Philadelphia University + Thomas Jefferson University

Connect-the-Docs: Facts of the White Matter

file
An example of the kind of MR tractography JIMRIC's researchers create

The brain is one of the great black boxes of science. Researchers have long known of its role in producing mental activity, but very little about how the parts fit together and give rise to consciousness. Until MRI arrived on the scene to offer a repeatable, non-invasive method of observation, lobotomies and case studies of patients with brain injuries were the only ways to peek inside the box.

Enter the newly-minted Jefferson Integrated Magnetic Resonance Imaging Center (JIMRIC), which has been making headway in understanding the vast networks between our ears and throughout our bodies.

“Our goal is to integrate all aspects of MR imaging, so clinicians and researchers across the Jefferson enterprise can have an accessible resource for scientific & technological consultation,” says Feroze Mohamed, PhD, a professor of radiology and the Center’s founding director. Spawned by a close partnership between the Department of Radiology and the Vickie & Jack Farber Institute for Neuroscience, JIMRIC is bringing advanced computational imaging out of the cloud and into the hands of clinicians and neuroscientists.

“Images are essential to the work we do. Because of how tightly connected brain structures are, we really need to know exactly where we’re going,” says Chengyuan Wu, MD, MSBmE, a neurosurgeon and JIMRIC’s clinical director. 

fileL-R: Drs. Feroze Mohamed, Chris Conklin, Chengyuan Wu

MRI takes advantage of quantum mechanical effects of the protons in the human body by applying electromagnetic field pulses in the presence of a strong, static magnetic field.  This excites the body’s hydrogen atoms and causes them to emit a radio frequency signal that is then collected by receivers near the region of interest. This information does not emerge as a photographic negative, but generates images representing specific anatomical features.

In the “post-processing” stage, biophysicists sort through the mathematical morass to create clear, quantitative, clinically useful images. Underlying the procedure is an arsenal of complex equations and algorithms, which analyze and show relationships between voxels, a kind of three-dimensional pixel.

In the case of JIMRIC, Drs. Mohamed and Wu are interested in white matter networks, the connective wiring between brain regions. “White matter is the body’s information superhighway,” says Chris Conklin, PhD, JIMRIC’s associate director and MR scientist working on characterizing spinal cord white matter. “It is found throughout the central nervous system and is the primary constituent of the spinal cord.”

The Center’s primary focus is in functional MRI and its biophysicists are adept at an fMRI technique called Diffusion Tensor Imaging. For this application, tensors are the geometric constructs that describe the magnitude and displacement of water molecules from multiple directions in the body. Water moves along the path of least resistance, which in the brain and spine means along the length of axons, the long, spindly appendages that grow from nerve cells.

Measuring the rate and direction of water diffusion can tell imagers a lot about a neuron’s function. For instance, fast diffusion could indicate a problem with myelin, the protein-lipid sheath that acts as insulation for axons. This information is correlated into models that enable neurosurgeons to decide in advance what course their interventions will take.

file
Drs. Chen and Mohamed use the lab's supercomputer to generate and interpret images

White matter tractography, the process of generating functional maps of the central nervous system, is computationally intensive, requiring imagers to precisely calculate millions of variables distributed not only over space but also through time in the case of fMRI. Previously, Wu relied on cloud computing services offered by third party vendors. This proved expensive and inefficient, so a new solution was sought—one that he built himself.

“The mini supercomputer was really born out of equal parts necessity and enjoyment,” he says of the custom-built post-processing machine he and Mahdi Alizadeh, one of the JIMRIC’s doctoral candidates, put together. “We want to use these techniques to actually help people, so we had to find a way to get all this processing done quickly.”

The hardware of a typical computer is run off of the central processing unit (CPU), which runs calculations in a linear, step-by-step way. But, the Center’s computer in based on the graphics processing unit (GPU), circuits designed to convert data into visual representations. While CPUs process things linearly, GPUs perform operations in parallel, breaking down a problem into many small parts and then solving them simultaneously.

JIMRC’s supercomputer, which runs 10,000 GPU cores, is able to process in a matter of hours images that would take days on a lesser machine. But speed isn’t an end in itself. “This is great because my interest in advanced imaging techniques is driven by watching people like Chen apply them in his clinic,” says Mohamed.

fileX-ray images of stereo EEG probs
file
Jefferson neurosurgeons used advanced imaging to plan their interventions

Wu and his Farber Institute colleagues can better understand brains of epilepsy patients and minimize procedures like stereoelectroencephalography (stereo EEG), which involves implanting electrodes into an epilepsy patient’s brain to measure electrical activity and determine the locus of seizures. Rather than take off the back of a patient’s skull, increasing the risk of infection, Wu can implant the sensory nodes through precisely placed millimeter-sized holes.

JIMRIC is also making possible new epilepsy treatments like asleep deep brain stimulation (DBS), a surgery of last resort for patients whose condition resists medication. In the past, neurosurgeons had to operate while patients were awake to avoid severing important neural connections, prodding carefully around the exposed brain, while listening to patients report on their perceptions.

Armed with stereo EEG results and an individualized MR map of a patient’s brain structure and function, Wu can plot a course to the area of interest without harming unrelated networks and vital anatomy. Plus, with sedation now an option, patients are less fearful and more likely to consent to the procedure, the essential first step in any treatment plan.

file
The End Result: Wu operates on an Epilepsy patient

An essential piece of medical practice is knowing what normal looks like. There must be a healthy standard to compare individual cases to. JIMRIC’s Dr. Mohamed is leading a partnership between Jefferson and the University of Montreal to create a pediatric spinal cord template. The model, a quantitative “average” of numerous MR studies of healthy children’s spinal cords, minimizes individual variation, revealing structural and functional commonalities.

Currently, no such standard exists, which makes assessment difficult, as diagnosing physicians have a limited number of cases to compare an injury to. However, this is changing, as JIMRIC and its collaborators are putting their heads—and data sets—together to increase their study population and provide a better resource to clinicians.

Like the white matter it studies, the Center’s purpose is to act as a connector—between research and the clinic, machine and patient, physicist and physician. As technology improves, MRI will become more applicable to more areas of medicine; JIMRIC is at the frontiers of this expansion. Already, it is going beyond the brain through nascent projects in support of cardiothoracic imaging efforts.

The old saying about a picture being worth a thousand words holds true here, but these hi-tech images are also worth a thousand man hours. While Jefferson researchers continually shorten the time it takes for an image to be processed by pouring vast expertise and hours into perfecting the craft that makes the pictures possible.