NUMBER FIVE: HARNESSING THE DATA IN NEUROIMAGING

Most engineers and clinicians at Johns Hopkins have a solid grasp of basic statistical techniques. But cutting-edge studies of medical imaging or robotics sometimes involve data sets that are so complex that they require guidance from top-level statisticians.

That’s where Brian Caffo, a professor of biostatistics at the Bloomberg School of Public Health, comes in.

Caffo’s lab has collaborated with physicians and engineers throughout Johns Hopkins on a huge range of problems—most recently on neuroimaging studies that aim to shed light on the structure and function of the brain.

“The people we work with are really smart,” Caffo says. “They know a lot of statistics. They know how to do most of this on their own. So what they come to us with are their most difficult data sets.”

In the last two years, Caffo has collaborated with clinicians and neuroscientists on brain-imaging studies involving autism, motor learning, attention disorders, and football-related concussions.

What makes brain imaging so statistically challenging, Caffo says, is the sheer scale of the data. “You have complex spatial relationships in three dimensions, because certain areas of the brain are functionally related to their counterparts in the opposite hemisphere. You also have complex temporal relationships. On top of that, these studies are beginning to involve larger numbers of subjects. Functional MRI studies used to include just a few patients, but now there can be hundreds.”

Neuroimaging is far from Caffo’s only clinical interest. He also has recently provided statistical muscle for studies of sleep-disordered breathing, infection prevention in ICUs, and retinal damage in multiple sclerosis.

Successful collaboration with clinicians, Caffo says, requires all parties to have a basic grasp of one another’s language. “I tell my graduate students, if you want to work with scientists and you don’t just want to be a tourist in the field, you’ve got to have at least a basic understanding of their clinical areas. For example, you can’t walk into a neuroimaging project without knowing basic brain anatomy and function.”

Much of Caffo’s work is in statistical method or theory. But that work is grounded in and motivated by his applied, collaborative work across Johns Hopkins. “I work alongside the best research hospital in the world and the best school of public health,” he says. “It would be a waste for me and the institution if I didn’t take advantage of that.”