Skip to content. Skip to navigation
CIM Menus

4D Graphs for Longitudinal Segmentation

Ipek Oguz
Department of Radiology University of Pennsylvania

March 23, 2017 at  2:00 PM
McConnell Engineering Room 437


An accurate, sensitive and reproducible method for longitudinal segmentation is crucial for a variety of medical imaging studies for assessing trajectories of healthy maturation, disease progression and response to treatment. This talk will present a novel 4D graph segmentation approach for joint segmentation of longitudinal datasets in a globally optimal manner. Our 4D approach produces highly accurate surfaces in a temporally consistent manner by leveraging all of the available longitudinal data for a given subject. The 4D approach enforces temporal coherence of the segmented surfaces but avoids over-regularization in the temporal domain. This leads to accurate quantification of the anatomy, which is less noisy and more robust than existing 3D or semi-4D approaches.

Our method will be illustrated in two applications: longitudinal choroid segmentation from retinal OCT images, and longitudinal cortical thickness analysis from MRI. The relatively flat geometry of the choroid allows for a variety of synthetic experiments. The more complex cortical surface reconstruction task illustrates how to apply this technique in more challenging applications. The results show that our joint 4D segmentation approach reduces measurement noise without bias. This leads to higher sensitivity and thus more statistical power in longitudinal studies, increasing the value of imaging studies for obtaining longitudinal biomarkers.


Dr. Ipek Oguz is a Research Associate in the Department of Radiology at the University of Pennsylvania, where she is a member of the Penn Image Computing and Science Laboratory (PICSL). She received her Ph.D. in Computer Science at the University of North Carolina at Chapel Hill. Her research is in the field of medical image analysis and specifically in the development of novel methodology for quantitative medical image analysis, with applications to neuroimaging, including Huntington’s disease and multiple sclerosis, as well as placenta imaging and ophthalmology. Her technical interests include graph-based segmentation methods and longitudinal studies. She has co-authored more than fifty peer-reviewed journal and conference publications. She is an executive in the Women in MICCAI Committee and a co-organizer of IPMI 2017.