Training methods, devices, systems, and electronic devices for cell detection models

Abstract

This application relates to a training method, apparatus, system, and electronic device for a cell detection model. The training method includes: acquiring a training calcium flow video targeting target tissue cells; extracting calcium signal features from the training calcium flow video to obtain calcium oscillation feature information; and inputting the calcium oscillation feature information into an initial random forest model for training to generate a target cell detection model. This target cell detection model is used to obtain feature prediction results for the target tissue cells. This application solves the problem of low efficiency in brain tumor cell detection and achieves a highly efficient and accurate method for detecting brain tumor cells.

Description

Technical Field

[0001] This application relates to the field of brain tumor cell detection technology, and in particular to training methods, devices, systems and electronic devices for cell detection models. Background Technology

[0002] Brain tumors are among the deadliest diseases today, resistant to conventional treatment strategies due to their unique cellular characteristics. Currently, the main diagnostic methods for tumors include tissue biopsy and in vivo imaging. Tissue biopsy, or tissue section examination, involves taking part or all of a sample from a patient's functionally damaged tissue to create a slide, which is then analyzed by a pathologist to determine the condition. In vivo imaging, compared to tissue biopsy, is less invasive and diagnoses the presence of tumors based on their imaging characteristics. However, among these technologies, in vivo imaging still requires further experiments on brain tumor cells, resulting in higher time and labor costs for brain tumor cell detection, thus reducing its efficiency.

[0003] Currently, no effective solution has been proposed to address the low efficiency of brain tumor cell detection in related technologies. Summary of the Invention

[0004] This application provides a method, apparatus, system, and electronic device for training a brain tumor cell detection model, in order to at least solve the problem of low efficiency in brain tumor cell detection in related technologies.

[0005] In a first aspect, embodiments of this application provide a method for training a brain tumor cell detection model, the method comprising:

[0006] Acquire training calcium flow videos targeting specific tissue cells;

[0007] Calcium signal features are extracted from the training calcium stream video to obtain calcium oscillation feature information;

[0008] The calcium oscillation feature information is input into an initial random forest model for training to generate a target cell detection model; wherein, the target cell detection model is used to obtain feature prediction results for the target tissue cells.

[0009] In some embodiments, the step of inputting the calcium oscillation feature information into an initial random forest model for training to generate a target cell detection model includes:

[0010] Training data is obtained based on the calcium oscillation feature information; wherein, the training data carries cell feature labels;

[0011] The training data is input into the initial random forest model for training to obtain feature prediction results for the target tissue cells;

[0012] The loss function is calculated based on the feature prediction results and the cell feature labels, and the initial random forest model is iteratively trained based on the loss function results to obtain the target cell detection model.

[0013] In some embodiments, the training data includes multiple datasets to be trained, and the step of inputting the training data into an initial random forest model for training to obtain feature prediction results for the target tissue cells includes:

[0014] Obtain the calcium signal sample features in each of the training datasets, and determine random partitioning features from the calcium signal sample features;

[0015] Based on the random partitioning feature, the dataset to be trained is input into multiple decision trees in the initial random forest model for training, to obtain sub-prediction results of multiple decision trees, and the feature prediction result is obtained based on all the sub-prediction results.

[0016] In some embodiments, the step of iteratively training the initial random forest model based on the loss function results to obtain the target cell detection model includes:

[0017] The verification dataset is obtained based on the calcium oscillation feature information;

[0018] Based on the loss function results, the initial random forest model is iteratively trained to obtain the cell detection model to be validated.

[0019] The validation dataset is input into the cell detection model to be validated to obtain the validation results. Based on the validation results, the cell detection model to be validated is optimized to obtain the target cell detection model.

[0020] In some embodiments, the step of extracting calcium signal features from the training calcium stream video to obtain calcium oscillation feature information includes:

[0021] Regions of interest are extracted from brain tumor cells in the training calcium flow video to obtain target cell regions, and calcium signal features are extracted based on the target cell regions to obtain feature extraction information;

[0022] The extracted features are filtered and preprocessed to obtain the calcium oscillation feature information.

[0023] In some embodiments, after generating the target cell detection model, the method further includes:

[0024] Obtain the calcium flow information of the target object;

[0025] The calcium flow information to be tested is input into the target cell detection model, and the cell detection results for the target object are output.

[0026] In some embodiments, the cell detection results include cell proliferation rate; after outputting the cell detection results for the object to be detected, the method further includes:

[0027] Obtain the mapping relationship between the preset cell proliferation range and the tumor malignancy level;

[0028] Based on the cell detection results, the actual cell proliferation range that matches the cell proliferation rate within the preset cell proliferation range is determined, and based on the mapping relationship, the actual tumor level corresponding to the actual cell proliferation range within the tumor malignancy level is determined;

[0029] The brain tumor prediction result of the subject to be tested is determined based on the actual tumor grade.

[0030] Secondly, embodiments of this application provide a training device for a cell detection model, the device comprising an acquisition module, a feature extraction module, and a training module;

[0031] The acquisition module is used to acquire training calcium flow videos for target tissue cells;

[0032] The feature extraction module is used to extract calcium signal features from the training calcium stream video to obtain calcium oscillation feature information;

[0033] The training module is used to input the calcium oscillation feature information into the initial random forest model for training to generate a target cell detection model; wherein, the target cell detection model is used to obtain feature prediction results for the target tissue cells.

[0034] Thirdly, embodiments of this application provide a training system for a cell detection model, the system including a terminal device and a training apparatus for the cell detection model as described in the second aspect above.

[0035] Fourthly, embodiments of this application provide an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the training method for the cell detection model as described in the first aspect above.

[0036] Compared to related technologies, the cell detection model training method, apparatus, system, and electronic device provided in this application acquire training calcium flow videos for target tissue cells; extract calcium signal features from the training calcium flow videos to obtain calcium oscillation feature information; input the calcium oscillation feature information into an initial random forest model for training to generate a target cell detection model; wherein, the target cell detection model is used to obtain feature prediction results for the target tissue cells, solving the problem of low efficiency in brain tumor cell detection and realizing an efficient and accurate brain tumor cell detection method.

[0037] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0038] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0039] Figure 1 This is an application environment diagram of a cell detection model according to an embodiment of this application;

[0040] Figure 2 This is a flowchart of a training method for a cell detection model according to an embodiment of this application;

[0041] Figure 3 This is a flowchart of another method for training a cell detection model according to an embodiment of this application;

[0042] Figure 4 This is a schematic diagram of a random forest model according to an embodiment of this application;

[0043] Figure 5 This is a structural block diagram of the training of a cell detection model according to an embodiment of this application;

[0044] Figure 6 This is a structural diagram of the internal structure of a computer device according to an embodiment of this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application. Furthermore, it is understood that although the efforts made in such a development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, modifications to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0046] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0047] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application means two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The terms “first,” “second,” “third,” etc., used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0048] The training method for the cell detection model provided in this application can be applied to, for example... Figure 1The application environment shown includes a terminal device 102 and a server 104. The terminal device 102 can communicate with the server 104 via a network. A data storage system can store the data that the terminal device 102 needs to process. The data storage system can be integrated on the server 104 or located in the cloud or on other network servers. The server 104 acquires training calcium flow videos of the target tissue cells from the terminal device 102, extracts calcium signal features from the training calcium flow videos to obtain calcium oscillation feature information; the server 104 inputs this calcium oscillation feature information into an initial random forest model for training to generate a target cell detection model; this target cell detection model is used to obtain feature prediction results for the target tissue cells. The terminal device 102 can be, but is not limited to, various smartphones, personal computers, laptops, and tablets, and the server device 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0049] This embodiment provides a training method for a brain tumor cell detection model. Figure 2 This is a flowchart of a training method for a brain tumor cell detection model according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps:

[0050] Step S210: Obtain training calcium flow video for target tissue cells.

[0051] The calcium flow video mentioned above was specifically designed for target tissue cells, namely, in vitro brain tumor cells. Specifically, the in vitro brain tumor cells were subjected to mechanical grinding, trypsin digestion, filtration, and centrifugation. The treated cells were then cultured in a medium containing 10% fetal bovine serum, 2.5 mmol / L L-glutamine, and 100 U / ml penicillin-streptomycin solution. Plasma cells were incubated at 37°C in a humid environment with 5% CO2, with the medium changed every one day. Cells were seeded onto sterile coverslips at a density of 4.5 × 10³ cells, and then supplemented with a specific level of chemotherapeutic agents, such as temozolomide (TMZ) or anlotinib hydrochloride (AL3818), according to experimental requirements, and incubated for another 48 hours. During this period, the drug is refreshed every 24 hours. For signal acquisition, a certain concentration of a calcium ion fluorescent probe, namely fluo-8 calcium indicator, is first added to the drug. After incubation for a period of time, the sample is placed under a fluorescence microscope to detect and record calcium oscillations. The sampling time is 5 minutes, thus obtaining the aforementioned training calcium flow video. Subsequently, the cell proliferation rate can be characterized using a cell proliferation assay kit, and the cell migration rate can be characterized using a cell scratch assay. It is evident that in this embodiment, only a small amount of fluo-8 calcium indicator is needed for incubation for 30-60 minutes to obtain the corresponding calcium oscillation signal, making the experiment simple, rapid, and low-cost.

[0052] Step S220: Extract calcium signal features from the training calcium flow video to obtain calcium oscillation feature information.

[0053] Specifically, in order to establish a calcium signal feature database, the features of the calcium signal of each cell are extracted from the brain tumor cells detected by the above-mentioned calcium flow video recognition to obtain the above-mentioned calcium oscillation feature information; for example, the calcium oscillation feature information includes at least the amplitude, frequency, duration, rise time, decay time, energy, and peak-to-peak interval of the calcium signal peak; it should be further noted that, in order to further improve the accuracy of cell detection, non-temporal features such as the center frequency of the calcium signal, the information entropy, the proportion of active cells, and the overall cell characteristics of overall cell synchronization can also be added.

[0054] Step S230: Input the calcium oscillation feature information into the initial random forest model for training to generate a target cell detection model; wherein, the target cell detection model is used to obtain feature prediction results for the target tissue cells.

[0055] The aforementioned target cell detection model indicates the mapping relationship between brain tumor tissue cells and their biological characteristics. Specifically, the model outputs feature detection results for the target tissue cells, including predicted cell proliferation rate and predicted cell migration rate. Furthermore, the target cell detection model can be trained using scikit-learn in Python 3.10. Model parameters are determined using a grid search method. To avoid overfitting, a 10-fold cross-validation method is used to construct the model. Each model is trained 10 times, and the correlation coefficient (R²) and root mean square error (RMSE) between predicted and actual values ​​are calculated to measure the final performance of the model. It should be noted that the target cell detection model generated by the random forest model in this embodiment performs better than neural network models trained using other regression models such as support vector machines and Lasso linear regression, thus improving the accuracy of predictive evaluation of the biological characteristics of brain tumor cells.

[0056] Compared to existing technologies that require multiple corresponding experiments to detect the migration and proliferation rates of brain tumor cells, the embodiments of this application, through steps S210 to S230, extract calcium signal features from the training calcium flow video to obtain calcium oscillation feature information, and input the calcium oscillation feature information into an initial random forest model for training to generate a target cell detection model. Thus, multiple biological characteristic parameters of brain tumor cells can be obtained through a simple calcium imaging experiment, eliminating the need for multiple experiments to detect cell biological characteristics. This effectively reduces the time and labor costs of brain tumor cell detection, solves the problem of low efficiency in brain tumor cell detection, and achieves a highly efficient and accurate method for detecting brain tumor cells.

[0057] In some embodiments, a method for training a brain tumor cell detection model is provided. Figure 3 This is a flowchart of another brain tumor cell detection method according to an embodiment of this application, such as... Figure 3 As shown, the process includes Figure 2 Steps S210 to S220 shown herein also include the following steps:

[0058] Step S310: Obtain training data based on the calcium oscillation feature information; wherein the training data carries cell feature labels. Further, to ensure comparability between the data and the model, z-score can be used to standardize the data, and training data extracted from the calcium signal feature database can be obtained.

[0059] Step S320: Input the training data into the initial random forest model for training to obtain the feature prediction results for the target tissue cells.

[0060] Step S330: Calculate the loss function result based on the feature prediction result and the cell feature label, and iteratively train the initial random forest model based on the loss function result to obtain the target cell detection model.

[0061] Specifically, the feature prediction results output by the model and the loss function results between the cell feature labels carried by the training data can be calculated using algorithms such as cross-entropy loss function. The gradient of the calculated loss function results is then backpropagated to the initial random forest model for iterative training until the number of iterations or the iteration time is reached, or the random forest model has converged, and a fully trained target cell detection model is obtained.

[0062] Through steps S310 to S330 above, the loss function result is calculated using the feature prediction results output by the initial random forest model and the cell feature labels carried by the training data. The loss function result is then used to iteratively train the model to generate an optimized target cell detection model, thereby effectively improving the accuracy of the target cell detection model in detecting the biological characteristics of the brain tumor cells to be tested.

[0063] In some embodiments, the training data includes multiple training datasets. The step of inputting the training calcium flow video into an initial random forest model for training to obtain feature prediction results for the training calcium flow video further includes the following steps:

[0064] Step S321: Obtain the features of calcium signal samples in each dataset to be trained, and determine random partitioning features from the features of the calcium signal samples.

[0065] Step S322: Based on the random partitioning feature, the dataset to be trained is input into multiple decision trees in the initial random forest model for training, to obtain multiple sub-prediction results of the decision trees, and the feature prediction result is obtained based on all the sub-prediction results.

[0066] In this model, the random forest regression model consists of multiple decision trees, and there is no correlation between the decision trees in the forest; the final output of the random forest model is determined by the collective decision of each decision tree in the forest. Specifically, Figure 4 This is a schematic diagram of a random forest model according to an embodiment of this application, as shown below. Figure 4As shown, in this embodiment, 11 features can be extracted from each original calcium signal sample, and these features are organized into an n×m training sample set, where n is the number of samples and m is the number of features. First, N samples are randomly selected from the sample set to obtain multiple new sub-training sets S1 to SN. Then, a CART regression tree is trained using these sub-training sets. During training, the splitting rule for each node is to first randomly select k features from all features, i.e., determine the aforementioned random splitting features, and then select the optimal splitting point from these k features to divide the left and right subtrees. Through this method, multiple classification and regression tree (CART) models can be obtained. The final prediction result of each CART regression tree is the mean of all leaf nodes for that sample point, and the final prediction result of the random forest is the mean of all CART regression tree prediction results. Furthermore, in this embodiment, to reduce the error caused by the imbalanced distribution of training samples, the samples are split 10 times using a random splitting method to obtain 10 prediction models. Finally, the mean of all 10 prediction models is taken as the final prediction result.

[0067] Through steps S321 to S322 above, the random forest model is used to train multiple decision trees based on random partitioning features to obtain sub-prediction results. Finally, the target cell detection model is trained and generated. This realizes the training method of constructing multiple sub-prediction models by random partitioning on the basis of general random forest model training. This can effectively reduce the error caused by the uneven distribution of training samples and further improve the accuracy of cell detection.

[0068] In some embodiments, the process of iteratively training the initial random forest model based on the loss function result to obtain the target cell detection model further includes the following steps:

[0069] Step S331: Obtain the verification dataset based on the calcium oscillation feature information.

[0070] Specifically, after extracting features of each cell's calcium signal from the training calcium flow video to establish a calcium signal feature database through the above steps, different sample datasets are obtained based on the calcium oscillation feature information in the calcium signal feature database according to a certain ratio; for example, all datasets can be randomly divided into training set, validation set, and test set in a 3:1:1 ratio to obtain the aforementioned validation dataset. Further, this validation dataset for model validation can include two parts: a validation set of patients in the same cohort and a clinical dataset of patients in a new cohort; where the same cohort refers to calcium oscillation data of patients in the same cohort as the training set, used to evaluate the training effect of the model; the new cohort refers to calcium oscillation data of patients not participating in training, used to evaluate the generalization performance of the model.

[0071] Step S332: Based on the loss function result, iteratively train the initial random forest model to obtain the cell detection model to be validated. Specifically, the gradient of the loss function result is backpropagated to the initial random forest model for iterative training, resulting in the optimized neural network model, which is the aforementioned cell detection model to be validated.

[0072] Step S333: Input the validation dataset into the cell detection model to be validated to obtain the validation result, and optimize the cell detection model to be validated based on the validation result to obtain the target cell detection model.

[0073] After obtaining the cell detection model to be validated, the obtained validation dataset can be input into the model. Based on the test results output by the model on the validation dataset, the hyperparameter settings of the model are modified to retrain the cell detection model using the adjusted hyperparameters, resulting in a hyperparameter-optimized target cell detection model. It should be noted that this implementation uses a random forest-based machine learning model, which provides significantly enhanced predictive performance compared to traditional linear models. Specifically, the determination coefficients for cell proliferation rate and cell migration rate obtained by evaluating the model's training effect using calcium oscillation data from patients in the same cohort as the training set are 0.93 and 0.81, respectively. The determination coefficients for cell proliferation rate and cell migration rate obtained by evaluating the model's generalization performance using calcium oscillation data from patients in a new cohort are 0.776 and 0.771, respectively.

[0074] Through steps S321 to S333 above, the cell detection model is evaluated and optimized based on the above validation dataset, thereby effectively avoiding the generalization ability of the cell detection model and further improving the accuracy of cell detection model training.

[0075] In some embodiments, the above-mentioned calcium signal feature extraction from the training calcium stream video to obtain calcium oscillation feature information further includes the following steps: extracting regions of interest (ROIs) from brain tumor cells in the training calcium stream video to obtain target cell regions, and extracting calcium signal features based on the target cell regions to obtain feature extraction information; performing filtering preprocessing on the feature extraction information to obtain the calcium oscillation feature information. Specifically, after obtaining the above-mentioned training calcium stream video, the maximum intensity projection algorithm can first be used to reduce each image stack in the video stream to a single frame to obtain processed calcium stream image frames; the cells in the above image frames are manually segmented into regions of interest to obtain target cell regions, and the features of calcium signals of each cell are extracted from the target cell region image to obtain the above-mentioned feature extraction information; next, algorithms such as adaptive iterative least squares and smoothing filtering can be used to preprocess the feature extraction dataset, including baseline removal, smoothing filtering, normalization, and feature extraction preprocessing, to correct the baseline and eliminate random noise.

[0076] Through the above embodiments, by extracting the region of interest (ROI) for brain tumor cells from the image frames in the training calcium flow video, subsequent processing of the image portion of the ROI can be performed, which can help improve the efficiency of model training. At the same time, data preprocessing such as baseline removal and smoothing filtering on the dataset further improves the efficiency and accuracy of model training.

[0077] In some embodiments, after generating the target cell detection model, the training method for the cell detection model further includes the following steps:

[0078] Step S241: Obtain the calcium flow information for the object to be tested.

[0079] Step S242: The calcium flow information to be tested is input into the target cell detection model, and the cell detection results for the target object are output. Specifically, after generating a fully trained target cell detection model through any of the above method embodiments, the target cell detection model can be deployed on hardware devices such as servers or computers. In actual application, the calcium flow information of the target object is input into the target cell detection model to obtain the cell detection results output by the model for the target object. This eliminates the need for multiple experiments to detect the biological characteristics of brain tumor cells, effectively reducing the time and labor costs of cell detection.

[0080] In some embodiments, the brain tumor cell detection results include cell proliferation rate; after outputting the brain tumor cell detection results of the target object, the method further includes the following steps:

[0081] Step S243: Obtain the mapping relationship between the preset cell proliferation range and the tumor malignancy level.

[0082] The mapping relationship between the preset cell proliferation range and the tumor malignancy level can be pre-set by the staff. It should be noted that in this embodiment, both cell proliferation range and cell migration range can be considered to determine the mapping relationship between cell biological characteristics and tumor malignancy level, thereby improving the accuracy of subsequent brain tumor prediction.

[0083] Step S244: Determine the actual cell proliferation range that matches the cell proliferation rate within the preset cell proliferation range, and based on the mapping relationship, determine the actual tumor level that corresponds to the actual cell proliferation range within the tumor malignancy level.

[0084] The aforementioned preset cell proliferation range can be pre-set by staff; for example, the cell proliferation rate can be divided into multiple intervals, that is, the cell proliferation rate can be divided into corresponding intervals according to high, medium and low levels. Then, based on the cell detection results output by the above model, it is determined whether the cell proliferation rate in the cell detection results falls into the interval of the above predicted cell proliferation range, thus obtaining the above actual proliferating cell range, and then determining the actual tumor grade based on the above mapping relationship.

[0085] Step S245: Determine the brain tumor prediction result for the subject to be tested based on the actual tumor level.

[0086] Through steps S243 to S245 above, the actual tumor grade is determined by the cell detection results output by the model, and then the brain tumor prediction result is determined. This allows the tumor biological characteristics predicted by the model to serve as the basis for tumor diagnosis and provide reference value for clinical tumor grading.

[0087] It should be noted that the steps shown in the above flowchart or the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be performed in a different order than that shown here. For example, Figure 2 Steps S210 to S220 can be performed before step S241, that is, the target cell detection model can be pre-trained and stored in a local or remote server before the calcium flow video to be tested is acquired; or, the model training process of steps S210 to S230 can be set to be performed after step S241 and before step S242, or the model training process of steps S210 to S230 can be set to be performed simultaneously with the step of acquiring the calcium flow information to be tested in step S241, etc., which will not be elaborated here.

[0088] This embodiment also provides a training device for a cell detection model. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the terms "module," "unit," "subunit," etc., can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0089] Figure 5 This is a structural block diagram of the training of a cell detection model according to an embodiment of this application, such as... Figure 5 As shown, the device includes: an acquisition module 52, a feature extraction module 54, and a training module 56; the acquisition module 52 is used to acquire training calcium flow videos for target tissue cells; the feature extraction module 54 is used to extract calcium signal features from the training calcium flow videos to obtain calcium oscillation feature information; the training module 56 is used to input the calcium oscillation feature information into an initial random forest model for training to generate a target cell detection model; wherein, the target cell detection model is used to acquire feature prediction results for the target tissue cells.

[0090] Through the above embodiments, the feature extraction module 54 extracts calcium signal features from the training calcium flow video to obtain calcium oscillation feature information. The training module 56 then inputs the calcium oscillation feature information into the initial random forest model for training to generate a target cell detection model. This allows for the acquisition of multiple biological characteristic parameters of brain tumor cells through a simple calcium imaging experiment, eliminating the need for multiple experiments to detect cell biological characteristics. This effectively reduces the time and labor costs of brain tumor cell detection, solves the problem of low efficiency in brain tumor cell detection, and realizes a highly efficient and accurate brain tumor cell detection device.

[0091] In some embodiments, the training module 56 is further configured to acquire training data based on the calcium oscillation feature information; the training data carries cell feature labels; the training module 56 inputs the training data into an initial random forest model for training to obtain feature prediction results for the target tissue cells; the training module 56 calculates a loss function result based on the feature prediction result and the cell feature labels, and iteratively trains the initial random forest model based on the loss function result to obtain the target cell detection model.

[0092] In some embodiments, the training data includes multiple training datasets. The training module 56 is further configured to obtain calcium signal sample features in each training dataset and determine random partitioning features from the calcium signal sample features. Based on the random partitioning features, the training module 56 inputs the training dataset into multiple decision trees in the initial random forest model for training, obtains multiple sub-prediction results of the decision trees, and obtains the feature prediction result based on all the sub-prediction results.

[0093] In some embodiments, the training module 56 is further used to obtain a validation dataset based on the training calcium flow video; the training module iteratively trains the initial random forest model based on the loss function result to obtain a cell detection model to be validated; the training module 56 inputs the validation dataset into the cell detection model to be validated to obtain the validation result, and optimizes the cell detection model to be validated based on the validation result to obtain the target cell detection model.

[0094] In some embodiments, the feature extraction module 54 is further configured to extract the region of interest of brain tumor cells in the calcium flow video to be tested, obtain the target cell region, and extract calcium signal features based on the target cell region to obtain the feature extraction information; the feature extraction module 54 is further configured to perform filtering preprocessing on the feature extraction information to obtain the calcium oscillation feature information.

[0095] In some embodiments, the training device for the cell detection model further includes a detection module; the detection module is used to acquire calcium flow information to be tested for the object to be tested; the detection module inputs the calcium flow information to be tested into the target cell detection model and outputs cell detection results for the object to be tested.

[0096] In some embodiments, the training device for the cell detection model further includes a prediction module; the prediction module is used to obtain a mapping relationship between a preset cell proliferation range and a tumor malignancy level; the prediction module is also used to determine the actual cell proliferation range that matches the cell proliferation rate within the preset cell proliferation range, and based on the mapping relationship, determine the actual tumor level corresponding to the actual cell proliferation range within the tumor malignancy level; the prediction module determines the brain tumor prediction result of the subject to be tested based on the actual tumor level.

[0097] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0098] This embodiment also provides a training system for a cell detection model, which includes a terminal device and a training device for the cell detection model as described in any of the above-described apparatus embodiments. The terminal device receives training calcium flow videos for target tissue cells and sends the training calcium flow videos to the training device for the cell detection model. The training device for the cell detection model can be deployed on a server, computer, or other device to execute the steps in any of the above-described method embodiments.

[0099] Through the above embodiments, calcium signal features are extracted from the training calcium flow video to obtain calcium oscillation feature information. This calcium oscillation feature information is then input into an initial random forest model for training to generate a target cell detection model. As a result, multiple biological characteristic parameters of brain tumor cells can be obtained through a simple calcium imaging experiment, eliminating the need for multiple experiments to detect cell biological characteristics. This effectively reduces the time and labor costs of brain tumor cell detection, solves the problem of low efficiency in brain tumor cell detection, and realizes a highly efficient and accurate brain tumor cell detection system.

[0100] In some embodiments, a computer device is provided, which may be a server. Figure 6 This is a structural diagram of the internal structure of a computer device according to an embodiment of this application, such as... Figure 6 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores target cell detection models. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the training method for the aforementioned cell detection model.

[0101] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0102] This embodiment also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0103] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0104] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0105] S1, acquire training calcium flow video for target tissue cells.

[0106] S2, extract calcium signal features from the training calcium flow video to obtain calcium oscillation feature information.

[0107] S3, input the calcium oscillation feature information into the initial random forest model for training to generate a target cell detection model; wherein, the target cell detection model is used to obtain feature prediction results for the target tissue cells.

[0108] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0109] Furthermore, in conjunction with the brain tumor cell detection method in the above embodiments, this application embodiment can provide a storage medium for implementation. This storage medium stores a computer program; when executed by a processor, the computer program implements the training method for any of the brain tumor cell detection models in the above embodiments.

[0110] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0111] Those skilled in the art should understand that the technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A training method for a cell detection model, characterized in that, The method includes: Acquire training calcium flow videos for target tissue cells, wherein the target tissue cells are brain tumor cells; the training calcium flow videos are obtained by the following method: brain tumor cells treated with temozolomide or anlotinib acid for 48 hours are incubated with fluo-8 calcium indicator for 30-60 minutes, and then a 5-minute calcium oscillation signal video is acquired using a fluorescence microscope. Calcium signal features are extracted from the training calcium flow video to obtain calcium oscillation feature information. This includes: compressing the video image stack to a single frame using the maximum intensity projection algorithm; manually segmenting the region of interest (ROI) of brain tumor cells from the single frame image to obtain the target cell region; extracting the features of the calcium signal for each cell in the target cell region image; and preprocessing the feature extraction dataset using an adaptive iterative least squares method and a smoothing filtering algorithm to obtain the calcium oscillation feature information. The calcium oscillation feature information includes the amplitude, frequency, duration, rise time, decay time, energy, and peak-to-peak interval of the calcium signal peak. A training sample set is constructed based on the calcium oscillation feature information, and after standardization, it is input into the initial random forest model for iterative training to generate a target cell detection model; wherein, the target cell detection model is used to obtain feature prediction results for the target tissue cells, and the feature prediction results include cell proliferation rate and cell migration rate.

2. The training method according to claim 1, characterized in that, The step of constructing a training sample set based on the calcium oscillation feature information, standardizing it, and then inputting it into an initial random forest model for iterative training to generate a target cell detection model includes: Training data is obtained based on the calcium oscillation feature information; wherein, the training data carries cell feature labels; The training data is input into the initial random forest model for training to obtain feature prediction results for the target tissue cells; The loss function is calculated based on the feature prediction results and the cell feature labels, and the initial random forest model is iteratively trained based on the loss function results to obtain the target cell detection model.

3. The training method according to claim 2, characterized in that, The training data includes multiple datasets to be trained. The training data is input into an initial random forest model for training to obtain feature prediction results for the target tissue cells, including: Obtain the calcium signal sample features in each of the training datasets, and determine random partitioning features from the calcium signal sample features; Based on the random partitioning feature, the dataset to be trained is input into multiple decision trees in the initial random forest model for training, to obtain sub-prediction results of multiple decision trees, and the feature prediction result is obtained based on all the sub-prediction results.

4. The training method according to claim 2, characterized in that, The step of iteratively training the initial random forest model based on the loss function results to obtain the target cell detection model includes: The verification dataset is obtained based on the calcium oscillation feature information; Based on the loss function results, the initial random forest model is iteratively trained to obtain the cell detection model to be validated. The validation dataset is input into the cell detection model to be validated to obtain the validation results. Based on the validation results, the cell detection model to be validated is optimized to obtain the target cell detection model.

5. The training method according to any one of claims 1 to 4, characterized in that, After generating the target cell detection model, the method further includes: Obtain the calcium flow information of the target object; The calcium flow information to be tested is input into the target cell detection model, and the cell detection results for the target object are output.

6. The training method according to claim 5, characterized in that, The cell detection results include cell proliferation rate; after outputting the cell detection results for the object to be detected, the method further includes: Obtain the mapping relationship between the preset cell proliferation range and the tumor malignancy level; Based on the cell detection results, the actual cell proliferation range that matches the cell proliferation rate within the preset cell proliferation range is determined, and based on the mapping relationship, the actual tumor level corresponding to the actual cell proliferation range within the tumor malignancy level is determined; The brain tumor prediction result of the subject to be tested is determined based on the actual tumor grade.

7. A training device for a cell detection model, characterized in that, The device includes an acquisition module, a feature extraction module, and a training module; The acquisition module is used to acquire training calcium flow videos for target tissue cells, wherein the target tissue cells are brain tumor cells; the training calcium flow videos are obtained by the following method: brain tumor cells treated with temozolomide or anlotinib acid for 48 hours are incubated with fluo-8 calcium indicator for 30-60 minutes, and then a 5-minute calcium oscillation signal video is acquired using a fluorescence microscope. The feature extraction module is used to extract calcium signal features from the training calcium stream video to obtain calcium oscillation feature information. This includes: compressing the video image stack to a single frame using the maximum intensity projection algorithm; manually segmenting the region of interest (ROI) of brain tumor cells from the single frame image to obtain the target cell region; extracting features of the calcium signal for each cell in the target cell region image; and preprocessing the feature-extracted dataset using an adaptive iterative least squares method and a smoothing filtering algorithm to obtain the calcium oscillation feature information. The calcium oscillation feature information includes the amplitude, frequency, duration, rise time, decay time, energy, and peak-to-peak interval of the calcium signal peak. The training module is used to construct a training sample set based on the calcium oscillation feature information, and after standardization, input it into the initial random forest model for training to generate a target cell detection model; wherein, the target cell detection model is used to obtain feature prediction results for the target tissue cells, and the feature prediction results include cell proliferation rate and cell migration rate.

8. A training system for a cell detection model, characterized in that, The system includes a terminal device and a training device for the cell detection model as described in claim 7.

9. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the training method of the cell detection model according to any one of claims 1 to 6.

Images