can be found at https://www.egr.msu.edu/~aalessio/.