Interview with Prof. Oliver Paul and Dr. Patrick Ruther
Prof. Paul is a professor at the Department of Microsystems Engineering (IMTEK) of the University of Freiburg. He holds the Chair for MEMS materials (MEMS: microelectromechanical systems). His research focuses on process technology and materials characterization for MEMS based on silicon technologies, in particular complementary metal oxide silicon (CMOS) technology, as well as silicon based MEMS devices and systems for physical measurements, automation, reliability testing and life science applications.
Dr. Ruther is a senior scientist and group leader with Prof. Paul and lecturer at the IMTEK. His research interests include CMOS-based sensor systems and innovative processing technologies. He served as the technology coordinator of the EU project NeuroProbes targeting multifunctional neural probe devices.
What is so special about silicon that makes it interesting for biological applications?
Paul: Silicon-based MEMS engineering means that you achieve incredible integration densities. What I mean by this is that modern microelectronic silicon chips carry many billions of transistors in a tiny area. The resulting high functionality can also be used in biological applications. The sizes of single cells are in the range of micrometers, that is one thousandth of a millimeter. If we now build smart structures as they are known from microelectronic chip design, and take advantage of them to investigate the activities of nerve cells with such high resolution, we are able to learn more than we possibly could with any other variant of microsystem technology.
With the construction of such densely packed sensor arrays, we also need filters for the gathered data, amplifiers, digital-analogue converters, for instance on a small implant within the brain. You couldn’t build these without microelectronics. Only this combination of sensors and processing technology enables us to construct this kind of small-scale systems useful for biology.
Silicon-based electrodes. Contrary to traditional types, they carry several recording sites along its length (from S. Kisban et al. (2010) Novel Method for the Assembly and Electrical Contacting of Out-of-Plane Microstructures, Tech. Dig. MEMS 2010 Conference, Hong-Kong, 484-487).
What risks come with the used materials and technology?
Paul: On one hand, we have to make sure that our machinery stays in the body, and more particularly in the brain if we are talking of neural probes. The organism will often try to digest an “alien” object, or to encapsulate it, or simply expel it. But there are interesting findings today that show that if an object is smaller than about 7 micrometers, the brain won’t recognize it as an intruder and will not try to get rid of it.
But we also have to make sure that the device does not carry a risk, for example by being irritating or even poisonous. In general, the materials on the surface, silicon or its oxidized form, are biologically harmless. Platinum as a metal is also a material that is widely used in medical technology. As for more exotic materials, which we don’t know that well, we are extremely reluctant to use them.
What are the challenges when you try to read out information from the brain?
Ruther: You always have to weigh how much information you actually wish to extract from the brain against how many probes you have to insert into it. Conventional probes basically consist of a wire that records at the tip. With those probes it is possible to read out information from different locations in the brain by moving the wire deeper or closer to the surface. However this movement can entail an irritation of the brain tissue. As far as our probes are concerned, we drive them into the brain once, and then we are able to record every 40 micrometers along the four millimeters of the probe length. That’s something you cannot do when you utilize other technologies.
Isn’t one problem nowadays that you can have many recording sites, but the number of sites that you can record simultaneously is much lower?
Paul: True. The probes we are working on right now have 188 channels per shaft, but you can only listen to eight of them at the same time. What the researchers usually do is to scan the activity along the shaft, and then focus on those channels that sound most promising. And if you combine these probe shafts into “combs”, you have 32 simultaneous channels. If you then stack these combs again in a sort of “pin cushion”, you have 144 sites already from which to record at the same time.
Neuronal probes with electrode arrays in two (above) or three dimensions (below). The arrays are combined with a cable made from polyimide, which is only 10 micrometers thick (IMTEK-MML).
Connecting the brain with a computer through a cable connection always carries the risk of infections in the area where the skin is breached. Are there possibilities to get the required data into or out of the body with wireless connections?
Paul: That is definitely the way to go and also a great chance for us here at Freiburg. All the necessary know-how is available here. We have engineers in microtechnology, signal processing, systems engineering, we have specialists in wireless communication, and finally data miners who find ways to decide which part of a recording is a meaningful signal, and which other part can be discarded right away. This is a big advantage that we have here through the presence of computer engineering, biology, medicine, and microsystems engineering.
Ruther: Right now, we cooperate with many colleagues worldwide. As an example, we developed in a cooperation with the University of Cambridge a neural probe systems that enables to read out 32 channels from the brain using a wireless connection. The next step is to double this to 64 channels, and with the know-how we have in Freiburg, this will definitely not be the limit.
How many nerve cells would neurobiologists like to record?
Two types of probes
Neurotechnologists will sometimes talk about "active" or "passive" probes. The difference between them is simple:
Paul: Ideally, they would like to record the activity of the entire brain! But more practically, there are quite a number of questions that can be answered already with data from a few places within the brain. However, we would be completely overwhelmed today if we were able to record simultaneously the signals of several thousand neurons. The sheer amount of data would be simply overwhelming. Nevertheless, we have the chance, especially here in Freiburg, to use the methods of computer science to decide what is useful information – and what isn’t.
Today, neuroscientists use their ears and say ‘Ah, I can recognize a signal here’. That is basically data processing by using the intelligence of the human brain. But if you look at many channels simultaneously, this requires data processing with a more modern approach. And there, fantastic opportunities open up by involving computer science in the neurosciences.
Ruther: We are in the process right now of developing a system with colleagues from the computer sciences that is based on machine learning and analyzes such recorded data in real-time. The system scans through the channels and automatically decides which input is worthwhile being analyzed further, and which recording sites can be discarded right away. In the next step, we want to bring in experienced neuroscientists and let them look at the same data and decide which channels carry useful information. We will sort of pitch the human and the computer-based system against each other in order to see how well the computer performs already. From the computer science point of view it amounts to extracting the right metrics able to measure information content. And I’m sure much can be learnt along this path.
The different parts of an assembled electrode array (from S. Herwiket al. (2009) Fabrication technology for silicon-based microprobe arrays used in acute and sub-chronic neural recording J. Micromech. Microeng. 19, 074008).
What new developments do you see coming up on the horizon?
Paul: Autonomous systems are the vision that many of us share. That is, clinically relevant applications that help people with neurological disabilities, as it is already being done in the case of Parkinson’s disease with deep brain stimulation probes. Of course, we as scientists are impatient, but a society also has to mature in its stance towards a technology. You simply cannot convince the public today that it’s necessary to have a data plug coming out of your brain. Many people would spontaneously oppose such a vision. But if you tell these people that there are tens of thousands of deep-brain stimulators in use already that ease the symptoms of Parkinson’s, you see surprised faces, and people say ‘Now, that’s interesting!’. Because it suddenly becomes related to their own lives. The paradigm that the human brain is the final, sacrosanct frontier in our body, will probably start to change if the public can see a tangible benefit coming from neurotechnology. Those working in the field clearly recognize this chance. It would have been unthinkable 50 years ago. Now it is. And in yet another period of 50 years, it may be seen as a natural thing.