Skip to main content

Big Data Science and Engineering

Big Data

Big Data, Big Goals

The answers to untold health questions lie in the routine scans Vanderbilt University Medical Center Radiology captures every day.

Think about it: Hundreds of patients checked per day, year after year, revealing what a perfectly healthy eye or chest looks like or providing an image that will be the first harbinger of a complicated diagnosis.

Bennett Landman’s big data work lies in capturing information that one day will allow physicians across the country to compare quickly the scans of a particular patient with scans of other patients, looking for similarities. With a list of close matches, the physicians could see the diagnosis and treatment results for others and learn what might work for their patient.

Big DataLandman, assistant professor of electrical engineering and computer science, has been laying the groundwork for this research—funded by a $436,000 National Science Foundation CAREER Award—since his arrival at Vanderbilt in 2010. He says that this kind of revolutionary medical technology is only possible because he can leverage the unique trans-institutional capabilities of Vanderbilt’s BioVU Program,  Department of Biomedical Informatics, Advanced Computing Center for Research and Education, Center for Computational Imaging and Vanderbilt University Institute of Imaging Science.

First, identifiers had to be stripped from thousands of radiological images of Vanderbilt patients. Those were replaced with anonymous avatars bearing information about health history, but no personally identifiable information.

In the next and current step, Landman and his team need to write reliable, open-source computer code that can be used by researchers across the globe and won’t break down when faced with minor discrepancies from scan to scan.

But a major hurdle remains: The sheer size of the scans, typically starting at 10 megabytes apiece. Vanderbilt collects countless terabytes of data per year in the course of regular health care.

Landman says that determining how to organize that information in a useful way is key. He has paired with the Medical Center’s Department of Radiology and the Department of Biomedical Informatics to construct systems that allow retrieval of that data.

Big DataWith his latest grant, he’ll start by focusing on the Medical Center’s littlest patients with some of the toughest diseases by enabling the analysis of tens of thousands of brain images from children and teenagers.

Analyzing brain scans of children is complicated. Because children change so rapidly and grow on such different timelines, it’s tougher to determine what’s abnormal and for doctors to find those abnormalities.

“Can we get all of those data in a space and examine what the different trajectories look like?” Landman asks. The new project will create the image processes that allow those computations to happen. To do that, his team will tap into a combination of the cloud computing structure Hadoop and a picture-archiving and communication system.

“We’re going to construct lowcost technologies that allow us to perform those types of queries,” he says. “We can’t just buy a machine big enough to store all that data on a research budget.”


This work is supported by NSF CAREER Award 1452485.

Top Photo: Bennett Landman and graduate student Yuankai Huo  work on a big data project that involves constructing processes that allow for the analysis and comparison of pediatric brain scans.