By Hugh Darrow
Excerpt from an article in Neuroprosthetics Journal Quarterly (Fall 2012).
Invasive Brain-Machine Interfaces (BMIs), which are most common in North American research, differ mainly from EEG-based interfaces by the method through which neural activity is recorded. While EEGs can record neural activity from outside the skull, invasive BMIs require the implantation of an electrode within the brain itself. In principle, invasive BMIs rely on the hypothesis that groups of neurons will modulate their activity in very similar ways each time a given action is performed, though individual neuron firing varies from trial to trial. A clinically viable BMI should be able to document these modulations and use these modulatory behaviors to reproduce motor actions in artificial actuators (Lebedev, 2006).
In the 1960s and 70s, the scientist Eberhard E. Fetz and his colleagues trained monkeys to coordinate activity of their cortical neurons through biofeedback readings from implanted electrodes. Later, Edward Schmidt proposed that cortical neural activity could move a prosthetic, such as a robotic arm. Due to limitations in technological development, such as the technology necessary to implant sufficient quantities of electrodes into the brain, it took nearly twenty years for researchers to test his hypothesis.
In the mid-90s, invasive BMI testing progressed from mice moving a device with a single degree of freedom to primate movement reproduction. In monkeys, researchers were able to produce ‘reaching’ actualization, in which the neural action of the primate was able to stimulate a robotic arm to feed itself, combining components such as reaching and grasping into a fluid movement (Lebedev, 2006 ).