Cell culture plate temperature measuring method, model training and deploying method and system

Abstract

The invention relates to the field of cell culture plate temperature measurement, and provides a cell culture plate temperature measurement method and a model training and deploying method and system.The training method comprises the steps that temperature time sequence data corresponding to the outer walls of the periphery of a cell culture plate and chamber temperature actual data of the designated position of the cell culture plate are obtained; performing normalization processing on the temperature time sequence data to obtain normalized temperature time sequence data; inputting the normalized temperature time sequence data into a cell culture plate temperature measurement model for iterative training to obtain corresponding chamber temperature sequence prediction data; performing reverse normalization processing on the chamber temperature sequence prediction data to obtain chamber temperature prediction data; and according to the chamber temperature prediction data and the corresponding chamber temperature actual data, the model parameters are adjusted until the model error reaches the preset precision requirement. According to the cell culture plate temperature measurement model, non-invasive accurate measurement of the chamber temperature of the cell culture plate can be realized.

Description

Technical Field

[0001] The present disclosure relates to the technical field of cell culture plate temperature measurement, and in particular to a cell culture plate temperature measurement method, a model training and deployment method and a system. Background Art

[0002] Cells are extremely sensitive to temperature. Too high or too low a temperature can significantly affect their activity, thus significantly affecting the experimental results. At present, when people observe cells under a microscope, they usually use small culture boxes to hold cells and place them on the microscope platform for observation. Small culture boxes are small in size and transparent from top to bottom, making it easy for people to observe the cells contained therein, but their sealing is often not as good as that of large culture boxes, which may cause changes in the ambient temperature of the cells. Therefore, when observing cells under a microscope, temperature control is a key factor in ensuring cell activity and the accuracy of experimental results.

[0003] The existing incubator temperature control method usually controls the power of one or more heaters in the incubator through one or more proportional-integral-derivative (PID) algorithms to adjust the incubator temperature. However, this temperature control method mainly obtains the ambient temperature in the incubator, rather than the actual temperature in the cell culture plate. Since there is usually a difference between the ambient temperature in the incubator and the temperature in the cell culture plate, accurately measuring and controlling the temperature in the cell culture plate is crucial to effectively ensure cell activity and experimental accuracy.

[0004] The prior art also provides some methods for measuring the temperature in a cell culture plate, which can be generally summarized into the following four types.

[0005] Method 1: Place the temperature sensor inside the incubator and indirectly measure the temperature of the well plate chamber by monitoring and controlling the temperature inside the incubator. The accuracy of this method depends on the sealing and thermal insulation of the incubator, but high sealing and good thermal insulation performance usually increase the size and weight of the incubator, making the incubator unsuitable for some scenarios.

[0006] Method 2: Use non-contact temperature measurement equipment such as thermal imagers to measure the temperature of the well plate surface through infrared technology. Although this method can provide accurate temperature data on the well plate surface, the equipment is expensive and there is a potential risk of thermal radiation damage to cells.

[0007] Method 3: Place a dedicated reference plate in the incubator. The reference plate is equipped with a temperature sensor. The temperature of the reference plate is used to indirectly represent the temperature of the actual culture plate. Although this method can improve the accuracy of cell culture plate temperature measurement, the installation and maintenance of the reference plate is relatively cumbersome, and it increases the complexity and volume of the equipment.

[0008] Method 4: directly bury the temperature sensor electrode inside the cell culture plate for in-situ temperature measurement. Although this method can directly obtain the temperature information inside the cell culture plate and ensure the measurement accuracy, the sensor electrode used is usually a disposable product with high cost, which is not suitable for large-scale experiments and will affect the transparency of the cell culture plate, thus adversely affecting the optical observation of the cells. Summary of the invention

[0009] The present disclosure aims to solve at least one of the problems existing in the prior art and provides a method for measuring the temperature of a cell culture plate, and a method and system for training and deploying a model.

[0010] In one aspect of the present disclosure, a method for training a cell culture plate temperature measurement model is provided, the training method comprising: Acquire temperature time series data corresponding to the four outer walls of the cell culture plate and actual chamber temperature data of a designated position of the cell culture plate; wherein the designated position of the cell culture plate includes positions of different areas of the cell culture plate; Normalizing the temperature time series data to obtain normalized temperature time series data; Inputting the normalized temperature time series data into a cell culture plate temperature measurement model for iterative training to obtain chamber temperature series prediction data corresponding to the normalized temperature time series data; Performing a denormalization process on the chamber temperature sequence prediction data to obtain chamber temperature prediction data; According to the chamber temperature prediction data and the corresponding chamber temperature actual data, the parameters of the cell culture plate temperature measurement model are adjusted until the error of the cell culture plate temperature measurement model reaches a preset accuracy requirement.

[0011] Optionally, the cell culture plate temperature measurement model includes a sequence input layer, a long short-term memory network layer, and a fully connected layer connected in sequence; The sequence input layer is used to receive the normalized temperature time series data and convert the normalized temperature time series data into corresponding tensor data; The long short-term memory network layer is used to receive the tensor data, process and predict the dependency relationship in the tensor data, and output corresponding sequence data; The fully connected layer is used to perform tensor compression and processing on the sequence data to obtain the corresponding chamber temperature sequence prediction data.

[0012] Optionally, the normalizing the temperature time series data to obtain normalized temperature time series data includes: Calculate the mean and standard deviation of the temperature time series data; The temperature time series data is normalized according to the average value and the standard deviation to obtain the normalized temperature time series data.

[0013] Another aspect of the present disclosure provides a method for deploying a cell culture plate temperature measurement model, the method comprising: Obtaining a cell culture plate temperature measurement model trained using the cell culture plate temperature measurement model training method described above; Quantizing the trained cell culture plate temperature measurement model, converting it from a floating point model to a fixed point integer model, to obtain a quantized cell culture plate temperature measurement model; Optimizing the quantified cell culture plate temperature measurement model, adjusting its memory usage and computing resource allocation, and obtaining an optimized cell culture plate temperature measurement model; Performing a benchmark test on the optimized cell culture plate temperature measurement model to evaluate the performance of the optimized cell culture plate temperature measurement model; converting the optimized cell culture plate temperature measurement model into target code; The target code is deployed to the chip.

[0014] Another aspect of the present disclosure provides a method for measuring the temperature of a cell culture plate, the method comprising: Obtaining actual temperature data corresponding to the four outer walls of the cell culture plate; Normalizing the actual temperature data to obtain actual temperature time series data; Inputting the actual temperature time series data into a trained cell culture plate temperature measurement model to obtain a corresponding model output result; wherein the trained cell culture plate temperature measurement model is trained using the training method of the cell culture plate temperature measurement model described above; The model output result is subjected to a denormalization process to obtain the chamber measurement temperature of the cell culture plate.

[0015] Optionally, the cell culture plate temperature measurement method further comprises: Performing environmental control on the incubator where the cell culture plate is located according to the temperature measured in the chamber; and / or, The measured temperature of the chamber is visually displayed.

[0016] Another aspect of the present disclosure provides a training system for a cell culture plate temperature measurement model, the training system comprising: An acquisition module, used to acquire temperature time series data corresponding to the four outer walls of the cell culture plate and actual chamber temperature data of a designated position of the cell culture plate; wherein the designated position of the cell culture plate includes positions of different areas of the cell culture plate; A normalization module, used for normalizing the temperature time series data to obtain normalized temperature time series data; An iteration module, used for inputting the normalized temperature time series data into a cell culture plate temperature measurement model for iterative training, and obtaining chamber temperature series prediction data corresponding to the normalized temperature time series data; A denormalization module, used for performing denormalization processing on the chamber temperature sequence prediction data to obtain chamber temperature prediction data; The adjustment module is used to adjust the parameters of the cell culture plate temperature measurement model according to the chamber temperature prediction data and the corresponding chamber temperature actual data, until the error of the cell culture plate temperature measurement model reaches a preset accuracy requirement.

[0017] Another aspect of the present disclosure provides a deployment system for a cell culture plate temperature measurement model, the deployment system comprising: A model acquisition module, used to acquire a cell culture plate temperature measurement model trained using the cell culture plate temperature measurement model training method described above; A deployment module is used to quantize the trained cell culture plate temperature measurement model, convert it from a floating point model to a fixed point integer model, and obtain a quantized cell culture plate temperature measurement model; optimize the quantized cell culture plate temperature measurement model, adjust its memory usage and computing resource allocation, and obtain an optimized cell culture plate temperature measurement model; perform a benchmark test on the optimized cell culture plate temperature measurement model to evaluate the performance of the optimized cell culture plate temperature measurement model; convert the optimized cell culture plate temperature measurement model into a target code; and deploy the target code to a chip.

[0018] Another aspect of the present disclosure provides a cell culture plate temperature measurement system, the cell culture plate temperature measurement system comprising: A temperature acquisition module is used to obtain actual temperature data corresponding to the four outer walls of the cell culture plate; A normalization processing module, used for performing normalization processing on the actual temperature data to obtain actual temperature time series data; A prediction module, used for inputting the actual temperature time series data into a trained cell culture plate temperature measurement model to obtain a corresponding model prediction result; wherein the trained cell culture plate temperature measurement model is trained using the training method of the cell culture plate temperature measurement model described above; The denormalization processing module is used to perform denormalization processing on the model prediction result to obtain the chamber measurement temperature of the cell culture plate.

[0019] Optionally, the cell culture plate temperature measurement system further comprises: an environmental control module, for performing environmental control on the incubator where the cell culture plate is located according to the temperature measured in the chamber; and / or, A visualization module is used to visualize the measured temperature of the chamber.

[0020] Compared with the prior art, the present invention designs a non-invasive measurement method to address the problem of high cleanliness requirements for cell culture. The cell culture plate temperature measurement model is used to model the heat energy transfer process from the inner chamber of the cell culture plate to the surrounding area, and deep learning technology is used to process the temperature time series data of the cell culture plate, which can effectively achieve accurate measurement of the chamber temperature of the cell culture plate. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0022] Figure 1 A flowchart of a method for training a cell culture plate temperature measurement model provided in one embodiment of the present disclosure; Figure 2 A schematic diagram of temperature collection and area division of a cell culture plate provided in another embodiment of the present disclosure; Figure 3 A schematic diagram of a process for obtaining chamber temperature prediction data using a cell culture plate temperature measurement model provided in another embodiment of the present disclosure; Figure 4 A schematic flow chart of a method for deploying a cell culture plate temperature measurement model provided in another embodiment of the present disclosure; Figure 5 A schematic flow chart of a method for measuring the temperature of a cell culture plate provided in another embodiment of the present disclosure; Figure 6A schematic structural diagram of a training system for a cell culture plate temperature measurement model provided by another embodiment of the present disclosure; Figure 7 A schematic structural diagram of a deployment system of a cell culture plate temperature measurement model provided by another embodiment of the present disclosure; Figure 8 A schematic structural diagram of a cell culture plate temperature measurement system provided by another embodiment of the present disclosure; Fig. 9 A schematic diagram of the workflow of a cell culture plate temperature measurement system provided in another embodiment of the present disclosure. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the embodiments of the present disclosure will be described in detail below in conjunction with the accompanying drawings. However, it can be understood by those skilled in the art that in each embodiment of the present disclosure, many technical details are proposed in order to enable readers to better understand the present disclosure. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed for protection in the present disclosure can also be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of the present disclosure. The various embodiments can be combined and referenced with each other without contradiction.

[0024] One embodiment of the present disclosure relates to a method for training a cell culture plate temperature measurement model, the process of which is as follows: Figure 1 As shown, it includes steps S110 to S150.

[0025] Step S110, obtaining the temperature time series data corresponding to the four outer walls of the cell culture plate and the actual chamber temperature data of the designated position of the cell culture plate; wherein the designated position of the cell culture plate includes the position of different areas of the cell culture plate.

[0026] Specifically, this embodiment does not limit the specific method of obtaining the temperature time series data and the actual chamber temperature data in step S110. The cell culture plate here can be any type of culture plate commonly used in the field of cell culture. For example, the cell culture plate can be a cell well plate commonly used in the prior art. The following embodiments of the present disclosure will be illustrated by taking the cell well plate as an example.

[0027] For example, combining Figure 2When the cell culture plate is a cell well plate, for a cell well plate 1 provided with multiple chambers, a person skilled in the art can place the cell well plate 1 into a fixture 16, and place the fixture 16 with the cell well plate 1 placed under an open incubator adapted to a microscope stage for observation. Among them, the outer walls of the cell well plate 1 are respectively in contact with the temperature sensor 2, the temperature sensor 3, the temperature sensor 4, and the temperature sensor 5 on the fixture 16, so as to obtain the temperature time series data corresponding to the outer walls of the cell well plate 1 through the temperature sensor 2 to the temperature sensor 5, realize non-invasive temperature measurement, avoid contamination of the cell culture process, and improve the cleanliness of the cell culture.

[0028] In particular, in order to make the temperature time series data corresponding to the four outer walls of the cell well plate 1 obtained by the temperature sensors 2 to 5 more representative, such as Figure 2 As shown, the temperature sensors 2 to 5 can be respectively contacted with the middle positions of the four edges of the cell well plate 1, so that the temperature time series data at the middle positions of the four edges of the cell well plate 1 can be used as the temperature time series data corresponding to the four outer walls of the cell well plate 1.

[0029] In order to obtain the actual chamber temperature data of a specified position of a cell culture plate, especially a cell well plate, such as Figure 2 As shown, a person skilled in the art can divide the cell well plate 1 into five areas, namely area 11, area 12, area 13, area 14, and area 15. These five areas completely cover all chambers in the cell well plate 1 to cope with the uneven temperature control of each chamber of the cell well plate that may occur during the culture process, and realize temperature monitoring of these five areas. Among them, area 15 is located at the center of the cell well plate 1, and areas 11 to 14 are respectively located at the upper left, lower left, lower right, and upper right of area 15. Then, the most representative chambers are selected from the above five areas as sampling points, so that the actual temperature data of each sampling point is used as the actual temperature data of the chambers corresponding to each area in the cell well plate. For example, the chamber in the upper left corner of area 11 can be used as sampling point 6, the chamber in the lower left corner of area 12 can be used as sampling point 7, the chamber in the lower right corner of area 13 can be used as sampling point 8, the chamber in the upper right corner of area 14 can be used as sampling point 9, and the chamber located in the middle and upper left of area 15, that is, in the second row and third column, can be used as sampling point 10.

[0030] In order to obtain enough data sets for model training, based on possible culture conditions, temperature sensors 2 to 5 and sampling points 6 to 10 can be used multiple times to obtain the temperature time series data corresponding to the four outer walls of the cell well plate 1 and the actual chamber temperature data at the specified position of the cell well plate.

[0031] For example, a person skilled in the art may collect 100 data sets including temperature time series data and actual chamber temperature data through a heating process and a heat preservation process.

[0032] A corresponding temperature rise data set can be obtained for each temperature rise process. In each temperature rise process, the incubator where the cell well plate 1 is located is first heated so that the temperature in the incubator rises from room temperature to the target temperature rise temperature, and the data is collected every 1 second within 2 hours when the incubator starts to rise from room temperature, that is, the readings of temperature sensor 2 to temperature sensor 5 and the temperature of sampling point 6 to sampling point 10 are read once, so that the temperature time series data of the cell well plate 1 and the temperature rise data set consisting of the actual chamber temperature data can be obtained, and then the incubator is stopped from heating for 2 hours and cooled to room temperature. Among them, the target temperature rise temperature range is 33℃-40℃, and each interval of 0.1℃ in this range corresponds to a target temperature rise temperature, so that 70 temperature rise processes are obtained through 70 target temperature rises, and then 70 temperature rise data sets are obtained. For example, a target temperature rise temperature of 33℃ corresponds to a temperature rise process, a target temperature rise temperature of 33.1℃ corresponds to a temperature rise process, a target temperature rise temperature of 33.2℃ corresponds to a temperature rise process, and so on. Since the interval is 0.1℃, 33℃-40℃ can be divided into 71 temperature points, namely 33℃, 33.1℃, 33.2℃, 33.3℃, ​​33.4℃, 33.5℃, 33.6℃, ​​33.7℃, 33.8℃, 33.9℃, 34℃, ..., 39℃, 39.1℃, 39.2℃, 39.3℃, 39.4℃, 39.5℃, 39.6℃, 39.7℃, 39.8℃, 39.9℃, 40℃, therefore, you can choose to ignore the temperature point of 33℃ or 40℃, and use the remaining 70 temperature points as the corresponding 70 target heating temperatures.

[0033] A corresponding insulation data set can be obtained for each insulation process. In each insulation process, first, the cell well plate 1 is taken out from the standard incubator in which the temperature is at the target insulation temperature and cultured in the incubator for 2 hours, and data is collected every 1 second within the 2 hours, that is, the readings of the temperature sensor 2 to the temperature sensor 5 and the temperature of the sampling point 6 to the sampling point 10 are read once, so that the insulation data set consisting of the temperature time series data of the cell well plate 1 and the actual data of the chamber temperature can be obtained. The incubator has been preheated to the target insulation temperature for 30 minutes before the cell well plate 1 is placed. Subsequently, the heating of the incubator is stopped for 2 hours and cooled to room temperature. Among them, the target insulation temperature ranges from 36°C to 37°C, and each interval of 0.1°C within the range corresponds to a target insulation temperature, so that 10 insulation processes are obtained through 10 target insulation temperatures, and these 10 insulation processes are repeated 3 times each, so that 30 insulation data sets can be obtained through 30 insulation processes. For example, a target insulation temperature of 36°C corresponds to one insulation process, a target insulation temperature of 36.1°C corresponds to one insulation process, a target insulation temperature of 36.2°C corresponds to one insulation process, and so on. Since 36°C-37°C can be divided into 11 temperature points, namely 36°C, 36.1°C, 36.2°C, 36.3°C, 36.4°C, 36.5°C, 36.6°C, 36.7°C, 36.8°C, 36.9°C, and 37°C, when the interval is 0.1°C, you can choose to ignore the temperature point of 36°C or 37°C, and use the remaining 10 temperature points as the corresponding 10 target insulation temperatures.

[0034] Step S120, normalizing the temperature time series data to obtain normalized temperature time series data.

[0035] Specifically, normalization can improve the training effect of the model, accelerate the convergence speed, and enhance the generalization ability of the model.

[0036] Exemplarily, step S120 includes: calculating the average value and standard deviation of the temperature time series data; and normalizing the temperature time series data according to the average value and the standard deviation to obtain normalized temperature time series data.

[0037] Specifically, assuming that a certain temperature value in the temperature time series data is X, μ is the average value of the temperature time series data, σ is the standard deviation of the temperature time series data, then the normalized temperature value corresponding to X It can be expressed as .

[0038] Step S130 , inputting the normalized temperature time series data into the cell culture plate temperature measurement model for iterative training, and obtaining chamber temperature series prediction data corresponding to the normalized temperature time series data.

[0039] Specifically, step S130 can utilize the existing toolkit in MATLAB software to implement iterative training of the cell culture plate temperature measurement model. During the iterative training process, if the actual number of iterations reaches the preset number of iterations, it can be considered that the cell culture plate temperature measurement model has been trained.

[0040] Exemplarily, the cell culture plate temperature measurement model includes a sequence input layer, a long short-term memory (LSTM) layer, and a fully connected layer that are connected in sequence.

[0041] The sequence input layer is used to receive normalized temperature time series data and convert the normalized temperature time series data into corresponding tensor data. The long short-term memory network layer is used to receive tensor data, process and predict the dependencies in the tensor data, and output the corresponding sequence data. The fully connected layer is used to perform tensor compression and processing on the sequence data to obtain the corresponding chamber temperature sequence prediction data.

[0042] Specifically, combined with Figure 3 , step S130 may specifically include steps S1 to S4.

[0043] Step S1 is used to Figure 2 The four-hole plate wall temperatures of the cell well plate 1, i.e., the normalized temperature time series data, are input into the sequence input layer in the cell culture plate temperature measurement model. Step S2 is used to receive the normalized temperature time series data through the sequence input layer in the cell culture plate temperature measurement model, and convert it into corresponding tensor data and pass it to the subsequent layer, i.e., the long short-term memory network layer. The long short-term memory network layer is a special recurrent neural network layer for processing and predicting dependencies in time series data or sequence data. For example, 128 LSTM units can be provided in the long short-term memory network layer. The LSTM unit can capture and retain long-term dependent information through its memory cells and gating mechanism, and is suitable for modeling the process of heat energy transfer from the inner chamber of a cell culture plate, such as a cell well plate, to the surrounding areas over time. Step S3 is used to receive tensor data through the long short-term memory network layer, process and predict dependencies in the tensor data, and output corresponding sequence data. Step S4 uses a fully connected layer to integrate the features extracted from the previous layer (i.e., the long short-term memory network layer) and generate the final output result, thereby obtaining the corresponding chamber temperature sequence prediction data by tensor compression and processing the sequence data. Each neuron in the fully connected layer is connected to all neurons in the previous layer, and the input features are processed through weighted summation and nonlinear activation functions.

[0044] Step S140 , performing a denormalization process on the chamber temperature sequence prediction data to obtain chamber temperature prediction data.

[0045] For example, combining Figure 3 Step S140 is used as step S5 to perform denormalization on the chamber temperature sequence prediction data output by the fully connected layer in the cell culture plate temperature measurement model to obtain Figure 2 The temperatures of the five well plate areas of the cell well plate 1 shown are the predicted data of the chamber temperature corresponding to the cell well plate 1 .

[0046] The denormalization process here is a restoration process corresponding to the normalization process in step S120, so as to restore the chamber temperature sequence prediction data output by the cell culture plate temperature measurement model into chamber temperature prediction data having actual physical meaning through the denormalization process.

[0047] Step S150, adjusting the parameters of the cell culture plate temperature measurement model according to the chamber temperature prediction data and the corresponding chamber temperature actual data, until the error of the cell culture plate temperature measurement model reaches a preset accuracy requirement.

[0048] Specifically, step S150 can first calculate the root mean square error (RMSE) between the chamber temperature prediction data and its corresponding actual chamber temperature data, and adjust the parameters of the trained cell culture plate temperature measurement model according to the calculated root mean square error until the root mean square error between the chamber temperature prediction data obtained by the cell culture plate temperature measurement model and its corresponding actual chamber temperature data reaches the preset accuracy requirement, thereby obtaining the final cell culture plate temperature measurement model.

[0049] Compared with the prior art, the training method of the cell culture plate temperature measurement model provided in the embodiment of the present disclosure adopts the cell culture plate temperature measurement model to model the heat energy transfer process from the inner chamber of the cell culture plate to the surrounding area, so that the trained cell culture plate temperature measurement model can be applied to the non-invasive and accurate measurement of the chamber temperature of the cell culture plate, thereby effectively avoiding the contamination of the chamber of the cell culture plate during the temperature measurement process, improving the accuracy of temperature measurement and reducing costs.

[0050] Another embodiment of the present disclosure relates to a method for deploying a cell culture plate temperature measurement model, comprising: obtaining a cell culture plate temperature measurement model trained by the cell culture plate temperature measurement model training method described in the above embodiment; quantizing the trained cell culture plate temperature measurement model, converting it from a floating-point model to a fixed-point integer model, and obtaining a quantized cell culture plate temperature measurement model; optimizing the quantized cell culture plate temperature measurement model, adjusting its memory usage and computing resource allocation, and obtaining an optimized cell culture plate temperature measurement model; benchmarking the optimized cell culture plate temperature measurement model to evaluate the performance of the optimized cell culture plate temperature measurement model; converting the optimized cell culture plate temperature measurement model into a target code; and deploying the target code to a chip.

[0051] Specifically, Figure 4 As shown, the deployment method of the cell culture plate temperature measurement model can be represented as steps S6 to S11.

[0052] Step S6 is used to obtain the trained cell culture plate temperature measurement model and convert it into the ONNX (OpenNeural Network Exchange) format to improve the efficiency and flexibility of model deployment. ONNX is an open file format used to represent machine learning models, which aims to provide a unified standard for model conversion between different deep learning frameworks. Through the ONNX format, developers can seamlessly migrate models between different frameworks, thereby improving development efficiency and flexibility.

[0053] Step S7 is used for model quantization, which converts the cell culture plate temperature measurement model from a floating point model such as a 32-bit floating point model to a fixed point integer model such as an 8-bit integer model to obtain a quantized cell culture plate temperature measurement model. The quantization process can significantly reduce the data size of the cell culture plate temperature measurement model, thereby improving the use efficiency and reasoning speed of the central processing unit (CPU) / microcontroller unit (MCU).

[0054] Step S8 is used for model optimization, and the memory usage and computing resource allocation of the quantified cell culture plate temperature measurement model are adjusted to optimize the model performance.

[0055] Step S9 is used to perform a benchmark test to evaluate the performance of the optimized cell culture plate temperature measurement model, including inference time, memory usage, accuracy, etc. By performing a benchmark test on the optimized cell culture plate temperature model, it can be ensured that the performance of the target code subsequently generated according to the model on the chip where it is deployed is consistent with the performance of the originally trained cell culture plate temperature measurement model, and the time it takes for the model to perform one operation can be obtained.

[0056] Step S10 is used to generate code. For example, step S10 can use the tool STM32Cube.AI to convert the optimized cell culture plate temperature measurement model into C code suitable for embedded processors such as STM32 microcontrollers, so as to use the C code to run artificial intelligence (AI) reasoning on the embedded processor. Of course, step S10 can also convert the optimized cell culture plate temperature measurement model into corresponding target codes suitable for other devices such as other types of processors, so as to apply the optimized cell culture plate temperature measurement model to other devices.

[0057] Step S11 is used to deploy the target code corresponding to the cell culture plate temperature measurement model to the chip, so as to use the chip to monitor and analyze the temperature of the cell culture plate in the corresponding culture scene in real time. For example, step S11 can compile the target code into the culture box code, so as to use the culture box to realize real-time monitoring and analysis of the temperature of the cell culture plate in the culture box during the culture process.

[0058] In particular, when the chip is an STM32 series microcontroller, the deployment method of the above-mentioned cell culture plate temperature measurement model is preferably implemented using STM32 official tools to improve deployment efficiency and reduce deployment difficulty.

[0059] Compared with the prior art, the deployment method of the cell culture plate temperature measurement model provided in the embodiment of the present disclosure enables the chip deployed with the cell culture plate temperature measurement model to independently and quickly perform non-invasive monitoring of the chamber temperature of the cell culture plate, while improving accuracy and reducing costs.

[0060] Another embodiment of the present disclosure relates to a method for measuring the temperature of a cell culture plate, comprising: obtaining actual temperature data corresponding to the four outer walls of the cell culture plate; normalizing the actual temperature data to obtain actual temperature time series data; inputting the actual temperature time series data into a trained cell culture plate temperature measurement model to obtain a corresponding model output result; wherein the trained cell culture plate temperature measurement model is trained using the training method for the cell culture plate temperature measurement model described in the above embodiment; and denormalizing the model output result to obtain the chamber measurement temperature of the cell culture plate.

[0061] Specifically, Figure 5 As shown, the method for measuring the temperature of a cell culture plate can be represented as steps S12 to S23.

[0062] Step S12 is used to initialize and create a neural network, load the network structure and parameters of the trained cell culture plate temperature measurement model into the memory, allocate the required memory for each part of the network structure including weights, biases, activation buffers, etc., and configure the parameters and status of the network structure to ensure that the network structure is correctly connected and ready.

[0063] Step S13 is used to obtain the actual temperature data corresponding to the four outer walls of the cell culture plate through temperature acquisition. Figure 2 As shown, step S13 can utilize temperature sensors 2 to 5 arranged on the outer walls of the cell well plate 1 to collect actual temperature data corresponding to the outer walls of the cell well plate 1 to achieve non-invasive temperature measurement, avoid contamination of the cell culture process, and improve the cleanliness of cell culture.

[0064] Step S14 is used for data normalization processing to normalize the actual temperature data.

[0065] Step S15 is used to complete the temperature sequence input, and input the actual temperature time series data corresponding to the actual temperature data obtained by normalization processing into the prepared cell culture plate temperature measurement model.

[0066] Steps S16 to S19 are all implemented by the sequence output layer in the cell culture plate temperature measurement model. Step S16 is used to receive the actual temperature time series data as input and output a 1-dimensional 64-bit signed number tensor containing the input tensor shape. This step is used to obtain the dimension information of the actual temperature time series data.

[0067] Step S17 receives a data tensor containing dimension information of actual temperature time series data and an index tensor set in initialization through an index collection operation, collects entries on a specified axis of the data tensor by index, and connects them to the output tensor. This step is used to extract data from the data tensor by a specified index.

[0068] Step S18 connects a list of tensors into a single tensor through a tensor connection operation.

[0069] Step S19 receives a tensor as input through a tensor tiling operation, and outputs a new tensor generated by repeating the input tensor a specified number of times along each dimension to the long short-term memory network layer in the cell culture plate temperature measurement model. This step is used to construct a new tensor generated by tiling a given tensor.

[0070] Step S20 is used to perform calculations through a long short-term memory network in the cell culture plate temperature measurement model. The long short-term memory network receives actual temperature time series data, captures and retains long-term dependencies, and outputs processed sequence data.

[0071] Step S21 is used for result processing. The result output by the long short-term memory network is tensor compressed and processed through the fully connected layer in the cell culture plate temperature measurement model to obtain the calculation result of the cell culture plate temperature measurement model.

[0072] Step S22 is used to output the calculation result of the cell culture plate temperature measurement model.

[0073] Step S23 is used to perform a denormalization process on the calculation result of the cell culture plate temperature measurement model to obtain the final result, that is, the chamber measurement temperature of the cell culture plate. Figure 2 When step S13 uses temperature sensors 2 to 5 arranged on the outer walls of the cell well plate 1 to collect actual temperature data corresponding to the outer walls of the cell well plate 1, the chamber measurement temperature obtained in step S23 includes the temperatures corresponding to the five areas of area 11, area 12, area 13, area 14, and area 15 in the cell well plate 1, thereby realizing non-invasive measurement of the cell well plate chamber temperature and avoiding contamination of the cell well plate culture process by the temperature measurement process.

[0074] Exemplarily, the method for measuring the temperature of a cell culture plate further includes: performing environmental control on an incubator where the cell culture plate is located according to the measured temperature of the chamber.

[0075] For example, if the measured chamber temperature is lower than the preset chamber temperature, the incubator where the cell culture plate is located can be heated to increase the temperature of each chamber in the cell culture plate. Conversely, if the measured chamber temperature is higher than the preset chamber temperature, the incubator where the cell culture plate is located can be cooled to reduce the temperature of each chamber in the cell culture plate.

[0076] By controlling the environment of the incubator where the cell culture plate is located according to the measured temperature of the chamber, the chamber temperature of the cell culture plate can be effectively and accurately controlled during the culture process to ensure that the chamber temperature meets the culture requirements.

[0077] Exemplarily, the cell culture plate temperature measurement method further includes: visually displaying the measured temperature of the chamber.

[0078] Specifically, the visual display can be implemented using a corresponding display screen or a visual operation screen, so that the user can view the chamber measurement temperature more intuitively.

[0079] Compared with the prior art, the cell culture plate temperature measurement method provided in the embodiment of the present disclosure is designed to solve the problem of high cleanliness requirements for cell culture, and a non-invasive cell culture plate temperature measurement method is designed. A cell culture plate temperature measurement model is used to model the heat energy transfer process from the inner chamber of the cell culture plate to the surrounding area, and deep learning technology is used to process the temperature time series data of the cell culture plate, thereby effectively realizing the accurate measurement of the chamber temperature of the cell culture plate.

[0080] Another embodiment of the present disclosure relates to a training system for a cell culture plate temperature measurement model, such as Figure 6 As shown, it includes an acquisition module 610, a normalization module 620, an iteration module 630, a denormalization module 640, and an adjustment module 650.

[0081] The acquisition module 610 is used to acquire the temperature time series data corresponding to the four outer walls of the cell culture plate and the actual chamber temperature data of the designated position of the cell culture plate; wherein the designated position of the cell culture plate includes the positions of different areas of the cell culture plate.

[0082] The normalization module 620 is used to perform normalization processing on the temperature time series data to obtain normalized temperature time series data.

[0083] The iteration module 630 is used to input the normalized temperature time series data into the cell culture plate temperature measurement model for iterative training to obtain chamber temperature sequence prediction data corresponding to the normalized temperature time series data.

[0084] The denormalization module 640 is used to perform denormalization processing on the chamber temperature sequence prediction data to obtain chamber temperature prediction data.

[0085] The adjustment module 650 is used to adjust the parameters of the cell culture plate temperature measurement model according to the chamber temperature prediction data and the corresponding chamber temperature actual data, until the error of the cell culture plate temperature measurement model reaches the preset accuracy requirement.

[0086] The specific implementation method of the training system of the cell culture plate temperature measurement model provided in the embodiment of the present disclosure can be found in the training method of the cell culture plate temperature measurement model provided in the embodiment of the present disclosure, and will not be repeated here.

[0087] Compared with the prior art, the training system of the cell culture plate temperature measurement model provided in the embodiment of the present disclosure adopts the cell culture plate temperature measurement model to model the heat energy transfer process from the inner chamber of the cell culture plate to the surrounding area, so that the trained cell culture plate temperature measurement model can be applied to the non-invasive and accurate measurement of the chamber temperature of the cell culture plate, thereby effectively avoiding the contamination of the chamber of the cell culture plate during the temperature measurement process, improving the accuracy of temperature measurement and reducing costs.

[0088] Another embodiment of the present disclosure relates to a deployment system for a cell culture plate temperature measurement model, such as Figure 7 As shown, it includes a model acquisition module 710 and a deployment module 720.

[0089] The model acquisition module 710 is used to acquire the cell culture plate temperature measurement model trained by the training method of the cell culture plate temperature measurement model described in the above embodiment.

[0090] The deployment module 720 is used to quantize the trained cell culture plate temperature measurement model, convert it from a floating-point model to a fixed-point integer model, and obtain a quantized cell culture plate temperature measurement model; optimize the quantized cell culture plate temperature measurement model, adjust its memory usage and computing resource allocation, and obtain an optimized cell culture plate temperature measurement model; benchmark the optimized cell culture plate temperature measurement model to evaluate the performance of the optimized cell culture plate temperature measurement model; convert the optimized cell culture plate temperature measurement model into target code; and deploy the target code to the chip.

[0091] The specific implementation method of the deployment system of the cell culture plate temperature measurement model provided in the embodiment of the present disclosure can be found in the deployment method of the cell culture plate temperature measurement model provided in the embodiment of the present disclosure, and will not be repeated here.

[0092] Compared with the prior art, the deployment system of the cell culture plate temperature measurement model provided in the embodiment of the present disclosure enables the chip deployed with the cell culture plate temperature measurement model to independently and quickly perform non-invasive monitoring of the chamber temperature of the cell culture plate, while improving accuracy and reducing costs.

[0093] Another embodiment of the present disclosure relates to a cell culture plate temperature measurement system, such as Figure 8 As shown, it includes a temperature acquisition module 810, a normalization processing module 820, a prediction module 830, and a denormalization processing module 840.

[0094] The temperature acquisition module 810 is used to acquire actual temperature data corresponding to the four outer walls of the cell culture plate.

[0095] The normalization processing module 820 is used to perform normalization processing on the actual temperature data to obtain actual temperature time series data.

[0096] The prediction module 830 is used to input the actual temperature time series data into the trained cell culture plate temperature measurement model to obtain the corresponding model prediction results; wherein the trained cell culture plate temperature measurement model is trained using the training method of the cell culture plate temperature measurement model described in the above embodiment.

[0097] The denormalization processing module 840 is used to perform denormalization processing on the model prediction results to obtain the chamber measurement temperature of the cell culture plate.

[0098] Exemplarily, the cell culture plate temperature measurement system further includes an environment control module, which is used to perform environment control on the incubator where the cell culture plate is located according to the temperature measured in the chamber.

[0099] Exemplarily, the cell culture plate temperature measurement system further includes a visualization module, which is used to visualize the measured temperature of the chamber.

[0100] The specific implementation method of the cell culture plate temperature measurement system provided in the embodiment of the present disclosure can be found in the description of the cell culture plate temperature measurement method provided in the embodiment of the present disclosure, and will not be repeated here.

[0101] Compared with the prior art, the cell culture plate temperature measurement system provided in the embodiment of the present disclosure is designed with a non-invasive cell culture plate temperature measurement method to address the problem of high cleanliness requirements for cell culture. The cell culture plate temperature measurement model is used to model the heat energy transfer process from the inner chamber of the cell culture plate to the surrounding area, and deep learning technology is used to process the temperature time series data of the cell culture plate, thereby effectively realizing the accurate measurement of the chamber temperature of the cell culture plate.

[0102] For example, Fig. 9As shown, a cell culture plate temperature measurement system includes a temperature acquisition module, a main control board and a visual operation screen. Among them, the temperature acquisition module is used to use the temperature sensors arranged on the outer walls of the cell culture plate such as the cell well plate in the incubator to collect temperature, obtain 4-way temperature data, i.e., the actual temperature data corresponding to the outer walls of the cell culture plate such as the cell well plate, and transmit the actual temperature data to the main control board, and the communication method can be RS485. The main control board is an STM32F407 main control board, which adopts a multi-task system, one of which is used for neural network computing tasks, specifically for normalizing the actual temperature data, obtaining the actual temperature time series data, and inputting the actual temperature time series data into the trained cell culture plate temperature measurement model containing the long short-term memory network to obtain the corresponding model prediction result, and then the model prediction result is denormalized to obtain the five-region temperature, i.e., the center and four corners of the cell culture plate such as the cell well plate The corresponding temperature is used as the chamber measurement temperature of the cell culture plate such as the cell well plate. In addition to the neural network computing tasks, other task systems of the main control board can be used to control the environment of the incubator based on the incubator control system. For example, the temperature in the incubator can be controlled based on the temperature measured in the chamber of a cell culture plate such as a cell well plate. After obtaining the five-zone temperature, the main control board can use a universal asynchronous receiver-transmitter (UART) to transmit the five-zone temperature to the visual operation screen for display.

[0103] right Fig. 9 The cell culture plate temperature measurement system shown in the figure is tested. Among them, 8 groups of data correspond to 8 heating processes of a cell culture plate such as a cell well plate, which are respectively heated from room temperature to the target heating temperature and then stopped for 2 hours after being cultured for 2 hours. Another 8 groups of data correspond to 8 insulation processes of a cell culture plate such as a cell well plate, which are respectively taken out of an incubator at 37°C and placed in a culture box for culture for 2 hours and then stopped for 2 hours. The root mean square error between the measured temperature of the chamber of a cell culture plate such as a cell well plate obtained by the cell culture plate temperature measurement system and the actual temperature of the chamber obtained by probing into the corresponding chamber of the cell culture plate such as a cell well plate is calculated, and the corresponding results are shown in Table 1 below.

[0104] Table 1 Root mean square error (RMSE) corresponding to different test conditions

[0105] It can be seen from Table 1 that the root mean square errors are all in the range of 0.22 to 0.38, indicating that the measured temperature of the chamber of a cell culture plate such as a cell well plate obtained by the cell culture plate temperature measurement system is not much different from the actual temperature of the corresponding chamber in the cell culture plate such as the cell well plate, and can meet the accuracy requirements.

[0106] Those skilled in the art will appreciate that the above-mentioned embodiments are specific embodiments for implementing the present disclosure, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present disclosure.

Claims

1. A training method for a cell culture plate temperature measurement model, characterized in that: The training method comprises: Acquire temperature time series data corresponding to the four outer walls of the cell culture plate and actual chamber temperature data of a designated position of the cell culture plate; wherein the designated position of the cell culture plate includes positions of different areas of the cell culture plate; Normalizing the temperature time series data to obtain normalized temperature time series data; Inputting the normalized temperature time series data into a cell culture plate temperature measurement model for iterative training to obtain chamber temperature series prediction data corresponding to the normalized temperature time series data; Performing a denormalization process on the chamber temperature sequence prediction data to obtain chamber temperature prediction data; According to the chamber temperature prediction data and the corresponding chamber temperature actual data, the parameters of the cell culture plate temperature measurement model are adjusted until the error of the cell culture plate temperature measurement model reaches a preset accuracy requirement.

2. The training method according to claim 1, characterized in that: The cell culture plate temperature measurement model comprises a sequence input layer, a long short-term memory network layer, and a fully connected layer connected in sequence; The sequence input layer is used to receive the normalized temperature time series data and convert the normalized temperature time series data into corresponding tensor data; The long short-term memory network layer is used to receive the tensor data, process and predict the dependency relationship in the tensor data, and output corresponding sequence data; The fully connected layer is used to perform tensor compression and processing on the sequence data to obtain the corresponding chamber temperature sequence prediction data.

3. The training method according to claim 1, characterized in that: The step of normalizing the temperature time series data to obtain normalized temperature time series data includes: Calculate the mean and standard deviation of the temperature time series data; The temperature time series data is normalized according to the average value and the standard deviation to obtain the normalized temperature time series data.

4. A method for deploying a cell culture plate temperature measurement model, characterized in that: The deployment method includes: Obtaining a cell culture plate temperature measurement model trained by the cell culture plate temperature measurement model training method according to any one of claims 1 to 3; Quantizing the trained cell culture plate temperature measurement model, converting it from a floating point model to a fixed point integer model, to obtain a quantized cell culture plate temperature measurement model; Optimizing the quantified cell culture plate temperature measurement model, adjusting its memory usage and computing resource allocation, and obtaining an optimized cell culture plate temperature measurement model; Performing a benchmark test on the optimized cell culture plate temperature measurement model to evaluate the performance of the optimized cell culture plate temperature measurement model; converting the optimized cell culture plate temperature measurement model into target code; The target code is deployed to the chip.

5. A method for measuring the temperature of a cell culture plate, characterized in that: The cell culture plate temperature measurement method comprises: Obtaining actual temperature data corresponding to the four outer walls of the cell culture plate; Normalizing the actual temperature data to obtain actual temperature time series data; Inputting the actual temperature time series data into a trained cell culture plate temperature measurement model to obtain a corresponding model output result; wherein the trained cell culture plate temperature measurement model is trained using the training method for the cell culture plate temperature measurement model according to any one of claims 1 to 3; The model output result is subjected to a denormalization process to obtain the chamber measurement temperature of the cell culture plate.

6. The method for measuring the temperature of a cell culture plate according to claim 5, characterized in that: The cell culture plate temperature measurement method further comprises: Performing environmental control on the incubator where the cell culture plate is located according to the temperature measured in the chamber; and / or, The measured temperature of the chamber is visually displayed.

7. A training system for a cell culture plate temperature measurement model, characterized in that: The training system comprises: An acquisition module, used to acquire temperature time series data corresponding to the four outer walls of the cell culture plate and actual chamber temperature data of a designated position of the cell culture plate; wherein the designated position of the cell culture plate includes positions of different areas of the cell culture plate; A normalization module, used for normalizing the temperature time series data to obtain normalized temperature time series data; An iteration module, used for inputting the normalized temperature time series data into a cell culture plate temperature measurement model for iterative training to obtain chamber temperature series prediction data corresponding to the normalized temperature time series data; A denormalization module, used for performing denormalization processing on the chamber temperature sequence prediction data to obtain chamber temperature prediction data; The adjustment module is used to adjust the parameters of the cell culture plate temperature measurement model according to the chamber temperature prediction data and the corresponding chamber temperature actual data, until the error of the cell culture plate temperature measurement model reaches a preset accuracy requirement.

8. A deployment system for a cell culture plate temperature measurement model, characterized in that: The deployment system comprises: A model acquisition module, used to acquire a cell culture plate temperature measurement model trained by the cell culture plate temperature measurement model training method according to any one of claims 1 to 3; A deployment module is used to quantize the trained cell culture plate temperature measurement model, convert it from a floating-point model to a fixed-point integer model, and obtain a quantized cell culture plate temperature measurement model; optimize the quantized cell culture plate temperature measurement model, adjust its memory usage and computing resource allocation, and obtain an optimized cell culture plate temperature measurement model; perform a benchmark test on the optimized cell culture plate temperature measurement model to evaluate the performance of the optimized cell culture plate temperature measurement model; convert the optimized cell culture plate temperature measurement model into a target code; and deploy the target code to a chip.

9. A cell culture plate temperature measurement system, characterized in that: The cell culture plate temperature measurement system comprises: A temperature acquisition module is used to obtain actual temperature data corresponding to the four outer walls of the cell culture plate; A normalization processing module, used for performing normalization processing on the actual temperature data to obtain actual temperature time series data; A prediction module, used for inputting the actual temperature time series data into a trained cell culture plate temperature measurement model to obtain a corresponding model prediction result; wherein the trained cell culture plate temperature measurement model is trained using the training method for the cell culture plate temperature measurement model according to any one of claims 1 to 3; The denormalization processing module is used to perform denormalization processing on the model prediction result to obtain the chamber measurement temperature of the cell culture plate.

10. The cell culture plate temperature measurement system according to claim 9, characterized in that: The cell culture plate temperature measurement system also includes: an environmental control module, used to perform environmental control on the incubator where the cell culture plate is located according to the temperature measured in the chamber; and / or, A visualization module is used to visualize the measured temperature of the chamber.

Images