Mind and Machine
Index:
Brain-computer interface and neuronal basis
1. Brain-computer interfaces
2. Real-time BCI and neurofeedback
3. Neuronal oscillation + mindspa (guestlecture)
4. Transcranial magnetic simulation (guestlecture)
Artificial intelligence and neural networks
5. IBM Watson
6. Artificial intelligence I
7. Artificial intelligence II
8. Neural networks
9. Live neural network simulation
10. Philosophy of mind
Modelling the brain
11. Neural modelling
12. Modelling the brain
Course learning objectives:
- An overview of the different forms of brain modeling, AI and BCI that exist.
- Explain the meaning of key concepts treated in the course. For example, what is a “mind”,
what is “artificial intelligence” and its different subtypes, what is “machine learning”, what
is “transhumanism”, what is the difference between “brain-computer” and “computer-
brain” interfaces, …
- Give examples of current applications of AI and BCI in services or products.
- Explain the principle of simulating neural systems and give examples of the different
levels of detail that such models may incorporate.
- Understand the principles, and practical implementation of BCI.
- Explain the relationship between brain activity and EEG signals, and how an EEG
measurement is performed.
- Explain the rationale behind neurofeedback therapy.
- Develop, present and defend a business proposal, i.e., an idea for a product or service that
exploits state-of-the-art technological advances within the themes of the course, or
advances that may be anticipated in the coming years.
- Formulate opinion about the ethics, and positive and negative aspects of AI and BCI.
Brain-computerinterface and neuronal basis
Mind and Machine Page 1
,Brain-computerinterface and neuronal basis
Brain-computer interfaces
Learning objectives:
- You can explain what types of BCI currently exist
- Which BCIs are already used by patients, outside the lab
- And areas where BCI is currently developed for future application
- The different steps in the BCI cycle
- The relationship between brain activity and EEG signals (in detail)
- The interpretations of EEG amplitude and topography
- The theory and interpretation of ERPs
- The units and typical scale of EEG events
- How EEG is used to research the human brain in health and disease, and how it could
potentially be used in clinical practice in the future.
Brain-computer interface
Brain-computer interface (BCI): a direct communication pathway between a brain and an
external device (which can be both ways)
- BCI can be used for patients (control and communication)
- BCI can be used with healthy users (gaming, brain training)
- BCI can be either non-invasive or invasive. Invasive technologies are often tested
throroughly on animals
- Bci can be used to allow the person for neurofeedback: self-regulate (or modulate) brain
activity: attention, relaxation (brain bar)
Overall aim for BCI is to assist, augment or repair cognitive or sensory-motor functions
Pioneering research by Nicolelis in BCI are neuroprosthetics: restoring damaged hearing, sight
or movement.
- For example, in 2011 his lab designed robotic prosthetic arms: and implemented this in a
study where monkeys 'move and feel' virtual objects using only their brain.
○ 'mind in motion'
○ For this BCI, not only the motor cortex was used (for moving) but also the sensory
cortex, enabeling for sensation!
- 3 years later, in 2014, this type of technology was implemented in humans: allowing
someone with an amputated arm to feel sensory signals.
Neuralink: private company has a lot more financial resources compared to science-funded
- The goal is to eventually develop brain implants for human use.
A remarkable succes of invasive BCI for human would be cochlear implants:
- When hearing is lost, often the nerve is still in tact (only the hair cells are lost)
○ Which means it is possible to stimulate the auditory nerve or (when nerve is also
defect) even implants in the brainstem directly
- However, this sound is nothing compared to what we are used to, but at least children
who are born without hearing are able to develop language, understand it and be able to
speak.
Another technology already developed and tested on people is implantable visual prosthetics
- However, its a big challenge to couple the readings of the prostethics to the millions of
cells
○ In 20-30 years it is likely a feasible possibility.
- Direct brain stimilation of the visual cortex is also possible.
How does a cochlear implant work?
- Electrode array is made of silicone rubber, while the electrodes are platinum or similar
highly conductive material.
- They are inserted into the cochlea deeper in the skull.
- The cochlea winds around the auditory nerve, which is tonotopically organized as is the
basilar membrane.
- When an electrical current is routed to an intracochlear electrode, an electrical fiels is
generated and auditory nerve fibers are stimulated.
Mind and Machine Page 2
, generated and auditory nerve fibers are stimulated.
- Cochlea is the part that converts sound waves into electrical signals
So, the nervous system is electric. Which allows us to read its activity and couple it external
devices (e.g. prostheses), stimulate neural tissue to induce sensations of touch, sound or light-
or normalize pathophysiological activities.
- Future developments rely on better understanding of the physiology, techniques to
measure and stimulate physiological systems, analysis and interpretation of physiological
signals.
Neuronal basis of EEG
Brain activity is spatio-temporally complex: how can we measure and interpret these patterns?
In the figure underneath, you can see what an EEG looks like:
EEG measures changes in brain activity:
Stages of sleep from top to bottom: awake brain, alert to drowsy/sleepy, to rem sleep where
different cycles are at play.
Why do we use EEG?
- First of all, we can use tools like EEG to broaden our fundemental knowledge about the
brain.
○ It allows us to study things like free will, brain disorders and use this knowledge for
prevention, diagnosis and possible therapies.
- EEG is an inexpensive and safe form of BCI (non-invasive)
- Advantage: temporal resolution: direct measurement of neuron activity
○ It actually measures electricity and does it without delay (instead of
bloodflow/oxygen which is seen in other measurements)
○ E.g., The evidence from EEG suggests that Gestalt binding in healthy controls is
mediated by synchrony of oscillations in beta- and gamma-frequency bands.
- Learning how to treat brain disorders is a very important application of EEG/BCI. 35% of
healthcare costs go to brain disorders, and only 50% are actually treated.
○ With BCI tech we can prevent them in earlier stages and make diagnosis way easier.
Why don't we use EEG in premature babies? They monitor pulse, temperature and heart rate,
but why not the brain?
- Even tough premature babies are very vulnerable to developing brain disorders --> babies
have a too complex of a signal. We still cannot fully translate the signals of babies into
change of therapy
Mind and Machine Page 3
, change of therapy
- Cycles of popularity in technology likes these, especially true for PET and EEG.
Adventages of EEG
- Direct reflection of brain activity
- High temporal resolution (1 ms) = basically no delay (in constrast to hemodynamic
response)
- Greater specificty: one voxel in fMRI generates multiple oscillations with distinct functions
○ An fMRI devides the brain into cubes and measures blood flow within these cubes.
And signal is merged
○ EEG can get multiple signals out of the same section, which means the layers of the
brain are more dissectable than an fMRI can measure.
- Non-invase, great availability, inexpensive, and increasingly portable.
Resolution = ability of a technique to seperate two events in space (spatial resolution) or time
(temporal resolution) or accuracy of localizing events in space or time.
Time-space trade-off: some techniques provide high temporal resolution of brain activity, while
others provide higher spatial resolution.
- For the exam, you have to know what the abbreviations stand for, the difference between
structural and functional methods, wether a method is electrical or metabolic, and the
approximate spatial and temporal resolution.
q: Why do you think the temporal resolution for fMRI has such a broad range?
a: sampling time, difference in cognitive tasks (some require more processing time>)
ECoG: Electrocorticography
LFP: Local field potential
EEG: Electroencephalography (structural)
MEG: Magnetoencephalography (combination of structural and functional)
fMRI: Functional magnetic resonance imaging (functional)
PET: Positron emission tomography (functional)
Difference between structual and functional methods:
Structural imaging focuses on visualization and analysis of anatomical parts of the brain
functional imaging is used to identify brain areas and underlying brain processes that are
associated with performing a particular cognitive or behavioral task.
How exactly does electricoencephalography work? (EEG)
- Neurons are like small batteries: the concentration of ions such as sodium is not the same
on the inside and on the outside of the cell (unequal ionconcentrations)
- When a neutotransmitter causes the opening of a transmembrane ion channel (gate in the
cellmembrane), a current of ions will flow from the outside to the inside of the cell.
Mind and Machine Page 4
