[make tea / coffee now]
MEG scanning is a hot topic in brain / computer interfacing at the moment. On seeing the scan results, even people who don't know much about science realise that it offers something a lot better than normal EEG and electrodes stuck to the person's head.
I believe brain / computer interfacing is the way we should be looking to extend our lives. Those who understand the full potential of this transistion, not limited to lifespan extension, will recognise why I believe this. I feel that things like cryogenics aren't really in the spirit of life extension, merely life perservation. Cryogenics offers the general public a way to sit back and wait while others get on with all the hard work of trying to find a way to defrost them and regenerate their defective tissue masses. Saying this of coarse, if I have the money, I'll be signing up as well. But I view it purely as a safety measure rather than a truely effective method; I would carry on my brain / computer work even after being defrosted.
The problems involved with firstly making the connection and then finding a means of 'swapping' a consciousness between forms is immense and I have limited time to find a solution before I'm likely to die. I'm 21 at the moment, but decades isn't long with a problem this big and I worry it's not enough.
With the worry in mind, I began cluster researching numerous methods of brain / computer interfacing, MEG / EEG / Direct Connection etc for feasibility. Direct interfacing seems the most distance on first appearance, the best electrodes used for patients with critical conditions or experimental research work at the moment have about a hundred penetrating channels. It doesn't take a scientist to realise a hundred is quite short of the number of neurons in the brain, which is counted in billions.
MEG and EEG (including derivatives, i.e. HEG etc) can produce real world 'wow factor' data, like being able to control a wheelchair or 'Pong' paddle in a game without moving, that makes the none scientific interested in them. For immortalists, these technologies, MEG in particular, provide a possible means of saving, relocating or transformer consciousness.
Since I have the advantage over a physicist in that I know what I'm looking for, where it'll be and how I can find it, I decided to investigate the possibility that we can't already achieve a 'mind reading machine' for the lack of simply spending enough money on an individual unit, not bothering with the economics manufacture and market, only results. That, like these immortalist hope, the effect could be achieved by just building better and better MEG scanners.
After a lot of reading into the systems used in MEG scanners, like the superconducting quantum interference detectors, one of the things that becomes apparent is the scale of the request. The brain, even as a whole, emits a painfully small magnetic field that's easily demoted to background noise by something like the Earth's own magnetic field fluctuating slightly. You need hugely elaborate pickup designs (Gradiometers) and magnetic shielding (Superconducting) to get rid of most of the noise. This is something the SQUID developers are working on, to create a SQUID that ignores anything but localised fields.
A much more concerning problem is that of ultimate sensitivity. Moving a Pong paddle is not a complicated task, you could do it with just one signal, biasing the paddle to return to one side of the screen on release and using a single input to temporarily move it in the other direction a bit. I decided to push this out towards the limits that would be needed for the kind of brain / computer interfacing we'd like, wherein there's some hope of a brain being 'downloaded', that being a sensitivity high enough to read a single neuron depolarisation and an equipment build / result processing power accuracy high enough to locate that event's exact position in both space and time.
The initial results of my researching don't seem promising at all.
I'm aware of almost no one trying to push magnetic sensitivites out this far. Most SQUID research seems to be being targeted at applications in industry requiring higher than normal magnetic sensitivity but ultimately involving higher field strengths than those found in the brain; increasing the SQUID's rejection of ambient noise for instance. The level of sensitivity required for single neuron scanning puts the SQUID elements into the domain of prototype, one off physics equipment and perhaps beyond.
Present day MEG scanners scan 'brain tissue' acitivity for analysis of conditions like epilepsi. For this activity to show up on the scanner, somewhere between ten and a hundred thousand neurons need to depolarise pseudo-synchronously, far from single neuron sensitivity. What makes this worse is that SQUID devices count discrete units of flux, fluxons, and the quoted threshold for a SQUID is only one order of magnitude below today's 'brain activity', that is, it's self, ten or a hundred thousand times below what it really needs to be for detailed brain / computer transfers.
To put the expected level of sensitivity into perspective, a fridge magnet you might have in your Kitchen will be around 0.1 Teslas. The level of brain activity being scanning for now is on the order of tens or hundreds of femto Teslas, that's 10^-15 or ten million times lower than the field around a fridge magnet. The level required for sensing individual neuron depolarisations is closer to atto or zepto, 10^-18 to 10^-21, that's about a billionth of a billionth of a Tesla or less, possibly extending towards yocto Teslas for deeper tissue, 10^-24. As far as I can trust my own memory, the world's most accurate clocks wouldn't be capable of measuring units this small and time is the dimension we can most accurately measure at present.
The requirement placed on present day MEG is not just to improve it's performance a few multiples, but thousands of times over. A big request for a technology that's already pushing the boundaries in it's own design as well as that of the equipment used to support it; helium refrigerators for near absolute zero performance / noiseless power supplies and amplifiers for the output signal / shielding etc.
I've been discussing the feasibility of improving this sensitivity with a Dr of phyics, who happens to have experience with superconductors and SQUIDs. While he mentioned that he wasn't an expert on SQUID design, only that he uses them and understands superconductivity, he, like myself now, believes that attempting to design a MEG scanner with this degree of sensitivity is questionable.
The main problem MEG scanners have, aside from being swamped in magnetic noise, is that they're attempting to measure fields that are already weak, from distance. Magnetic fields decay rapidly over distance, moving just a few mm away can make the difference between it being registered or not.
SQUID detectors are superconducting devices, indeed, they must be for the effect they rely on to occur, and we do not possess room temperature superconductors. This means the devices must be positioned physically separated from the human head and cooled in a cryostat if one does not wish to simultaneously freeze the individual, ceasing any neural activity that was there to begin with. This has a benefit because the scanning is a totally passive affair, there's no need to expend time or money positioning, implanting, then removing and sterilising or disposing of and rebuying the field detectors before the next implantation.
Moving the detectors closer the individuals brain tissue may give MEG a helping hand towards individual neuron depolarisation sensitivity but it's still unlikely to be enough. This may also mean implanting the detectors, requiring body temperature superconductors, and doesn't take into account the fact that the detectors are physically quite large; although the SQUID element can be made quite small, the pickup coils can't, as making them smaller also decreases their efficacy, something MEG really doesn't need! Such detectors may even be probe like and carefully positioned within the tissue it's self, not just sitting on top of it.
The Dr I spoke to suggested making the pickup coil area bigger. This is an idea, but more coil area per channel also means fewer channels in the same total area, meaning lower spatial resolution. Making the field stronger isn't an option, since we can't 'power up' our brain tissue. Nore can we roll our brains out to make the scanning area bigger.
All these improvements might be enough, but they rely on seriously big developments; body temperature superconductors, pickup coils small enough to be implanted etc.
Then we have more problems. Humans being scanned will be alive and living humans have a tendancy to twitch or move ever so slightly from time to time. They also produce a strong, continuous vibration in the form of a heartbeat that's mechanically coupled to the brain tissue, via the excellent blood flow from the heart that supplies it. None of these are a major problem if you're looking for macro tissue activity, but to transfer data between two points you need to have some form of spatial and temporal lock between the two such that the transfer will be from a known point to a known point and at a known time index. If you're trying to move that data between individual neurons any movement, even on a micrometer scale, could be enough to corrupt the transfer. MEG scanner's can't 'lock onto' tissue and actively track it's movement. Theoretically they could, but remember that you'd be expecting it to track billions of separate targets that have the potential for growth and reorienting themselves morphologically.
Now we move on to one of the biggest problems. Lets assume that to be sure that we, the conscious being, isn't wiped out during the transfer we want to be able to communicate back and forth as the transfer takes place, to remain conscious and ensure continuity. MEG can't stimulate brain tissue. Magnetic fields have been shown to produce some form of result in brain tissue but at macro levels, entire half hemispheres of the brain, or at least very large fractions of them.
To focus a magnetic field down to a single neuron accuracy, especially over these distances, is beyond our ability. To focus a field to simultaneously stimulate billions of neurons as individual units is an unbelievably big request. Also remember that this field will function like a gamma knife would, with the potential for 'side stimulations' as the field gets to the point in the volume it's targeted at. Now this stimulating field, something that produces lots of noisey magnetic field modulation, and MEG scanner, something that is incredibly sensitive to magnetic fields, need to put in the same space as each other.
All theoretically possible. But, I believe, painfully distant.
Unlike a lot of scientists, immortalists need to be very aware of timescales. Something may be theoretically possible, but if it's not realistically possible within their lifespan then it's not really an option.
It's for these reasons that I would suggest MEG is a less than great option for other immortalists looking at brain / computer interface technologies to place too much hope on for the near future. I would, of coarse, actively encourage you to keep up to date with what is happening with SQUIDs etc and to make an attempt to learn how they work, I myself am ready to reconsider MEG if I see a substantial development emerge. But expecting the scanners to go from controlling Pong games to downloading a consciousness is, while theoretically possible, a remote option at present.
A number of similar problems occur for EEG and there is no electric field sensitive variant of the SQUID. Although, if such a device were to appear, it may proove interesting. Neurons depolarise at voltages that are high enough for a normal audio amplifier's input to pick them up if there were a direct connection between the two; a depolarisation occurs at tens of millivolts. However, the depolarisation involves only minimal amounts of current, a few thousand ions or nano amps. Neurons would appear to be 'high voltage / low current' devices. Magnetic field strength is set by current, if the current of a neuron is low, it's magnetic flux creating ability will also be low. Electric fields strength is set by voltage. Thusly, if the neurons depolarise at substantially higher voltages than they allow current to flow through them, and we have an electric field detector as sensitive as a SQUID is with regards to magnetic fields, capacitively coupling to them would be appear to present a more positive option.
Personally, I believe that direct connection is the best route to follow. Once a direct connection is made everything else will be much easier. Detecting and measuring depolarisations will be a joke it'll be so simple! Off the shelf IC's can achieve nano volt noise performance at their input, whereas neurons depolarise at tens of millivolts. Bidirectional communication would also be easier since there would be fixed connections from each neuron to each interface address, no need to actively track each neuron as it moves with vibration. Lastly, this is an area that has a lot of drive already behind it, all be it indirectly, from the electronics industry, providing it with things like microscopic amplifiers with excellent noise performance, and ever increasing data storage and handling capacities via holographic drives and solid state semiconductor blue lasers.
These are just some thoughts and conclusions I've been thinking over for the last few weeks whilst I've been putting in ten or so hours of reading a day towards answering this. I thought it would be best to summarise some of these here for other immortalists interested in brain / computing interfacing to read before they 'waste' time having to reread the information in it's long form and piece it together for themselves.
My reference for a successful neural interface is one that is capable of addressing each individual neuron of the brain tissue simultaneously, although fractions may proove to be perfectly acceptable given that large percentages of the brain tissue seem to be below their maximum potential for activity a lot of the time, and supplying a pathway for bidirectional communication between those neurons.
Any thoughts, questions, ideas, discussion?
Hugs and kisses, [thumb]
John
Edited by johnuk, 17 December 2005 - 04:35 PM.