Microfluidic biological depth reserve tank preparation method and blood cell concentration prediction method

Abstract

The invention provides a micro-fluidic biological depth reservoir preparation method and a blood cell concentration prediction method.The method comprises the steps that based on a preset micro-fluidic chip mold, a micro-fluidic biological depth reservoir structure is made of a PDMS material, and the micro-fluidic biological depth reservoir structure comprises a plurality of sub-treatment ponds which are directionally connected; the sizes of the sub-treatment ponds are the same, a plurality of directional connection channels are arranged between the directionally connected sub-treatment ponds, and each sub-treatment pond corresponds to one storage pond of the depth storage ponds; extracting primary nerve cells from a pregnant ICR mouse, and implanting the primary nerve cells into the sub-treatment pond of the micro-fluidic biological deep storage pond structure; and culturing the inoculated micro-fluidic biological deep reservoir structure in an incubator, and infecting neurons by jointly applying CheRiff and jRCaMP1b viruses in the culture process, so that the neurons simultaneously express optical genetic protein and a calcium ion indicator, and finally culturing to obtain the micro-fluidic biological deep reservoir.

Description

Technical Field

[0001] This invention relates to the field of biological neural network technology, and in particular to a method for preparing a microfluidic biological deep reservoir and a method for predicting blood cell concentration. Background Technology

[0002] In recent years, with the development of emerging interdisciplinary fields such as biomedical engineering and materials science, microfluidic technology has demonstrated rich research value and broad application prospects in fields such as biosensors, clinical testing, and organ-on-a-chip construction, especially showing advantages in the long-term culture of in vitro biological neural networks. Microfluidic technology refers to the technology of processing or controlling fluids in micro- and nano-scale spaces by using specially designed microchannels. Microfluidic chips constructed using microfluidic technology are mainly made of polydimethylsiloxane (PDMS), whose transparent material and good biocompatibility provide favorable conditions for the culture and survival of nerve cells and optical observation.

[0003] Pooled computation is a type of recurrent neural network (RNN) with low training costs, widely used in temporal information processing. Deep pools, through increasing the number of directed connection layers within the pool, achieve hierarchical information processing capabilities, a richer number of pool states, larger memory capacity, and more complex dynamic characteristics. In vitro cultured biological neural networks utilize microfluidic technology to achieve modular and directed neural cell culture, thereby constructing a biological deep pool capable of signal stimulation and acquisition. Compared to artificial neural networks, biological deep pools offer advantages such as low energy consumption, strong generalization ability, good robustness, and in-memory computation.

[0004] In recent years, numerous studies on in vitro cultured biological neural networks have demonstrated the superior characteristics of biological reservoirs in performing some typical tasks. For example, Tohoku University in Japan used optogenetic stimulation to stimulate multi-module biological reservoirs in vitro to perform speech-digit classification tasks, and the biological reservoirs exhibited good generalization ability. Similarly, Indiana University in the United States used in vitro cultured brain organoids to perform tasks such as speech recognition, achieving high accuracy while also studying the plasticity of biological neural networks.

[0005] All of the above studies have verified the feasibility of applying biological reservoirs to information processing, but there are still some limitations. The layers of the existing biological reservoirs are all ordinary connections, and ordinary connections will lead to bidirectional growth of neurons between two layers. The existing technology does not simulate the complex connection characteristics in the brain through directional connections. Therefore, the data processing accuracy of the final biological reservoir is low. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a method for preparing a microfluidic biological deep reservoir to eliminate or improve one or more defects existing in the prior art.

[0007] One aspect of the present invention provides a method for preparing a microfluidic biological deep reservoir, the method comprising the steps of: Based on a pre-set microfluidic chip mold, a microfluidic biological deep storage pool structure is fabricated using PDMS material. The microfluidic biological deep storage pool structure includes multiple directionally connected sub-processing pools, each of the same size. Multiple directionally connected channels are provided between the directionally connected sub-processing pools. The directionally connected channels are provided with an arc-shaped blind zone structure. Each sub-processing pool corresponds to one storage pool of the deep storage pool. Fetal rat cortical tissue was extracted from pregnant ICR mice, and primary nerve cells were dissociated from the fetal rat cortical tissue. The primary nerve cells were then implanted into a sub-treatment tank of a microfluidic biological deep reservoir structure. The inoculated microfluidic biological deep reservoir structure was cultured in an incubator. During the culture process, CheRiff and jRCaMP1b viruses were used to infect neurons, so that the neurons could simultaneously express optogenetic proteins and calcium ion indicators, and finally the microfluidic biological deep reservoir was obtained.

[0008] Using the above scheme, when a neuron's axon encounters a physical barrier, such as a microfluidic structure, the axon tends to grow along the edge of the structure when the angle between the structure's edge and the axon's existing growth direction is small. However, when the angle is too large (e.g., obtuse), the axon tends to grow linearly along its original growth direction. Based on this principle, this scheme designs a directional connection channel with an arc-shaped blind zone structure. The angle between the arc-shaped blind zone structure and the axon growing in the opposite direction of the directional extension is smaller than the angle at which the axon continues to grow along the normal channel, thus inducing the axon to enter the arc-shaped structure for growth and avoiding connection formation. For axons growing in the directional extension direction, the angle at which they may turn is exactly the opposite of that of the axon growing in the opposite direction. The angle at which they enter these arc-shaped and triangular regions is much larger than the angle of their forward growth channel, thus allowing them to grow smoothly and form a directional connection with the next target compartment. Therefore, directional connections between different sub-processing chambers can be achieved through the structural and angle design of the arc-shaped blind zone and the triangular structure of the exit. This scheme can simulate the complex connectivity characteristics in the brain through the arc-shaped blind zone.

[0009] In some embodiments of the present invention, the microfluidic biological deep storage tank structure includes four directionally connected sub-treatment tanks, and three directionally connected channels are provided between adjacent directionally connected sub-treatment tanks.

[0010] In some embodiments of the present invention, the arc-shaped blind zone structure is provided with a plurality of sub-arc-shaped blind zones extending in an arc shape in the opposite direction to the directional extension direction of the directional connection channel.

[0011] In some embodiments of the present invention, the radius of the sub-arc blind zone is R, 0.015mm≤R≤0.025mm, and the angle range of the sub-arc blind zone is δ, 250°≤δ≤290°.

[0012] In some embodiments of the present invention, the directional connection channel is further provided with a channel opening structure, the channel opening structure being provided with a plurality of symmetrically arranged right-angled triangle structures, the right-angled triangle structures pointing in the same direction as the directional extension direction of the directional connection channel.

[0013] In some embodiments of the present invention, the channel opening structure is provided with three symmetrical right-angled triangle structures spaced apart.

[0014] Another aspect of the present invention provides a method for predicting blood cell concentration, the method comprising the steps of: Blood samples were collected and analyzed to obtain blood cell concentration values. Based on the blood cell concentration values, simulation calculations were performed to obtain blood cell concentration values ​​at multiple time points, and a blood cell concentration sequence was constructed. The blood cell concentration sequence was then encoded into a three-channel sequence. Based on the three-channel sequence, the light stimulation signals of the first, second, and third channels are determined, and the light stimulation signals are applied to the first sub-treatment pool of the multiple sub-treatment pools that are directionally connected in the microfluidic biological deep storage pool. The photostimulation response signal of the neutron treatment cell in the microfluidic biological deep storage pool is acquired using an image acquisition device. The photostimulation response signal is then input into a prediction module. The prediction module is equipped with an output layer, which outputs the predicted blood cell concentration value at the predicted time position.

[0015] Using the above scheme, this scheme applies the aforementioned microfluidic biological deep reservoir, which serves as the deep reservoir for the computer. The raw data is encoded into a light stimulation signal, light stimulation is applied to the microfluidic biological deep reservoir, and the response of the microfluidic biological deep reservoir is collected. The response of the microfluidic biological deep reservoir is processed through the final output layer to determine the predicted blood cell concentration value at the final predicted time position.

[0016] In some embodiments of the present invention, the method further includes pre-training a prediction module, calculating a blood cell concentration prediction result for each blood cell concentration value in the preset training data, calculating a loss function based on the output result and the corresponding label data, and pre-training the prediction module based on the loss function.

[0017] In some embodiments of the present invention, in the step of simulating and calculating blood cell concentration values ​​based on blood cell concentration values ​​to obtain blood cell concentration values ​​at multiple time points and constructing a blood cell concentration sequence, the currently measured blood cell concentration value is used as the blood cell concentration value at the first time point, the blood cell concentration value at the next time point is simulated and calculated based on the blood cell concentration value at the first time point, and the same simulation and calculation method is used to calculate the blood cell concentration values ​​at multiple time points, and finally the blood cell concentration values ​​at all time points are constructed into a blood cell concentration sequence.

[0018] In some embodiments of the present invention, in the step of calculating blood cell concentration values ​​at multiple time points using the same simulation calculation method, the change in blood cell concentration is calculated using the following formula, and the blood cell concentration value at the next time point is determined based on the original blood cell concentration value and the change in blood cell concentration. in, This represents the blood cell concentration value at the current time point. This indicates the change in blood cell concentration at the current point in time. This represents the preset delay parameter, when hour, As a preset constant value, when At that time, the calculated historical values ​​were used.

[0019] In some embodiments of the present invention, in the step of encoding the blood cell concentration sequence into a three-channel sequence, the range of blood cell concentration variation is determined based on the blood cell concentration value at the previous time point and the change threshold in the blood cell concentration sequence. In the blood cell concentration sequence, if the blood cell concentration value at the current time point is greater than the maximum value of the range of blood cell concentration variation, it is encoded as 1 in the first channel; if the blood cell concentration value at the current time point is less than the minimum value of the range of blood cell concentration variation, it is encoded as 1 in the second channel; if the blood cell concentration value at the current time point is within the range of blood cell concentration variation, it is encoded as 0 in both the first and second channels. In the third channel, rate encoding is used. The blood cell concentration value at the current time point is compared with an accumulation threshold. If it is greater than the accumulation threshold, it is encoded as 1 at the current time point; if it is not greater than the accumulation threshold, the blood cell concentration value at the current time point is accumulated with the blood cell concentration value at the next time point, and compared with the accumulation threshold again. If it is greater than the accumulation threshold, it is encoded as 1 at the next time point; if it is still not greater than the accumulation threshold when compared with the accumulation threshold again, the accumulation continues.

[0020] In some embodiments of the present invention, in the step of determining the photostimulation signals of the first, second, and third channels based on the three-channel sequence, and applying photostimulation to the first sub-processing pool of a plurality of directionally connected sub-processing pools in the aforementioned microfluidic biological deep reservoir based on the photostimulation signals, the first, second, and third channels are respectively provided with stimulation positions in the first sub-processing pool of the microfluidic biological deep reservoir, and photostimulation is applied to the first sub-processing pool of the microfluidic biological deep reservoir based on the encoding results of the first, second, and third channels.

[0021] In some embodiments of the present invention, in the step of acquiring photostimulation response signals of sub-processing cells in a microfluidic biological deep storage pool using an image acquisition device and inputting the photostimulation response signals into a prediction module, a preset number of response images are acquired from each sub-processing cell, the response images acquired from each sub-processing cell are superimposed to obtain a superimposed image, the target neuron location of each sub-processing cell is selected based on the superimposed image, and the pixel value of the target neuron location is acquired for each response image of each sub-processing cell to construct an input vector; the input vectors of multiple sub-processing cells are combined to obtain the photostimulation response signal.

[0022] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the text, or may be learned by practice of the invention. The objects and other advantages of the invention will become apparent from the description and the accompanying drawings.

[0023] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0024] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to limit the scope of the invention.

[0025] Figure 1 This is a schematic diagram of one embodiment of the microfluidic biological deep reservoir preparation method of this scheme; Figure 2 This is a schematic diagram of one implementation of the blood cell concentration prediction method of this scheme; Figure 3 This is a schematic diagram of the overall architecture of this solution; Figure 4 This diagram illustrates the microfluidic biological deep storage tanks with 2, 3, and 4 sub-treatment tanks used in this scheme. Figure 5 This is a schematic diagram of the coding process for this scheme; Figure 6 A schematic diagram showing the comparison between predicted and reference values ​​for a single sub-treatment tank's microfluidic biological deep reservoir and microfluidic biological deep reservoirs in different sub-treatment tanks, along with the calculation results of correlation coefficient and root mean square error. Figure 7 A schematic diagram comparing the correlation coefficients of a microfluidic biological deep reservoir for a single sub-treatment tank and microfluidic biological deep reservoirs for different sub-treatment tanks; Figure 8 A schematic diagram comparing the root mean square error of a microfluidic biological deep reservoir for a single sub-treatment tank and microfluidic biological deep reservoirs for different sub-treatment tanks. Figure 9 Heatmap of neuronal activity correlation coefficients in microfluidic biological deep reservoirs with different sub-treatment cells; Figure 10 A schematic diagram comparing the modularity of microfluidic biological deep storage tanks with different sub-treatment tanks; Figure 11 A schematic diagram comparing the average clustering coefficients of microfluidic biological deep storage tanks with different sub-treatment tanks; Figure 12 A schematic diagram comparing the network density of microfluidic biological deep storage tanks with different sub-treatment cells; Figure 13 This is a schematic diagram of the fabrication process for a microfluidic chip mold. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0027] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0028] like Figure 1 As shown, this invention proposes a method for preparing a microfluidic biological deep reservoir, the method comprising the following steps: Step S100: Based on the preset microfluidic chip mold, a microfluidic biological deep storage pool structure is fabricated using PDMS material. The microfluidic biological deep storage pool structure includes multiple sub-processing pools that are directionally connected. Each sub-processing pool is the same size. Multiple directional connection channels are set between the directionally connected sub-processing pools. The directional connection channels are provided with an arc-shaped blind zone structure. Each sub-processing pool corresponds to one storage pool of the deep storage pool. In specific implementation, the microfluidic chip mold can be as follows: Figure 13 As shown, in the steps of fabricating a microfluidic biological deep reservoir structure using PDMS material, colloids A and B in the PDMS raw material were prepared in a 10:1 ratio. After mixing and stirring, the PDMS mixture was vacuum-sealed to remove air bubbles. Once no air bubbles remained, a 1 mL pipette with a volume adjusted to 800 μL was used to extract the PDMS mixture. An appropriate amount of the PDMS mixture was added to the edge of the mold structure, allowing the PDMS to diffuse freely into the gaps of the structure, while preventing the PDMS from submerging the structure. After the PDMS colloid completely covered the edge of the structure, no bright spots reflecting fluid could be seen in the gaps of the structure under a microscope, and the bottom layer structure could be clearly seen through the PDMS colloid. After the PDMS colloid completely stopped flowing and diffusing, it was placed in an oven at 80°C for drying for 1.5 hours. After drying, the PDMS structure was peeled off, sterilized, and directly bonded to a glass slide coated with poly-L-lysine to obtain a microfluidic chip for cell culture, thus constructing a biological deep reservoir structure.

[0029] Specifically, a mold is made using photolithography. Photoresist is evenly coated onto the substrate surface, and then ultraviolet light is irradiated through a mask, causing a chemical change in the exposed areas of the photoresist. After irradiation, the uncured photoresist is removed, leaving the desired pattern.

[0030] like Figure 4 As shown, in specific implementation, the sub-processing pools can be 2, 3, or 4, such as... Figure 4 As shown in (A), each of the two sub-treatment tanks has a compartment size of 1.2mm × 0.6mm, and the two sub-treatment tanks are connected by five directional channels; Figure 4 (B) and Figure 4 As shown in (C), in the microfluidic biological deep reservoir structure with 3 or 4 sub-processing cells, the compartment size of each sub-processing cell is 0.4 mm × 0.6 mm, and every two sub-processing cells are connected by 3 directional channels. In the microfluidic biological deep reservoir structure with different numbers of sub-processing cells, the channel size used to achieve directional connection is the same, with a length of 0.4 mm and a width of 0.01 mm, which allows axons to pass through while preventing the entry of neuronal cell bodies; at the outlet end, there are three symmetrical right-angled triangle structures spaced 0.04 mm apart to induce reverse-growing biological neurons, and the length of the right-angled side opposite the 30° angle of the right triangle is 0.01 mm; 0.09 mm from the entrance end of the directional channel, there are three arc-shaped blind zone structures spaced 0.05 mm apart, with a radius of 0.02 mm and an angle of 270°, which are used to guide and restrict the axonal structures that reverse-growing to this point.

[0031] Step S200: Fetal rat cortical tissue is extracted from pregnant ICR mice, and primary nerve cells are dissociated from the fetal rat cortical tissue. The primary nerve cells are then implanted into a sub-treatment tank of a microfluidic biological deep reservoir structure. In step S300, the inoculated microfluidic biological deep reservoir structure is cultured in an incubator. During the culture process, CheRiff and jRCaMP1b viruses are used to infect neurons so that neurons simultaneously express optogenetic proteins and calcium ion indicators, and finally the microfluidic biological deep reservoir is obtained.

[0032] Specifically, pregnant ICR mice (approximately day 15.5 of embryonic development) were euthanized through anesthesia, cervical dislocation, and decapitation. Subsequently, the skin of the mouse abdomen was disinfected and dissected using 75% alcohol, and the uterus was transferred to pre-cooled HBSS buffer. Next, the embryo's head was separated from the uterus and placed in another pre-cooled HBSS buffer. The cortical tissue was carefully separated in the HBSS buffer using medical forceps. After brain region separation, it was immersed in DMEM buffer containing DNase and papain for 15 minutes for dissociation. Finally, the brain tissue in the buffer was gently dispersed using a 1000 μL pipette tip, and the supernatant was centrifuged at 100g for 5 minutes. The extracted primary neural cell solution was used to add the mesocellular cell suspension to culture medium, resuspend it in Neurobased Plus medium, centrifuge, wash, and resuspend the cells. The suspended cells were counted using a hemocytometer, and the cell suspension was adjusted to the desired concentration using culture medium. After sterilizing poly-L-lysine-coated culture dishes or chips with UV light for 1 hour, cell suspension was dripped into the chips for cell seeding. The seeded chips were then placed in a CO2 incubator to promote cell growth. After 24 hours, the culture medium was completely replaced to remove any unattached dead cells. Once the nerve cells had adhered, the medium was changed halfway every 3 days. Three days after each medium change, neurons were infected with a combination of CheRiff and jRCaMP1b viruses, enabling them to simultaneously express optogenetic proteins and calcium ion indicators. Photoactivation and recording were then performed in the in vitro cultured neural network. Four to five days after viral infection, the neuronal cell status and infection effect were observed under a fluorescence microscope.

[0033] Using the above scheme, when a neuron's axon encounters a physical barrier, such as a microfluidic structure, the axon tends to grow along the edge of the structure when the angle between the structure's edge and the axon's existing growth direction is small. However, when the angle is too large (e.g., an obtuse angle), the axon tends to grow linearly along its original growth direction. Based on this principle, this scheme designs a directional connection channel with an arc-shaped blind zone structure. The angle between the arc-shaped blind zone structure and the axon growing in the opposite direction of the directional extension is smaller than the angle at which the axon continues to grow along the normal channel, thus inducing the axon to enter the arc-shaped structure for growth and avoiding connection formation. For axons growing in the directional extension direction, the angle at which they may turn is exactly the opposite of that of the axon growing in the opposite direction. The angle at which they enter these arc-shaped and triangular regions is much larger than the angle of their forward growth channel, thus allowing them to grow smoothly and form a directional connection with the next target compartment. Therefore, directional connections between different compartments can be achieved through the structural and angle design of the arc-shaped blind zone and the triangular structure at the exit. This scheme can simulate the complex connection characteristics in the brain through the arc-shaped blind zone.

[0034] like Figure 4 As shown in (C), in some embodiments of the present invention, the microfluidic biological deep storage tank structure includes four directionally connected sub-treatment tanks, and three directionally connected channels are provided between adjacent directionally connected sub-treatment tanks.

[0035] like Figure 4 As shown in (D), Figure 4 (D) The directional extension direction is from left to right. In some embodiments of the present invention, the arc-shaped blind zone structure is provided with a plurality of sub-arc-shaped blind zones that extend in an arc shape in the opposite direction to the directional extension direction of the directional connection channel.

[0036] In some embodiments of the present invention, the radius of the sub-arc blind zone is R, 0.015mm≤R≤0.025mm, and the angle range of the sub-arc blind zone is δ, 250°≤δ≤290°.

[0037] In some embodiments of the present invention, the directional connection channel is further provided with a channel opening structure, the channel opening structure being provided with a plurality of symmetrically arranged right-angled triangle structures, the right-angled triangle structures pointing in the same direction as the directional extension direction of the directional connection channel.

[0038] In some embodiments of the present invention, the channel opening structure is provided with three symmetrical right-angled triangle structures spaced apart.

[0039] like Figure 2 and 3 As shown, another aspect of the present invention provides a method for predicting blood cell concentration, the method comprising the steps of: Step S400: Collect and analyze blood samples to obtain blood cell concentration values. Perform simulation calculations based on the blood cell concentration values ​​to obtain blood cell concentration values ​​at multiple time points and construct a blood cell concentration sequence. Encode the blood cell concentration sequence into a three-channel sequence. Step S500: Based on the three-channel sequence, determine the light stimulation signals of the first channel, the second channel, and the third channel, and apply light stimulation to the first sub-treatment pool of the multiple sub-treatment pools directionally connected in the microfluidic biological deep storage pool based on the light stimulation signals. In step S600, an image acquisition device is used to acquire the photostimulation response signal of the neutron processing cell in the microfluidic biological deep storage pool. The photostimulation response signal is input into the prediction module. The prediction module is equipped with an output layer, which outputs the predicted blood cell concentration value at the predicted time position.

[0040] In some embodiments of the present invention, the photostimulation response signal of each sub-treatment cell in the microfluidic biological deep storage tank is collected.

[0041] In practice, the output layer of the prediction module is a fully connected layer.

[0042] In the specific implementation process, the blood cell concentration value prediction result is the blood cell concentration value prediction result after a preset time period, which can be 1 hour, 3 hours or 5 hours, etc.

[0043] Using the above scheme, this scheme applies the aforementioned microfluidic biological deep reservoir, which serves as the deep reservoir for the computer. The raw data is encoded into a light stimulation signal, light stimulation is applied to the microfluidic biological deep reservoir, and the response of the microfluidic biological deep reservoir is collected. The response of the microfluidic biological deep reservoir is processed through the final output layer to determine the predicted blood cell concentration value at the final predicted time position.

[0044] In some embodiments of the present invention, the method further includes pre-training a prediction module, calculating a blood cell concentration prediction result for each blood cell concentration value in the preset training data, calculating a loss function based on the output result and the corresponding label data, and pre-training the prediction module based on the loss function.

[0045] In some embodiments of the present invention, during the pre-training process of the prediction module, the neuronal response signals of the biological deep reservoir acquired after optogenetic stimulation are processed, and the fluorescence signal changes of 60 neuronal nodes within a 1000-frame time window are selected as the network state input. The first 90% of the data in the time series is fed into the output layer of the reservoir calculation, and the output layer weights are trained using the least squares method to establish a mapping relationship between the biological neural network state and the concentration of mature blood cells. After training is completed, the obtained output layer weight parameters are used to predict and analyze the remaining 10% of the time series data. The prediction adopts a single time step prediction method, that is, based on the network state at the current moment, the concentration of mature blood cells corresponding to the next time step is predicted. The predicted concentration time series is compared with the corresponding reference concentration series, and the modeling ability and prediction effect of the microfluidic biological deep reservoir in the mature blood cell concentration prediction task are evaluated by analyzing the deviation between the predicted value and the reference value.

[0046] In some embodiments of the present invention, in the step of simulating and calculating blood cell concentration values ​​based on blood cell concentration values ​​to obtain blood cell concentration values ​​at multiple time points and constructing a blood cell concentration sequence, the currently measured blood cell concentration value is used as the blood cell concentration value at the first time point, the blood cell concentration value at the next time point is simulated and calculated based on the blood cell concentration value at the first time point, and the same simulation and calculation method is used to calculate the blood cell concentration values ​​at multiple time points, and finally the blood cell concentration values ​​at all time points are constructed into a blood cell concentration sequence.

[0047] In practice, blood cells are regulated by multi-level biological feedback mechanisms during their generation and maturation, thus affecting the concentration of mature cells in the blood at any given moment. Their dynamic evolution typically exhibits significant time delays, nonlinear regulation, and irregular oscillations. To simulate the aforementioned maturation dynamics with delayed feedback characteristics under controllable conditions, this scheme employs a typical delayed nonlinear dynamic model to generate an equivalent time series of mature blood cell concentration states, the mathematical expression of which is as follows: in, This represents the blood cell concentration value at the current time point. This indicates the change in blood cell concentration at the current point in time. This represents the preset delay parameter, when hour, As a preset constant value, when At that time, the calculated historical values ​​were used.

[0048] In the specific implementation process This time delay, introduced during maturation regulation, describes the feedback relationship between the current change in mature cell concentration and the previous stage state. With... As the value increases, the dynamic complexity of the system gradually increases. When When the value is 17, the system exhibits obvious non-periodic oscillation characteristics, which can be used to simulate the complex dynamic behavior caused by delayed feedback regulation during blood cell maturation. The aforementioned delayed feedback dynamic model has been widely used in studies such as physiological regulation process modeling and complex system signal analysis. The time series it generates has a high similarity to actual biological regulatory signals in terms of statistical characteristics and dynamic behavior, making it suitable as an equivalent simulation input for blood cell maturation dynamic signals. Figure 5 As shown, in this method, since blood regulation is a time-delay system, it first occurs in [- During the time period [0], the blood concentration is assumed to remain constant, simulating the basic steady state of a mature regulatory system in the initial stage, i.e., when hour, The preset constant value is 0.5; starting from time 0, the measured value is substituted into the equation to obtain the change in the concentration of mature cells in the blood at this time, and then the value of y1 at the next time is obtained. This process is repeated to obtain 1000 discrete time series values, which represent the concentration of mature blood cells over a period of time, i.e., the blood cell concentration sequence.

[0049] like Figure 5 As shown, in some embodiments of the present invention, in the step of encoding the blood cell concentration sequence into a three-channel sequence, the range of blood cell concentration variation is determined based on the blood cell concentration value at the previous time point and the change threshold in the blood cell concentration sequence. In the blood cell concentration sequence, if the blood cell concentration value at the current time point is greater than the maximum value of the range of blood cell concentration variation, it is encoded as 1 in the first channel; if the blood cell concentration value at the current time point is less than the minimum value of the range of blood cell concentration variation, it is encoded as 1 in the second channel; if the blood cell concentration value at the current time point is within the range of blood cell concentration variation, it is encoded as 0 in both the first and second channels. In the third channel, rate encoding is used. The blood cell concentration value at the current time point is compared with the accumulation threshold. If it is greater than the accumulation threshold, it is encoded as 1 at the current time point; if it is not greater than the accumulation threshold, the blood cell concentration value at the current time point is accumulated with the blood cell concentration value at the next time point, and compared with the accumulation threshold again. If it is greater than the accumulation threshold, it is encoded as 1 at the next time point; if it is still not greater than the accumulation threshold when compared with the accumulation threshold again, the accumulation continues.

[0050] like Figure 5As shown, in the specific implementation process, in the step of determining the range of blood cell concentration change based on the blood cell concentration value and change threshold of the previous time point in the blood cell concentration sequence, the change threshold and the initial blood cell concentration value are preset. The initial blood cell concentration value + change threshold = the maximum value of the initial blood cell concentration change range; the initial blood cell concentration value - change threshold = the minimum value of the initial blood cell concentration change range; compared with y(0) of the blood cell concentration sequence, when comparing y(1), y(0) is used as the blood cell concentration value of the previous time point.

[0051] In the specific implementation process, a blood cell concentration sequence of 1000 discrete sequence values ​​is generated according to the equation and encoded into 3 channels. Two of the channels use step-forward encoding. The encoding method is as follows: first, a dynamic baseline and threshold are set to record the "reference level" of the current signal; if the signal exceeds the baseline + threshold, a rising pulse is output (channel 1 outputs 1); if the signal is below the baseline - threshold, a falling pulse is output (channel 2 outputs 1). Whenever a pulse is triggered, the baseline "follows" the signal change. When a rising pulse is output, the baseline = baseline + threshold; when a falling pulse is output, the baseline = baseline - threshold. This prevents frequent repeated triggering. SF encoding is very suitable for capturing the time structure of complex time series and expressing nonlinear transition features. Another channel uses rate coding, simulating the characteristic of biological neurons that they only fire signals when they receive a threshold amount of stimulation. During the encoding process, the input temporal signal is accumulated (integrated) over time. When the accumulated amount exceeds the threshold of 1.78, a pulse is fired (corresponding to a 1 in channel 3), then the output is reset to zero, waiting for the next threshold to be met before firing again. The threshold selection considers the neuron's response rate, avoiding frequent stimulation that could affect the activity of the biological sample. This encoding method allows for faster accumulation and more frequent firing of high-intensity signals, emphasizing the intensity of accumulated signals. Rate coding effectively extracts global intensity patterns, such as average energy and low-frequency trends, making it suitable for supplementing local information beyond SF coding and providing global clues to "signal intensity distribution." The resulting 0 / 1 sequence corresponds to optogenetic stimulation being on / off. This sequence is then fabricated into a pattern the same size as the observation field of view of the biological depth reservoir, where 1 corresponds to a white area and 0 corresponds to a black area. This achieves the goal of converting the current concentration information of blood cells into optogenetic stimulation patterns on three channels.

[0052] In some embodiments of the present invention, in the step of determining the photostimulation signals of the first, second, and third channels based on the three-channel sequence, and applying photostimulation to the first sub-processing pool of a plurality of directionally connected sub-processing pools in the aforementioned microfluidic biological deep reservoir based on the photostimulation signals, the first, second, and third channels are respectively provided with stimulation positions in the first sub-processing pool of the microfluidic biological deep reservoir, and photostimulation is applied to the first sub-processing pool of the microfluidic biological deep reservoir based on the encoding results of the first, second, and third channels.

[0053] like Figure 3 and 5 As shown, in the specific implementation process, the microfluidic biological depth reservoir was removed from the incubator, disinfected, and placed in a laboratory observation device with a constant temperature and gas environment. The light source, stage, camera, and digital micromirror array used for optogenetic stimulation and fluorescence imaging were turned on, and the position of the microfluidic biological depth reservoir was adjusted so that all sub-processing cells were within the microscope's field of view. Spontaneous network activity was recorded for 20 minutes. If neurons emitted signals (changes in fluorescence intensity) during this period, 1000 encoded patterns were input into the optical system. Specifically, if the encoded value was 1, light stimulation was applied at that time point in that channel. After DMD processing, the blue light source that could induce neuronal signal changes was discrete onto a ring composed of micron-level light spots. The on / off state of the illumination corresponded one-to-one with the encoding results of the three channels. Compared with a solid circle of the same area, the large-diameter ring of stimulation can more effectively induce network responses and reduce direct and prolonged irradiation of large areas of neurons, thus minimizing potential phototoxicity and cell damage. The stimulation frequency is 10Hz, one set lasts 100s, and after stimulation, rest for 100s. If the stimulation is repeated more than 5 times, rest for 5 minutes.

[0054] In some embodiments of the present invention, in the step of acquiring photostimulation response signals of sub-processing cells in a microfluidic biological deep storage pool using an image acquisition device and inputting the photostimulation response signals into a prediction module, a preset number of response images are acquired from each sub-processing cell, the response images acquired from each sub-processing cell are superimposed to obtain a superimposed image, the target neuron location of each sub-processing cell is selected based on the superimposed image, and the pixel value of the target neuron location is acquired for each response image of each sub-processing cell to construct an input vector; the input vectors of multiple sub-processing cells are combined to obtain the photostimulation response signal.

[0055] In the specific implementation process, the responses of each neuron in the neural network to optogenetic stimuli differ, reflected in changes in calcium ion concentration within the cell body, which are captured by the camera in the form of fluorescence signals. After placing the sample and starting the relevant equipment, the microscope focus was adjusted to ensure clear boundaries of the biological depth reservoirs within the field of view, allowing for observation of changes in fluorescence intensity of neurons within the network over time. The imaging frequency was 10 Hz. The network response was recorded simultaneously with the start of the stimulus pattern playback. During recording, the overall field of view was ensured to cover all layers of different biological depth reservoirs. Under the premise of ensuring normal neuronal activity, the responses of 60 neurons were collected as outputs. For sub-processing pools 2, 3, and 4, responses of 30, 20, and 15 neurons were randomly selected from each sub-processing pool, respectively, to represent the output of each sub-processing pool for subsequent analysis. The response time for each stimulus was recorded for 100 seconds, followed by 100 seconds of spontaneous network activity.

[0056] In the specific implementation process, in the step of selecting the target neuron location for each sub-processing pool based on the superimposed image, the acquired signal is first identified and extracted using suite2p software. The main process is as follows: input the original imaging data; load and initialize the recognition parameters; perform correction to ensure alignment of all frames; estimate the background and generate background images with average brightness and correlated brightness to obtain the response image, and then superimpose the response images; cell detection (ROI detection) detects candidate neurons by using the inter-pixel correlation and spatial filtering of the superimposed image, while subtracting local background signals. The signal extracted using suite2p needs to be normalized: (F-F0) / F0, abbreviated as ∆F / F, where F0 is the baseline fluorescence.

[0057] The beneficial effects of this plan include: This scheme realizes the modeling and prediction of biological maturation dynamics signals with delayed feedback and nonlinear characteristics. Experimental verification confirms the completeness and feasibility of the system in terms of structure construction, signal encoding, network response acquisition, and predictive analysis. To quantitatively evaluate the performance of the microfluidic biological deep reservoir in the task of predicting mature blood cell concentration, the predicted mature blood cell concentration sequence is compared and analyzed with the corresponding reference mature blood cell concentration time series. First, the correlation coefficient is introduced as an evaluation index to measure the ability of the prediction results to fit the trend of concentration state signal changes. Its calculation method is as follows: in, These represent the standard deviations of the predicted mature blood cell concentration sequence and the reference mature blood cell concentration sequence, respectively. The correlation coefficient r represents the covariance between two sets of data. A larger absolute value (closer to 1) indicates a stronger linear correlation between the variables; conversely, a smaller absolute value indicates a weaker linear correlation. A zero correlation coefficient indicates no linear correlation between the variables. To further analyze the cumulative error between the predicted mature blood cell concentration and the reference mature blood cell concentration, the root mean squared error (RMSE) is introduced for calculation. RMSE is a commonly used indicator to measure the difference between predicted and actual values. The representative represents the predicted concentration of mature blood cells at time t. This represents the corresponding reference mature blood cell concentration. Its core idea is to measure the magnitude of the prediction error and provide a concise numerical value for understanding the prediction accuracy. Unlike the ordinary "mean error," it emphasizes larger errors through the "square" and "square root." This approach makes RMSE particularly sensitive to larger errors; therefore, it reflects the performance of the prediction result when there is a large deviation. Figure 6 As shown, Figure 6 The results of a single sub-processing pool, as well as 2, 3, and 4 sub-processing pools, are compared with the reference state in the mature blood cell concentration prediction task. Correlation coefficients and root mean square errors are calculated based on the aforementioned evaluation indicators. Experimental results show that, compared to sub-processing pools, the modular, multi-layered biological deep reserve pool can more effectively characterize the complex features in blood cell maturation dynamics. Its prediction consistency and accuracy are significantly improved, intuitively demonstrating the technological improvements brought about by the evolution from non-deep structures to deep structures and from single-layer to multi-layered structures.

[0058] like Figure 7 As shown, by substituting the predicted and reference values ​​into the formula, the correlation performance of biological reservoirs at different depths in prediction was calculated after 10 repeated sample experiments. The highest correlation coefficient for a single-layer non-deep biological reservoir was 0.713, the lowest was 0.493, and the average was 0.615. For two sub-treatment reservoirs, the highest correlation coefficient was 0.942, the lowest was 0.632, and the average was 0.787. For three sub-treatment reservoirs, the highest correlation coefficient was 0.937, the lowest was 0.716, and the average was 0.871. For four sub-treatment reservoirs, the highest correlation coefficient was 0.955, the lowest was 0.799, and the average was 0.893. Figure 8As shown, by substituting the predicted and reference values ​​into the formula, the root mean square error (RMSE) of the biological reservoirs at different depths was calculated after 10 repeated sample experiments. The RMSE values ​​for a single sub-treatment pool were highest (0.453), lowest (0.161), and average (0.238); for two sub-treatment pools, the RMSE values ​​were highest (0.346), lowest (0.101), and average (0.196); for three sub-treatment pools, the RMSE values ​​were highest (0.207), lowest (0.085), and average (0.12); and for a four-layer biological reservoir, the RMSE values ​​were highest (0.246), lowest (0.09), and average (0.136).

[0059] Furthermore, to reflect the modularity of the deep reservoir, this scheme uses a correlation coefficient heatmap for visualization. Specifically, it calculates the correlation coefficient of activity between every two neurons in the network, encoding the strength of synchronicity among a group of neurons into a visually appealing image using color. If distinct "color blocks" appear on the heatmap, it indicates the existence of a group of neurons whose activity is highly synchronized. This typically means that these neurons belong to the same functional circuit, jointly processing specific information or performing a specific task. Figure 9 A heatmap showing the correlation coefficients of biological storage ponds with different depths was displayed.

[0060] To better evaluate the network state, the following parameters were further analyzed to more accurately and comprehensively reflect the characteristics of the neural network.

[0061] Modularity measures the rationality and strength of dividing a network into communities (modules): in These are elements in the adjacency matrix A of the network, where k is the node degree and k is the degree of the i-th node. Let i be the sum of the correlation coefficients (adjacency matrix values) between i and other nodes. m is the total number of sides. . This is an indicator function for community partitioning. If i and j belong to the same community, the value is 1; otherwise, it is 0. Community partitioning is achieved using the Louvain algorithm, which continuously adjusts the community of each neuron node to maximize the modularity of the entire network. A higher modularity Q-value indicates denser connections within a module and sparser connections between modules. The modular structure reflects the functional partitioning of the neural network; for example, certain neuron groups may undertake specific tasks or synchronized activities. In in vitro calcium imaging data, high modularity indicates that neurons form relatively independent functional communities, potentially corresponding to different functional circuits or processing units.

[0062] The average clustering coefficient defines the tightness of connections between a node's neighbors. The average clustering coefficient of a network is the average of the clustering coefficients of all nodes. in This represents the actual number of triangles existing between the neighbors of node i. It is calculated by taking the set of all neighbors of node i and calculating their connection count, as shown in the following formula: ,also Here, is the node degree, and N is the number of nodes. A higher clustering coefficient indicates that neurons tend to form locally dense clusters of connections, which may be related to local processing and synchronized firing. In calcium imaging, high clustering reflects frequent collaborative activity of local neuronal groups, suggesting the existence of possible microcircuits. The analytical method calculates the clustering coefficient of each node in the undirected network and then averages it to reflect the local connectivity characteristics of the network.

[0063] Network density is the ratio of the actual number of edges to the theoretical maximum number of edges. Where E is the number of edges in the network, E The adjacency matrix is ​​calculated as half the sum of all values ​​in the adjacency matrix to remove duplicates, where N is the number of nodes. Network density reflects the overall tightness of connections between neurons. In calcium imaging data, high density indicates a greater number of functionally related pairs of neurons, suggesting a more "compact" network and potentially higher activity; low density indicates relative sparseness. The analysis method constructs a binary network using thresholding and then calculates the density to help assess the overall connectivity strength of the neural network.

[0064] Figures 10-12 The relevant parameters of biological storage ponds with different depths are shown. In addition, Figures 7-12 Statistical analysis was performed using a significance test method (paired t-test). Significance values ​​in the figure are marked with an asterisk. This indicates that p < 0.05. This indicates that p < 0.01. This indicates that p < 0.001. This means p < 0.0001.

[0065] Figure 7 The correlation coefficients of single-layer biological treatment pools and biological deep reservoirs with different numbers of sub-treatment pools were compared when performing the mature blood cell concentration prediction task (n = 10). As the number of reservoir layers increased, the correlation of the prediction results improved, and the multi-layer biological deep reservoir performed better than the single-layer biological non-deep reservoir.

[0066] Figure 8The study compares the root mean square error (RMSE) of a single sub-treatment pool and multiple sub-treatment pools of biological deep reservoirs when performing a mature blood cell concentration prediction task (n = 10). It shows that the errors of biological deep reservoirs with different sub-treatment pools are all at a low level when performing the prediction task, with three sub-treatment pools showing the best performance. However, the errors of single sub-treatment pools of biological non-deep reservoirs are relatively large, and they do not perform as well as multi-layer biological deep reservoirs.

[0067] Figure 9 The AD diagrams show the correlation coefficient heatmaps of all neurons in the network during spontaneous activity for a single sub-processing pool (non-deep reservoir) and for 2, 3, and 4 sub-processing pools (deep reservoirs). It can be seen that compared to the overall correlation within the network of a single sub-processing pool (non-deep reservoir), the deep reservoirs with multiple sub-processing pools exhibit good modularity, providing richer network characteristics for processing complex temporal information. Furthermore, the number of modules increases with the increase in the number of deep reservoir layers.

[0068] Figure 10 This study compares the modularity of a single sub-processing pool (non-deep biological reservoir) and biological reservoirs with different numbers of layers when performing a mature blood cell concentration prediction task (n = 10). The single sub-processing pool (non-deep biological reservoir) exhibits relatively low modularity, which is influenced by its structural limitations and internal connectivity, affecting its ability to complete complex tasks. In contrast, the other sub-processing pools (deep biological reservoirs) maintain a high degree of modularity (Q > 0.1), more accurately reflecting the modular characteristics of the deep biological reservoirs from a data perspective.

[0069] Figure 11 This paper compares the average clustering coefficients (n = 10) of a single sub-processing pool (non-deep biological reservoir) and biological reservoirs with different numbers of layers when performing a mature blood cell concentration prediction task. For a single sub-processing pool (non-deep biological reservoir), the high average clustering coefficient reflects the high connectivity within the module. Structural constraints prevent information from propagating between layers, thus reducing the reservoir's computational performance. For biological reservoirs with multiple sub-processing pools, the average clustering coefficient remains at a moderately high level, reflecting the tight connectivity within each sub-processing pool and the relatively sparse connectivity between sub-processing pools.

[0070] Figure 12The study compared the network density of a single sub-processing cell in a non-deep biological reservoir and biological reservoirs with different numbers of layers when performing a mature blood cell concentration prediction task (n = 10). For a single sub-processing cell in a non-deep biological reservoir, neurons are not restricted by microfluidic structures, thus forming a high-density network. However, as the number of layers in the biological reservoir increases, i.e., the number of culture modules increases and the area of ​​the compartment structure decreases, the connections between modules become sparse, thus slightly reducing the network density, but it still remains at a high level, reflecting the high level of network response to stimuli.

[0071] In summary, through systematic analysis and result verification of the biological neural network state, the feasibility and effectiveness of the biological deep reservoir constructed based on the microfluidic platform described in this scheme for predicting the concentration of mature blood cells are fully demonstrated. The biological deep reservoir, relying on a modular and targeted network structure design, can form richer and more stable dynamic response patterns. Compared with a single-layer non-deep biological reservoir, its network state characterization ability and prediction performance are significantly improved, thus demonstrating a technical advantage in processing biological maturation dynamic signals with delayed feedback characteristics. This structural advantage makes this invention, compared with traditional unstructured biological reservoirs, more promising in terms of prediction accuracy, stability, and system scalability.

[0072] This invention also provides a microfluidic biological deep reservoir preparation device, which includes a computer device, a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the device implements the steps of the method described above.

[0073] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned microfluidic biological deep reservoir preparation method. The computer-readable storage medium can be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, floppy disk, hard disk, removable storage disk, CD-ROM, or any other form of storage medium known in the art.

[0074] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave.

[0075] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0076] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for preparing a microfluidic biological deep reservoir, characterized in that, The steps of the method include: Based on a pre-set microfluidic chip mold, a microfluidic biological deep storage pool structure is fabricated using PDMS material. The microfluidic biological deep storage pool structure includes multiple directionally connected sub-processing pools, each of the same size. Multiple directionally connected channels are provided between the directionally connected sub-processing pools. The directionally connected channels are provided with an arc-shaped blind zone structure. Each sub-processing pool corresponds to one storage pool of the deep storage pool. Fetal rat cortical tissue was extracted from pregnant ICR mice, and primary nerve cells were dissociated from the fetal rat cortical tissue. The primary nerve cells were then implanted into a sub-treatment tank of a microfluidic biological deep reservoir structure. The inoculated microfluidic biological deep reservoir structure was cultured in an incubator. During the culture process, CheRiff and jRCaMP1b viruses were used to infect neurons, so that the neurons could simultaneously express optogenetic proteins and calcium ion indicators, and finally the microfluidic biological deep reservoir was obtained.

2. The method for preparing a microfluidic biological deep reservoir according to claim 1, characterized in that, The arc-shaped blind zone structure is provided with multiple sub-arc-shaped blind zones that extend in an arc shape in the opposite direction to the directional extension direction of the directional connection channel.

3. The method for preparing a microfluidic biological deep reservoir according to claim 2, characterized in that, The radius of the sub-arc blind zone is R, 0.015mm≤R≤0.025mm, and the angle range of the sub-arc blind zone is δ, 250°≤δ≤290°.

4. The method for preparing a microfluidic biological deep reservoir according to claim 1 or 2, characterized in that, The directional connection channel is also provided with a channel opening structure, which has multiple symmetrically arranged right-angled triangle structures, and the direction of the right-angled triangle structures is the same as the directional extension direction of the directional connection channel.

5. A method for predicting blood cell concentration, characterized in that, The steps of the method include: Blood samples were collected and analyzed to obtain blood cell concentration values. Based on the blood cell concentration values, simulation calculations were performed to obtain blood cell concentration values ​​at multiple time points, and a blood cell concentration sequence was constructed. The blood cell concentration sequence was then encoded into a three-channel sequence. The first, second and third channel light stimulation signals are determined based on the three-channel sequence, and light stimulation is applied to the first sub-treatment pool of a plurality of sub-treatment pools oriented and connected in the microfluidic biological deep storage pool as described in any one of claims 1 to 4 based on the light stimulation signals. The photostimulation response signal of the neutron treatment cell in the microfluidic biological deep storage pool is acquired using an image acquisition device. The photostimulation response signal is then input into a prediction module. The prediction module is equipped with an output layer, which outputs the predicted blood cell concentration value at the predicted time position.

6. The blood cell concentration prediction method according to claim 5, characterized in that, In the step of simulating and calculating blood cell concentration values ​​based on blood cell concentration values ​​to obtain blood cell concentration values ​​at multiple time points and constructing a blood cell concentration sequence, the currently measured blood cell concentration value is used as the blood cell concentration value at the first time point. The blood cell concentration value at the next time point is simulated and calculated based on the blood cell concentration value at the first time point. The same simulation and calculation method is used to calculate the blood cell concentration values ​​at multiple time points. Finally, the blood cell concentration values ​​at all time points are constructed into a blood cell concentration sequence.

7. The method for predicting blood cell concentration according to claim 6, characterized in that, In the step of calculating blood cell concentration values ​​at multiple time points using the same simulation calculation method, the change in blood cell concentration is calculated using the following formula, and the blood cell concentration value at the next time point is determined based on the original blood cell concentration value and the change in blood cell concentration. in, This represents the blood cell concentration value at the current time point. This indicates the change in blood cell concentration at the current point in time. This represents the preset delay parameter, when hour, As a preset constant value, when At that time, the calculated historical values ​​were used.

8. The method for predicting blood cell concentration according to claim 5, characterized in that, In the step of encoding the blood cell concentration sequence into a three-channel sequence, the range of blood cell concentration variation is determined based on the blood cell concentration value at the previous time point and the change threshold in the blood cell concentration sequence. In the blood cell concentration sequence, if the blood cell concentration value at the current time point is greater than the maximum value of the range of blood cell concentration variation, it is encoded as 1 in the first channel; if the blood cell concentration value at the current time point is less than the minimum value of the range of blood cell concentration variation, it is encoded as 1 in the second channel; if the blood cell concentration value at the current time point is within the range of blood cell concentration variation, it is encoded as 0 in both the first and second channels. In the third channel, rate encoding is used, comparing the blood cell concentration value at the current time point with an accumulation threshold. If it is greater than the accumulation threshold, it is encoded as 1 at the current time point. If the value is not greater than the accumulation threshold, the blood cell concentration value at the current time point is accumulated with the blood cell concentration value at the next time point, and then compared with the accumulation threshold again. If the value is greater than the accumulation threshold, it is encoded as 1 at the next time point; if the value is still not greater than the accumulation threshold when compared with the accumulation threshold again, the accumulation continues.

9. The method for predicting blood cell concentration according to claim 5, characterized in that, In the step of determining the photostimulation signals of the first, second, and third channels based on the three-channel sequence, and applying photostimulation to the first sub-processing pool of the multiple sub-processing pools oriented and connected in the microfluidic biological deep reservoir based on the photostimulation signals, the first, second, and third channels are respectively set with stimulation positions in the first sub-processing pool of the microfluidic biological deep reservoir, and photostimulation is applied to the first sub-processing pool of the microfluidic biological deep reservoir based on the encoding results of the first, second, and third channels.

10. The method for predicting blood cell concentration according to claim 5, characterized in that, In the step of acquiring the photostimulation response signals of the sub-processing cells in the microfluidic biological deep storage pool using an image acquisition device and inputting the photostimulation response signals into the prediction module, a preset number of response images are acquired from each sub-processing cell. The response images acquired from each sub-processing cell are superimposed to obtain a superimposed image. Based on the superimposed image, the target neuron location of each sub-processing cell is selected. The pixel value of the target neuron location is acquired for each response image of each sub-processing cell and constructed as an input vector. The input vectors of multiple sub-processing cells are combined to obtain the photostimulation response signal.

Images