Deep learning-based stem cell differentiation monitoring system

Abstract

The invention relates to the technical field of biology, in particular to a stem cell differentiation monitoring system based on deep learning, which comprises a culture module, an environment regulation module, a data acquisition module, a differentiation prediction module and an analysis module, the environment adjusting module is used for adjusting the culture environment of the culture module, the data acquisition module is used for acquiring various parameters and images in the stem cell differentiation process, and the differentiation prediction module is used for predicting the stem cell differentiation result. And the analysis module is used for acquiring an accuracy index of differentiation prediction and optimizing the differentiation prediction module according to the accuracy index. In combination with the deep learning technology, stem cell differentiation is monitored, and the differentiation result of the stem cells after a period of time is predicted, so that a worker can take corresponding environment adjustment measures according to the prediction result, and the intelligent degree of the system is improved.

Description

Technical Field

[0001] The present invention relates to the field of biotechnology, and in particular to a stem cell differentiation monitoring system based on deep learning. Background Art

[0002] Stem cell technology is one of the most popular cutting-edge technologies in the biomedical field today. However, monitoring the differentiation process of stem cells is a challenging task, and traditional methods are often limited by resolution, accuracy, and efficiency. Deep learning, as one of the important technologies in the field of artificial intelligence, has powerful pattern recognition and learning capabilities. In the field of biotechnology, deep learning has been used to monitor stem cell differentiation.

[0003] For example, the prior art disclosed in CN111986802A discloses an auxiliary determination system and method for the pathological differentiation grade of lung adenocarcinoma, which relates to the field of deep learning technology and includes: an image acquisition module for acquiring digital pathological images of several early lung adenocarcinoma patients; an image annotation module for annotating each digital pathological image to obtain a digital pathological annotation image; a model training module for training to obtain a lung adenocarcinoma tissue growth pattern recognition model; an image prediction module for inputting the digital pathological image to be predicted into the lung adenocarcinoma tissue growth pattern recognition model to obtain the predicted lung adenocarcinoma tissue growth pattern corresponding to each lesion area; and a pathological differentiation module for calculating the tumor cell proportion of each predicted lung adenocarcinoma tissue growth pattern, so as to provide an auxiliary reference for doctors in determining the pathological differentiation grade of lung adenocarcinoma.

[0004] Another typical example is a deep learning-based iPSCs directional differentiation control system and control method disclosed in the prior art of CN117409867A. In the process of somatic cell induction and differentiation, the present invention counts the amount of stem cell apoptosis in real time to determine the effective differentiation amount of stem cells, and provides corresponding data support for the execution of the evaluation model. After the somatic cells are induced to differentiate to the required amount of stem cells, the biomechanical factors applied to them will be removed, and the effective differentiation amount of stem cells in the natural state will be counted to avoid errors in the output results of standard parameters caused by shortening the stem cell apoptosis cycle during the induction process of biomechanical factors. After that, the optimal biomechanical factor is determined based on the effective differentiation amount of somatic cells after the induction is completed.

[0005] Let's look at the deep learning-based neural stem cell differentiation direction prediction system and method disclosed in the prior art CN111666895A. The system combines the high-throughput processing capability of flow cytometry technology for cells, creatively uses experimental means to collect cell images of neural stem cell differentiation culture collected by panoramic flow cytometer, establishes a cell training data set for the three-line differentiation of neural stem cells, and then uses convolutional neural network for model training optimization.

[0006] At present, in the process of studying cell differentiation through deep learning, the existing technology is often unable to independently judge the accuracy of the prediction results, and is unable to optimize the model in time, which is prone to prediction errors and lags. In order to solve the common problems in this field, the present invention has been made. Summary of the invention

[0007] The purpose of the present invention is to propose a stem cell differentiation monitoring system based on deep learning to address the current deficiencies.

[0008] In order to overcome the shortcomings of the prior art, the present invention adopts the following technical solutions:

[0009] A stem cell differentiation monitoring system based on deep learning includes a culture module, an environmental regulation module, a data acquisition module, a differentiation prediction module and an analysis module. The culture module is used to culture stem cells and provide a place for stem cell differentiation. The environmental regulation module is used to regulate the culture environment of the culture module. The data acquisition module is used to collect various parameters and images of the culture module and the environmental regulation module during the stem cell differentiation process. The differentiation prediction module is used to predict the differentiation results of stem cells based on the data collected by the data acquisition module. The analysis module is used to obtain the accuracy index of differentiation prediction and optimize the differentiation prediction module based on the accuracy index.

[0010] Furthermore, the culture module includes a culture container, a culture medium supply unit, a lighting unit, and a disinfection unit. The culture container is used to accommodate stem cells and their culture medium, the culture medium supply unit is used to add culture medium to the culture container at a regular interval, the lighting unit is used to illuminate the culture container, and the disinfection unit is used to disinfect the culture container.

[0011] Furthermore, the environmental adjustment module includes a gas exchange unit, a temperature control unit, a pH adjustment unit and a data transmission unit. The gas exchange unit is used to adjust the concentration of various gases in the culture container, the temperature control unit is used to control the ambient temperature of the culture container, the pH adjustment unit is used to adjust the pH value of the culture medium in the culture container, and the data transmission unit is used to send the adjustment parameters of the environmental adjustment module to the data acquisition module.

[0012] Furthermore, the differentiation prediction module includes a data input unit, a differentiation stage identification unit and a differentiation result prediction unit. The data input unit is used to receive the data collected by the data acquisition module, the differentiation stage identification unit is used to identify the current differentiation stage or differentiation time of the stem cells according to the data received by the data input unit, and the differentiation result prediction unit is used to predict the future differentiation image of the stem cells through a prediction model according to the data received by the data input unit.

[0013] Furthermore, the analysis module includes an image comparison module, a calculation module and a prediction model optimization module. The image comparison module is used to compare cell differentiation images of different periods and obtain image similarity, the calculation module is used to calculate the accuracy index of differentiation prediction, and the prediction model optimization module is used to optimize the prediction model in the differentiation result prediction unit based on the various data obtained from the differentiation prediction module, the calculation module and the image comparison module.

[0014] Furthermore, the workflow of the differentiation monitoring system includes the following steps:

[0015] S1, the culture module is used to culture stem cells.

[0016] S2, the data acquisition module collects various parameters and images of stem cells during the differentiation process.

[0017] S3, the differentiation stage recognition unit recognizes the current differentiation stage of stem cells.

[0018] S4, the environment adjustment module adjusts the culture environment of some culture containers in the culture module.

[0019] S5, the differentiation result prediction unit predicts the differentiation image of each stem cell in the future.

[0020] S6, the analysis module optimizes the differentiation result prediction unit according to the acquired data.

[0021] Furthermore, the differentiation result prediction unit predicts the differentiation image of each stem cell in the future period of time, including the following steps:

[0022] S51, inputting the environmental parameters of the culture environment of each stem cell, the set time and the current differentiation image of each stem cell.

[0023] S52, the prediction model processes each input and obtains a predicted differentiation image after a set time.

[0024] Furthermore, the analysis module optimizes the differentiation result prediction unit according to the acquired data, including the following steps:

[0025] S61, the image comparison module compares the actual differentiation image and the predicted differentiation image after a set time, and obtains an image similarity index between the actual differentiation image of each stem cell and its corresponding predicted differentiation image.

[0026] S62, the calculation module calculates the secondary accuracy index corresponding to each actual differentiation image and the accuracy index of differentiation prediction based on the data collected by the data collection module and the actual differentiation image.

[0027] S63, determine whether the accuracy index of differentiation prediction is greater than 0. If not, the system reports an error and ends the optimization. Otherwise, the secondary accuracy index and image corresponding to each experimental group are sent to the prediction model optimization module; the secondary accuracy index is used to characterize the accuracy of the prediction model corresponding to each experimental group; the accuracy index of differentiation prediction is used to characterize the accuracy of the prediction model corresponding to all experimental groups; the larger the index, the higher the accuracy.

[0028] S64, the prediction model optimization module labels the images of each experimental group according to the secondary accuracy index, and sends the labeled images to the differentiation result prediction unit, which adds the labeled images to the training set.

[0029] The beneficial effects achieved by the present invention are:

[0030] 1. Combined with deep learning technology, the differentiation of stem cells can be monitored and the differentiation results of stem cells after a period of time can be predicted, which is conducive to the staff to take corresponding environmental adjustment measures according to the prediction results, thereby improving the intelligence level of the system.

[0031] 2. By setting secondary accuracy indicators, the system can autonomously judge the accuracy of different stem cell differentiation predictions. By setting accuracy indicators, it is helpful to judge the accuracy of the prediction of the differentiation result prediction unit and screen the results suitable as training sets. By marking the image with the indicator value and using it as a training set, the model can be optimized in time, which is conducive to continuously improving the model's capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate the same parts.

[0033] Figure 1 It is a schematic diagram of the structure of the present invention.

[0034] Figure 2 It is the work flow chart of the present invention.

[0035] Figure 3 A flow chart of optimizing the differentiation result prediction unit by the analysis module of the present invention.

[0036] Figure 4 This is a work flow chart of the second embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following is an explanation of the embodiments of the present invention through specific embodiments. Those skilled in the art can understand the advantages and effects of the present invention from the contents disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. In addition, the drawings of the present invention are only simple schematic illustrations and are not depicted according to actual sizes. It is stated in advance. The following embodiments will further explain the relevant technical contents of the present invention in detail, but the disclosed contents are not intended to limit the scope of protection of the present invention.

[0038] Embodiment 1: According to Figure 1 , Figure 2 and Figure 3 This embodiment provides a stem cell differentiation monitoring system based on deep learning, including a culture module, an environment adjustment module, a data acquisition module, a differentiation prediction module and an analysis module. The culture module is used to culture stem cells and provide a place for stem cell differentiation. The environment adjustment module is used to adjust the culture environment of the culture module. The data acquisition module is used to collect various parameters and images of the culture module and the environment adjustment module during the stem cell differentiation process. The differentiation prediction module is used to predict the differentiation results of stem cells according to the data collected by the data acquisition module. The analysis module is used to obtain the accuracy index of differentiation prediction and optimize the differentiation prediction module according to the accuracy index.

[0039] Furthermore, the culture module includes a culture container, a culture medium supply unit, a lighting unit, and a disinfection unit. The culture container is used to accommodate stem cells and their culture medium, the culture medium supply unit is used to add culture medium to the culture container at a regular interval, the lighting unit is used to illuminate the culture container, and the disinfection unit is used to disinfect the culture container.

[0040] Furthermore, the environmental adjustment module includes a gas exchange unit, a temperature control unit, a pH adjustment unit and a data transmission unit. The gas exchange unit is used to adjust the concentration of various gases in the culture container, the temperature control unit is used to control the ambient temperature of the culture container, the pH adjustment unit is used to adjust the pH value of the culture medium in the culture container, and the data transmission unit is used to send the adjustment parameters of the environmental adjustment module to the data acquisition module.

[0041] Specifically, the adjustment parameters include but are not limited to temperature, pH value and various gas concentrations.

[0042] Furthermore, the differentiation prediction module includes a data input unit, a differentiation stage identification unit and a differentiation result prediction unit. The data input unit is used to receive the data collected by the data acquisition module, the differentiation stage identification unit is used to identify the current differentiation stage or differentiation time of the stem cells according to the data received by the data input unit, and the differentiation result prediction unit is used to predict the future differentiation image of the stem cells through a prediction model according to the data received by the data input unit.

[0043] Specifically, the differentiation stage identification unit performs identification through a deep learning model, and the differentiation result prediction unit predicts the differentiation result through a prediction model. The above model can be trained by technicians in this field by inputting cell differentiation images of different time periods or directly using an existing model.

[0044] Furthermore, the analysis module includes an image comparison module, a calculation module and a prediction model optimization module. The image comparison module is used to compare cell differentiation images of different periods and obtain image similarity, the calculation module is used to calculate the accuracy index of differentiation prediction, and the prediction model optimization module is used to optimize the prediction model in the differentiation result prediction unit based on the various data obtained from the differentiation prediction module, the calculation module and the image comparison module.

[0045] Furthermore, the workflow of the differentiation monitoring system includes the following steps:

[0046] S1, the culture module is used to culture stem cells.

[0047] S2, the data acquisition module collects various parameters and images of stem cells during the differentiation process.

[0048] S3, the differentiation stage recognition unit recognizes the current differentiation stage of stem cells.

[0049] S4, the environment adjustment module adjusts the culture environment of some culture containers in the culture module.

[0050] S5, the differentiation result prediction unit predicts the differentiation image of each stem cell in the future.

[0051] S6, the analysis module optimizes the differentiation result prediction unit according to the acquired data.

[0052] Furthermore, the differentiation result prediction unit predicts the differentiation image of each stem cell in the future period of time, including the following steps:

[0053] S51, inputting the environmental parameters of the culture environment of each stem cell, the set time, and the current differentiation image of each stem cell.

[0054] Specifically, the environmental parameters include but are not limited to culture medium pH value, temperature value, oxygen concentration and carbon dioxide concentration; specifically, the set time is set by a person skilled in the art according to the desired prediction results.

[0055] S52, the prediction model processes each input and obtains a predicted differentiation image after a set time.

[0056] Furthermore, the analysis module optimizes the differentiation result prediction unit according to the acquired data, including the following steps:

[0057] S61, the image comparison module compares the actual differentiation image and the predicted differentiation image after a set time, and obtains an image similarity index between the actual differentiation image of each stem cell and its corresponding predicted differentiation image.

[0058] Specifically, the image similarity index of the actual differentiated image and its corresponding predicted differentiated image can be the SSIM value of the two images. The range of SSIM value is 1 to -1. The closer the SSIM value is to 1, the more similar the two images are. When the SSIM value is 1, it means that the two images are exactly the same. The method of obtaining the SSIM value belongs to the prior art and will not be elaborated here.

[0059] S62, the calculation module calculates the secondary accuracy index corresponding to each actual differentiation image and the accuracy index of differentiation prediction based on the data collected by the data collection module and the actual differentiation image. The secondary accuracy index is used to characterize the accuracy of the prediction model corresponding to each experimental group, which can characterize the accuracy of the prediction model to a certain extent, and is also used to select sample data of the training set. The accuracy index of differentiation prediction is used to characterize the accuracy of the prediction model corresponding to all experimental groups. The larger the index, the higher the accuracy.

[0060] The impact indicators can be obtained by following the steps below:

[0061] The actual differentiation image and its corresponding predicted differentiation image corresponding to each SSIM value greater than the SSIM threshold are regarded as an experimental group; the actual differentiation image and its corresponding predicted differentiation image corresponding to each SSIM value less than the SSIM threshold are regarded as an error group.

[0062] Specifically, the SSIM threshold is set by those skilled in the art between 0 and 1 according to the required prediction accuracy;

[0063] A predicted differentiation image of a time period near the predicted differentiation image of each experimental group is extracted from the differentiation result prediction unit as a secondary predicted differentiation image of the predicted differentiation image.

[0064] Obtain the SSIM values ​​of each secondary predicted differentiation image and its corresponding actual differentiation image of each experimental group, select the predicted differentiation image with the largest corresponding SSIM value between the secondary predicted differentiation image and the predicted differentiation image as the matching differentiation image, and add the secondary predicted differentiation image to the experimental group where its corresponding predicted differentiation image is located.

[0065] The secondary accuracy index for each experimental group was calculated according to the following formula:

[0066] ZB=k1*CS1+(1-k1)*CS2

[0067]

[0068] Among them, ZB is the secondary accuracy index. The larger the index value, the higher the accuracy. k1 is the ideal accuracy weight, CS1 is the ideal accuracy parameter, CS2 is the average accuracy parameter, and SSIM is the average accuracy parameter. best is the SSIM value of the matching differentiation image and the corresponding actual differentiation image, ssim is the SSIM threshold, e is a natural constant, t1 is the cell differentiation time corresponding to the predicted differentiation image, t2 is the cell differentiation time corresponding to the matched differentiation image, A is the number of secondary predicted differentiation images, and SSIM α is the SSIM value of the ath secondary predicted differentiation image and the corresponding actual differentiation image, t a is the differentiation time corresponding to the ath secondary predicted differentiation image; the differentiation time involved in this scheme is the time experienced by differentiation.

[0069] The accuracy index of differentiation prediction is calculated according to the following formula:

[0070]

[0071] Among them, zb is the accuracy index, C is the number of error groups, B is the number of experimental groups, SSIM c is the SSIM value corresponding to the cth error group, ZB b is the secondary accuracy index corresponding to the bth experimental group.

[0072] S63, determine whether the accuracy index of differentiation prediction is greater than 0. If not, the system reports an error and ends the optimization. Otherwise, the secondary accuracy index and image corresponding to each experimental group are sent to the prediction model optimization module.

[0073] S64, the prediction model optimization module labels the images of each experimental group according to the secondary accuracy index, and sends the labeled images to the differentiation result prediction unit, which adds the labeled images to the training set.

[0074] The beneficial effects of this program are: 1. Combined with deep learning technology, it monitors stem cell differentiation and predicts the differentiation results of stem cells after a period of time. It is reasonable for staff to take corresponding environmental adjustment measures based on the prediction results, thereby improving the intelligence of the system.

[0075] 2. By setting secondary accuracy indicators, the system can autonomously judge the accuracy of different stem cell differentiation predictions. By setting accuracy indicators, it is helpful to judge the accuracy of the prediction of the differentiation result prediction unit and screen the results suitable as training sets. By marking the image with the indicator value and using it as a training set, the model can be optimized in time, which is conducive to continuously improving the model's capabilities.

[0076] Embodiment 2: Figure 4 This embodiment should be understood to include all the features of any of the above embodiments, and further improve on the above embodiments, and also include a method for determining the influence of environmental parameters on the accuracy of differentiation prediction results, including the following steps:

[0077] STEP 1, place multiple stem cells at the same differentiation stage or the same differentiation time in different culture modules, and culture and differentiate the stem cells under different environments;

[0078] STEP2, predict the differentiation image of stem cells within a period of time through the differentiation result prediction unit.

[0079] STEP 3, obtain the actual differentiation images of stem cells over a period of time, and obtain the secondary accuracy index corresponding to each stem cell.

[0080] STEP4, use the following formula to obtain the impact indicators of different environmental parameters on differentiation prediction results.

[0081]

[0082] Among them, EF m is the impact index of environmental parameter m, T is the time from the beginning of prediction to the end of prediction, SSIM mt is the SSIM value of the actual differentiation image and the predicted differentiation image at time t corresponding to the environmental parameter m, ZB(t) is the secondary accuracy index based on the image data from the beginning of prediction to time t, CS m (t) is the parameter value of the environmental parameter m at time t, ZB(t+Δt) is the secondary accuracy indicator based on the image data from the start of prediction to time t+Δt, CS m (t+Δt) is the parameter value of the environmental parameter m at time t+Δt, and Δt is the interval time between each prediction.

[0083] Specifically, the larger the index value of the influencing index is, the greater the influence of the environmental parameter on the differentiation prediction result is.

[0084] The beneficial effects of this embodiment are as follows: by obtaining the impact indicators of different environmental parameters on the differentiation prediction results, it is helpful to understand the impact of different environments on the differentiation prediction results. By marking the corresponding images of the indicator values ​​and inputting them into the differentiation result prediction unit as a training set, it is helpful to improve the accuracy of the differentiation result prediction.

[0085] The above disclosed contents are only preferred feasible embodiments of the present invention, and do not limit the protection scope of the present invention. Therefore, all equivalent technical changes made by using the contents of the present invention specification and drawings are included in the protection scope of the present invention. In addition, the elements therein can be updated with the development of technology. The above units are only examples. Those skilled in the art can use corresponding units according to different designs according to actual needs when implementing this solution.

Claims

1. A stem cell differentiation monitoring system based on deep learning, characterized in that: It includes a culture module, an environment adjustment module, a data acquisition module, a differentiation prediction module and an analysis module. The culture module is used to culture stem cells and provide a place for stem cell differentiation. The environment adjustment module is used to adjust the culture environment of the culture module. The data acquisition module is used to collect various parameters and images of the culture module and the environment adjustment module during the stem cell differentiation process. The differentiation prediction module is used to predict the differentiation results of stem cells according to the data collected by the data acquisition module. The analysis module is used to obtain the accuracy index of differentiation prediction and optimize the differentiation prediction module according to the accuracy index.

2. A stem cell differentiation monitoring system based on deep learning according to claim 1, characterized in that: The culture module includes a culture container, a culture medium supply unit, a lighting unit, and a disinfection unit. The culture container is used to accommodate stem cells and their culture medium. The culture medium supply unit is used to add culture medium to the culture container at a regular interval. The lighting unit is used to illuminate the culture container. The disinfection unit is used to disinfect the culture container.

3. A stem cell differentiation monitoring system based on deep learning according to claim 2, characterized in that: The environment adjustment module includes a gas exchange unit, a temperature control unit, a pH adjustment unit and a data transmission unit. The gas exchange unit is used to adjust the concentration of various gases in the culture container. The temperature control unit is used to control the ambient temperature of the culture container. The pH adjustment unit is used to adjust the pH value of the culture medium in the culture container. The data transmission unit is used to send the adjustment parameters of the environment adjustment module to the data acquisition module.

4. A stem cell differentiation monitoring system based on deep learning according to claim 3, characterized in that: The differentiation prediction module includes a data input unit, a differentiation stage identification unit and a differentiation result prediction unit. The data input unit is used to receive the data collected by the data collection module. The differentiation stage identification unit is used to identify the current differentiation stage or differentiation time of the stem cells according to the data received by the data input unit. The differentiation result prediction unit is used to predict the future differentiation image of the stem cells through a prediction model according to the data received by the data input unit.

5. A stem cell differentiation monitoring system based on deep learning according to claim 4, characterized in that: The analysis module includes an image comparison module, a calculation module and a prediction model optimization module. The image comparison module is used to compare cell differentiation images at different periods and obtain image similarity. The calculation module is used to calculate the accuracy index of differentiation prediction. The prediction model optimization module is used to optimize the prediction model in the differentiation result prediction unit based on the data obtained from the differentiation prediction module, the calculation module and the image comparison module.

6. A stem cell differentiation monitoring system based on deep learning according to claim 5, characterized in that: The workflow of the differentiation monitoring system includes the following steps: S1, the culture module cultures stem cells; S2, data acquisition module collects various parameters and images of stem cells during differentiation; S3, the differentiation stage recognition unit recognizes the current differentiation stage of the stem cells; S4, the environment adjustment module adjusts the culture environment of some culture containers in the culture module; S5, the differentiation result prediction unit predicts the differentiation image of each stem cell in the future; S6, the analysis module optimizes the differentiation result prediction unit according to the acquired data.

7. A stem cell differentiation monitoring system based on deep learning according to claim 6, characterized in that: The differentiation result prediction unit predicts the differentiation image of each stem cell in the future, including the following steps: S51, inputting the environmental parameters of the culture environment of each stem cell, the set time, and the current differentiation image of each stem cell; S52, the prediction model processes each input and obtains a predicted differentiation image after a set time.

8. The stem cell differentiation monitoring system based on deep learning according to claim 7, characterized in that: The analysis module optimizes the differentiation result prediction unit according to the acquired data, including the following steps: S61, an image comparison module compares the actual differentiation image and the predicted differentiation image after a set time, and obtains an image similarity index between the actual differentiation image of each stem cell and its corresponding predicted differentiation image; S62, the calculation module calculates the secondary accuracy index corresponding to each actual differentiation image and the accuracy index of differentiation prediction according to the data collected by the data collection module and the actual differentiation image; the secondary accuracy index is used to characterize the accuracy of the prediction model corresponding to each experimental group; the accuracy index of differentiation prediction is used to characterize the accuracy of the prediction model corresponding to all experimental groups; the larger the index, the higher the accuracy; S63, determining whether the accuracy index of differentiation prediction is greater than 0, if not, the system reports an error and ends the optimization, otherwise, the secondary accuracy index and image corresponding to each experimental group are sent to the prediction model optimization module; S64, the prediction model optimization module labels the images of each experimental group according to the secondary accuracy index, and sends the labeled images to the differentiation result prediction unit, which adds the labeled images to the training set.

Images