In the United States, a child born with cataracts would have a simple surgery to correct the condition soon after birth. But infants born to India's poorest families don't have access to that surgery — and so they grow up blind, unable to attend school and, in many cases, sentenced to a life of begging.
A decade ago, MIT neuroscientist Pawan Sinha, PhD, realized that giving these children sight-restoring surgeries could also advance science. By studying how these older children learn to see after a decade or more in the dark, he could elucidate the mechanisms that underlie vision in all of us.
Today, Sinha's Project Prakash ("light" in Sanskrit) has treated more than 400 children and helped answer a 300-year-old question about how our senses of vision and touch interact. Sinha spoke to the Monitor about his research and his plans to reach thousands more children in the years to come.
How did you come up with the idea for Project Prakash?
I think the seed of the scientific idea was already in my mind when I was starting my faculty position at MIT.
This question — how does the brain learn to see? — has been extremely difficult to address experimentally. The only approach that we have is to work with either very young animals or, in the human case, to work with infants. But by the time infants are even minimally able to respond — say they are a few months old — we have missed many very important changes. So I was struggling with this question.
Then on a trip to India in 2002 to visit my father, who lives in New Delhi, I saw these little children begging on the streets. That itself is a very distressing sight, but what was even more disturbing was the fact that some of these children had treatable disabilities. A couple of the children I saw were blind because of cataracts, and even though I'm not an ophthalmologist, I knew that this is a treatable condition.
I realized that if I were to try to do something just on a personal scale, I would be able to provide funding for the surgeries of a handful of these children, which would be satisfying but would not really begin to make a dent in the larger scope of the problem.
And that's when I realized that the scientific question I'd been struggling with found almost a perfect approach in the treatment of these children. If you have a child who's, say, 10 years old, who has been blind since birth, and in this child you are able to initiate sight, then you have an unprecedented opportunity to examine visual development, right from point zero.
Once I had that realization, then the path forward became reasonably clear. As a scientist I know how to define a scientific problem and how to approach scientific grant-making bodies. So I could describe to them the humanitarian crisis and the scientific benefits to be derived by addressing this humanitarian crisis. Funding bodies like NIH, and some private foundations, responded very well to that combination.
What have you learned from your research about the development of human sight?
Some of the first questions that we addressed had to do with a child's ability to interpret an image as a collection of objects. If I give you a marker and an image, you would have no problem delineating the boundaries of the objects [in the picture].
But even that very basic task of breaking up an image into distinct entities is shrouded in mystery. How does the brain do that? If you try to get a computer to do the same thing, it fails catastrophically. And that says something about just how difficult this problem is, notwithstanding the fact that we are able to do it so easily.
What we found was that in the initial stages of visual learning — let's say you have a child who has been treated just a few days ago — that child tends to break up the visual image into many more pieces than a normally sighted person would. Little things like shadows become distinct objects. It's only over time that the child begins to understand the correct parsing of an image.
We found that motion seems to be of profound significance for this early learning. It has a teaching role in the early stages of visual development, to inform the brain what regions in an image ought to go together and form complete objects.
So we learned two lessons from this. One lesson is that motion might be important not just in the moment, but rather might truly wire up the brain system in a way that subsequently, even when motion is absent, the brain is able to correctly interpret static imagery.
In addition, on a slightly higher level, this result showed that even children who had suffered [sight] deprivation for many years, 10 years or 14 years, these children were able to acquire significant visual function.
We had been worried about the idea that some neuroscientists had posited, that beyond the first few years of life the brain simply does not have the capability to learn from new visual information. And we found that was not true. These children, even though they were gaining sight late in childhood, were acquiring significant visual function.
How long did it take for the children to learn this image parsing?
In our experience, that ranged from about eight months to 18 months. And we're seeing a rough correlation between the age at which a person is treated and how long this process of learning takes.
So knowing that learning can happen has opened the door to studying specific aspects of visual perception in greater detail. So we can ask: How is face recognition developing, or how is the ability to name colors developing?
Can you give an example of how these vision studies would work?
For one study (published in Nature Neuroscience), our starting point was a question that had been framed over 300 years ago by William Molyneux, an Irish philosopher/scientist who was interested in the empiricism versus nativism debate, which is simply, does the brain come prepared with knowledge to interact with the world, or does the brain have to acquire that knowledge through experience? Empiricism says you need experience, and nativism says you come natively prepared for interacting with the world.
Molyneux framed a very elegant question to get at this issue. He said, imagine that you have a person who has been blind since birth but has learned to distinguish between different objects, such as a cube and a sphere, by touch. Now, imagine that a miracle happens and the person gains sight. And the first thing that you do after the person gains sight is to show him a cube and sphere. But you don't let him touch these two objects. Would he be able to tell them apart by sight? Would there be an immediate transfer of knowledge that he has about these shapes from the touch modality to the visual task? Because if he can, then it would argue for some sort of nativist structures in the brain that allow all sensory modalities to communicate with each other. If on the other hand he could not, then it would argue more in favor of empiricism because maybe experience across the two senses is needed to begin to build a map between them.
To test this, we presented children who had just gained sight with pairs of objects. They would feel one object in the pair, without looking at it — we hid it under a bed sheet. Then I would take the object and a distractor object, and put these two objects in front of the child, and ask them, "Which was the object you were just touching?"
We found that immediately after surgery, these children could not perform this task of transferring knowledge from touch to vision. So that finding on its own allows us to answer Molyneux's question. The answer is no — there is no transfer right after surgery.
However, to our surprise, when these children came back for a follow-up — some about a week after the first surgery, some after a month and a half — and we tested them again, they showed almost perfect transfer. So, even though the mapping between vision and touch is not available right from the outset, it's something that the brain can learn rapidly, in a matter of just a week or so.
And during this period, they had been experiencing the world without our intervention, we were not tutoring them on what to do.
That's much quicker than the eight or 18 months it took to learn image parsing.
Exactly. And that brings up this interesting question: What are the timelines of learning different visual tasks? The image parsing task seems to be reasonably difficult. It seems to take several months. The cross-modal task seems to come about very quickly. How would the trajectory of face learning compare with this? Object characterization?
So, this is all very interesting work that we hope we will eventually be able to flesh out.
How do the ways in which these children learn to see compare with the ways infants learn to see?
That's a very deep question, and a very difficult one to definitively answer. I can take guesses, but we don't have data to back up our speculations. My instinct, partly because of my training as an engineer — before I became a neuroscientist I was a computer scientist — would be that some of the programs of learning would share basic similarities, whether they are unfolding in the infant brain or in the brain of a child.
One example of such a similarity would be the reliance on motion information. One of the first things that a human infant becomes sensitive to is motion. You shake something in front of a baby and the baby is far more interested in that than in any other thing. And I don't think that is happenstance. I think it serves a very important purpose, perhaps the same kind of training purpose that it serves in the brain of a Prakash child. From the perspective of mechanistic parsimony — why invent two different mechanisms when you already have a good one? — it would make sense for the infant brain as well as a later learning system to make use of the same kinds of algorithms.
But as I said, these are speculations.
And almost certainly there are some differences. The Prakash children already have information about the world. They know that the world is made up of objects. They even know the names of some of these objects. So for them, when they're trying to make sense of the visual world, the task is associating the visual information with the labels that they already possess. For a baby, the task is twofold. Babies have to make sense of the visual information and then they also have to discover the labels of these objects. So, the tasks are different, but I think we will find that some of the basic components of the mechanisms are shared.
How many children has Project Prakash treated?
We have treated 440 children surgically, and about 1,400 non-surgically. To identify these children, we've had to screen over 40,000.
Even though these numbers seem impressive, when you put them in the context of the true need in a developing country like India, you realize what a minuscule amount of the problem we have really tackled. There are probably around 200,000 or 300,000 children who can be treated, so for us to derive much happiness from having treated 440 of those — it's premature.
What are your future plans?
We need to be able to expand out, to reach out to many, many more of these children. So that's one consideration, scaling up our efforts.
The other is education. When we started this project, we had a somewhat Pollyanna-ish view of how things would proceed. We thought that these children would gain sight, and then the world would open up for them — they would be able to get admission into a regular school, get a good education and then move into the workforce. But we have found to our disappointment that many of them do not go into the schooling system because even though they now have sight, the system thinks they're too old to start in grade one.
For these children to fully benefit from the new sight, [we need] to provide them the beginnings of an education, essentially a compressed scholastic course that brings them to an age-appropriate level. After that, they can be mainstreamed into a regular school.
We have started doing this in a very small way with volunteers who work one-on-one with the Prakash children. And it's been amazing to see just how quickly a Prakash child, starting with zero education, is able to progress up to, say, a grade five level, in a matter of six months.
So there are three aspects: the expanded health-care aspect, the inclusion of education and then certainly research, expanding the research and working with many more of these children.
Our hope is that we would be able to integrate all three of these aspects, by having one campus — that we call the Prakash Center for Children — where the pediatric hospital, the research facility and the school for the children will all be co-located. This is the dream for us. We want to make this Prakash Center for Children happen.