Fine-Tuning MRI Techniques to Help Patients
July 25, 2014
Advances in neuroimaging have revolutionized the way we look at and understand neurodegenerative diseases and have laid the foundation for developing new ways to help patients who suffer from them.
These advances have given us unique insight into the development and progression of brain lesions, brain atrophy and more targeted therapeutic responses. Detecting such brain changes and unraveling patterns in them using different MRI techniques is the focus of Vasiliki Ikonomidou's work.
Ikonomidou is an assistant professor in Mason's Bioengineering Department and has a joint appointment in the Electrical and Computer Engineering Department in the Volgenau School of Engineering.
She came to Mason in 2009 from the National Institute of Neurological Disorders and Stroke (NINDS) where she had worked under Jeff Duyn in the Laboratory for Functional and Molecular Imaging, Advanced MRI Section, developing high-contrast anatomical MRI techniques. From 2006 on — as part of the Neuroimmunology Branch of NINDS — she specialized in brain imaging, using MRI and positron-emission tomography, in patients with multiple sclerosis (MS).
Ikonomidou calls her general research area "Data Mining in Neuroimaging Datasets" — the extraction of patterns from neuroimaging datasets that are indicative of disease or its progression.
Her research involves big data on many different levels — myriad images consisting of gigabytes of data from which features are extracted, which, in turn, are broken down into categories such as volume, intensity values and levels/growth of lesions.
Using image and signal processing, as well as data mining techniques, she and her students sift through these endless numbers to find algorithms, hoping that these searches will at some point be automated and programmed to help doctors deliver fast and reliable diagnoses.
But big data is not the only challenge in her research: "MS is a kind of random disease," says Ikonomidou. Not only is it hard to project outcomes, but there are always unexpected surprises in the way MS takes its course in patients, although "there is some order there," she says.
Using dimensionality reduction techniques, she and her colleagues at the University of Maryland were able to detect progression patterns independent of the more random lesion development. And using data from a clinical trial performed by collaborators at the University of Padova, students at the Neuroimaging Laboratory are working on developing automated algorithms that characterize lesion progression over time.
Ikonomidou is currently partnering with Mason's Sports Medicine Assessment Research and Testing Laboratory to conduct research on concussions. She is also working with her students on data mining in neuroimaging datasets. In collaboration with the U.S. Army's Night Vision Lab, she is using the same data mining techniques for human signature detection. Ikonomidou hopes to come up with techniques for better images, which could lead to automated pattern detection for various diseases.
Ikonomidou teaches BENG 220 Biomedical Systems and Signals, BENG 499/ECE 590 Medical Image Processing, and ECE 699 Medical Image Analysis.
This article, written by Claudia Borke, originally appeared in the "Bioengineering at Mason" newsletter.