UVM Neuroscience Ph.D. Student, Roman Popov, writes:
We, seasoned neuroscience pundits, aspiring students, and science enthusiasts, are all enthralled with the most sophisticated and intriguing organ of our bodies: the brain. Understanding its inner workings holds the potential for a multitude of advances in theoretical and applied areas of science. What is consciousness and how does the brain produce it? How do brains age and how do they acquire pathology, and how are these pathologies alleviated? What is the language of neurons and can we make devices that will communicate with the neural tissue seamlessly? These and many other questions await answers, but the brain’s sophistication has a dark side as well: the brain is exceptionally hard to study. In order to understand the brain on a functional level, researchers employ several imaging methods, but most of them have substantial limitations that restrict the interpretation and understanding of the data. Ideally, neuroscientists need high-precision and quick-responding methods to understand what neurons are doing on the surface of the brain and in deep structures at the same time. The closest to this description is the implantation of multiple-electrode arrays directly into the brain tissue of a living human. While such procedures yield far more accurate and informationally dense data, it predictably comes with a number of potential risks. There are also serious ethical issues to consider. Finally, no one in their right mind would even contemplate asking human study participants to donate their brain tissue for further investigation.
Is this obstacle surmountable? Yes! Firstly, there are cases where elderly people with neurological conditions choose to donate their bodies for scientific research. Secondly, electrocorticography (ECoG, more here: https://en.wikipedia.org/wiki/Electrocorticography) – placing a mat with recording electrodes directly on the brain – is performed on patients that have to undergo neurosurgery, typically during treatment of epilepsy. ECoG, unlike its less invasive relative electroencephalography (EEG), provides much better special resolution and higher signal-to-noise ratio, although it still fails to pick up activity in the deep structures of the brain. Lastly, the use of animals in contemporary neuroscience allows pinpointing the hard questions thanks to more invasive methods (e.g., multiple-electrode arrays), as well studying brain tissues after the animal is sacrificed.
However, there is insufficient availability of donated tissues, as well as lack of patients that are willing to leave a foreign object in their head (which is understandable), and, finally, there are areas of science where use of animal models is highly constrained. Consider the following example: gait initiation and locomotion are both greatly affected in a number of neurological disorders, as in Parkinson’s disease, cerebral palsy, Huntington’s disease, etc. (“Cerebral Palsy: Hope Through Research,” 2015; Haddad & Cummings, 1997; Jankovic, 2008). Animals used to study impaired locomotion are typically quadrupeds, i.e. walking on four limbs (cats, mice, rats), while humans are bipeds. The use of quadrupedal animal models is partially justified due to apparent similarity of basic spinal neuronal mechanisms (V. Dietz, 2002). However, in humans, an additional neural mechanism is required to maintain upright position. In addition, the human nervous system seems to appropriate a higher measure of control over locomotion to the brain itself rather than the spinal cord, as in cats or rats (V Dietz, Colombo, Jensen, & Baumgartner, 1995; Schubert, Curt, Jensen, & Dietz, 1997). Therefore, an important part of the story about how the brain controls the body in healthy and pathological conditions remains obscured in the animal studies. Thus, humans themselves are invited into the laboratories. However, human studies are also subject to some serious limitations. Just to name a few, participant recruitment is challenging; most people with motor disorders are elderly who have additional neurological conditions that would make interpreting results difficult. They also get tired easily, are reluctant to travel far (or even unable to travel without their caregiver who might have a busy schedule), and, finally, may have medical contraindications that may restrict their ability to follow an experimental protocol. Considering this, can we find an additional experimental model? More specifically, can we use computer models to study how the brain works?
The answer to this question, in the opinion of a growing group of scientists, is “yes”. First, contemporary computers enable the recreation of physical reality with a high degree of precision (Coumans). Think for a second about how realistic contemporary video games are. Thus, thousands of human-like two-legged robots can be created simultaneously and put into simulated environments (Figure 1). Letting computers search for the best solution to make these robots, for example, walk and analyzing these solutions may provide interesting insights into the process of locomotion. Then, established knowledge about neural control of movement in humans and new findings from computer simulations can be compared for similarities and dissimilarities. If robots independently (i.e. without researcher explicitly giving movement instructions to robots) find a way to produce gait similar to human, then it may be inferred that body shape and environment dictate the type functional control necessary to exhibit human-like locomotion. Conversely, if robots achieve a very different control strategy, it may illustrate that some tasks (i.e. locomotion) have several viable solutions. In addition, such an unexpected result (computer model consistently finding a solution that is contradicting current theoretical paradigm) can provide a base for novel hypotheses, especially in the areas of science, where theoretical framework is yet to be completed. Finally, computer simulations can be performed relatively fast and inexpensively; they are malleable to changes, and can be scaled up enabling experiments with very large sample sizes.
To illustrate scientific merit of computer models, let us consider an example from the book “Vehicles: Experiments in synthetic psychology” by Valentino Braitenberg (Braitenberg, 1986), Italian neuroscientist and cyberneticist. He proposed a series of thought experiments using simple creatures he called “vehicles”. These creatures consisted of a body with two wheels in the back and two sensors in the front; the only difference between species of “vehicles” was the internal wiring from sensors to wheels’ motors (Figure 2).
Here, sensors react to light and send more excitation to motors if the light source is closer. This very simple model of the neural system (just two neurons connecting sensors to effectors) produces behaviors similar to those frequently seen in the animal world: fearful avoidance and aggressive pursuit. Vehicle “a” will always avoid, or “fear”, the source of light as the sensor from the side closer to the source will put more excitation to the motor on the same side. Thus, the wheel closer to the light source will spin faster making the vehicle turn away from the source. Conversely, when sensors are wired to wheels on the opposite sides (vehicle “b”), behavior observed strongly resembles aggressive pursuit of the target. The sensor closer to the source of light will force the wheel on the opposite side spin faster until the vehicle orients itself toward the target. Subsequently, vehicle “b” will speed up on the collision course with the source of light.
Figure 3 displays behaviors of analogous vehicles that have inhibitory connections instead of excitatory. Vehicle “a” will slow down as it approaches the source of light, and any deviation will be corrected by slowing down the wheel on the side, which turns the vehicle away from the source of light. Vehicle “a” will follow the light as a devoted lover speeding up if the source of light moves away and slowing down again in its vicinity, as if fearing the full contact. Vehicle “b” on the other hand may also slowly approach the source of light, but only if facing it directly. Any small perturbation will inhibit one of the oppositely located wheels making the vehicle hastily turn away from the source of light. In the same paradigm, vehicle “b” is a superficial lover that easily abandons its object of adoration to explore better options.
Of course, none of the vehicles really “feel” anything; they have just two “neurons,” so to speak. However, they illustrate that somewhat complex behaviors maybe achieved with relatively simple neuronal mechanisms. More importantly, a neuroscientist exposed to such models will have an advantage over naïve one when studying real, live animals exhibiting behaviors of avoidance and pursuit. How so? As we have just learned from Figures 2 and 3, avoidance, for example, can be achieved by connecting excitatory sensors to motors on the same side (2a) or by connecting inhibitory sensors to motors on the opposite side of the body (3b). The neuroscientists who modeled such behavior will know what connections and tracts to look for while dissecting the animal’s nervous system. Finally, if to her surprise she finds none of these theoretical neural mechanisms, it will still gain her knowledge. Indeed, she will learn that computer models are missing some important detail, and discrepancy between model and live animal might hint where to look for missing parts of the puzzle.
In conclusion, there is a growing need for more insightful tools to study how the brain works. While these tools are in development and current experiments on animals and humans may provide limited results, computer models of neural networks can provide relatively cheap and fast means of generating hypotheses about potential neural architectures and hidden connections between the brain and the body.
Braitenberg, V. (1986). Vehicles: Experiments in synthetic psychology: MIT press.
Cerebral Palsy: Hope Through Research. (2015). http://www.ninds.nih.gov/disorders/cerebral_palsy/detail_cerebral_palsy.htm Retrieved from http://www.ninds.nih.gov/disorders/cerebral_palsy/detail_cerebral_palsy.htm
Coumans, E. Bullet Collision Detection & Physics Library. Retrieved from http://bulletphysics.org/
Dietz, V. (2002). Proprioception and locomotor disorders. Nat Rev Neurosci, 3(10), 781-790. doi:10.1038/nrn939
Dietz, V., Colombo, G., Jensen, L., & Baumgartner, L. (1995). Locomotor capacity of spinal cord in paraplegic patients. Annals of neurology, 37(5), 574-582.
Haddad, M. S., & Cummings, J. L. (1997). Huntington’s disease. Psychiatr Clin North Am, 20(4), 791-807.
Jankovic, J. (2008). Parkinson’s disease: clinical features and diagnosis. Journal of Neurology, Neurosurgery & Psychiatry, 79(4), 368-376.
Schubert, M., Curt, A., Jensen, L., & Dietz, V. (1997). Corticospinal input in human gait: modulation of magnetically evoked motor responses. Experimental brain research, 115(2), 234-246.