Fear Recognition in Mice based on Neurochat Implantable BCI

Wenbin Qu

, Fangcai Mai

and Minmin Luo

School of Biomedical Engineering, Southern Medical University, Guangzhou, GuangDong, China

Keywords: Brain-Computer Interface, Neural Signal Acquisition, Fear Response.

Abstract: Establishing the connection between the animal brain and external equipment through the brain-computer

interface and realizing the exchange of information between the brain and the outside world is the basis for

many imaginations of future science and technology. This project improved the production of the brain-

computer interface collection component-Neurochat series, which realized the signal collection from brain

waves, to local field potentials and neuron spikes, and successfully identified the fear response of mice.

1 INTRODUCTION

1.1 Brain-Computer Interface

In 2008, neurobiologists at the University of

Pittsburgh claimed that monkeys could manipulate

mechanical arms to feed themselves by using brain

computer interface (BCI) (Velliste 2008). In April

2021, Neuralink, a BCI company owned by Elon

Musk, showed the world their practical brain-

computer interface technology and automatic

implantation of surgical equipment so that a monkey

can play video games with his mind. This also

provides unlimited possibilities for the future of BCI

(Vourvopoulos 2019, Wu 2020, Shi 2018, Tomislav

2018, Chai 2017). Through interdisciplinary

research such as neuroscience, signal detection and

machine learning, it is popular in the medical and

entertainment industries, especially in the field of

virtual manipulation (Patil 2008).

At present, the methods of obtaining information

by brain computer interface include invasive and non-

invasive. Non-invasive is safe for humans and

animals, but the acquired EEG signals are not

accurate. Invasive type damages the animal brain, but

the potential of a single brain cell can be accurately

recorded. We improved and fabricated a set of

invasive brain computer interface elements neurochat

series, which were implanted into the superior

colliculus (SC) nucleus of mouse brain through

https://orcid.org/0000-0002-8364-1289

https://orcid.org/0000-0002-6907-4580

https://orcid.org/0000-0003-2971-2311

electrodes. In the experimental environment, sound

and visual stimuli were used to verify and stably

trigger the animal's instinctive defense behavior.

According to the collected EEG signals, we can judge

whether the mice have instinctive fear. The

experimental results prove the effectiveness of the

brain-computer interface system made in this project,

and provide a basis for the next theoretical research

and practical application.

1.2 Fear and Instinctive Defensive

Response

In nature, in order to survive, animals have an innate

fear of danger signals from the external environment

and induce them to make innate behaviors. The

generation of this instinctive fear defense depends on

the animal's sensory nervous system basically, such

as using smell to perceive predator's scent

information, using vision to observe the predator's

figure, and using the auditory system to perceive

predator's sound information.

The instinctive defense response of animals has

three main manifestations: Startle, Flight and

Freezing. Startle is a short-term startle response

caused by high-intensity sound, which is mainly

regulated by the cochlear nucleus (CN) located in the

lower brainstem and the specific loop is CN-pontine.

Flight can be directly induced by noise or light

stimulation in the awake state. Freezing mainly

490

Qu, W., Mai, F. and Luo, M.

Fear Recognition in Mice based on Neurochat Implantable BCI.

DOI: 10.5220/0011372700003438

In Proceedings of the 1st International Conference on Health Big Data and Intelligent Healthcare (ICHIH 2022), pages 490-495

ISBN: 978-989-758-596-8

occurs under the condition of auditory fear. It is rare

for mice to induce freezing behavior by simple sound

stimulation. The superior colliculus has many

advanced functions and is one of the important nuclei

in the midbrain. However, for SC research, most of

the predecessors focused on vision-related fields.

Through the Neurochat brain-computer interface

system, we conducted a preliminary study on the role

of SC in the fear and instinctive defense response

loops in terms of vision and hearing.

2 MATERIALS AND METHOD

2.1 Neurochat BCI System

The overall system architecture diagram is shown in

Figure 1.

Figure 1: Overall system design diagram.

This project uses NeuroCollector to complete the

signal collection with a single fine-tuning electrode.

The structure diagram is shown in Figure 2 and the

electrode driving operation logic is shown in Figure

3. We will describe the design and production of the

Neurochat overall system separately.

Figure 2: Single fine-tuning electrode structure diagram.

Figure 3: Single fine-tuning electrode driving operation

logic.

The recording system we chose is a self-built

system based on RHD2000 (Intan, USA).It is a multi-

channel recording system and the Intan Technologies

RHD2000 Interface software is used for signal

acquisition and control. Its software supports

monitoring signals from any channel and transfers the

collected data to the front-end PC in binary format.

Finally, the data file is imported into MATLAB

(MathWorks, USA) for subsequent processing and

analysis.

2.2 Auditory Stimulation Method

The mouse is placed in a behavior box containing two

chambers as shown in Figure 4 and the mouse is free

to explore in the behavior boxes on both sides. Two

cameras are used to capture the behavior information

of the mouse at the same time. When the mice moved

freely to the central area of the chamber, they were

randomly and non-repeatedly given sound

stimulation with an intensity of 30-80db each time,

and the behavioral results were recorded and

analysed.

Figure 4: Stimulating behavior paradigm.

2.3 Visual Stimulation Method

For visual stimuli, we adopt the same behavioral

paradigm. When the mice move freely to the central

area of the chamber, randomly and non-repetitively

give looming stimuli with a contrast of 25%-100%

each time, record and analyze the behavioral results.

2.4 Surgery

We choose mouse with a body weight of about 20 g,

normal hearing, and good condition for operation

preparation. Mouse was prepared for surgery within

3-5 days before the experiment. The animal is

prepared for surgery as follows: anesthetize with the

equipped 1.5% sodium pentobarbital solution or

isoflurane. After anesthesia, the mouse head is

Fear Recognition in Mice based on Neurochat Implantable BCI

491

depilated with a shaving device and shaving cream,

then wiped with alcohol and then smeared with

iodophor for disinfection. Use sterile scissors to cut

the scalp to an appropriate size to expose the bregma

of the mouse. Use a stereotaxic instrument for

leveling to ensure that the front and rear fontanelles

of the mouse are in a horizontal state. Four skull nails

are symmetrically implanted on the surface of the

mouse skull for fixation and then locate the SC.

Figure 5: Mouse implanted with NeuroCollector

The SC location coordinates of the mouse are 2.8-

4.5mm behind the bregma, and the maximum

distance beside the midline is 1.75mm. After

marking, the skull is drilled to open the window, the

dura mater and pia mater are removed and stop

bleeding. After the exposed area is clean and free of

blood clots, the electrode is slowly lowered at a speed

of 10 micrometers per second and inserted into the

nucleus of SC. Use biological silica gel and white

wax to seal the skull window, and then use dental

cement to paste a shallow circle on the skull nail and

the surface of the skull. After drying, the whole

adjustable electrode device NeuroDrive or the single

adjustable electrode device NeuroCollector is pasted

and fixed on the mouse skull with dental cement to

ensure that the device can remain stable and does not

produce relative displacement with the skull due to

the free movement of the mouse. Figure 5 shows a

mouse successfully implanted with NeuroCollector.

3 RESULTS

3.1 Auditory Response Verification

Experiment

Since the SC receives a large number of axon inputs

from the auditory nucleus, we first verified the

mouse's auditory stimulus response. In a sound-

shielded room, we perform sound stimulation on

mice: give different frequencies of pure tones or

noises, and record neurophysiological activities at the

same time

As shown in Figure 6, A is the recorded real-time

data, which reflects the action potentials evoked by

10 auditory stimuli. B is the sum of action potentials

accumulated by all stimuli (PSTH diagram:

histogram of time distribution after stimulation). C is

a scatter plot, the horizontal axis is the time and the

vertical axis is the length of the stimulus sound, and

the corresponding scatter plot of the complete sound

sequence stimulus is drawn.

3.2 Auditory Response Verification

Experiment

We found that high-intensity noise could very stably

induce the instinctive defense behavior flight of

awake mice. When the mouse is out exploring, giving

80db and 70db sound stimulation can stably induce

the mouse's flight behavior. When the sound intensity

is 60db, the probability of flight is 67.5%; when the

sound intensity is 50db and 40db, the proportion of

flight is 37.5%; when the sound intensity is 30db, the

mouse will not show the defensive behavior of flight,

but maintain the state of free movement.

At the same time, as the stimulus intensity

decreases, the maximum speed of the mouse in the

process of generating a flight also decreases in a

stepwise manner. This indicates that as the intensity

of the stimulus increases, the mice show stronger and

stronger defensive behaviors, and auditory

stimulation is more likely to induce the mouse to

produce flight. Figure 7 is an analysis diagram of the

trajectory and speed of the mouse after hearing 80db

sound stimulation. In subsequent experiments, we

used 80db sound intensity for stimulation.

ICHIH 2022 - International Conference on Health Big Data and Intelligent Healthcare

492

Figure 6: An example of data analysis of neuron response

to noise in SC.

We record the changes of SC neuron activity

signals after mouse received auditory stimulation and

count the number of action potentials within 3ms, as

shown in Figure 8. During the recording process, the

intensity of the sound stimulation we gave is 80db and

the duration is 5s. In the sound interval, the mice

developed instinctive fear, which caused defensive

behaviors, and a large number of neurons were fired.

After the sound was over, the mouse completed their

defense. The calcium signal quickly weakened and

returned to the baseline level, that is, when the mouse

heard noise stimulation and produced defensive

behaviors, SC neurons fired in large numbers, which

was related to the instinctive fear emotion.

Figure 7: The speed analysis graph of the mouse after

hearing 80db sound stimulation.

Figure 8: Cell firing and change rate in SC after sound

stimulation to mouse.

3.3 Visual Instinct Fear Experiment

For the visual system, giving different speeds and

contrasts of looming (visual approximation) will also

induce the flight and freezing behavior of mice. When

a mouse is out exploring, no matter what the contrast

of the looming stimulus is given, it will make the

mouse have flight behavior. When the mouse is in a

corner, given the looming stimulus, there is a 60%

probability that the mouse will have a freezing

behavior, 40% of mice will have flight behavior. At

the same time, as the intensity of the stimulus

decreases, the maximum speed during which the

mouse generates a flight also linearly decreases. This

shows that as the stimulation intensity increases, the

mice show stronger and stronger defensive behaviors,

among which the flight behavior tendency is more

pronounced. Figure 9 is an analysis diagram of the

trajectory and speed of the mouse after being

stimulated by looming with a contrast of 100%. In

subsequent experiments, we gave a looming stimulus

with a contrast of 75%.

Fear Recognition in Mice based on Neurochat Implantable BCI

493

Figure 9: Analysis of the trajectory and speed of the mouse

after being stimulated by looming with a contrast of 100%.

Next, we recorded the changes in the SC neuron

activity signals after the mouse is stimulated by visual

looming and the results are shown in Figure 10.

During the recording process, we give looming

stimulation with a contrast of 75%, similar to the

result of auditory stimulation. In the stimulation

interval, the mouse produced flight/freezing defense

behavior, and the signal rose rapidly. After the

stimulation, the mouse completed their defense. In

behavior, the calcium signal quickly weakened and

returned to the baseline level. That means, when the

mouse felt visual stimulation and produced defensive

behavior, SC neurons were fired in large numbers,

which indicates that SC is related to visually evoked

defensive behaviour.

Figure 10: Cell fire and change rate in SC after looming

stimulation to mouse.

In the next step, we will establish a Support Vector

Machine (SVM) model, take neuro-

electrophysiological signals as the input, and take

whether the mouse produces defense response as a

criterion for fear, trains SVM and analyzes the

mouse’s fear emotions. Using f1 score as the

standard, evaluate the analytical effect of the model,

complete the two-classification problem, and realize

whether the mouse has fear or not and predict the

subsequent response.

4 CONCLUSIONS

Through the self-improved brain-computer interface

Neurochat, this project realizes the signal acquisition

requirements of brain computer interface from EEG

to local field potential and then to neuron spike

potential, and successfully analyzes the instinctive

fear of mice. It is proved that our system scheme is

feasible and effective.

The various methods integrated by the project

system have mature theoretical foundations, and there

is a huge market space for applications in the fields of

neurocognitive science, electrophysiology, and brain-

computer interface. The field of brain-computer

interface is known as the highway for communication

between the human brain and the outside world

(

Belwafi 2018)

. It is the key core technology of the

latest human-computer interaction and human-

computer hybrid intelligence, and its application

prospects are unlimited. Using Neurochat series can

provide experimental evidence in a multi-faceted,

multi-layered, and humanized manner, and provide

help for the further application of brain-computer

interfaces (Lee 2010, Gao 2020).

ACKNOWLEDGEMENTS

Thanks to the Neuroinformation Engineering

Laboratory of the School of Biomedical Engineering,

Southern Medical University.

REFERENCES

Belwafi, K., Romain, O., Gannouni, S., Ghaffari, F.,

Djemal, R., & Ouni, B. (2018). An embedded

implementation based on adaptive filter bank for brain-

computer interface systems.J. Journal of Neuroscience

Methods, S016502701830116X.

Chai, R., Naik, G. R., Ling, S. H., & Nguyen, H. T. (2017).

Hybrid brain–computer interface for biomedical cyber-

physical system application using wireless embedded

eeg systems. J. Biomedical Engineering Online, 16(1),

Gao, Z., Y Li, Y Yang, Dong, N., Yang, X., & Grebogi, C.

(2020). A coincidence-filtering-based approach for

cnns in eeg-based recognition. J. IEEE Transactions on

Industrial Informatics, 16(11), 7159-7167.

Lee, P. L. , Sie, J. J. , Liu, Y. J. , Wu, C. H. , & Shyu, K. K. .

(2010). An ssvep-actuated brain computer interface

using phase-tagged flickering sequences: a cursor

system. J. Annals of Biomedical Engineering, 38(7),

2383-2397.

ICHIH 2022 - International Conference on Health Big Data and Intelligent Healthcare

494

Patil, P. G., & Turner, D. A. (2008). The development of

brain-machine interface neuroprosthetic devices. J.

Neurotherapeutics, 5(1), 137-146.

Shi, M. H. , Zhou, C. L. , Xie, J. , Shao-Zi, L. I. , Hong, Q.

Y. , & Jiang, M. , et al. (2018). Electroencephalogram-

based brain-computer interface for the chinese spelling

system: a survey. J. Front Inform Tech El, 19(3), 423-

436.

Tomislav, M., Sarma, A. A., Daniel, B., Simeral, J. D., Jad,

S., & Chethan, P., et al. (2018). Stable long-term bci-

enabled communication in als and locked-in syndrome

using lfp signals. J. Journal of

Neurophysiology, 120(1), 343-360.

Velliste, M., Perel, S., Spalding, M. C., Whitford, A. S., &

Schwartz, A. B. (2008). Cortical control of a prosthetic

arm for self-feeding. J. Nature, 453(7198), 1098-1101.

Vourvopoulos, A., Jorge, C., Abreu, R., Figueiredo, P.,

Fernandes, J. C., & Badia, S. (2019). Efficacy and brain

imaging correlates of an immersive motor imagery bci-

driven vr system for upper limb motor rehabilitation: a

clinical case report. J. Frontiers in Human

Neuroscience, 13.

Wu, Q., Yue, Z., Ge, Y., Ma, D., & Wang, J. (2020). Brain

functional networks study of subacute stroke patients

with upper limb dysfunction after comprehensive

rehabilitation including bci training. J. Frontiers in

Neurology, 10.

Fear Recognition in Mice based on Neurochat Implantable BCI

495