Failure Analysis on Neural Implants

I've spent pretty much my entire career working on one instrument, the Focused Ion Beam (FIB). I'm a FIBber! FIB's are really useful for failure analysis. And when I talk about failure analysis, I'm not referring to sitting around and thinking about my life...No, I'm talking about failure analysis on microelectronics! FIB's allow you to cut tiny things open and look inside of them with an electron microscope. Lately I've been doing some failure analysis on neural implants. We'd really like to understand how and why they fail, as well as how long it takes before they start failing so that we know what to expect before we stick them in humans. 

Ideally, if you are going to go through all the trouble of opening up your head and putting some electronics in your brain, you want it to work for the rest of your life. And when you think about it, your brain is a pretty harsh environment for electronics. I mean think about it, would you expect your iPhone to work right if you put it inside of a 98° F salty, pulsating, gooey liquid for 50 years? Probably not, shit is going to corrode, dissolve, swell up, or fall apart in other ways.  

Because neural implants are such "new" (1) devices, there aren't a lot of standards for how they should be tested and characterized. This review article posted in Nature proposes some unified standards for testing neural probes. I'll take a look at what these nerds think and summarize it here.  

WOW! Ok, so now I'm actually trying to do what I said I was going to do in the first paragraph and I became really overwhelmed. This paper is dense AF! I guess that's the nature (haha) of a review paper, it distills the current state of the art of the entire field into one ~20 page bundle. Ok, biting it off chunk by chunk.

Let's start here: 

When we talk about neural interfaces, we are talking about a device that either "reads" signal from a neuron or "writes" signal to a neuron. It does that by stimulating the cell with electrical impulses or recording the electrical impulses that the cell is outputting.

This is accomplished with the use of an electrode (ideally a very small electrode) that can either be laid on top of the brain or stabbed into the brain.

When neurons fire, it is electrical signal, but not really the same as how we understand it in our big world of power lines and power outlets. The charge carriers are different. What I mean by this is it's not electrons traveling through a copper wire, it's ions like sodium and calcium moving through brain juice.  That means when you want to stimulate a brain cell or record from a brain cell, you need to translate from electronic to ionic signal and vice versa. This is what electrochemistry is all about. The voltage change that results from ions moving in and out of the cell moves electrons around on the electrode and that's what our recording electronics receive. When we stimulate, we move electrons onto or away from the electrode and it pushes ions in or out of the cell. Essentially, we are using an electrode to make neurons fire when they otherwise wouldn't.

I'm working on a project that seeks to characterize how the electrodes are going to change and break down as they age.

So, in order to test electrodes outside of a brain, we simulate the brain as a beaker of salt solution. 

 

 

Oh, what? You think your brain is way more complex than a beaker of salt solution? Well, it turns out it's not! (in this context). When we are characterizing just the performance of the electrode alone, all we care about is the immediate liquid environment surrounding the electrode surface, which is actually quite similar to a beaker of salt solution...at least for short term studies. 

You see the "WE" and the "CE" in those images above? They stand for Working Electrode and Counter Electrode. The working electrode is what delivers the pulse and the counter electrode has to be there to complete the circuit. 

When we simulate the brain as a beaker of salt, it allows the use of a reference electrode (RE). This allows us to measure the charge (the electrical potential) of the liquid alone. We can't do this with the working electrode alone because when we run current through the WE, it changes it's potential. The RE is made of a material with an enormous resistance so no current will flow through it, it's just there to passively measure charge. 

So what do we do next? We try out a bunch of electrical signals of different frequencies and see how the system responds. Remember all that stuff about sine waves you learned in high school? Watch out, here it comes again! So, we send in pulses of alternating current that vary in frequency and then we measure the phase and amplitude of whatever is coming out, this allows us to measure the impedance of the electrode. This is called Electrode Impedance Spectroscopy. 

Impedance is kind of like electrical resistance but it's more complicated. We use the concept of impedance when we have things like capacitors in a circuit. In the case of these electrodes, they do behave like capacitors. The surface gets charged up with electrons from the pulse we are sending in and that causes the positively charged ions in the solution to flow over and form a layer on top of the electrode. A plane of positive charge and a plane of negative charge is all you need to make a capacitor, so we can model the electrode as a circuit with a capacitor and some other mystery box that contains all the other electrochemical processes going on. 

That's what is going on in the leftmost part of the diagram below, the electrochemical processes that are happening when you submerge an electrode in a saline solution are modeled as a capacitor and some other resistances.  

 

 

People testing neural probes love using EIS because it's a way to separate the performance of just the electrode alone from the performance of the entire circuit (which would include the oscilloscope or whatever device you are using to generate the input signal and measure the output signal.) 

When you do EIS, you end up making a bode plot - which is actually two plots, one of them plots frequency against impedance and one plots frequency against phase. We already talked about what impedance tells us. Phase tells us how much the response is lagging the signal. Like, I send in a pulse but it might take some time before the ions in the solution flow over to the electrode. This will result in a phase change. 

There are a bunch of design choices we can make to change the electrochemical characteristics of the electrode. For one thing, how much charge a capacitor can hold (it's capacitance) is determined by its surface area. If you make your electrode rough like Velcro instead of smooth, that will increase the surface area and thus increase the capacitance. You can also use all kinds of different materials that yield a variety of different chemical reactions. BUT, if a chemical reaction takes place as a result of a charge pulse from the electrode, we want to make sure that an equal and opposite reaction takes place. Otherwise, we are changing the chemistry of the implant and the surrounding tissue, which is no bueno.

When we run current pulses into these systems, we want to do whatever we can to prevent the electrode or its surrounding solution from changing. This means we don't want any electrons to come off the device. If electrons come off the electrode and flow into the solution, this could change the pH of the solution and induce chemical reactions. This is called an irreversible process, and it's the last thing we want. It means either the electrode or the surrounding tissue is turning into something else. Usually, this looks like corrosion on the electrode surface. We are shooting for reversable processes here! Ideally, ions move toward or away from the electrode when we charge and discharge it with external pulses with no chemical reactions taking place. OR, if a chemical reaction is taking place…we want it to be reversible.

One thing you REALLY don't want to do is exceed what's called "the water window". This means you are driving the electrode at a voltage that pumps so many electrons into the tissue that it starts breaking water apart into it's constituent components and making hydrogen gas and oxygen gas. This damages both the electrode and the surrounding tissue. AND, since there is so much water available in the body, it’s a run away process. We usually determine the voltage at which water splitting occurs and set this as an upper boundary for safe electrode use. 

So, what we’re doing is putting some electrodes into a solution and heating them up. The heating is supposed to simulate accelerated aging. Elevated temperatures speed up the kinetics and make it so that whatever chemical reactions are going to take place will take place faster.

At certain time intervals, the electrodes get taken out of the solution and EIS data is collected. There must have been previous experiments on these devices to determine the right voltages to apply in order to stay out of the water window. I’m not sure, I wasn’t involved in this part of the experiment. Hopefully someone tells me!

 Ok, that's enough nerding out for now!

1.) They aren't that new and I have a blog coming soon on the history of neural implants! Stay tuned.