We spoke in a previous blog that we want in the future to carry out research on our apps to make sure they really do work. One of the techniques we want to use is functional magnetic resonance imaging. This was developed as a way of 'seeing' what parts of the brain are active when people are carrying out a task. It has been used widely since the 1990's in many areas and has helped us to understand how the 'normal' brain works and what may be going wrong when people have illnesses such as depression and schizophrenia.
It's all about magnets!
As the name suggests it uses magnetism. A basic explanation is that blood is magnetic and differs in its level of magnetism if it is oxygenated or not. When a particular part of the brain is working hard it gets more blood and uses up the oxygen to burn glucose for energy. The scanner creates an immensely strong magnetic field, which can detect these tiny changes. It can then map where they are in the brain and that gives you a picture of which areas of the brain were working harder when a person was doing a particular action.
As you may imagine there is rather more to it than this and---if you are interested in more detail---here is a link to Oxford University, UK, for a more detailed, but still comprehensible explanation. Other more detailed, and---to me at least---increasingly less comprehensible explanations, are available.
It gets trickier
Although there is a vast amount of computer analysis probably the trickiest bit is designing the experiment. This gets more difficult the more complex the task or concept you are investigating. Designing a study to see areas of the brain involved in pressing a button can be difficult. I have been involved in projects looking at empathy and deception where we spent months trying to get the experiment right. The problem is you want a 'clear picture' of the brain areas involved in the task you are interested in, for example looking at a spider.
The issue is the brain is doing lots of things all the time---even mine is doing quite a lot! So, if you simply put me in the scanner and show me a spider (the stimuli) you get the brain activation associated with me seeing a spider and being scared. But also brain activation associated with the colour black (the spiders black), me tracking a moving object, me seeing the tree in the background, me listening to a noisy scanner, and lots of other stuff making the picture very, very blury. This is called 'noise' in the trade. There are other sources of noise such as people moving and physical limitations of the scanner.
So what you have to do is create another 'stimuli' which has everything apart from the thing you're interested in; in this case the spider. So if the first spider stimuli has a tree, movement, the colours black green and red you need to match these as close as possible in the second one. Itsy our friendly spider might be good she has 8 legs, moves, is the right size but is designed not to have the characteristics most people find frightening in spiders.
The fMRI equation
So we show me stimuli A (scary spider), scan, process and get picture A. We then show me stimuli B (itsy, a spider but with the scary bits removed ), scan and get picture B. Both pictures have 'noise'. So we get the computer to subtract picture B from picture A which cancels out the noise and everything else apart from the difference; the brain activations associated with the scary aspects of spiders.
Here is the equation as promised:
(Scary+Spider+Noise) - (Spider+Noise)=Scary
The system isn't perfect as you can never get rid of all the noise. One of the ways to try to improve the picture is to take lots of picture A's and B's by getting lots of people to do the task 20 or 30 times.
Could you try this at home?
So there you are. Now all you need is an MRI scanner, a team of psychiatrists, psychologists, radiologists, statisticians, some money and you too can do fMRI!