Transcranial Magnetic Stimulation

[Enthalpy]

Hello everybody!

 

Neurology uses Transcranial Magnetic Stimulation (TMS) for research and sometimes diagnostic and treatment.

Introduction there: http://en.wikipedia....tic_stimulation

Some documentation there, but other manufacturers exist: http://www.magstim.com/

 

As I understand it (but beware I'm bad on biology and neurology), not the magnetic field, but the gradient of electric potential and current it induces, creates the desired effect on the brains. The action resembles partially an electro-shock, but:

• The skull is tranparent to the magnetic field, while it hampers the electro-shock's current;

• Hence the skin isn't as brutalized by TMS;

• The path of the current and the intensity are better controlled;

• This lets concentrate the effect to some cm².

Since the gradient of electric potential is the time derivative of the magnetic vector potential A, the average of the gradient over a cycle is zero; I understand a net electrochemical effect is obtained because electrochemical reactions are non-linear (they often show a threshold to voltage) and the pulse is made asymmetric, with a sharp current rise that induces a strong short gradient, and a long current tail that induces a weak long gradient.

 

A pulse can have 100µs rise time and 1ms fall time. Series of pulses are also used, sometimes of alternate polarity, but always with pulses of asymmetric transition times.

 

A good gradient of electric potential needs a strong magnetic induction, and in these non-permeable materials, this means kA in a coil, many kW pulse power, and as a result, the coil is very loud and gets warm, which limits the duration of a session or demands active cooling.

 

-----

 

I propose to split each pulse into N shorter ones, scaling like that for understanding:

• Individual pulses are N times shorter, both at rise and fall time;

• The reactive volts per coil turn are kept the same;

• Hence the rate: amps*turns per time unit is unchanged;

• And the amps*turns are divided by N.

The effect on the brains should keep unchanged when introducing N, because:

• The electric potential gradient, which results from the rate of change of the magnetic potential, is unchanged; (if you prefer, it varies as the volts per coil turn, which are unchanged)

• The cumulated duration of the electric potential gradient is unchanged;

• The proportion between rise and fall time, hence the ratio of electric potential gradient is unchanged.

Beware a few things can go wrong here... For instance if some electrochemical reaction needs a consolidation time before it withstands the inversion of polarity of the electric potential gradient. As usual, I didn't check if this was proposed, tried and possibly abandoned - sorry for that - so interested readers should.

 

Provided the desired effect is still present, splitting the pulse provides big advantages to the apparatus:

• The current is divided by N, but the duration remains;

• Consequently, forces in the coil are divided by N²;

• The coil's deformation speed after N shorter pulses is divided by N, and the noise power by N²;

• The energy dissipated in the coil is lowered. By less than N² because losses increase with frequency;

• Electronics can deliver N times less current if the voltage is kept, and the energy is N times smaller.

Transition times divided by N may cool down the enthusiasm of electronics designers. As well, it requires probably to brake the coil current actively, for instance with an H bridge, or braking diodes reinjecting current in the power supply, possibly by a secondary winding somewhere. A transformer before the coil can bring advantages: match the cable's and coil's impedance to the power supply and switching components' possibilities; have a separate winding to brake the current; produce inverted pulses by an other primary winding and two switches instead of an H-bridge.

 

Coils should be made of so-called Litz wire (probably a wrong translation for "braided wire"). In fact, if this isn't already done with 100µs pulses, it should have been: this limits the heat in the coil but may demand a current braker. The current braker could even be a series resistor, easier to cool than the coil. Shorter pulses demand wire braided more finely, which reduce the wire's section filled by copper, so the energy lost in the coil is divided not quite as quickly as N².

 

Marc Schaefer, aka Enthalpy

[iNow]

Interesting thoughts. I'd be curious to see data when this is tested empirically, but what did you want to discuss here about it?

[Enthalpy]

A well designed TMS coil can have 1ms time constant, real ones have a bit less, but operation at N=10 or N=100, with fall time shortened accordingly, needs to brake the current actively. The brake circuitry (a lossy freewheel if you wish) is less than obvious, because the brake voltage is (say 10 times) less than the supply voltage, and the brake circuitry shall not misfire during pulse rise, despite users want pulses of both polarities with few ms between inversions.

 

Here's a circuit example:

An IGBT opened long before the pulse decides the polarity. The TMS coil could have been connected between these collectors, but the second transformer insulates the TMS coil and permits a direct gate control. A big E ferrite core suffices for such a transformer, so I feel this combination easier.

 

The left IGBT defines the rise time of the current pulse, and the supply voltage defines the rise rate.

 

On this sketch the resistor sets the current fall time in the coil. Operation at big N is meant to reduce losses in the TMS coil, but here most supply power is dissipated in the resistor. While this power is smaller than at N=1, and the resistor easier to cool than the TMS coil, one may prefer to reinject the energy in the supply. A third winding can help this, but since the voltage and current must already challenge the existing main switch technology, this winding shouldn't step up the voltage, so the power dump rail must have a lower voltage than the supply.

 

An adjustable power dump rail would control the fall time independently, and a braking voltage constant over the pulse may improve the TMS operation by limiting to a uniform plateau the induced electric potential gradient. The rail must accept and transform power but be pre-loaded before the first pulse. Several braking windings of different voltage ratios may help. I preferred to draw a resistor on the sketch

 

At high N operation, hence with shorter commutation of a smaller current, MOS must be better than IGBT. Soft commutation (resonance) would be nice, except for the clarity of the sketch

 

Marc Schaefer, aka Enthalpy

Edited November 3, 2012 by Enthalpy

[iNow]

Again, what precisely did you want to discuss here? You are treating this site like a blog, from what I can tell.

[Enthalpy]

If you know TMS apparatus that already work like this, don't hesitate to tell.

Or if you know reasons why the brains shouldn't react the same way when the pulse is split.

Just as examples of discussion subjects.

[iNow]

What about sQuid?

[Enthalpy]

Up to now, supplied power increases the current quickly in a TMS coil, which must be the neuro-active phase when the induced electric potential gradient is strong, and a passive circuit lets the current decrease more slowly, in which phase I postulate the electric potential gradient must be below a threshold so it doesn't cancel out the effect obtained in the active phase.

 

If this model holds, an other operation mode must produce the same result, where the powered phase increases the current slowly in the TMS coil, inducing a limited electric potential gradient, and the neuro-active phase brakes the coil current quickly to induce a strong electric potential gradient. (This must apply as well when the pulse is not split.)

 

In this other operation mode, the circuitry first supplies a moderate voltage to the coil to accelerate the current, then interrupts the current brutally, which results in a strong electric potential in the coil as well. Very similar to an engine's spark ignition circuit.

 

This operation mode seems advantageous:

• The smaller electric potential gradient, which I suppose is the more delicate phase, is better controlled as it results from the supplied voltage;
• The brutal phase, when both U and I are strong, can flow in the circuitry through a diode, which is more robust than a transistor;
• The circuit example here is simpler.

 

The diodes and transistors must withstand the inductive overvoltage when cutting the current; the supply voltage is much lower, roughly in the same ratio as current fall and rise times in the TMS coil, for instance 10 times lower. The inductive overvoltage is defined through the brake voltage in this example.

 

The power supply provides less peak power now, but the brake circuit absorbs more, and the transistors must cut the full I at full U, uncomfortable. If the brake windings have as many turns as the accelerating windings, these windings can be coupled very closely to protect the transistors, and can even be the same winding - but with the same number of turns, the brake voltage is much bigger than the supply voltage.

 

A different choice uses fewer turns at the brake windings, and injects the brake current directly in the supply. This limits to a fixed ratio between the current rise and fall time, which can be switched if additional brake windings provide different numbers of turns. The transistors demand a separate protection then. Anyway, circuitry to protect the components (and preferably soften the transitions) is necessary, though not drawn on the sketch.

 

Marc Schaefer, aka Enthalpy

Edited November 4, 2012 by Enthalpy

[iNow]

Okay, yeah. You're just blogging. Have fun then.

[Enthalpy]

Here are eventually some wave forms, click to magnify:

 

 

I've sketched N=5 for clarity, but real apparatus will probably use a bigger N.

 

Wave forms are idealized; natural ones are smoother, and shall be to protect the components.

[Wilmot McCutchen]

So what is the purpose of this, and its advantage over what is known? Maybe less math and abstraction would make that clear. Your proposal seems similar to the magnetohydrodynamic circulation stimulator (Bemer 3000). Do you intend to make a more focused electrode for near field effects on brain targets? Is modulation of the effect important and why? You say there is a problem with noise in the known devices, so what is the source of the noise and how does your proposal fix that? Are present devices operating at sonic frequencies?

[Enthalpy]

As shorter pulse times reduce the induction far below the present 1-2T, a ferrite core gets possible. Only outside the skull, but it can cut by two the magnetic path length and the current needed.

 

Smaller current and induction also enable coil shapes not circular. Though, neither induction nor the vector potential can be focussed at distance; at best, a subtractive pattern of currents could make the vector potential sharper. As creating the desired fields gets easier, we may consider subtractive patterns...

 

In every case (also at 1ms and 10ms), the turns of the coils should be spread apart where a concentrated field is not desired. This reduces the self-inductance hence the necessary voltage - or current if increasing the number of turns.

 

Marc schaefer, aka enthalpy

 

=============================================

 

Even shorter pulse times seem possible.

 

With a single thick turn, an 8-shaped coil with both D=70mm loops in series shows about 240nH. It still needs about 200V per turn, but limiting this to 5ns reduces the peak current to 4A approximately, and this is accessible to some RF transmitter transistors. Or have on transistor per loop, at 100V and 4A each.

 

Then, the transistor(s), capacitor (or a cable´s capacity) and flywheel can sit in the coil head, and the cable transmit DC power instead of the strong pulse. Much gets simpler.

 

Can we exaggerate the pulse duration further? Yes, but with an electromagnetic pulse then, not just magnetic. It must have unsymmetric durations if my explanation holds. This one would fully enable to concentrate the field. More later, maybe.

 

Marc Schaefer, aka Enhalpy

Edited November 11, 2012 by Enthalpy

[Enthalpy]

Here are the losses I computed for eddy currents in wires, because these predominate if thick wire is used at moderate or high frequency.

 

Omega is 2*pi*F

B the RMS induction

rho the resistivity (18e-9 ohm*m for cold copper)

d the diameter of individual wires

D the diameter of the bundle

l the length and V the volume of the conductors

eta the filling factor

mu the total permeability (pi*4e-7 H/m for vacuum).

 

A wire, thin enough to let an external field pass through nearly unchanged, dissipates as (provided I didn't botch it, of course):

 

 

A round bundle of thin wires that creates its own induction dissipates as below. This holds approximately for an annular coil BUT beware it's for 1 turn... Multiply by the turns squared. Note losses don't depend on the bundle's diameter.

 

 

The limit where eddy currents lose as much power as the DC resistance does is below - still when the bundle creates its own induction. It can be worse with a magnetic core, especially near an air gap.

 

 

The parry to eddy currents in conductors is a braided wire ("Litz wire") composed of insulated thinner wires twisted together. Round wires can occupy this fraction of the cylindrical section at best:

7/9 ~78% for a 7-wire Litz

19/25 ~76% for a 19-wire Litz

49/81 ~60% for a 7*7-wire Litz, and so on

knowing that a cylindrical section occupies at best 79% of the winding area of a coil.

 

This is how typical sections of Litz wire look like, with 7, 19, 37 and 7*7 twisted wires:

 

 

Marc Schaefer, aka Enthalpy

Edited November 17, 2012 by Enthalpy

[Enthalpy]

Here is a coil enabled by the smaller current and induction: neither round nor flat. The turns are packed close where the effect is sought, and spread apart elsewhere to reduce the inductance.

 

 

Usual 8-shaped coils induce undesired potential gradients nearly half as strong under their return paths as under the central zone; the coil sketched here shall minimize them by placing the return paths farther away from the cortex, but near enough to reduce the effect of the main current path outside the central zone.

 

The angles of the turns (possibly more than 4, possibly of different sizes and shapes) should hence be optimized. I won't do it; a small software needs only to sum over the current path the contributions to the vector potential A to various positions at the target. The Biot and Savart formula is here, and in

http://de.wikipedia....t-Savart-Gesetz

 

The active zone being ~45 mm long and the return paths ~85 mm, I estimate the inductance to 560nH with for instance 4 turns, using 4*500 nH/m where the turns are grouped and 1*500 nH/m where they're spread. To achieve the same 2,9 GA*turn/s as the 2* D70mm coil from Magstim fed with 2800 V, the design example takes only 725 MA/s and 406 V, and if the pulse's active duration is just 1µs, the peak current is only 725 A. Power components may prefer more turns.

 

The coil's wire consists for instance of commercial Litz wire of 33*Awg41 or 71µm individual wires. 7 such threads are twisted to a strand (for 2m in a lab, fasten them to the ceiling, put weights, and turn) and 7 strands to a conductor of 6.4 mm2 in D~4.5 mm, filled to 40%. Over the coil, 1.5 mohm DC resistance lose 2.8 mJ over one pulse of 1+10 µs * 725 A, while eddy currents lose roughly 1.6 mJ. Copper weighs 30g, so if hundred 1+10 µs pulses are one 0.1+1 ms pulse worth, the coil can absorb 2000 equivalents in 80K heating.

 

The cable may use the same 7*7*33* D=71µm conductors; 4+4 of them, insulated and braided, plus shielding and mechanical protection, make a 2 m bipolar cable. Copper weighs 0.91 kg; 2.8 mohm DC resistance loses 5.4 mJ and eddy currents some 1.8 mJ. The coil plus the cable lose 15mJ from a 150mJ pulse. Aluminium Litz wire would be highly welcome.

 

A pair of Toshibas's MG300Q1US51 - and certainly others - look capable of switching the pulse: 1200 V, 300 A, tr=50ns tf=100ns. The forward mode would reduce conduction losses in the IGBT. Regenerative braking reduces the supply's capacitors to ~1 dm3.

 

Marc Schaefer, aka Enthalpy

[Enthalpy]

Greater numbers of shorter pulses need even less power. Here the current shall increase for 5ns and decrease for adjustable 50ns.

 

The coil can keep the same shape as for 1µs but now thick rod or tube is good enough. The turns are now in parallel to reduce the voltage to 100V; or several generators can feed each a part of the coil; or the coil can consist of one single very broad conductor, preferably broader at the return leg, and accept a slightly smaller voltage. Over 5ns, the total current increases to 3.6A only.

 

The outer turns will swallow more current than the inner ones, which isn't bad for the induction pattern; to avoid this, put the turns on a square pattern (or hexagonal, etc) near the target zone, instead of side-by-side.

 

The fast main switch is a dual MOS meant for RF transmitters, like the BLF884P. Their supply is at most 50V, brought to 100V by a 1:2 transformer which also produces bipolar pulses. Braking through the diodes reinjects power in two special lines; the one that shall not brake is just fed over twice the supply voltage - this better design works for longer pulses as well. Some diode-capacitor pump can preload the braking lines, while voltage regulation defines the braking time. A few Zener chosen by transistors could dissipate the small power.

 

The diodes could be SiC Schottky like the C3D08060A. These would contribute 4W of the 7W switching losses at 70ns period, and to nearly all 4W conduction losses. These diodes begin at 600V in a LF package; a better fit would be welcome.

 

 

The HF electronics is at the coil, and the power and control cable is now easy and lightweight.

 

Marc Schaefer, aka Enthalpy

[iNow]

Very engaging dialog you have going here, Marc. I can see why you've chosen a discussion forum instead of a blog to post about your thoughts.

[Enthalpy]

I just haven't seen how the squid, which I understand as a sensor, relates to this actuator.

[iNow]

Much like I haven't understood how your posts encourage discussion of any sort. You're here blogging. It's annoying.

[Enthalpy]

Usually on a forum, I don't need to tell participants "I'd like to discuss this and that point": they find alone what they want to discuss or comment.

 

And on this forum and others, I do get useful feedback that is not available on a blog. Just one example there:

http://www.scienceforums.net/topic/70340-reactor-for-liquid-and-gas/

 

Maybe electromagnetism is just too difficult - perfectly arguable. Neurologists as well are scarce on a science forum, alas.

 

I understand you good right, iNow, to find this thread annoying. Hopefully other readers are more interested. Just consider that the improvements you're reading here were intensely sought for 25 years by the users and developers of TSM apparatus. Several threads on an other forum even look like badly disguised queries for technology enabling precisely this. If the great numbers of shorter pulses show the good will to be one longer pulse worth, the present discussion will be remembered in neurology as what made TMS practical.

[iNow]

I understand you good right, iNow, to find this thread annoying. Hopefully other readers are more interested.

Not to put too fine a point on it, but the complete and utter lack of participation from other readers here... and the fact that you're doing little more than serial posting, basically talking to yourself... should give you some insight into the likelihood of that.

[Wilmot McCutchen]

Not to put too fine a point on it, but the complete and utter lack of participation from other readers here... and the fact that you're doing little more than serial posting, basically talking to yourself... should give you some insight into the likelihood of that.

 

I for one am interested, and have posted here. The stream-of-consciousness format is not a turnoff to me, as I enjoy watching an expert grapple with a problem. I have learned a lot from reading this thread.

[iNow]

Good. I'm glad. That doesn't negate my point that this would be a better avenue for this given approach --> http://www.scienceforums.net/index.php?app=blogs

[Enthalpy]

With pulses like 70 ps long we leave the simpler near-field operation and magnetic coils. This is electromagnetic wave with antennas.

A strong short pulse followed by a weaker and longer compensation needs a wide band antenna, nevertheless directive and powerful. Here's one possibility I suggest (click to enlarge), where the rectangles are conducting elements meant to radiate:



The antenna comprises many dipoles side by side and butt, together with driving electronics. The shape should better approximate a section of a sphere; the dipoles can be driven with a slight time difference only, for fine steering, to compensate small tolerances, or to compensate diffraction at the target. The antenna has no reflector but can be backed by an absorber at some distance.

The dipoles are charged before each pulse; here resides the emitted energy. Each dipole has locally its own fast transistors to discharge it, for instance two BFG425W. The switch could also be behind a line, and possibly shared among several dipoles, but bigger transistors are uncommon at that speed.

Take dipoles 2* 20 mm long, wide to have 80 ohm wave impedance, charged at +-4 V. The transistors can reach the 50 mA in 30 ps; this constant current widens to 2* 20 mm within 70 ps, which defines the duration of the strong short plateau of induced electric field at the target.

10*12 dipoles at 200 mm distance would create in air the same field as present TMS apparatus with 3 GA/s. Permittivity at the brains and the cranial liquid reduces the field, and so do reflections at the scalp and the skull, but R² times more dipoles, R times more distant, create a field R times stronger.

When the current pulse reaches the ends of a dipole, fast diodes (for instance three BAT62 per dipole) allow the current to continue flowing, due to the antenna's inductance, this time between butt dipoles. The current decreases more slowly, first due to losses in the transistors and diodes, say 1.2 V versus 8 V accelerating voltage, and second due to the finite length cumulated by butt dipoles. This defines the weaker longer plateau of induced electric field, maybe 6 times weaker here than the strong plateau. Less asymmetric than before, but still better than present TMS apparatus.

-----

The dipole width or diameter can adjust the wave impedance a bit to match the components' current capability. They can't be too close, or their interaction will limit the current. The examples given are not optimum, especially the old diode is capacitive and slow (which cancels out partially). Stronger voltages would be very useful but MOS seem unavailable at this speed. I haven't checked other FET.

Driving the bipolars is difficult. I expect no charge gain per stage at 30ps, so the gain of the driving tree shall result from impedance transformation, but on a wide band and with floating voltages... Striplines similar to a gamma match, with ferrite for the bandwidth, and tapered to match the impedance?

-----

70 ps make a pulse 20 mm long in air, which relates to the best field concentration at the target. This speed is difficult for components, and isotonic water has only 49 mm penetration depth at 5 GHz and 22 mm at 10 GHz.

Operation immersed in a water-like (or brain-like) material would improve a lot. It could use a liquid gel at the hair and a solid gel on the way from the antenna. 2 to 5 GHz limit the losses; the ~150 ps pulse is easier to produce even if stronger and is only 5.5 mm long in water, which improves the field concentration at the target and the size of the dipoles.

Immersed operation also reduces reflections, as only the skull has strong interfaces, and reduces imprecision due to refraction. Subtle beam synthesis by electronic steering can compensate the aberrations due to the skull, but this difficult option is more futuristic.

If the antenna doesn't touch the skull, the beam can follow the optically observed head's movements, by limited electronic steering, or by automatic control of the antenna's position.

Marc Schaefer, aka Enthalpy

Edited December 9, 2012 by Enthalpy

[iNow]

I see you're spamming other sites with the same copy/pasted text, too.

 

http://www.physicsforums.com/showthread.php?p=4189728

[MonDie]

I didn't gather much from the OP, but it seems like they want to influence brains with an external device. If that's the case, they could use it to trick people who think they're telepathic. If it works, the "telepathics" weren't actually doing it, they were just attributing the effects to themselves rather than an outside variable.

[iNow]

Transcranial magnetic stimulation is a very real and very cool technology. It's been used for everything from treating depression, to seizures, to triggering a sense of oneness with the universe and feeling like you're standing before god.

 

It's just that enthalpy is blogging, not discussing. He doesn't even respond to pointed questions. He's just here to have an audience, not to chat. That's why I recommended his style would be better suited to a blog. My point has never been to imply that TMS is crap. It's not. It's really quite cool stuff. It's that he doesn't seem to grasp the nature of a discussion forum or community like this.

 

He may as well be installing brakes on his car and posting the measurements between the rotor and the caliper or posting ingredients for his favorite ice cream or lyrics from Dolly Parton songs.

Edited December 9, 2012 by iNow