1 transistor per neuron recording device

ScienceDaily: Semiconductor Brain: Nerve Tissue Interfaced With A Computer Chip

From the article:

16384 transistors on an area of one square millimeter record the neural activity in the brain.

Hmmm, that sounds like a lot of transistors… what kind of voltage sensing resolution can a device like that provide? Well, that works out to 1.6 transistors per 10 square microns, which is arguably the relevant area for a neuron. Although these are extracellular signals, this high-resolution tool is going to have quite a large impact.

From the abstract:

We report on the recording of electrical activity in cultured hippocampal slices by a multi-transistor array (MTA) with 16384 elements. Time-resolved imaging is achieved with a resolution of 7.8 µm on an area of 1 mm2 at 2 kHz. A read-out of fewer elements allows an enhanced time resolution. Individual transistor signals are caused by local evoked field potentials. They agree with micropipette measurements in amplitude and shape. The spatial continuity of the records provides time-resolved images of evoked field potentials and allows the detection of functional correlations over large distances. As examples, fast propagating waves of presynaptic action potentials are recorded as well as patterns of excitatory postsynaptic potentials across and along cornu ammonis.

M. Hutzler, A. Lambacher, B. Eversmann, M. Jenkner, R. Thewes, and P. Fromherz: High- resolution multi-transistor array recording of electrical field potentials in cultured brain slices. Journal of Neuropyhsiology. Preprint online (May 10, 2006).

The original article (whichs seems to online in a preprint form) has excellent photos of the array (showing how it can cover a lot of a hippocampal slice), the tight correspondence between the transistor signal and a microelectrode field signal, and some cool readouts of the “whole hippocampus” with various blockers. I doubt anyone has ever been able to simultaneously do such fine scale electrophysiology on such a large portion of the mammalian brain ever before.

3 thoughts on “1 transistor per neuron recording device

  1. This is quite a impressive development indeed. It looks like extracellular resolution will increase in a Moore’s law like fashion in the coming years, but this still leaves the complexity of the intracellular matrix a relative black box. Until that is figured out in more detail the development of a “functional” hippocampal neuroprosthetic will remain very difficult.


  2. I’ve been discussing this device a bit more with my friend Alex Shalek, a fabrication guy also doing work with neural recording devices. Here’s some of our discussion about potential limitations of the technique, which I’ve posted in its original dialogue format.

    > >>>Yep, I know the work well … it’s a really cool idea with a couple
    > >>>limitations at present … primarily, the heterogeneous distribution of
    > >>>ion channels present across a neuron makes it hard to interpert signals
    > >>>(e.g., you might have capacitive coupling or a leak current or anything).
    > >>>The videos are cool as well – anyway, it would be better if they could
    > >>>make an array of eos-fets rather then via-ed mosfets but they’re harder to
    > >>>scale … enjoy your retreat and let me know about the cells … I
    > >>>appreciate everything,

    > >>Alex,
    > >>
    > >>Thanks for the interesting stuff to think about. Is the heterogeneity
    > >>really an
    > >>issue? Are the MOSFETs really limited to sensing such a small area that they
    > >>would be susceptible to ion channel inhomogeneities? (I’m thinking of the FETs
    > >>as really tiny bipolar electrodes here, which I think it correct…) Are the
    > >>source and base terminals for each FET very close together in a chip
    > >>like this?

    > > Hey Neville,
    > >
    > > So, in short, the answer is yes – some of their older papers document this
    > > well … If you compare a patch recording (x) with what the FET sees (y),
    > > sometimes x=y, sometimes dx/dt = y, sometimes -dx/dt = y, and everything
    > > in between. The real issue arises from the manner in which the FETs are
    > > gated – basically, you’re using gradients in the ion concentration as the
    > > gating mechanism … if you have a distribution, and those channels have
    > > different polarities and opening times, it becomes very hard to figure
    > > out what’s going on … and even without ion channels, you can get a
    > > capacitive signal from the membrane. In CMOS, there’s only one carrier
    > > (the electron), here everything is much more complicated … finally, to
    > > help you think about size, remember that the FETs are at 7 um pitch –
    > > that means that within 7 um, you have a source, a gate, and a drain for
    > > the FET and a spacer area (between it and the adjacent FET) … this
    > > means that the gate is relatively small … I hope this helps,

    >> I see what you mean… I can imagine that the terminals
    >> could be small enough that you would have issues with
    >> spatial inhomogeneity. I guess the best solution would be to
    >> pack the chip at an even higher density. Then, you’d hope
    >> that certain spatially continguous sets of FETs covary
    >> together enough that you’d believe that they were controlled
    >> by a single cell…

    > Neville,
    > I should point out that it’s a brilliant idea – it just hasn’t proven
    > itself yet … you’re right about generating algorithms to interpret the
    > data – the big problem then is the size of the data stream – 2khz*10^4
    > transitors*experiment time = a lot of data to store and then correlate
    > and then analyze … it’s daunting to say the least … we thought about
    > modifying Fromherz’s FET design originally (and probably will do something
    > like it in the not so distant future), but opted not to based on the fact
    > that, for large data sets, you want to be positive from which neuron
    > you’re recording … also, we wanted to study intrinsic changes (e.g.,
    > resting membrane potential) …


  3. Pingback: BiPolar Disorder blog

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s