Prof. Lars Linsen, Int'l University Bremen & Institut für DiagnostischeRadiologie und Neuroradiologie, EMAU Greifswald
By way of introduction the article offers background on magnetic resonance imaging (MRI) - assumed the current imaging gold standard as far as answering the question - where in the brain activity is taking place.
In previous studies, scientists have used magnetic resonance imaging to identify peaks and valleys of neural activity when people are doing various things, like making decisions, reacting to frightening images or reliving painful memories. But these studies, while provocative, revealed virtually nothing about the underlying neural networks involved - about which brain regions speak to one another and when. Previous estimates of network structure, based on such imaging, have been sketchy.
The new findings, while not conclusive, give scientists what is essentially a wiring diagram that they can test and refine.
1. When "scientists have used magnetic resonance imaging to identify peaks and valley of neural activity when they are doing various things..." they use fMRI (functional MRI) but that is a quibble. I suppose it is accurate to say that scientists have used MRI for such a purpose, but it would be more accurate to say that making this claim is itself controversial and then describe what MRI actually measures. Scientists who love fMRI love to say that the pretty colored images of the brain that one can see everywhere nowadays gives us a picture of which brain areas are 'on' and which are 'off' during a given task, but that is not a fair picture of what MRIs depict. Authors describe the brain "lighting up" and make a direct link between the brain's relative use of oxygen in some parts of the brain as compared to others and its activation. THAT is what fMRI measures - which regions of the brain are metabolizing the most oxygen at a certain moment in time relative to others- one must make a leap from the measure of metabolism to the claim that that shows us which regions in the brain are most 'active.' It's not an impossible leap but there are problems with it. For one thing, neuronal activity can be excitatory or inhibitory - that is the electro-chemical messages traveling down the neurons can facilitate OR inhibit a behavior. When thousands of neurons are active at once- some excitatory and some inhibitory - it is their net activity which determines whether a given behavior occurs. A great deal of activity (which would presumably require a great deal of metabolism, and would hence be measured as highly active by fMRI) could sum to zero. I.e., nothing could happen. Scientists and science writers need to be very careful about the inferences they make about that activity. Much is dependent on well-designed experiments. I have ranted in greater detail about fMRI before so I'll spare you more now - here's a link if you're into it. Here's a link to another critique of fMRI by my favorite pedant, Jake Young. And yet another by literary science phenom Jonah Lehrer. My point regarding this article is, if you are going to start an article about an innovative scientific technique by introducing its forerunner, at least be specific about what it does.
2. "...these studies, while provocative, revealed virtually nothing about the underlying neural networks involved..." Not true. Scientists have done their best using available techniques to measure "functional connectivity" between brain regions, examining the correlation of activity in disparate regions at the same time with fMRI, and by combining fMRI which is specific about where activity occurs with event related potentials (ERPs) which is very specific about when. But wait a minute, you say, you just said that fMRI was suspect. Well, yes I did, but it seems to me that to claim in one breath that fMRI is an undisputed picture of brain area function without preamble and in the same paragraph critcize it for an inability to measure connectivity is more about creating a dramatic story for the sake of glorifying DSI than it is giving an accurate picture about fMRI, which is the current state-of-the-art measure, whatever you think of it. Both rely on inferences. If you are going to be critical about it, be critical about it. But don't criticize the aspects of fMRI that are going to give you a sexy story about DSI, which brings me to my third point - glorifying DSI.
3. "The new findings, while not conclusive, give scientists what is essentially a wiring diagram that they can test and refine...The technique allows scientists to estimate the density and orientation of the connections running through specific brain locations." Ok, he does say 'not conclusive' and 'estimate,' but from what I understand about diffusion tensor imaging (DTI, the basis of DSI, which is not much, which is why I was looking forward to a more informative article) it measures the probability of the diffusion of water molecules in brain tissue. There is a physical phenomenon known as Brownian Motion - which is the random movement of particles suspended within a liquid or gas. Mathematics allows one to predict the probability of the location of such a particle at any moment in time. Normally water molecules would diffuse isotropically - the probability is equal that they could go in any direction. However, the architecture of brain fiber tracts (white matter) as opposed to regions of cortex or other brain areas (grey matter) is that they impose physical boundaries on the movement of these molecules, changing the probability from isotropic to anisotropic - it is more likely that the molecule will travel in a given direction than equally probably that they could travel in all directions. It is evidently four times more probably that the water molecules will move parallel to the fiber than perpendicular to it. The myelin sheathing the axons is thought to direct the diffusion of water molecule along their path. The imaging technique compares diffusion of water molecules, looking for those regions where anisotropic diffusion in more prevalent than random - and creates an image of those areas by brightening or color coding those regions. If you follow the logic and trust the inferences behind this technique - that this mathematical model of the placement of water molecules is indeed accurately depicting fiber tracts - then you could say, as this article does, that it 'creates a wiring diagram to test and refine.' But this model of connectivity is easily as inferential as other functional connectivity models made via fMRI at this point. Testing and refining is precisely what much of the current DTI and DSI research is doing now - comparing the measure taken with actual brain tissue samples to confirm the accuracy of the technique. There are problems with the leap made from water diffusion to 'picture of white matter,' just as their are controversies surrounding the leap from BOLD measures to 'brain activity' in fMRI. DTI voxels (3-dimensional pixels) are 2mm - this is not great resolution given that nerve fibers are microscopic in scale. There is lots of room for error in "tractography." Points where multiple fibers cross can be particularly tricky, which is where DSI is making an improvement over DTI in ways that I understand too little to describe. My point is, when introducing this technique to the public, it would be more responsible to say what it does and doesn't do (and some of its possibilities are very exciting) than to offer as the first information we receive (and therefore the best remembered) hyperbole and spin. Especially when the pictures these techniques create are so alluring. Pretty pictures can be awfully persuasive, but it is useful to think critically and certainly it is the job of scientists and science writers to encourage that.
I'll add just one other thought. Recently I read a criticism of imaging techniques, I can't remember where, it was that fMRI couldn't accomplish a certain measurement it claimed by itself. For example, I've explained that current research is comparing direct examination of brain tissue with the images generated by DSI - please don't interpret that as a criticism of the technique. Understanding in the sciences is always improved by multiple measures. We must begin with models based on inference and we forever improve our measures and deepen our understanding by testing and resting from multiple perspectives. Our understanding is improved not reduced when we have behavioral studies, brain lesion studies, fMRI, and ERP all converging (or not) around the same neuroscientific phenomena.
No comments:
Post a Comment