Wednesday 21 May 2014

Processing speed and white matter: Mark Bastin

 

You may recall that, from time to time, I have tut-tutted when researchers take scans of a few brains, get very excited, climb high up a long ladder of assumptions based on small samples, and then fall precipitously into a smelly cesspit of neuro-bollocks.

http://drjamesthompson.blogspot.co.uk/2013/04/not-all-neuroscience-is-rubbish-just-92_18.html

Tim Shallice has pointed out the three main problems: small sample sizes, inconsistent methods and measures, and a lack of theory against which to test findings.

In the preparations for the Processing Speed conference I checked out Mark Bastin, about whose presentation I was supposed to make some comments. He graduated from the Universities of York, Edinburgh and Oxford with degrees in Theoretical Physics (BSc (Hons)), Remote Sensing and Image Processing Technology (MSc) and Geophysical Fluid Dynamics (DPhil). Given that I did one year of undergraduate physics, I felt we would probably be able to chat over a few things which had been bothering me at the proton level.

I read Mark’s work with some trepidation, and listened to his presentation with great pleasure. He showed data on 581 subjects which, by my calculations, was sufficient to have written 29 neuro-bollocks papers. Although, as you would expect, I have a finely tuned mind, I found that his explanations on a number of matters, although intended for a more general audience, were of help even to me in brushing up on a few concepts. A helpful exposition, clearly presented.

Full presentation here: https://docs.google.com/file/d/0B3c4TxciNeJZeVBTN0lnVDVXOUU/edit

Slide 3 contains several explanations required when looking at the later results: Diffusion tensor MRI (DT-MRI) measures the random motion of water molecule protons due to thermal energy. In brain white matter, the presence of coherently ordered axonal membranes and myelin causes water molecules to diffuse preferentially along the fibre direction rather than across it. The random motions of water protons can be quantified to generate biomarkers of white matter integrity, such as mean diffusivity D which represents the magnitude of diffusion;  and fractional anisotropy (FA) which represents the directional coherence.

Healthy, structurally intact white matter has low magnitude of diffusion 〈D〉and high directional coherence FA. Tidy connections. Structurally compromised white matter has high magnitude of diffusion and low directional coherence FA. Frayed wires. I would not stoop to populism, of course, but cognitive ageing may be simply the difference between tidy connections and frayed wires.

It is also possible to follow the principal diffusion direction across the brain to generate 3D representations of the major tracts (tractography) of which more later.

Slide 5 brings you into the inner sanctum of scanning methods. The first technique involved drawing round features by hand which is given the better sounding jargon of ROI analysis, which relates to Regions Of Interest. Call it the clinician’s approach. The more standardized automatic voxel-based methods are called  (TBSS) or Tract Based Special Statistics. The automatic segmentation of tracts has (tractography) is an art done in various, entirely defensible, ways which vary according to the lab and the person doing the automated segmentation. Rather than “Computer says Here be Dragons” it is “My approach to segmentation leads Computer to see Dragons in the following Ways and Locations”. Good to know that neuro-phrenology still requires human intervention. I should add that his slowly rotating 3D tractographic picture of the connection in the brain is very beautiful. How the hell does the brain do all this marvellous stuff, which even allows some people to understand theoretical physics?

Slide 8 shows the tests used, covering general intelligence, general processing and general memory.

Slide 9 shows the main result: A general factor of Information Processing Speed, g speed, has a very strong brain-wide association with white matter directional coherence (tidy connections).

Slide 10 shows that Mark Bastin and colleagues have developed a novel tractography approach, termed probabilistic neighbourhood tractography (PNT), which allows the same tract to be segmented from subject to subject using single seed point tractography. This gives them a more solid basis for the fabulous 3D pictures.

Slide 14 has the final conclusions: Older age changes in white matter integrity affect major tracts simultaneously. Both voxel-based and tractography analyses show that Information Processing Speed tasks require connectivity at the whole brain rather than the individual tract level. Associations between general intelligence and white matter directional coherence are mediated by Information Processing Speed.

Just as Mark felt he had done his work for the day I decided to press him to tell us where the problems were: namely, are these sorts of results replicable? In his view it was mostly the personal style of the scan analyst which generated the pretty pictures in tractography, and the techniques used by one analyst were rarely followed by another. Hence the need to try to get PNT techniques evaluated and used by other researchers.

Ever the mischief maker, I asked Mark to compare the development of theoretical physics with that of psychology. My impression, in line with Tim Shallice, is that psychologist’s conceptions of brain function will always be, ahem, impressionistic, if not thoroughly Turner-esque. Being a typically cautious scientist Mark demurred, shyly admitting that both his boss and the main grant giver were present in the room, so he didn’t feel he had the basis to venture an opinion. Undaunted, I kept heckling him (egged on by both his boss and his grant giver) until I managed to drag out from him that at least psychology had worked some things out, and had some theories, but in his view there was another profession which knew nothing about their patients, and even less about the pills they prescribed them. He was probably too circumspect to give Voltaire quotation in full: “Doctors give drugs of which they know little, into bodies, of which they know less, for diseases of which they know nothing at all”. I have no idea whether he is still in employment, but if not he would be an asset to any department which wants to study brain function.

We need to talk about Protons. I did once measure the elementary charge on the electron (which is carried by a single proton) using Millikan's oil-drop experiment.  Strange to see odd drops float upwards against gravity because a very much smaller particle was driving it in the induced electric field. I had no idea it would be useful 50 years later. Worth a Saturday morning in a lab.

No comments:

Post a Comment