What would it take to make my own fMRI?

So my brain is not quite firing on all cylinders at the moment and being a geek one of my thoughts is how to debug the problem. The inability to insert breakpoints, no console or printf() facilities and the lack of a decent IDE stumped me to begin, but I've debugged systems with these challenges before. It's normally hardware or embedded systems that present these challenges and some of the primary tools to understand the problems are logic probes and oscilloscopes - which got me thinking: what's the equivalent tool for a brain and can I make a cheap version for my own home use?

The first thing to spring to mind was the EEG (electroencephalograph) which records the electrical activity across the scalp which is indicative of the neurons firing on the surface of the brain (gyri for the Scrabble fans out there). There are indeed plans and instructions for building your own EEG out there on the 'net including the OpenEEG project which looks quite good and is within my technical capabilities (mainly limited by my soldering capabilities). But as a debugging tool, indeed an information tool, I think there are severe limitations.

An EEG measures the potential difference between synchronous synaptic activity of thousands of neurons on the surface of the brain. This produces two issues:

• only shows an average potential across groups of neurons which are physically separated. Its very coarse grain and doesn't necessarily reflect the potentials between the actual synaptic routes in the brain. It's a bit like measuring the straight-line distance between two towns to estimate the driving time without taking the roads into account. Anyone who's driven the coast of Norway or Iceland understands why this is a problem
• doesn't effectively measure activity deeper in the brain, especially any activity which is tangential to the surface of the skull (or even at an oblique angle)

Now there's no denying the value of EEG for the diagnosis of a wide variety of brain related problems, however, as a debugging tool it's a bit crude. It's like trying to debug a computer system using the flashing of LEDs, disk noises and fan sounds (we've all been there!). This can provide a very gross level debug (e.g. its not powering up) but is not much good for most software bugs.

To get information about the deeper goings on in my head in a non-invasive manner, that does not require the use of controlled radioactive isotopes, leaves me two options: the MEG (magnetoencephalograph) and the fMRI (functional magnetic resonance imager).

The MEG works by measuring the magnetic effects of the electrical currents in the brain. Unfortunately, these are only on the femotesla range (10^-15 T) - given ambient magnetic flux is in microtesla (10^-6 T) this not only requires a magnetically shielded room but the use of SQUIDs (Superconducting QUantum Interference Devices). While convincing my wife to allow me to paper a room with tin-foil and chicken wire might be possible, allowing me to have large quantities of liquid helium is probably out for a number of reasons.

So all I'm left with is the fMRI.

So firstly I had a look on eBay for a second hand MRI - but no joy (so you can't "buy anything on eBay" as some party-bores insist). However, I did find the DOTmed site which does have auction listings for used MRIs!!  Sadly, these things are huge, mainly in the US, expensive and also require the use of cryogens. So its back to thinking "home-made".

Conventional MRIs use field strengths of around 1-2 Tesla (that's ~20,000 earth's magnetic field) which requires superconducting magnets, and hence cryogens. Equally, they are usually designed for whole body, detailed scanning. With fMRI scanning of brains you need a smaller scanner (head sized) and you need to rapidly scan a 'slice' (within 1-2 seconds) at low resolution to pick-up the changes in blood flow within the brain.

In recent years there's been quite of bit of work on Ultra Low Field MRIs which reduce the need to have superconducting magnets for imaging. For example, these guys at Los Alamos built the fMRI in the picture, which operates without superconducting magnets. It does use a SQUID detector array which requires cryogens and a coil immersed in liquid nitrogen, but its a step in the right direction for me. Even better, one of their papers explains the construction and even lists where to get some of the parts!

Is it possible to remove the need for cryogens all together?? There are some reasonable results of images generated using "conventional room temperature receivers" with Ultra Low Field MRIs, such as these ones from Aberdeen University. Images of this quality would be a result! In fact, these guys have also shown results using permanent magnets for small samples (why do I need to pay for access to a scientific paper?? What has happened to science publishing?)

However, it looks like Ultra Low Field MRIs need SQUIDs which = cryogens. So unless I can convince my wife to let me have liquid helium, it looks like I won't be building an fMRI. Of course, I then need to convince her to let me spend £8000 on parts...