Biomedical Engineering Reference
In-Depth Information
110 CHAPTER 10. APPLICATIONS
10.5 BRAINMACHINE INTERFACE
The concept of a Brain Machine Interface (BMI) is an extension of the old idea that a human and a
machine can work together to perform a task. When neuroscientists and neuroengineers refer to BMI,
however, they mean the direct connection of a human brain to an external machine in a feedback loop.
Signals from the brain are recorded either from extracellular electrodes implanted directly into the cortex
or from the surface EEG. Although implanting electrodes in the cortex is invasive, it is permanent and
allows for many hundreds of simultaneous recordings instead of the 30-50 recordings of the EEG. The
signals must then be decoded to parse out relevant information. Many of the techniques of decoding
in Ch. 9 are used to sort spikes and generate frequency maps. These data must then be translated into
commands to be sent to the external machine. Perhaps the most difficult aspect of BMI is to design a
system that performs the translation of the neural code tomachine commands.These machine commands,
may be the opening of a valve, rotation of a motor or opening of an electrical switch.
The hope of a BMI is that it will be able to adapt to new situations. Similar to neural networks, the
BMI must first be trained by sending in inputs to which the desired outputs are known.Using these known
input-output pairs, the parameters of the neural-machine translation are adjusted. Next, the model, now
tuned, can be sent an arbitrary input and generate an output that is close to desired. In the ideal situation,
the person can use their own senses to monitor the progress of the external device. Corrections to the
device can then be made by simply thinking (changing neural firing patterns). In a somewhat less ideal
situation, the effect may need to be transduced into a signal that the person can evaluate.
BMI was originally envisioned as a therapy to restoremotor control to disabled patients, particularly
those suffering from ALS, spinal cord injuries, stroke, cerebral palsy, or amputees. The sick patient would
be capable of controlling a device using neural firings that correlate to some thought. It is also possible that
the “external” device would in fact be contained within the body. For example, patients with incontinence
could have a small valve installed on their urethra that would be directly controlled by thoughts from the
brain. The concept of thought controlled devices has many other applications. For example, dangerous
jobs (fighter pilot, underwater bridge building), exploration (space missions) or rescue missions could be
performed by a machine under direct control from a human brain.
There are a number of hurdles to overcome for BMI to become widely applicable. For complex
tasks, such as the delicate finger movements needed to tie a shoe, the amount of information processing
and coordination in time is well beyond the present capabilities of a wearable device. Information can
only be sent at the speed of light, so exploration of space using BMI will always contain fundamental
delays.These delays may also cause some series problems in performing tasks that require fast action such
as flying a fighter jet. Despite these obstacles, there are some encouraging finds. In an amazing display of
the brain's plasticity, early studies using BMI in animals have shown that after prolonged use, the cortex
dedicates some area to the external device. In other words, from the perspective of the brain, the external
device has become an additional part of the body.
10.6 GROWINGNEURALTISSUE
The goals of tissue engineering are to: (1) stimulate the natural growth of new tissue inside the body; and
(2) artificially grow tissue outside the body. Below we highlight two examples of the application of tissue
engineering in neuroscience.
