Logo IMG


Hiding in Plain Sight

New technology allows us to see neural networks more clearly

Catherine Clabby

A brain is an organ of connectivity. Neural fibers that are only microns wide—but sometimes meters long—connect its highly specialized regions. The anatomy of those connections is largely unknown, especially in humans. That doesn’t mean it is complex; it may be simple. But we haven’t had a means to observe it. Diffusion spectrum magnetic resonance imaging (DSI) technology may change that. Van Wedeen, an associate professor of radiology at Harvard Medical School who is based at the Athinoula A. Martinos Center for Biomedical Imaging, is developing and testing that tool. He explained its promise to American Scientist associate editor Catherine Clabby.

2010-09SightingsFA.jpgClick to Enlarge ImageA.S. What is diffusion spectrum imaging?

V.W. Diffusion MRI refers to a body of methods that map, noninvasively, the fiber architecture of a living organism’s brain. All these methods use the three-dimensional patterns of the random diffusion of water to trace microscopic tissue structure. The first, diffusion tensor imaging, looked promising but did not record crossing pathways. The brain has extensive systems of crossing fibers, which are essential for efficient circuits. To address this, we invented several more methods 10 years ago, including diffusion spectrum MRI (DSI). DSI acquires a detailed image of the three-dimensional pattern of water diffusion by measuring diffusion in dozens to hundreds of directions. This yields a six-dimensional image of the brain: three dimensions of anatomical location and an additional three dimensions for the distribution of water flux at each location. Just as they do in high-energy physics, extra dimensions can make a problem simpler. We’re now assessing the methods and what they reveal about brain structure and design.

A.S. With colleagues, you used DSI to explore this connectivity in a cat, 10 days and then three months after its birth. Why?

V.W. Cats are very unusual. They have large complicated brains that, like the brains of primates, including humans, have extensive folding, or gyri, to increase the surface area of the cortex. They have an especially large visual cortex. But unlike primates, cats are born with extremely simple, non-folded, immature brains. Most of their brain development occurs postnatally. This makes cats an ideal model to study brain development, especially development of the cerebral cortex. Ellen Grant and Emi Takahashi, my colleagues at Boston Childrens’ Hospital, are using such studies to explore how development goes wrong in children and how that may produce diseases such as epilepsy.

A.S. How much certainty is there in this imaging?

V.W. Diffusion MRI has two steps. First we perform a diffusion MRI scan in which we obtain images of the rate of diffusion in hundreds of directions. That is the six-dimensional data set. Second, we compute the simplest or most probable model of fiber pathways consistent with the measurements. The problem of the relative certainty about the results of these methods is very real. In rhesus monkeys, we can compare the DSI results with results of invasive tract-tracer studies that inject dyes directly into their brains. There, we’ve found the MRI to be reasonably accurate. We can’t verify the human studies in this way, so uncertainty remains a problem. Over the past year or so, I’ve come to the view that the true structure of the brain may be highly constrained, so much so that the difference between accurate and inaccurate data reconstructions should be evident within each study.

A.S. Have these images changed the way you think about brain structure?

V.W. Diffusion MRI studies suggest that the brain is far more orderly—better organized in three dimensions—than others methods have suggested. Often, this order can’t be explained. For example, in every mammalian species we’ve looked at—in rats, cats, humans or New and Old World monkeys—the fibers pathways of the hippocampus form sheets of counter-rotating spirals in two layers. This seems far too well conserved, and too beautiful, to be accidental. If you ask an engineer why a bird wing, an eye or a heart ventricle has its specific structure, you get a fairly good explanation. In the case of the brain, we have little idea. But MRI may lay bare the logic of brain structure.

A.S. What must be overcome before you can scan the brains of living people?

V.W. Diffusion makes very unusual demands on scanners compared with previous MRI methods. Today’s scanners are far from optimized. An optimal diffusion scanner would incorporate the strongest possible gradient coils, which are the heart of an MRI scanner. What field strength is to other MRI techniques, gradient strength is to diffusion. This is a very challenging problem. You need to inject megawatts of current inches from someone’s head without producing excessive heat, noise or vibration and with accuracy better than 1 part in 100,000 over thousands of repetitions. We foresee tenfold improvement over today’s technology in the near future. On the algorithmic side, we are only partway to understanding and harvesting the full structural order within the brain. This may have as great or even greater impact than imaging technology. A long-range goal is to map brain structure variability among people and the possible relation between that and differences in people’s character, plasticity, adaptability, vulnerabilities and mental, behavioral, and emotional aptitudes. The National Institutes of Health this year sponsored a “Human Connectome Project” with these aims. The imaging community expects great progress within a decade.

In Sightings, American Scientist publishes examples of innovative scientific imaging from diverse research fields.

comments powered by Disqus


Subscribe to American Scientist