CN111406706A - A method for constructing a behavioral model of brain-computer interface - Google Patents

Abstract

The invention relates to a method for constructing a brain-computer interface behavioural model. The construction method comprises the following steps: an operational behavior phase and a mind control phase. And training the mouse to obtain reward through specific behaviors in different time periods, and completing the construction of a brain-computer interface behavioural model. The construction method has high success rate and stable model, and provides a new research carrier for brain-computer interface learning.

Description

A method for constructing a behavioral model of brain-computer interface

technical field

The invention relates to the technical field of brain-computer interface, in particular to a method for constructing a behavioral model of a brain-computer interface.

Background technique

Brain-machine interfaces (BMI) establish a new type of information exchange and control channel between the brain and external devices, enabling direct interaction between the brain and the outside world. This technology does not rely on the conventional spinal cord/peripheral neuromuscular system. . In 1969, Fetz, E.E. detected the activity of cortical neurons and constructed volitional control by increasing the firing of cortical neurons through operant conditioning. In 1999, by detecting and decoding rat cortical M1 neurons and using the electrical signals of M1 cortical neurons to control an external robotic arm to obtain rewards, a real direct interaction between the brain and the outside world was established. In the past ten years, the research on brain-computer interface using rats, mice, non-human primates and paralyzed patients has flourished. Using real-time control of external equipment, successfully controlled nerves in mice, apes, and paralyzed patients. Prosthetic limbs, and clinical pre-experimental studies of brain-computer interfaces are underway. Although current research has made great strides in signal detection, signal decoding, and feedback for BMI learning, greatly improving BMI learning, the interaction between machine and brain is not only about restoring broken connections, but also involving neuroprosthetics. Learning and Neural Adaptation. Controlling and controlling with neuroprosthetics does not come close to normal movement. Neuroprosthetic learning is to control neuroprosthetic limbs by mind in a goal-oriented way through compulsive learning theory. Therefore, the learning of brain-computer interface is a long process. At present, there is no better way to promote the learning of brain-computer interface except to strengthen training and improve the decoding and recording of neural signals through engineering techniques.

SUMMARY OF THE INVENTION

Based on this, the present invention provides a new construction method of a brain-computer interface behavior model, the construction method has a high success rate, and the constructed model is stable.

The specific technical solutions are:

A method for constructing a behavioral model of a brain-computer interface, comprising the following steps:

Operant behavior stage: train mice to obtain rewards through operant behavior within T1 after the cue appears, and the number of successful operant behaviors to obtain rewards is N times. The calcium signal F ₁ value of the mouse M1 cortical neurons, the calcium signal of the mouse M1 cortical neurons when the cue appeared was recorded as F ₀ , and the change rate of the calcium signal in the N times of operant behavior was calculated ( The average value of F ₁ -F ₀ )/F ₀ , and the average value is set as the threshold value;

Mind control stage: train mice to control the calcium signal of mouse M1 cortical neurons by mind in T2 after the cue appears, record the calcium signal _F2 value of mouse M1 cortical neurons in the T2, wait for calcium If the rate of change of the signal (F ₂ -F ₀ )/F ₀ exceeds the threshold value, the mouse receives a reward, which is recorded as a successful test, and the success rate of the mouse in the daily test is counted, and the completion of M times of success is recorded every day. The total time required for the test;

The construction of the brain-computer interface behavioral model is completed by taking the success rate in the daily trials and the total time required to complete M successful trials as evaluation indicators.

In one embodiment, the T1 is within 0-30s or 0-20s after the prompt appears.

In one embodiment, the T2 is within 0-30s or 0-20s after the prompt appears.

In one embodiment, the mind control stage further includes:

Train mice to control the calcium signal of mouse M1 cortical neurons through thought in T3 before the cue, and record the calcium signal F ₃ value of mouse M1 cortical neurons in the T3, until the F ₃ < F ₁ , and the rate of change of the calcium signal (F ₂ -F ₀ )/F ₀ exceeds the threshold value, the mouse is rewarded, and a successful trial is recorded.

In one embodiment, the T3 is within 15s before the prompt appears.

In one of the embodiments, it is judged whether the behavioral model of the brain-computer interface is successfully constructed by comparing the success rate in the daily experiments and the total time required to complete M successful experiments.

In one embodiment, the training period of the operant behavior phase is 5-8 days.

In one embodiment, the training period of the mind control phase is 8-15 days.

In one embodiment, the N times are 40-60 times.

In one embodiment, the M times are 40-60 times.

In one embodiment, the operative action is a lever action.

In one of the embodiments, the prompt is a sound prompt.

Compared with the prior art, the present invention has the following beneficial effects:

The invention provides a method for constructing a new brain-computer interface behavior model, and the successful construction of the model provides a new research carrier for brain-computer interface learning. Through tests, the construction method of the present invention has a high success rate, especially in the stage of mind control, the success rate is kept relatively high, and the model is stable.

Description of drawings

Fig. 1 is the identification diagram of GCaMP6f expressing cells;

Figure 2 is a schematic diagram of a method for building a behavioral model of a brain-computer interface;

Fig. 3 is the test result schematic diagram of embodiment 1;

Fig. 4 is the test result schematic diagram of embodiment 2, embodiment 3 and embodiment 4;

FIG. 5 is a schematic diagram of the test method and test results of Example 5. FIG.

Detailed ways

The new application of the adenosine A _2A receptor antagonist drugs of the present invention will be described in further detail below with reference to specific examples. The present invention may be embodied in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that a thorough and complete understanding of the present disclosure is provided.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terms used herein in the description of the present invention are for the purpose of describing specific embodiments only, and are not intended to limit the present invention. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.

Example 1 Constructing a behavioral model of brain-computer interface using the calcium signal of mouse M1 cortical neuron population cells

Brain-computer interface can help paralyzed patients to restore motor function, and the construction of behavioral model of brain-computer interface (neural prosthesis) is the core of adapting to brain-computer interface. In this example, the calcium signal of the mouse M1 cortical group neurons is used as a "controller" to construct a behavioral model of brain-computer interface in freely moving mice (Fig. 3A-D), and the specific method is as follows:

First, AAV9-syn-GCaMP6f-WPRE-SV40 (AAV virus) was used to express the genetically encoded Ca2 ⁺ indicator GCaMP6f in mouse M1 cortical neurons, and optical fibers were inserted at the site of virus injection. It was shown that GCaMP6f is selectively expressed in L4-6 (L4/L5/L6) neurons of the cortex and co-expressed with the neuronal marker NeuN, but not with the microglial marker (GFPA+) and microglia A cell marker (IBA1) was co-expressed (Figure 1A-L). The changes of calcium signals in the L5 layer of M1 cortical neurons were recorded in real time using an optical fiber recording system (Fig. 2A), and the calcium signal was used as a "controller" to build a behavioral model of the brain-computer interface (Fig. 2B).

Secondly, since it is very difficult for mice to learn the non-existent work structure, a two-step transformation learning method (model construction of operant behavior stage and model construction of idea control stage) is applied to help mice learn brain-computer interface learning (Fig. 2C-D).

(1) Operant behavioral stage: Mice were trained to obtain reward through operant behavior (bar pressing) for a specific time, accompanied by the appearance of specific calcium signal changes in M1 cortical neurons.

The specific method is:

Mice with M1 cortical neurons expressing GCaMP6f were given unconditional sugar water at fixed intervals (half an hour). Then, they were forced to study for three days, and sugar water was obtained by pressing the bar 50 times a day (if not completed 50 times, but the total training time reached 1 hour, is also considered to complete the daily test). Then, enter the 6-day operant behavior training, adding voice prompts, within 30s after the voice prompts appear, the mice can get sugar water by pressing the bar (Fig. 2C), and the success rate of the mice in the daily test in this stage is calculated. and the total time required to obtain 50 times of sugar water per day (if 50 times are not completed, but the total training time reaches 1 hour, it is also regarded as the completion of the daily test, and the total time is recorded as 1 hour), it can be seen that the mice press the bar to obtain the sugar water The success rate increased (Fig. 3A, n=8: one way ANOVAP < 0.0001), and the daily training time (total time to obtain 50 sugar water sessions) decreased (Fig. 3B, n=8; one way ANOV P < 0.0001). The study found that the success rate plateaued on the second day of training, and remained above 95 percent for the next four days.

The calcium signal of mouse M1 cortical neurons at the time of the sound cue was recorded as the baseline of calcium signal (F ₀ ); ΔF was defined as the change value of calcium signal F ₁ on the basis of F ₀ (F ₁ -F ₀ ). By analyzing the changes of ΔF/F[(F ₁ -F ₀ )/F ₀ ] 2 s before and 5 s after pressing the bar, it was found that the calcium signal started to increase before the bar was pressed and began to decrease after the bar was pressed (Fig. 2C). Therefore, It is concluded that the change of calcium signal is closely related to the rod pressing behavior, indicating that the change of neuronal activity is triggered by the rod pressing behavior. To further demonstrate the correlation between calcium signal changes and rod pressing behavior, EMG changes in mouse forelimbs were also recorded (Fig. 2C). It was confirmed that the rise or fall of the M1 calcium signal in the cortex was exactly the same as the rise or fall of the mouse forelimb EMG (Fig. 2C; P<0.05), which further proved that the change of the calcium signal of the M1 neuron was related to the pressure bar of the mouse forelimb. Behavior is related.

The calcium signal F1 value of the mouse M1 cortical neurons was recorded during the daily operant behavior stage when the rod was pressed successfully for 50 times. The mean value of the rate of change of calcium signal (F ₁ -F ₀ )/F ₀ was calculated, and the mean value of the rate of change of calcium signal (F ₁ -F ₀ )/F ₀ in the 6-day operant behavioral stage was set is the domain value;

(2) Thought control stage: train mice to obtain rewards through the appearance of mind-controlled M1 cortical neurons and calcium signals similar to operant behavior at a specific time.

The specific method is: after the 6-day training of operant behavior, on the 7th day, the rod is withdrawn, and the mice are switched to the 10-day training of the mind-control phase of the brain-computer interface. Within 30 s after the sound cue appeared, the mice controlled the calcium signal of M1 cortical neurons through thought, and recorded the calcium signal of the mouse M1 cortical neurons within 0-30 s after the sound cue appeared as F ₂ . Mice need to increase ΔF ₂ [(F ₂ -F ₀ )/F] to the set threshold within 30s after the sound prompt appears, in order to obtain sugar water, which is recorded as a successful experiment, and the mice are counted. The success rate in daily trials, and record the total time required to complete 50 successful trials per day (if 50 trials are not completed, but the total training time reaches 1 hour, it is also regarded as the completion of daily trials, and the total time is recorded as 1 hour ).

Baseline F ₀ is important for each successful mouse reward by increasing the calcium signal above a set threshold and completing each experiment, because each sound cue appears, indicating the start of the experiment, and we define the baseline (F ₀ ) as : The value of the calcium signal of mouse M1 cortical neurons at the 10ms of the sound cue, so F ₀ is uncertain in each experiment.

Mice can successfully complete the experimental task only when the calcium signal is reduced before the start of the experiment (before the sound prompt appears), and the calcium signal is increased after the sound prompt. Calcium signals in M1 cortical neurons in mice were denoted as F ₃ , and mice were trained to reduce calcium signal F ₃ before the sound cue.

In the initial stage of mindfulness training, there is a large variation in calcium signals, and with the increase of training, the changes of calcium signals gradually converge. It was shown that the calcium signal started to decrease 10 s before the sound, and then started to increase after the sound cue (Fig. 3C-D; n=7). Ultimately, the results showed that the mice not only conditioned mindfully to increase calcium signals above a set threshold after the sound cue, but also mindfully decreased calcium signals before the sound cue.

After training, there are two indicators to judge the success of the BCI behavior model: 1. Whether there is a significant difference in the success rate after 10 days of training; 2. Whether the completion time is significantly reduced after 10 days of training.

The results showed that the success rate increased with the number of days of training (Fig. 3E; One-way ANOVA, P<0.05). During the 10-day mindfulness training period, the duration of each session was also reduced (Fig. 3F; One-way ANOVA, P<0.05). On the first day of training, the mice did not have a significant increase in calcium signal after the sound cue (Fig. 3C), but on the tenth day, there was a clear trend (Fig. 3D), and, on the tenth day (Fig. 3G), the completion time of each training was significantly lower than on the first day (Fig. 3H). It shows that the mice gradually learn to regulate the changes of calcium signals to complete the construction of the behavioral model of the brain-computer interface, and the model is relatively stable.

Finally, in order to confirm that the above-mentioned brain-computer interface behavioral model is controlled by the mouse mind, the following verifications were performed:

(1) During the thought control stage, the EMG of the mouse left forelimb and the changes of the mouse M1 calcium signal were recorded simultaneously. Separation of EMG and calcium signal production was observed (Fig. 2D).

(2) The calcium signal in the vicinity of F ₀ of the mice that successfully obtained the sugar water was lower than that of the mice that failed, indicating that the increase in the success rate of the mice's mind control was the conscious reduction of the calcium signal at the F ₀ of the mice.

(3) After the sound prompt appeared, the ΔF/F of the mice that successfully obtained the sugar water increased, and the ΔF/F of the mice that failed did not increase.

(4) The video results showed that the free movement and head-up of the mice did not reach the set threshold. In the video, we can see that there is no obvious change in the calcium signal during the free movement and head-up of the mouse, indicating that the change of the calcium signal has nothing to do with these movements, and only when the mouse wants to drink water. An increase in calcium signal occurs, indicating that calcium signal is generated by thought.

Through the above experimental verification, it is proved that the above-mentioned brain-computer interface behavior model is controlled by the mind of the mouse.

Example 2

In this example, in the operant behavior stage and the idea control stage of the behavioral model of the brain-computer interface, the mice participating in the training were injected with the adenosine A _2A receptor antagonist KW6002 (5mg/kg) intraperitoneally, half an hour before the training started. Start the injection once a day.

The results showed that, in the operant behavior stage, the success rate of mice obtaining sugar water was not significantly different from the five-day training success rate except for the obvious increase in the first day (Fig. 4A; two-way ANOVA, P<0.050, ), the daily training time was not significantly different from the four-day training time except that it decreased significantly in the first two days (Fig. 4B). In the mind control stage, the success rate of the mice in obtaining sugar water all increased (Fig. 4C; two-way ANOVA, P<0.051, P=0.05). However, the time to complete the training did not change significantly each time (Fig. 4D).

Example 3

In this example, only in the mind control stage of building the behavioral model of brain-computer interface, the mice participating in the training were injected with the adenosine A _2A receptor antagonist KW6002 (5mg/kg) intraperitoneally, and the injection was started half an hour before the start of training. once a day.

The results show that: in the stage of mind control, the success rate of mice obtaining sugar water increases. Compared with the results of Example 2, it is found that whether KW6002 is injected in the stage of operant behavior does not affect the success rate of training in the stage of mind control. The training success rate of KW6002 was only related to whether KW6002 was injected during the mind control phase (Fig. 4E, two-way-ANOVA, P<0.05), meanwhile, the injection of KW6002 had no effect on the completion time of each training (Fig. 4F).

Example 4

In this example, two groups of mice were trained to build a behavioral model of a brain-computer interface.

The training method of the first group is basically the same as that of Example 1, the difference is that in the operant behavior stage, the training condition is changed to: after adding the sound prompt, the mice can only obtain sugar water by pressing the bar within 20s after the sound prompt appears. In the mind control stage, the training condition was changed to: within 20s after the sound cue appeared, the mice needed to raise _ΔF2 to the set threshold within 20s after the sound cue appeared, in order to obtain the sugar water. The mice in this group were not injected with the adenosine A _2A receptor antagonist KW6002 during the operant behavioral stage and the thought control stage.

The training method of the second group was basically the same as that of the first group, except that the mice in this group were injected with the adenosine A _2A receptor antagonist KW6002 during the operant behavior stage and the idea control stage.

The results showed that the mice in the first group could not successfully complete the construction of the behavioral model of the brain-computer interface. In the stage of mind control, the training success rate of the mice in the first group was significantly different from that of the mice in the second group. (Fig. 4G, two-way-ANOVA, P<0.05), under the same training conditions, mice injected with adenosine A _2A receptor antagonist KW6002 had a higher training success rate. There was also a significant difference in the time to completion of training (Fig. 4H, two-way-ANOVA, P<0.05), with mice injected with the adenosine A _2A receptor antagonist KW6002 taking less time to complete training under the same training conditions. The second group of mice successfully completed the construction of the behavioral model of the brain-computer interface, while the first group failed to successfully complete the construction of the model.

Example 5

In this example, the Cre enzyme-mediated gene knockout of floxed A _2A receptor is used to verify the effect of blocking A _2A receptor on the construction of the behavioral model of the brain-computer interface. The specific method is:

_A2A -flox transgenic mice were taken, and AAV-Cre was injected into the dorsal striatum to knock down the expression of _A2A receptors (Fig. 5A). Behavioral models of brain-computer interfaces.

The results showed that knockout of _A2A receptors in the dorsal striatum could also promote operant behavior (Fig. 5B, two-way-ANOVA, P<0.05) and ideation control stages (Fig. 5D, two- way-ANOVA, P<0.05), also had no significant effect on training completion time (Figure 5C and Figure 5E; two-way-ANOVA, P>0.05).

The above studies have shown that adenosine _{A2A receptor} antagonist drugs can be used as drugs to promote the learning of brain-computer interface and improve the success rate of the construction of the behavioral model of brain-computer interface. A drug that promotes the construction of a model of the mind control stage in brain-computer interface learning, and improves the success rate of completing tasks in the mind control stage.

Further, in addition to the adenosine A _2A receptor antagonist KW6002, the adenosine A _2A receptor antagonist drugs can also be selected from other adenosine A _2A such as the adenosine A _2A receptor antagonist CPI-444. One or more of the receptor antagonists.

The adenosine A _2A receptor antagonist KW6002 can also be combined with pharmaceutically acceptable excipients to prepare medicines in various dosage forms, which can be used in the learning of brain-computer interface.

Further, through research, it was found that the striatum-pallidus pathway (introduction pathway) is involved in the learning of the brain-computer interface, and can affect the learning of the brain-computer interface by regulating the _A2A receptors that are highly expressed and the indirect pathway.

The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity, all possible combinations of the technical features in the above-described embodiments are not described. However, as long as there is no contradiction between the combinations of these technical features, All should be regarded as the scope described in this specification.

The above-mentioned embodiments only represent several embodiments of the present invention, and the descriptions thereof are specific and detailed, but should not be construed as a limitation on the scope of the patent of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can also be made, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention should be subject to the appended claims.

Claims (10)

1. The method for constructing the brain-computer interface behavioral model is characterized by comprising the following steps of:

an operational behavior phase: training a mouse to obtain reward through an operative behavior in T1 after the prompt appears, wherein the number of the operative behaviors for successfully obtaining the reward is N, and simultaneously, respectively recording a calcium signal F of a mouse M1 cortical neuron when the N operative behaviors occur₁Values, calcium signal of mouse M1 cortical neurons at the time of the cue was recorded as F₀Calculating the rate of change (F) of said calcium signal₁-F₀)/F₀Setting the average value to a threshold value;

a idea control stage: training the mice to control the calcium signal of the mouse M1 cortical neuron in the T2 after the prompt appears through the idea, and recording the calcium signal F of the mouse M1 cortical neuron in the T2₂Value, rate of change of calcium signal (F)₂-F₀)/F₀When the threshold value is exceeded, the mouse obtains reward, the reward is recorded as a successful test, the success rate of the mouse in the test every day is counted, and the total time for completing M successful tests every day is recorded;

and finishing the construction of the brain-computer interface behavioural model by taking the success rate in the test every day and the total time required for finishing M successful tests as evaluation indexes.

2. The method of claim 1, wherein the interval T1 is 0-30s or 0-20s after the prompt appears.

3. The method of claim 1, wherein the interval T2 is 0-30s or 0-20s after the prompt appears.

4. The build method of claim 1, wherein the idea control phase further comprises:

training mice to control calcium signals of mouse M1 cortical neurons in the T3 before the suggestion appears through idea, and recording calcium signals F of mouse M1 cortical neurons in the T3₃Value of F to be₃＜F₁And the rate of change of the calcium signal (F)₂-F₀)/F₀Beyond the threshold, the mouse received a reward, which was recorded as a successful trial.

5. The building method according to claim 4, wherein the time T3 is within 15s before the prompt appears.

6. The construction method according to any one of claims 1 to 3, wherein the training time of the operational behavior phase is 5 to 8 days.

7. The construction method according to any one of claims 1 to 3, wherein the training time of the will control phase is 8 to 15 days.

8. The method of constructing according to any one of claims 1 to 3, wherein the N times are 40 to 60 times; and/or

The number of M times is 40-60 times.

9. A construction method according to any one of claims 1-3, wherein said operational behaviour is a strut behaviour.

10. A construction method according to any one of claims 1 to 3, wherein the prompt is an audible prompt.

Images