Document Actions

You are here: Home / Research / Meet the Scientist / Ad Aertsen

Ad Aertsen

Bernstein Center Freiburg,
Bernstein Focus Neurotechnology Freiburg-Tübingen

As a young physics student at Utrecht, Ad Aertsen was not particularly thrilled by the “classical” areas of his field. So he started roaming adjacent fields such as philosophy of science and biophysics. Probably thanks to the fact that much of the neurophysiological research in Holland was conducted in physics departments, he finally found his favorite research subject in the brain. This topic offered plenty of open and exciting questions to which he could apply the tools of physics and math.

ad aertsen

Ad Aertsen (© Gunnar Grah/BCF)

During his doctoral studies with Peter Johannesma in Nijmegen, Aertsen began to study hearing in cats and frogs. The researchers followed an approach that actually nowadays still appears quite up-to-date: they sought to relate the response properties of neurons in the auditory system to the animals’ natural environment—their “acoustic biotope”. Already in those early days they applied complex and natural stimuli, next to conventional technical stimuli like pure tones and noise. This allowed them to study how the brain translates properties of environmental stimuli into neuronal activity, i.e., how stimuli are “encoded” in the brain. On the other hand, it also offered the opportunity to “decode” the neural activity, i.e., to reconstruct the sensory environment from the neuronal activity.

Ad Aertsen found out that the response properties of neurons (their “receptive fields”) were not static but dynamic, i.e., they could change with time and with stimulus conditions. The nerve cells apparently did not just linearly add up the incoming stimuli. Rather, there had to be stimulus-dependent, non-linear effects, most likely originating from interactions between nerve cells in the neural network. This insight made clear that, to understand brain function, one could not limit oneself to studying single nerve cells, but rather had to study many neurons simultaneously.

But that strategy brought new problems. Instead of trains of action potentials (‘spikes’) of individual nerve cells, the researchers now had to deal with a whole mess of multiple, simultaneous neuronal spike sequences. With his colleague Michael Erb, Aertsen invented a “device” to make this problem accessible to the senses: Their “neurophone” (a present for the 60th birthday of Valentino Braitenberg) translated the spikes of different neurons to tones of different pitch, such that one could listen to a whole choir of neurons and, maybe, might identify repeating patterns of brain activity as recurring melodies.

As a subjective experience, that was all quite nice, but how would it be possible to evaluate and quantify these patterns more objectively? To this end, as a postdoc with George Gerstein, Philadelphia, USA, Aertsen started to develop appropriate data analysis methods such as the “joint-PSTH” and “gravitational clustering”. These methods could not only deal with the activity of individual neurons, but also assessed the behavior of whole groups of neurons and neuronal interactions, and by that made a whole new dimension of the nervous system accessible.

Simulations of neuronal networks proved to be an enormously helpful tool for testing and calibrating these methods. By simulating networks with known connectivity and subjecting the simulated activity to the analysis tools, one could efficiently check whether and under what conditions the methods arrived at the correct conclusions regarding network structure and interactions.

Neural network simulations have a further advantage: they can be used as an “electronic playground” to systematically vary the structure of a network and observe the effects on network activity and dynamics. This made it possible to design computer experiments to address fundamental questions in neurophysiology such as: How does the propagation of neuronal signals depend on network structure? What can one learn from this about the function of these networks and the neural codes presumably entailed therein? Can the functions of different brain areas be deduced from differences in their structure? What impact do activity dependent learning mechanisms have on the development of neural networks and their functions?

In spite of all his enthusiasm for realistic detail, though, Aertsen reminds us that researchers must not forget one basic principle. “Models and simulations must be as simple as possible and only as complex as necessary, not the other way around”. Otherwise, one runs the risk of creating monstrosities like Salman Rushdie described them in his book “Haroun and the Seas of Stories” as “M2C2D for P2C2E: machines too complicated to describe for processes too complicated to explain”. That would not get us any further in our quest for understanding brain function.

Research has shown that “slim” simulations incorporating the key biological principles can not only be used for investigating the healthy brain, but may also provide hints regarding the causes of and even possible therapies for neurological diseases. Once you figured out which critical characteristics of a network give rise to pathological changes in brain activity, one can test in the same simulations which possible interventions could guide the activity back into the normal regime—precisely what Arvind Kumar, Ad Aertsen and colleagues are now trying for the basal ganglia in Parkinson’s disease.

ad aertsen - bci

Dekodierung von Greifbewegungen für motorische Neuroprothesen: Die in blau dargestellte Bewegung geht mit einer deutlich anderen Hirnaktivierung einher als die rot dargestellte Bewegung. (© BCF/Uni Freiburg)

Aertsen and his group also achieved major progress in applying their data analysis methods towards medical application. They recognized early on that the questions on encoding and decoding of brain activity they started asking decades ago opened up exciting new medical options. If one is good enough at decoding environmental stimuli, or, in the case of the motor system, intended movements from neuronal activity, one should also be able to develop neural prostheses that can replace lost body functions by reading out brain activity. That is also why Aertsen participates in the Freiburg Brain-Machine Interface Initiative that investigates how motor neuroprostheses may provide paralyzed people with some motor abilities. Although still under development, such systems are already being tested in first clinical trials in various locations worldwide and could, in the future, significantly increase the quality of life of many patients.

Twenty years ago, no one would have predicted that the seemingly academic question for the neural code would in a relatively short time give rise to such concrete perspectives for medical applications—a perfect example of how basic science keeps opening up new vistas that turn out to be of immense practical use for our daily lives.