AI Health Spark Seminar Series: Multi-site studies of aging with diffusion weighted MRI: Challenges and opportunities with harmonization

December 6, 2022
12:00 pm to 1:00 pm
Virtual

Event sponsored by:

AI Health
+DataScience (+DS)
Biostatistics and Bioinformatics
Computer Science
Department of Neurology
Department of Radiology
Duke Institute for Brain Sciences (DIBS)
Electrical and Computer Engineering (ECE)
Pratt School of Engineering
Psychology and Neuroscience

Contact:

Duke AI Health

Share

Speaker:

Bennett Landman, Ph.D., Professor and Department Chair, Electrical and Computer Engineering (primary), Computer Science, Biomedical Engineering, Vanderbilt University, Principal Scientist of ImageVU, Vanderbilt University Institute of Image Science, Radiology and Radiological Sciences, Vanderbilt Brain Institute, Psychiatry and Behavioral Sciences, Biomedical Informatics, Neurology, Vanderbilt University Medical Center Biomarker Core Co-Lead, Vanderbilt Alzheimer's Disease Research Center with host Maciej Mazurowski, PhD; Associate Professor in Radiology, Duke University
Patterns of altered brain white matter strongly correlate with neurological disease progression and may serve as early markers for future cognitive changes. To study these patterns, diffusion weighted magnetic resonance imaging (DW-MRI) has been routinely included in national-scale neuroimaging studies of Alzheimer's Disease and aging. Yet, quantitative investigation of DW-MRI data is hindered by a lack of consistency and difficulty of interpretation of potential biomarkers. We approach the challenge of biomarker estimation at the intersection of image acquisition, signal modeling, and image analysis fields. We will discuss image analysis approaches to robustly characterize microstructural, macrostructure, and connectivity. There are consistent advantages in data-driven approaches for interpretation of DW-MRI at all levels. For pre-processing, generative networks substantially improves over state-of-the-art techniques for geometric correction. For signal modeling, contrastive learning extraordinarily robust model-free interpretation of voxel-wise DW-MRI data. For connectomics, data driven approaches provide consistent anatomical context. We find that emerging diffusion MRI deep learning techniques offer the promise of biomarkers for white matter state that are robust and reproducible across scanners, sites, and studies. This session is a part of the monthly seminar series organized by Spark: AI Health Initiative for Medical Imaging. The seminar will highlight outstanding work in medical imaging at Duke and beyond. The seminar recordings will be publicly available. The Spark initiative focuses on development, validation, and clinical implementation of artificial intelligence algorithms for broadly understood medical imaging by bringing together the technical and clinical expertise across Duke campus. For more information please contact Dr. Maciej Mazurowski (maciej.mazurowski@duke.edu).