The Autistic Brain: Thinking Across the Spectrum (6 page)

Heterogeneity of behaviors.
Conversely, when researchers find an anomaly in the brain, they can’t be sure that that anomaly will have the same behavioral effect in a different brain. Or any effect, for that matter. Just because you have an enlarged amygdala doesn’t mean that you’re autistic.

But what if it did?

Not necessarily an enlarged amygdala. But what if some neuroanatomical finding or combination of them could serve as a reliable diagnostic tool? A diagnosis based not on behaviors alone but on biology as well would make a big difference in predicting deficits and targeting treatments. Doctors and researchers could:

• Apply early intervention, even in infancy, when the brain is still highly susceptible to being rewired.
  
• Target areas in the brain more locally, rehabilitating parts of the brain that they think they can help and not wasting time on parts that are unrecoverable.
  
• Test new therapies and monitor existing therapies more narrowly.
  
• Tailor a prognosis to an individual patient on a case-by-case basis.
  

For the patient, such a diagnosis would have a tremendous psychological benefit as well, by allowing him or her to know what’s actually unusual. Personally, I
like
knowing that my high level of anxiety might be related to having an enlarged amygdala. That knowledge is important to me. It helps me keep the anxiety in perspective. I can remind myself that the problem isn’t
out there
—the students in the parking lot under my bedroom window. The problem is
in here
—the way I’m wired. I can medicate for the anxiety somewhat, but I can’t make it go away. So as long as I have to live with it, I can at least do so secure in the knowledge that the threat isn’t real. The
feeling
of the threat is real—and that’s a huge difference.

Given the obstacles to investigating autism from a neurological perspective—the homogeneity of brains, the heterogeneity of behaviors and causes—you might ask whether finding a biomarker is a realistic goal. Yet in recent years, researchers have made tremendous progress toward reaching that goal, and now many speak of
when,
not
if.

“We still don’t have a litmus test for autism,” the neuroscientist Joy Hirsch said. “But we have a basis for it.”

As the director of the Functional MRI Research Center at the Columbia University Medical Center in New York City, Hirsch has tried to build that foundation in the search for a litmus test. In a study
her group conducted between 2008 and 2010, fifteen autistic subjects ranging in age from seven to twenty-two and twelve control children ranging from four to seventeen underwent fMRI scans of the superior temporal gyrus—the part of the auditory system that processes the sounds of speech into meaningful language. “The most obvious disability in autism is the disability of speech,” she said, regarding the rationale behind the experiment. “Our hypothesis was that at the first stage we could begin to see differences.” And they felt they did: Their measures of activity in that region could identify fourteen out of fifteen of the autistic subjects, a sensitivity rate of 92 percent. (Other researchers have questioned the reliability of comparing subjects who were awake and subjects who were sedated—factors that Hirsch’s team felt they accounted for. As always in science, further tests will or will not reinforce the validity of the findings.)

Another way that research groups are searching for a biomarker is by taking a sample of autistic and control subjects, focusing on one aspect of the brain that the researchers have reason to believe is associated with autistic behavior, and seeing if they can create an algorithm that can tell one kind of brain from another. Jeffrey S. Anderson, from the University of Utah, offers this simplified description: “We use a whole bunch of normal brains and brains of individuals with autism, and we make a template of each one”—of autistic brain and neurotypical brain—”and we take a new subject in and just ask, ‘Well, which one does it match more?’”

The point isn’t to identify this brain or that brain as belonging to an autistic person or a neurotypical. It’s to find an aggregate that could help identify areas of potential interest that might be biomarkers.

In a major study
that Anderson’s group published in 2011, the aspect of the brain under consideration was connectivity. The earlier studies indicating that autistic brains tend to have local overconnectivity and long-distance underconnectivity had focused on a small number of discrete brain regions. Anderson and his colleagues instead studied the connectivity of the entirety of the gray matter. Using a variation of fMRI called functional connectivity MRI, they obtained connectivity measurements among 7,266 “regions of interest.” In a group of forty male adolescents and young adults with autism and a like sample of forty typically developing subjects, Anderson found that the connectivity test could identify whether a brain was autistic or typical with 79 percent accuracy overall and 89 percent accuracy for subjects who were under the age of twenty.

That level of accuracy is consistent with results from other research groups. A 2011 MRI study
from the University of Louisville found that in a sample of seventeen autistic and seventeen neurotypical subjects, the length of the centerline of the corpus callosum could be used to distinguish between the two types of brains with a level of accuracy ranging from 82 percent to 94 percent, depending on statistical confidence levels.

In another MRI study
from 2011, researchers at the Stanford University School of Medicine and Lucille Packard Children’s Hospital looked not at the size of an individual part of the brain, as structural MRI studies usually do, but at the topology of the gray matter’s folds—the brain’s cliffs and valleys. In a sample of twenty-four autistic children and twenty-four typically developing children (all aged eight to eighteen), they identified differences between the two groups in the default mode network, a system associated with daydreaming and other brain-at-rest, nontask activities. The study subjects whose brains showed the greatest deviations from the norm also exhibited the most severe communication deficits. Volume measurements of the posterior cingulate cortex in particular achieved an accuracy rate of 92 percent in telling one kind of brain from the other.

Accuracy rates in the 80 to 90 percent range are not high enough for researchers to claim they’ve discovered a marker for autism, but it’s progress of a sort that would have been difficult to imagine only a decade ago. And it’s certainly high enough to inspire confidence in the algorithmic approach.

One of the goals for further research is to adapt these techniques to younger subjects. As Utah’s Anderson says, “It’s not really helpful to diagnose a teenager with autism, because we already know it.” The younger the subject, the earlier the possibility of intervention. The earlier the intervention, the greater the potential effect on the trajectory of an autistic person’s life.

Just how young a person in the scanner can be depends in part on the technology. Functional MRI, for instance, requires responses to stimuli that create brain activity, so children need to be old enough (and, of course, to possess the neurological capacity) to understand the stimuli. Structural MRI, including DTI, doesn’t rely on brain activity, so it allows researchers to study subjects who are even younger—so young, in fact, that they might not exhibit behavioral signs of autism yet.

That was the case in a 2012 DTI study
led by researchers from the University of North Carolina at Chapel Hill. The participants were ninety-two infants who all had older siblings diagnosed as autistic and therefore were thought to be at high risk themselves. Researchers scanned the subjects’ brains at six months, then followed up with a behavioral assessment at twenty-four months (as well as further scanning in most cases). At that point, twenty-eight of the subjects in the study met behavioral criteria for ASD, and sixty-four did not. Did the white-matter fiber tracts of one group exhibit any differences from the tracts of the other group? Researchers concluded that in twelve of the fifteen tracts under investigation, they did. At the age of six months, the children who later developed autistic symptoms showed higher fractional anisotropy (or FA, the measure of the movement of water molecules through the white-matter tracts) than the rest of the children. Usually that would be a good sign; a higher FA indicates a stronger circuit. But by age twenty-four months, those same children were showing lower FA, a sign of a weaker circuit. Why were those same circuits stronger at six months than those of the children who were developing typically? Were they even stronger even earlier? The researchers don’t have an answer, but they do have a new goal: three-month-olds.

Another goal for further research is to look at the brain in even finer detail. Fortunately, the future is already here. I know, because I’ve seen it.

Actually, I’ve been
inside
the future—a radically new version of DTI called high-definition fiber tracking. HDFT was developed at the Learning Research and Development Center at the University of Pittsburgh. Walter Schneider,
senior scientist at the center, explains that HDFT was underwritten by the Department of Defense to investigate traumatic brain injuries: “They came to me saying, we need something that can do for brain injury what X-rays do for orthopedic injury.”

When the research team posted a paper
on the
Journal of Neurosurgery
’s website in March of 2012, the technology got a fair amount of media attention. The paper reported on the case of a thirty-two-year-old male who had sustained a severe brain injury in an all-terrain-vehicle accident. (No, he wasn’t wearing a helmet.) HDFT scans revealed the presence and location of fiber loss so precisely that the research team accurately predicted the nature of the lasting motor deficit—severe left-hand weakness—“when other standard clinical modalities did not.”

“Just like there are 206 bones in your body, there are major cables in your brain,” Schneider says. “You can ask most anybody on the street to create a drawing of what a broken bone looks like, and they would be able to draw something somewhat sensible. If you ask them, ‘So what does a broken brain look like?’ most people—including researchers in the field—can’t give you the details.”

Including researchers in the field? Really?

“A fuzzy image of bones doesn’t give you a clean diagnosis,” Schneider says. “We took diffusion tensor imaging, and made it so it can.”

While the focus of HDFT research so far has been on traumatic brain injuries, Schneider’s long-range plan is to map the information superhighways of the brain. For years I’ve compared the circuitry of the brain to highways, and I’m hardly alone. But the
high-definition
part of HDFT technology has revealed just how apt the superhighways reference is.

Regular DTI technology shows the highways and off-ramps and crossroads of your brain as if they were all on a two-dimensional map. That kind of map is useful if you want to know whether a fiber gets from here to there. It can show you where I-94 and County Road 45 are in close proximity to each other. It can show you that they crisscross. But it can’t show you how they crisscross. Do they intersect, like a crossroads? Or does one road go over the other, like an overpass? The old technology can’t answer that question. HDFT can.

And it
tracks
the fibers. It keeps them individualized over long stretches.

And it tracks the fibers
farther
than any previous technology—all the way to the end of the road.

It even shows if a damaged circuit still has continuity or if it’s stopped transmitting. (As a biologist, I’m just freaking out, it’s so cool.)

I don’t want to overhype HDFT. It’s incredibly important, but it’s not going to solve all the mysteries of the brain. As Schneider says, “One of my favorite lines of neuroscience is if you can think of five ways for the brain to do something, it does it in all ten. The five you’ve thought of, and the five you haven’t thought of yet.” Still, HDFT is going to have a major impact on diagnoses involving brain trauma.

First, the diagnoses are going to be more precise. The existing state-of-the-art DTI scanner collects data from 51 directions. HDFT collects data from 257 directions. As a result, HDFT doesn’t just tell you what section of the brain has been damaged. It tells you what specific fibers have been damaged, and how many.

Second, the diagnoses are going to be more persuasive. You know how athletes sometimes collapse and die? Everybody makes the connection between cause and effect—between overexertion and a strain on the heart—because the tragedy is visible and vivid and immediate. There’s no mistaking it. And then the autopsy comes back, and it’s unambiguous. The high-school football player died of a heart attack. The college basketball player died of a coronary aneurysm. But brain injuries have lacked a similar sort of clarity and immediacy, and therefore they’ve also lacked a similar sort of urgency. When a football player suffers a concussion or when a boxer takes multiple punches to the head, the effects of an injury might not be evident for years or decades. Not anymore. HDFT will show what the blows to the head have done to the brain, and I’m telling you, it’s not going to be pretty. You won’t need a medical degree to compare a concussed brain and a control brain and go, “Oh
no.
”