Methods, apparatuses, devices, and media for predicting immune cell activation status

Abstract

A method for predicting an immune cell activation state is provided, comprising: obtaining a plurality of single immune cell images and a plurality of multiple immune cell images, each immune cell object in the plurality of single immune cell images and the plurality of multiple immune cell images being selected from a set of immune cell objects subjected to a same perturbation condition; processing the plurality of single immune cell images by using a first activation state prediction model to obtain a plurality of first activation state prediction values corresponding to the plurality of single immune cell images respectively, which are output by the first activation state prediction model; processing the plurality of multiple immune cell images by using a second activation state prediction model to obtain a plurality of second activation state prediction values corresponding to the plurality of multiple immune cell images respectively, which are output by the second activation state prediction model; and determining an activation state of the set of immune cell objects based on the plurality of first activation state prediction values and the plurality of second activation state prediction values.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of cell activation state determination, and in particular to a method, device, electronic device, computer readable storage medium and computer program product for predicting an immune cell activation state. BACKGROUND

[0002] At present, the biological fluorescence brightness value measured by an enzyme label instrument is usually used as a standard for determining the immune cell activation state. In a biological experiment for determining the immune cell activation state by using the biological fluorescence brightness measured by an enzyme label instrument, the immune cells need to contain or introduce a luciferase gene, and the experimental conditions are relatively complicated. In addition, in the measurement experiment of biological fluorescence brightness, professional experimenters need to operate, and the actual measured biological fluorescence brightness has errors due to experimental factors and human operation factors, and a large amount of manpower and material resources are consumed, which leads to that large-scale experimental verification cannot be performed in the task of screening drugs, and the experimental scale and efficiency of drug screening are affected. SUMMARY

[0003] It would be advantageous to provide a mechanism that alleviates, mitigates or even eliminates one or more of the above problems.

[0004] According to an aspect of the present disclosure, a method for predicting an immune cell activation state is provided, including: obtaining a plurality of single immune cell images and a plurality of multiple immune cell images, wherein each single immune cell image in the plurality of single immune cell images contains a single immune cell object, each multiple immune cell image in the plurality of multiple immune cell images contains a plurality of immune cell objects, and each immune cell object in the plurality of single immune cell images and the plurality of multiple immune cell images is selected from a set of immune cell objects subjected to a same perturbation condition; processing the plurality of single immune cell images by using a first activation state prediction model to obtain a plurality of first activation state prediction values corresponding to the plurality of single immune cell images respectively and output by the first activation state prediction model; processing the plurality of multiple immune cell images by using a second activation state prediction model to obtain a plurality of second activation state prediction values corresponding to the plurality of multiple immune cell images respectively and output by the second activation state prediction model; and determining an activation state of the set of immune cell objects based on the plurality of first activation state prediction values and the plurality of second activation state prediction values.

[0005] According to another aspect of the present disclosure, a method for predicting immune cell activation ability of a compound is provided, comprising: subjecting an immune cell object set to a perturbation condition, the perturbation condition comprising exposure to a predetermined concentration of the compound for a predetermined duration; taking images of the immune cell object set using a high-content cell imaging system to obtain a plurality of high-content images; generating a plurality of single immune cell images and a plurality of multi-immune cell images from the plurality of high-content images; processing the plurality of single immune cell images using a first activation state prediction model to obtain a plurality of first activation state prediction values corresponding to the plurality of single immune cell images respectively as output by the first activation state prediction model; processing the plurality of multi-immune cell images using a second activation state prediction model to obtain a plurality of second activation state prediction values corresponding to the plurality of multi-immune cell images respectively as output by the second activation state prediction model; and determining an activation state of the immune cell object set based on the plurality of first activation state prediction values and the plurality of second activation state prediction values, wherein the activation state indicates the immune cell activation ability of the compound.

[0006] According to another aspect of the present disclosure, an apparatus for predicting immune cell activation state is provided, comprising: a first module configured to obtain a plurality of single immune cell images and a plurality of multi-immune cell images, wherein each single immune cell image in the plurality of single immune cell images contains a single immune cell object, each multi-immune cell image in the plurality of multi-immune cell images contains a plurality of immune cell objects, and each immune cell object in the plurality of single immune cell images and the plurality of multi-immune cell images is selected from an immune cell object set subjected to a same perturbation condition; a second module configured to process the plurality of single immune cell images using a first activation state prediction model to obtain a plurality of first activation state prediction values corresponding to the plurality of single immune cell images respectively as output by the first activation state prediction model; a third module configured to process the plurality of multi-immune cell images using a second activation state prediction model to obtain a plurality of second activation state prediction values corresponding to the plurality of multi-immune cell images respectively as output by the second activation state prediction model; and a fourth module configured to determine an activation state of the immune cell object set based on the plurality of first activation state prediction values and the plurality of second activation state prediction values.

[0007] According to another aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory storing instructions executable by the processor, the instructions, when executed by the processor, causing the processor to perform the above method for predicting immune cell activation state.

[0008] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing instructions is provided, the instructions, when executed by a processor, causing the processor to perform the above method for predicting immune cell activation state.

[0009] According to another aspect of the present disclosure, there is provided a computer program product comprising instructions which, when executed by a processor, cause the processor to perform the method for predicting an immune cell activation state described above.

[0010] These and other aspects of the present disclosure will be apparent from the embodiments described below and will be elucidated with reference to those embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0011] In the following description of example embodiments in conjunction with the attached drawings, further details, features and advantages of the present disclosure are disclosed, in which:

[0012] Figure 1 is a flowchart of an example process of acquiring a plurality of single immune cell images in a method for predicting an immune cell activation state according to embodiments of the present disclosure;

[0013] Figure 2 is an example image of a color image according to embodiments of the present disclosure;

[0014] Figure 3 is an example image of a single immune image according to embodiments of the present disclosure;

[0015] Figure 4 is a flowchart of an example process of acquiring a plurality of multi immune cell images in a method for predicting an immune cell activation state according to embodiments of the present disclosure;

[0016] Figure 5 is an example image of a multi immune image according to embodiments of the present disclosure;

[0017] Figure 6 is a flowchart of a method for predicting an immune cell activation state according to embodiments of the present disclosure;

[0018] Figure 7 is a flowchart of an example process of determining an activation state of a set of immune cell objects in a method for predicting an immune cell activation state according to embodiments of the present disclosure; Figure 6

[0019] Figure 8 is a flowchart of a method for predicting an immune cell activation ability of a compound according to embodiments of the present disclosure;

[0020] Figure 9 is a block diagram of an apparatus for predicting an immune cell activation state according to embodiments of the present disclosure; and

[0021] Figure 10 is a block diagram of an electronic device for predicting an immune cell activation state according to embodiments of the present disclosure. DETAILED DESCRIPTION

[0022] ​In the present disclosure, the terms "first", "second", etc. are used to describe various elements only for the purpose of distinguishing one element from another, and the terms are not intended to imply a relative position, a temporal sequence, or a hierarchy of importance of the elements. In some examples, a first element and a second element can refer to the same example of an element, and in some cases, they can also refer to different examples of the element based on the context of description.

[0023] The terminology used in the description of the various described examples in the present disclosure is only for the purpose of describing particular ones of the examples and is not intended to be limiting. Unless specifically defined otherwise, an element that is a single in number can also be implemented in plural in number. As used herein, the term "plurality" means two or more, and the term "based on" shall be construed as "based at least in part on." Furthermore, the terms "and / or" and "at least one of" encompass any and all possible combinations of the listed items.

[0024] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and / or the present specification and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0025] As used herein, the term "immune cell" or "immune cell object" is commonly known as a leukocyte object, including lymphocytes and various phagocytes, etc., and also specifically refers to lymphocytes and the like that can recognize antigens and produce specific immune responses. The term "high content cell imaging system" refers to a high-resolution microscope imaging system for taking images of cells. Accordingly, the term "high content image" refers to a microscopic image acquired using such a microscope imaging system

[0026] In the related art, the biofluorescence brightness value measured by an enzyme label meter is usually used as the standard for determining the activation state of immune cells. This method includes culturing immune cells and detecting luciferase activity. Culturing immune cells includes: amplifying Jurkat activated T cell nuclear factor-fluorescent reporter cells (Jurkat NFAT-Lucia Reporter Cell) to a certain order of magnitude. After the immune cells are cultured in the well plate for a certain period of time, an immune stimulating factor (a compound with a predetermined concentration) is added to the well with the Jurkat activated T cell nuclear factor-fluorescent reporter cells and treated for a certain period of time. Detecting luciferase activity includes: after the experimental treatment time is met, the plate used to measure luciferase activity is taken out, the reagent used to detect luciferase activity is added to each well, and it is placed in the dark for a certain period of time. The value read by the enzyme label meter is used as the determination result of the activation state intensity of the immune cells.

[0027] Figure 1 is a flowchart of an example process of acquiring a plurality of single immune cell images in a method for predicting an activation state of immune cells according to embodiments of the present disclosure. As shown in Figure 1 , the method 100 includes steps 110-130.

[0028] In step 110, a plurality of high-content images taken by a high-content cell imaging system of a set of immune cells is acquired. Each high-content image in the plurality of high-content images contains a plurality of immune cell objects and has N color channels. Each color channel corresponds to a type of organelle or cellular component. N is an integer greater than 3. Organelles are microstructures or microorgans with certain morphology and function scattered in the cytoplasm, such as mitochondria, endoplasmic reticulum, centrosome, chloroplast, Golgi body, ribosome. Cellular components are various chemical components that make up cells, such as myoglobin.

[0029] In one example, the set of immune cell objects can be subjected to the same perturbation condition and undergo fluorescent staining treatment. In one example, the perturbation condition can be exposure to a predetermined compound at a predetermined concentration for a predetermined duration. In one example, the fluorescent staining treatment can include staining each organelle and cellular component with a different color dye.

[0030] In one example, a plurality of high-content images taken by a high-content cell imaging system of the set of immune cell objects subjected to the same perturbation condition and undergoing fluorescent staining treatment is acquired.

[0031] In one example, the number of color channels N can be 6. Each color channel can be used to observe an organelle or cellular component, such as 6 color channels that can be used to observe mitochondria, endoplasmic reticulum, Golgi body, ribonucleic acid, nucleus, and actin, respectively.

[0032] At step 120, each of the plurality of high-content images is pre-processed to generate a plurality of color images respectively corresponding to the plurality of high-content images.

[0033] In one example, the pixel value range of the plurality of high-content images can be 0~65535, and the color channel is 6. The plurality of high-content images need to be pre-processed to be converted into color images with pixel value range of 0~255 and channel number of 3, so as to facilitate direct observation and subsequent algorithm analysis. An example of such color images is shown in FIG. 1 (although for the convenience of illustration, it is shown as a grayscale image). Figure 2

[0034] In one example, the pre-processing can include normalizing the high-content images. For example, the pixels of the high-content images can be subjected to maximum-minimum normalization, so that each pixel has a value range of [0, 1], and then each pixel is multiplied by 255, so that the pixel value range of the normalized high-content images is 0~255.

[0035] Further, the pre-processing can also include merging the N color channels of the normalized high-content images into red-green-blue color channels. For example, the N color channels can include a red channel, a green channel, a blue channel, a red-green channel, a red-blue channel, and a green-blue channel. In one example, the average value of the red components in the N color channels is calculated as the merged red channel. The average value of the green components in the N color channels is calculated as the merged green channel. The average value of the blue components in the N color channels is calculated as the merged blue channel.

[0036] In the example, the merged red channel is R, the merged green channel is G, and the merged blue channel is B, R=(R1+R2+R3) / 3, G=(G1+G2+G3) / 3, B=(B1+B2+B3) / 3. Wherein R1 is the red component in the red channel, R2 is the red component in the red-green channel, and R3 is the red component in the red-blue channel. G1 is the green component in the green channel, G2 is the green component in the red-green channel, and G3 is the green component in the green-blue channel. B1 is the blue component in the blue channel, B2 is the blue component in the green-blue channel, and B3 is the blue component in the red-blue channel.

[0037] Further, before normalizing the high-content images, the pre-processing can also include median filtering the high-content images to remove a small amount of Gaussian noise points present in the high-content images. While filtering out noise, the edge information is protected, so that the high-content images are smoother. Median filtering is a nonlinear smoothing technique that sets the gray value of each pixel point to the median value of all pixel points in a certain neighborhood window.​

[0038] Further, before normalizing the high-content images, the preprocessing can further include removing over-exposed pixels in the high-content images. An over-exposed pixel has a brightness value exceeding a brightness threshold. In one example, the brightness threshold can be set according to an empirical value. For example, in a case where the pixel value ranges between 0 and 255, the brightness threshold is taken as 243. In one example, the pixels in the high-content images can be sorted according to the brightness value from high to low, and then the pixels of a preset proportion (e.g., the top 3‰ of brightness values) in the high-content images are removed as over-exposed pixels.

[0039] In step 130, a plurality of single immune cell images are segmented from the plurality of color images.

[0040] In one example, the color images can be first subjected to target object (immune cell object) detection using, for example, a target detection algorithm YOLO (You Only Look Once) to obtain a target detection region. Then the target detection region is subjected to semantic segmentation using, for example, a fully convolutional neural network (FCN) to obtain a target pixel region. The target pixel region is then mapped to the color image to obtain a mapping image. Finally, the mapping image is subjected to segmentation processing to obtain a plurality of single immune cell images.

[0041] In one example, a pixel-based segmentation algorithm can also be used to segment a plurality of single immune cell images from the plurality of color images. For example, an instance segmentation algorithm, such as a Mask R-CNN model, can be first used to perform multi-classification of immune cell object pixel points on the plurality of color images, with each cell object being a class, to obtain a plurality of target pixel regions of immune cell objects. Each target pixel region is then mapped to the color image to obtain a mapping image. Finally, the mapping image is subjected to segmentation processing to obtain a plurality of single immune cell images.

[0042] In one example, a contour-based segmentation algorithm can also be utilized to segment multiple single immune cell images from the multiple color images. In one example, an instance segmentation algorithm, such as a Deep Snake model, can first be utilized to detect immune cell object edge points from the multiple color images to obtain immune cell object edge points. The immune cell object edge points can then be sequentially connected to obtain immune cell object contours. The immune cell object contours can then be mapped to the color images to obtain mapped images. Finally, the mapped images can be segmented to obtain the multiple single immune cell images. Since the activation state of an immune cell object is strongly correlated with the morphology of the immune cell object, the contour-based instance segmentation algorithm can more accurately segment the immune cell object contours and preserve more information about the morphology of the immune cell object. An example of such a single immune cell object is shown in FIG. 4B (although it is shown as a grayscale image for ease of illustration). Figure 3

[0043] Figure 4 is a flowchart of an example process 400 of obtaining multiple multi-immune cell images in a method for predicting an activation state of immune cells according to embodiments of the present disclosure. As shown in Figure 4 Process 400 includes steps 410-450.

[0044] At step 410, corresponding foreground regions are extracted from the multiple single immune cell images, each foreground region being defined by a contour of a corresponding single immune cell object.

[0045] In one example, the foreground region (single immune cell object body region) is extracted from the corresponding single immune cell image according to the contour of the single immune cell object obtained from step 130.

[0046] At step 420, single immune cell images having a maximum aspect ratio of a minimum bounding rectangle of a foreground region greater than an aspect ratio threshold are filtered from the multiple single immune cell images to obtain multiple selected single immune cell images.

[0047] In one example, the maximum aspect ratio of the minimum bounding rectangle of the foreground region in each single immune cell image is first calculated, i.e., the ratio of the longer side to the shorter side of the minimum bounding rectangle. Single immune cell images having a maximum aspect ratio greater than an aspect ratio threshold (e.g., 1.25) are then filtered. Since the multiple single immune cell images can include images containing adherent cell objects, this example can filter images containing adherent cell objects.

[0048] At step 430, the multiple selected single immune cell images are scaled to the same size.

[0049] ​In one example, to keep the image aspect ratio unchanged, the longest side of each of the selected single immune cell images can be uniformly scaled to 72 pixels first. Then the shortest side of the single immune cell image can also be uniformly scaled to 72 pixels by filling the image with pixels (using the same pixel value as the background region of the image), resulting in an image size of 72 72 pixels.

[0050] In one example, after scaling the selected single immune cell images, a preset number of pixels (using the same pixel value as the background region of the image) can be filled at the edges of the selected single immune cell images. For example, 4 pixels are filled at the edges of the image, filling the image from an image size of 72 72 pixels to an image size of 80 80 pixels.

[0051] In one example, before scaling the selected single immune cell images, the foreground regions of a first number of single immune cell images in the selected single immune cell images can also be rotated to obtain a plurality of rotated single immune cell images. In an example, to keep the integrity of the image information, the first number of single immune cell objects can be rotated at different rotation angles (e.g., 15 degrees) to establish new single immune cell images, achieving the effect of data augmentation.

[0052] In step 440, a second number of single immune cell images are selected multiple times from the scaled selected single immune cell images.

[0053] In one example, when the number of single immune cell images is relatively sufficient, a non-replacement sampling method can be used for multiple selections. In one example, when the number of single immune cell images is relatively small, a replacement sampling method can be used for multiple selections. In one example, a replacement sampling method can be used for selection first, and then a non-replacement sampling method can be used for selection. In one example, the second number (sampling number) can be set to 144.

[0054] In step 450, the second number of single immune cell images selected each time are arranged into an array to obtain a corresponding multi-immune cell image in the plurality of multi-immune cell images. In one example, the 144 single immune cell images selected each time can be arranged into a 12 12 array of multi-immune cell objects. In one example, in each multi-immune cell image in the plurality of multi-immune cell images, the second number of single immune cell images are spaced apart from each other, so that the immune cell objects in the multi-immune cell image are isolated from each other, avoiding the immune cell objects from sticking to each other. An example of such multi-immune cell objects is shown in Figure 5are shown (although for the convenience of illustration, they are shown as gray-scale images).

[0055] Figure 6 is a flowchart of a method 600 for predicting an activation state of an immune cell according to an embodiment of the present disclosure. As shown in Figure 6 The method 600 includes steps 610-640.

[0056] At step 610, a plurality of single immune cell images and a plurality of multiple immune cell images are obtained. Each immune cell object in the plurality of single immune cell images and the plurality of multiple immune cell images is selected from a set of immune cell objects subjected to a same perturbation condition. In one example, the plurality of single immune cell images can be obtained by the process 100. The plurality of multiple immune cell images can be obtained by the process 400. In one example, the plurality of single immune cell images and the plurality of multiple immune cell images can be read from a local storage device or downloaded from a remote storage device.

[0057] At step 620, the plurality of single immune cell images are processed by a first activation state prediction model to obtain a plurality of first activation state prediction values corresponding to the plurality of single immune cell images respectively as a first activation state prediction model output.

[0058] In one example, the first activation state prediction model can be a pre-trained machine learning model. In turn, the single immune cell images can be directly input into the trained machine learning model to obtain the corresponding first activation state prediction values. The training process of the first activation state prediction model can include the following steps:

[0059] First, a first training sample set is obtained. The first training sample can include a sample single immune cell image and a first labeled activation state detection result. The first labeled activation state detection result can be used to represent the activation degree of a single immune cell object contained in the sample single immune cell image. In one example, the sample single immune cell image can be obtained by the process 100.

[0060] In one example, different types of data augmentation methods can be used to expand the sample single immune cell image to different copies for expanding the sample richness.

[0061] In one example, the first labeled activation state detection result can be obtained by a professional using a microplate reader to measure the actual activation degree of a single immune cell object contained in the sample single immune cell image under the perturbation condition. Alternatively, the first labeled activation state detection result can be obtained by a professional according to the corresponding relationship between the perturbation condition and the activation state.

[0062] In one example, the first labeled activation state detection result can include a plurality of activation state levels, such as a first activation state level to a sixth activation state level. The first activation state level corresponds to a first perturbation condition including normal cell object culture medium and no additional stimulation. The second activation state level corresponds to a second perturbation condition including OKT3 drug coated on the surface of the culture plate + concentration 0.01 ug / ml + culture time 6 hours experiment. The third activation state level corresponds to a third perturbation condition including soluble OKT3 drug + concentration 0.1 ug / ml + culture time 6 hours experiment. The fourth activation state level corresponds to a fourth perturbation condition including OKT3 drug coated on the surface of the culture plate + concentration 1 ug / ml + culture time 6 hours experiment. The fifth activation state level corresponds to a fifth perturbation condition including PHA-p (phytohemagglutinin) drug + concentration 5 ug / ml + culture time 6 hours experiment. The sixth activation state level corresponds to a sixth perturbation condition including PHA-p drug + concentration 10 ug / ml + culture time 6 hours experiment.

[0063] In addition, the more and finer the activation state levels are set, the more accurate the prediction result of the immune cell activation state will be.

[0064] In a second step, the single immune cell image in the training sample is input into the first initial model to obtain an output first actual activation state prediction result. The first initial model can include various appropriate types of machine learning models. In an example, the first initial model can be a convolutional neural network, such as an Alexnet model, a ResNet (Residual Network) model.

[0065] In a third step, the model parameters of the first initial model are adjusted according to the first difference between the obtained first actual activation state prediction result and the first labeled activation state detection result until a preset training end condition is met, and thus a first activation state prediction model can be obtained. In one example, adjusting the model parameters of the first initial model according to the first difference can be implemented by using a gradient descent method. The preset training end condition can include any of the following: the first difference is less than a first preset difference threshold, the number of training times reaches a preset number, and the training time reaches a preset length.

[0066] In step 630, the second activation state prediction model is used to process a plurality of multi-immune cell images to obtain a plurality of second activation state prediction values corresponding to the plurality of multi-immune cell images respectively, which are output by the second activation state prediction model.

[0067] In one example, the second activation state prediction model can be a pre-trained machine learning model. In turn, the plurality of immune cell images can be directly input into the pre-trained machine learning model to obtain the corresponding second activation state prediction values. The training process of the second activation state prediction model can include the following steps:

[0068] Secondly, a second training sample set is obtained, and the second training sample can include a sample plurality of immune cell image and a second labeled activation state detection result. The second labeled activation state detection result can be used to represent the activation degree of the plurality of immune cell objects contained in the sample plurality of immune cell image. The sample plurality of immune cell image can be obtained through the operation of the process 400.

[0069] In one example, different types of data augmentation methods can be used to expand the sample plurality of immune cell images into different copies for the purpose of expanding the sample richness.

[0070] In one example, the second labeled activation state detection result can be obtained by a professional measuring the actual activation degree of the plurality of immune cell objects contained in the sample plurality of immune cell image under the disturbance condition.

[0071] In one example, the second labeled activation state detection result can be obtained by a professional measuring the actual activation degree of the plurality of immune cell objects contained in the sample plurality of immune cell image under the disturbance condition using an enzyme label instrument. Alternatively, the second labeled activation state detection result can be obtained by a professional according to the corresponding relationship between the disturbance condition and the activation state.

[0072] In one example, the second labeled activation state detection result can include a plurality of activation state levels, such as a first activation state level to a sixth activation state level.

[0073] Secondly, the sample plurality of immune cell image in the training sample is input into the second initial model to obtain the output second actual activation state prediction result. The second initial model can include various types of machine learning models. In an example, the second initial model can be a convolutional neural network, such as an Alexnet model or a ResNet model.

[0074] Thirdly, the model parameters of the second initial model are adjusted according to the second difference between the obtained second actual activation state prediction result and the second labeled activation state detection result until a preset training end condition is met, and then the second activation state prediction model can be obtained. In one example, adjusting the model parameters of the second initial model according to the second difference can be achieved by using a gradient descent method. The preset training end condition can include any one of the following: the second difference is less than a second preset difference threshold, the number of training times reaches a preset number, and the training time reaches a preset length.

[0075] In one example, to adapt to different complexities of data, parameters such as the depth (the number of layers of the convolutional neural network), the width (the number of channels of the convolutional neural network), the resolution (the input image size of the convolutional neural network) of the convolutional neural network can be set. Generally speaking, the complexity of multiple immune cell images is higher than that of single immune cell images, and a more complex convolutional neural network needs to be adapted. For example, the depth of the first activation state prediction model is 11, and the depth of the second activation state prediction model is 20. The average width of the second activation state prediction model is 1.5 times that of the first activation state prediction model. The resolution of the first activation state prediction model is 224x224, and the resolution of the second activation state prediction model is 960x960.

[0076] In step 640, the activation state of the immune cell object set is determined based on the plurality of first activation state prediction values and the plurality of second activation state prediction values.

[0077] The method 600 predicts the activation state of the immune cell object set by using two activation state prediction models to process the prediction results of two types of activation states output by two types of immune images respectively. First, a plurality of single immune cell images are processed using a first activation state prediction model to obtain a plurality of first activation state prediction value cell images corresponding to the plurality of single immune cell images output by the first activation state prediction model; a plurality of multiple immune cell images are processed using a second activation state prediction model to obtain a plurality of second activation state prediction values corresponding to the plurality of multiple immune cell images output by the second activation state prediction model; and finally, the activation state of the immune cell object set is determined using the plurality of first activation state prediction values and the plurality of second activation state prediction values.

[0078] Since the first activation state prediction model is trained based on single immune cell image data, the first activation state prediction model pays more attention to the color features and shape contour features of the single immune cell object region in the single immune cell image. Since the second activation state prediction model is trained based on multiple immune cell image data, the second activation state prediction model pays more attention to the statistical distribution features of each immune cell object as a whole in the multiple immune cell object regions in the multiple immune cell image. That is, the two models pay attention to different image features and different types of image features. Therefore, the activation state of the immune cell object set can be predicted based on the plurality of first activation state prediction values and the plurality of second activation state prediction values, and the activation state of the immune cell object set can be more comprehensively and accurately determined.

[0079] Figure 7 is an example process of determining the activation state of the immune cell object set in the method 600 according to an embodiment of the present disclosure. As shown in FIG. 6, the method 600 includes the following steps. Figure 6 is an example process of determining the activation state of the immune cell object set in the method 600 according to an embodiment of the present disclosure. As shown in FIG. 6, the method 600 includes the following steps.Figure 7 As shown, determining the activation state of the immune cell object set (step 640) further includes steps 710 to step 730.

[0080] At step 710, a first probability vector is determined according to the plurality of first activation state prediction values. The first probability vector indicates respective probabilities of the immune cell object set being at each of a plurality of activation state levels. The activation state is one of the plurality of activation state levels. For example, the plurality of activation state levels can include a first activation state level, a second activation state level, a third activation state level, a fourth activation state level, a fifth activation state level, and a sixth activation state level. The activation state level corresponds to an indication of the degree of activation of the immune cell object set. For example, the higher the activation state level, the greater the degree of activation of the immune cell object set.

[0081] In one example, each of the plurality of first activation state prediction values is a vector. The vector includes respective probabilities of the plurality of activation state levels. The maximum probability value among the respective probability values indicates the activation state level predicted by the first activation state prediction model. For example, the first activation state prediction value can be [0.4, 0.3, 0.2, 0.1, 0, 0], where the maximum probability value 0.4 corresponds to the first activation state level as the activation state level predicted by the first activation state prediction model.

[0082] In one example, the first probability vector can be determined by the following steps.

[0083] Firstly, for each of the plurality of activation state levels, the number of times the activation state level is predicted by the first activation state prediction model is counted. For example, the 10 activation state levels predicted by the first activation state prediction model are the first activation state level, the second activation state level, the first activation state prediction value, the first activation state prediction value, the first activation state prediction value, the second activation state level, the third activation state level, the first activation state prediction value, the first activation state prediction value, and the fourth activation state level. It can be seen that the first activation state level is predicted 6 times, the second activation state level is predicted 2 times, the third activation state level is predicted 1 time, the fourth activation state level is predicted 1 time, and the fifth activation state level and the sixth activation state level are not predicted.

[0084] Secondly, the ratio of the number of the plurality of activation state levels to the number of the plurality of single immune cell images is calculated as the first probability vector. For example, the ratio of the number of the plurality of activation state levels [6, 2, 1, 1, 0, 0] to the number of the plurality of single immune cell images 10 is [0.6, 0.2, 0.1, 0.1, 0, 0], i.e. the first probability vector is [0.6, 0.2, 0.1, 0.1, 0, 0].

[0085] At step 720, a second probability vector is determined according to the plurality of second activation state prediction values. The second probability vector indicates respective probabilities of the set of immune cell objects being in each of the plurality of activation state levels. In one example, each of the plurality of second activation state prediction values is a vector. The vector includes respective probability values of the plurality of activation state levels. A maximum probability value among the respective probability values indicates the activation state level predicted by the second activation state prediction model. For example, the second activation state prediction value can be [0.5, 0.2, 0.2, 0.1, 0, 0], where the maximum probability value 0.5 corresponds to the first activation state level as the activation state level predicted by the second activation state prediction model.

[0086] In one example, an average vector of the plurality of second activation state prediction values can be calculated as the second probability vector. For example, the two second activation state prediction values are [0.5, 0.2, 0.2, 0.1, 0, 0] and [0.6, 0.2, 0.1, 0.1, 0, 0] respectively, then the average vector is [0.55, 0.2, 0.15, 0.1, 0, 0], i.e., the second probability vector is [0.55, 0.2, 0.15, 0.1, 0, 0].

[0087] At step 730, a weighted sum of the first probability vector and the second probability vector is calculated as a third probability vector. A maximum probability value in the third probability vector indicates the activation state level corresponding to the activation state of the set of immune cell objects.

[0088] In one example, the third probability vector is P, the first probability vector is P1, and the second probability vector is P2, which can be represented as P = a1P1 + a2P2. Wherein a1 is a first weight coefficient corresponding to P1, a2 is a second weight coefficient corresponding to P2, and a1 + a2 = 1. The specific values of a1 and a2 can be set according to the influence degree of the first probability vector and the second probability vector on determining the activation state of the set of immune cell objects. For example, the first probability vector P1 is [0.6, 0.2, 0.1, 0.1, 0, 0], the second probability vector is [0.55, 0.2, 0.15, 0.1, 0, 0], a1 is 0.5, and a2 is 0.5, then the third probability vector P is [0.575, 0.2, 0.125, 0.1, 0, 0]. The maximum probability value 0.575 corresponds to the first activation state level as the activation state level corresponding to the activation state of the set of immune cell objects.

[0089] Since the plurality of first activation state prediction values are obtained based on individuals (single immune cell objects), more attention is paid to reflecting the specificity of individuals, and therefore the first probability is calculated by counting the prediction result that appears most frequently in the plurality of model prediction results. In contrast, since the plurality of second activation state prediction values are obtained based on the population (a plurality of immune cell objects), more attention is paid to reflecting the distribution of the population, and therefore the second probability is calculated by accumulating the plurality of model prediction results.

[0090] Figure 8 is a flowchart of a method 800 for predicting the immune cell activation ability of a compound according to an embodiment of the present disclosure. As shown in Figure 8 , the method 800 includes steps 810 to 860.

[0091] Step 810, subjecting a set of immune cell objects to a perturbation condition. The perturbation condition includes exposure to a predetermined concentration of a compound for a predetermined duration.

[0092] Taking the exposure of PHA (phytohemagglutinin) at a predetermined concentration for a predetermined duration as an example of the perturbation condition, in one example, Jurkat cells (human T lymphocyte leukemia cells) can be plated in a 96-well plate for a certain period of time, then PHA (phytohemagglutinin) at a concentration of 50 ug / ml is added, and then incubated for 6 hours. In one example, Jurkat cells can be plated in a 96-well plate for a certain period of time, then PHA at a concentration of 0.5 ug / ml is added, and then incubated for 6 hours.

[0093] Step 820, taking images of the set of immune cell objects using a high-content cell imaging system to obtain a plurality of high-content images. In one example, the immune cells in the 96-well plate can be placed in the high-content cell imaging system, and the set of immune cells can be taken. Each well can take multiple frames of high-content images. Each frame needs to be taken for 7 minutes.

[0094] In one example, before taking the images, different organelles or cell components can be stained using the staining method commonly used in the art. For example, a dye for staining mitochondria can be added in advance, and the staining is performed for 0.5 hours. Then, polyformaldehyde is added in different wells to fix the cells for 10 minutes. Then, most of the culture medium, the dye for staining mitochondria, and polyformaldehyde are washed away, and only a small volume of the bottom is retained to prevent the cells from being sucked away. Then, a buffer containing another 5 dyes (staining endoplasmic reticulum, Golgi, ribonucleic acid, nucleus, and actin, respectively) is added. After washing, the high-content imaging is performed.

[0095] Step 830, generating a plurality of single immune cell images and a plurality of multiple immune cell images from the plurality of high-content images. In one example, the plurality of single immune cell images can be generated from the plurality of high-content images by the operations of the process 100. In one example, the plurality of multiple immune cell images can be generated from the plurality of single immune cell images by the operations of the process 400.

[0096] Step 840, processing the plurality of single immune cell images by the first activation state prediction model to obtain a plurality of first activation state prediction values corresponding to the plurality of single immune cell images respectively, which are output by the first activation state prediction model. In one example, the plurality of multiple immune cell images can be generated from the plurality of single immune cell images by the operations of the process 400. The specific operations of step 840 are substantially the same as those of step 620, which will not be repeated here.

[0097] Step 850, processing the plurality of multiple immune cell images by the second activation state prediction model to obtain a plurality of second activation state prediction values corresponding to the plurality of multiple immune cell images respectively, which are output by the second activation state prediction model. In one example, the specific operations of step 850 are substantially the same as those of step 630, which will not be repeated here.

[0098] Step 860, determining the activation state of the immune cell object set based on the plurality of first activation state prediction values and the plurality of second activation state prediction values, wherein the activation state indicates the immune cell activation ability of the compound. In one example, the specific operations of step 860 are substantially the same as those of step 640, which will not be repeated here.

[0099] In one example, for example, the activation state is the fifth activation level, which indicates that the immune cell activation ability of the compound (e.g., 50 ug / ml of PHA) is strong. In one example, for example, the activation state is the second activation level, which indicates that the immune cell activation ability of the compound (e.g., 0.5 ug / ml of PHA) is weak.

[0100] Since different perturbation conditions have different degrees of activation of immune cells, the immune cells present different phenotypic changes, that is, the single immune cell image and the multiple immune cell image can present the phenotypic characteristics of the immune cells, and thus the single immune cell image and the multiple immune cell image are associated with the activation state of the immune cells. The method 800 determines the activation state of the immune cell object set based on the plurality of first activation state prediction values predicted based on the plurality of single immune cell images and the plurality of second activation state prediction values predicted based on the plurality of multiple immune cell images. Thus, the activation state of the immune cell object set determined by the method 800 can indicate the immune cell activation ability of the compound in the perturbation condition. Compared with the error prone by manual operation in the traditional drug screening task, the method 800 can improve the prediction accuracy of the activation state of the immune cell object set and the screening efficiency. Compared with the large amount of manpower and material resources required in the traditional drug screening task, the method 800 can save a large amount of experimental manpower and cost.

[0101] Figure 9 is a block diagram of an apparatus 900 for predicting an activation state of immune cells according to an embodiment of the present disclosure. As shown in Figure 9 The prediction apparatus 900 for the activation state of immune cells includes a first module 910, a second module 920, a third module 930, and a second module 940.

[0102] The first module 910 is configured to obtain a plurality of single immune cell images and a plurality of multiple immune cell images, wherein each single immune cell image in the plurality of single immune cell images contains a single immune cell object, each multiple immune cell image in the plurality of multiple immune cell images contains a plurality of immune cell objects, and each immune cell object in the plurality of single immune cell images and the plurality of multiple immune cell images is selected from an immune cell object set subjected to the same perturbation condition.

[0103] The second module 920 is configured to process the plurality of single immune cell images by using a first activation state prediction model to obtain a plurality of first activation state prediction values corresponding to the plurality of single immune cell images respectively and output by the first activation state prediction model.

[0104] The third module 930 is configured to process the plurality of multiple immune cell images by using a second activation state prediction model to obtain a plurality of second activation state prediction values corresponding to the plurality of multiple immune cell images respectively and output by the second activation state prediction model.

[0105] The fourth module 940 is configured to determine the activation state of the immune cell object set based on the plurality of first activation state prediction values and the plurality of second activation state prediction values.

[0106] It should be understood that Figure 9 The modules of the apparatus 900 shown in Figure 6The various steps in the described method 600 correspond. Thus, the operations, features and advantages described above for the method 600 apply equally to the apparatus 900 and its included modules. For the sake of brevity, certain operations, features and advantages are not repeated here.

[0107] While specific functions are discussed above with reference to particular modules, it should be noted that the functions of the various modules discussed herein can be split among multiple modules and / or at least some of the functions of multiple modules can be combined into a single module. A particular module discussed herein performing an action includes that particular module itself performing the action, or alternatively, that particular module invoking or otherwise accessing another component or module that performs the action (or performs the portion of the action that is relevant to that particular module). Thus, a particular module performing an action can include that particular module itself performing the action, and / or another module accessing or otherwise performing the action.

[0108] It should also be understood that various techniques described herein can be described in the general context of software hardware elements or program modules. The various modules described above with respect to the apparatus 900 can be implemented in hardware or in hardware combined with software and / or firmware. For example, these modules can be implemented as computer program code / instructions configured to be executed in one or more processors and stored in a computer-readable storage medium. Alternatively, these modules can be implemented as hardware logic / circuitry.

[0109] According to an aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and at least one memory communicatively connected with the at least one processor, the at least one memory storing instructions which, when executed by the at least one processor, cause the at least one processor to perform the method described above.

[0110] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing instructions is provided, the instructions, when executed by at least one processor of a computer, causing the computer to perform the method described above.

[0111] According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the method described above.

[0112] Figure 10 An example configuration of an electronic device 1000 that can be used to implement the methods described herein is shown.

[0113] The electronic device 1000 can be various different types of devices. Examples of the electronic device 1000 include, but are not limited to, a desktop computer, a server computer, a notebook or netbook computer, a mobile device (e.g., a tablet computer, a cellular or other wireless phone (e.g., a smart phone), a notepad computer, a mobile station), a wearable device (e.g., glasses, a watch), an entertainment device (e.g., an entertainment appliance, a set-top box communicatively coupled to a display device, a game console), a television or other display device, a car computer, and the like.

[0114] The electronic device 1000 can include at least one processor 1002, memory 1004, communication interface(s) 1006, a display device 1008, other input / output (I / O) devices 1010, and one or more mass storage devices 1012, which are capable of communicating with each other, such as over system bus 1014 or other appropriate connection.

[0115] The processor 1002 can be a single processing unit or a plurality of processing units, all of which can include single or multiple computing cores or processing units. The processor 1002 can be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuitries, and / or any devices that manipulate signals based on operational instructions. Among other capabilities, the processor 1002 can be configured to fetch and execute computer-readable instructions stored in the memory 1004, the mass storage device 1012, or other computer-readable media, such as program code for an operating system 1016, program code for applications 1018, program code for other programs 1020, and the like.

[0116] The memory 1004 and the mass storage device 1012 are examples of computer-readable storage media for storing instructions which are executed by the processor 1002 to implement the various functionalities described above. By way of example, the memory 1004 can include both volatile memory and non-volatile memory (e.g., RAM, ROM, etc.). In addition, the mass storage device 1012 can include a hard disk drive, a solid state drive, removable media, including external and removable drives, memory cards, flash memory, floppy disks, optical disks (e.g., CD or DVD), storage arrays, network-attached storage, a storage area network, and the like. Both the memory 1004 and the mass storage device 1012 can be collectively referred to herein as the memory or computer-readable storage media, and can be non-transitory media capable of storing the computer-readable, processor-executable program instructions as computer program code which can be executed by the processor 1002 as a particular machine configured to implement the operations and functionalities described in the examples herein.

[0117] A number of programs can be stored on the mass storage device 1012. These programs include an operating system 1016, one or more application programs 1018, other programs 1020, and program data 1022, and they can be loaded into the memory 1004 for execution. Examples of such application programs or program modules can include, for example, computer program logic (e.g., computer program code or instructions) for implementing the first module 910, the second module 920, the third module 930, the fourth module 940, the method 6 (including any suitable steps of the method 600), and / or the additional embodiments described herein, for example, to implement the following components / functions.

[0118] While illustrated in the Figure 10 memory 1004 of the electronic device 1000, the modules 1016, 1018, 1020, and 1022, or portions thereof, can be implemented using any form of computer-readable media that is accessible by the electronic device 1000. As used herein, "computer-readable media" includes both computer-readable storage media and communication media.

[0119] Computer-readable storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer-readable storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information for access by an electronic device. In contrast, communication media can embody computer-readable instructions, data structures, program modules or other data in a modulated data signal, such as a carrier wave or other transport mechanism. As defined herein, computer-readable storage media does not include communication media.

[0120] One or more communication interfaces 1006 are used to exchange data with other devices, such as over a network, direct connection, or the like. Such communication interfaces can be one or more of: any type of network interface (for example, a network interface card (NIC)), a wired or wireless (such as IEEE 802.11 wireless LAN (WLAN)) wireless interface, a Worldwide Interoperability Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth™ interface, a near field communication (NFC) interface, or the like. The communication interfaces 1006 can facilitate communications within a variety of networks and protocol types, including wired networks (for example, LAN, cable, or the like) and wireless networks (for example, WLAN, cellular, satellite, or the like), the Internet, or the like. The communication interfaces 1006 can also provide for communication with external storage (not shown), such as in a storage array, network attached storage, storage area network, or the like.

[0121] In some examples, a display device 1008, such as a monitor, can be included for displaying information and images to a user. Other I / O devices 1010 can be devices that receive input from and / or provide output to a user, and can include touch input devices, gesture input devices, cameras, keyboards, remote controls, mice, printers, audio input / output devices, or the like.

[0122] The technology described herein can be supported by these various configurations of the electronic device 1000 and is not limited to the specific examples described herein. For example, the functionality can also be implemented all or in part through use of a distributed system, such as over a "cloud." The cloud includes and / or comprises a platform of resources. The platform abstracts underlying functionality of hardware (for example, servers) and software resources of the cloud. Resources can include applications and / or data, that can be utilized while computer processing is executed on servers that are remote from the electronic device 1000. Resources can also include services provided over the Internet and / or through a subscriber network, such as a cellular or Wi-Fi network. The platform can abstract resources and functionality to connect the electronic device 1000 with other electronic devices. Hence, the implementation of functionality described herein can be distributed throughout the cloud. For example, the functionality can be implemented in part on the electronic device 1000 and in part by the platform that abstracts the functionality of the cloud.

[0123] While the disclosure has been illustrated and described in detail in the drawings and foregoing description, such illustration and description is to be considered illustrative or exemplary and not restrictive; the disclosure is not limited to the disclosed embodiments. Variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed subject matter, from a study of the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps not listed in the claims, the indefinite article "a" or "an" does not exclude a plurality, and the term "plurality" means two or more. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Claims

1. A method for predicting an activation state of immune cells, comprising: obtaining a plurality of single immune cell images and a plurality of multiple immune cell images, wherein each single immune cell image of the plurality of single immune cell images contains a single immune cell object, each multiple immune cell image of the plurality of multiple immune cell images contains a plurality of immune cell objects, and each immune cell object in the plurality of single immune cell images and the plurality of multiple immune cell images is selected from a set of immune cell objects subjected to a same perturbation condition; processing the plurality of single immune cell images using a first activation state prediction model to obtain a plurality of first activation state prediction values corresponding to the plurality of single immune cell images respectively output by the first activation state prediction model; processing the plurality of multiple immune cell images using a second activation state prediction model to obtain a plurality of second activation state prediction values corresponding to the plurality of multiple immune cell images respectively output by the second activation state prediction model; and determining an activation state of the set of immune cell objects based on the plurality of first activation state prediction values and the plurality of second activation state prediction values; wherein the obtaining a plurality of single immune cell images comprises: obtaining a plurality of high-content images of the set of immune cell objects taken by a high-content cell imaging system, wherein each high-content image of the plurality of high-content images contains a plurality of immune cell objects and has N color channels, each color channel corresponding to a type of organelle or cellular component, N being an integer greater than 3; preprocessing each high-content image of the plurality of high-content images to generate a plurality of color images corresponding to the plurality of high-content images respectively; and segmenting the plurality of single immune cell images from the plurality of color images; wherein the preprocessing each high-content image of the plurality of high-content images comprises: normalizing the high-content image; and merging the N color channels of the normalized high-content image into a red-green-blue color channel; wherein the first activation state prediction model and the second activation state prediction model are convolutional neural networks, wherein the first activation state prediction model has a depth of 11 layers and the second activation state prediction model has a depth of 20 layers, wherein the second activation state prediction model has an average number of channels that is 1.5 times of the first activation state prediction model.

2. The method of claim 1, wherein, the activation state is one activation state level selected from a group of a plurality of activation state levels, and wherein the determining an activation state of the set of immune cell objects based on the plurality of first activation state prediction values and the plurality of second activation state prediction values comprises: determining a first probability vector indicating respective probabilities of the set of immune cell objects being in each activation state level of the plurality of activation state levels according to the plurality of first activation state prediction values; and determining an activation state of the set of immune cell objects according to the first probability vector and the plurality of second activation state prediction values. determining a second probability vector according to the plurality of second activation state prediction values, the second probability vector indicating respective probabilities of the set of immune cell objects being in each of the plurality of activation state levels; and calculating a weighted sum of the first probability vector and the second probability vector as a third probability vector, wherein a maximum probability value in the third probability vector indicates an activation state level corresponding to an activation state of the set of immune cell objects.

3. The method of claim 2, wherein, each of the plurality of first activation state prediction values is a vector comprising respective probability values of the plurality of activation state levels, wherein a maximum probability value in the respective probability values indicates an activation state level predicted by the first activation state prediction model, and wherein the determining a first probability vector according to the plurality of first activation state prediction values comprises: counting, for each of the plurality of activation state levels, a number of times that the activation state level is predicted by the first activation state prediction model; and calculating a ratio of the plurality of activation state levels to a number of the plurality of single immune cell images as the first probability vector.

4. The method of claim 2, wherein, each of the plurality of second activation state prediction values is a vector comprising respective probability values of the plurality of activation state levels, wherein a maximum probability value in the respective probability values indicates an activation state level predicted by the second activation state prediction model, and wherein the determining a second probability vector according to the plurality of second activation state prediction values comprises: calculating a mean vector of the plurality of second activation state prediction values as the second probability vector.

5. The method of claim 1, wherein, the segmenting the plurality of single immune cell images from the plurality of color images comprises: segmenting the plurality of single immune cell images from the plurality of color images using an instance segmentation algorithm.

6. The method of claim 1, wherein, the pre-processing each of the plurality of high-content images further comprises: performing at least one of the following before normalizing the high-content image: median filtering the high-content image; and removing overexposed pixels in the high-content image having a brightness value exceeding a brightness threshold.

7. The method of claim 1, wherein, the merging N color channels of the normalized high-content image into red-green-blue color channels comprises: calculating a mean value of red components in the N color channels as a merged red channel; calculating a mean value of green components in the N color channels as a merged green channel; and calculating a mean value of blue components in the N color channels as a merged blue channel.

8. The method of claim 7, wherein, the N color channels comprise a red channel, a green channel, a blue channel, a red-green channel, a red-blue channel, and a green-blue channel.

9. The method of claim 1, wherein, the obtaining a plurality of multi-immune cell images comprises: extracting respective foreground regions from the plurality of single immune cell images, each foreground region being defined by an outline of a respective single immune cell object; filtering out single immune cell images of the plurality of single immune cell images whose maximum aspect ratio of immune cell objects is greater than an aspect ratio threshold, to obtain a plurality of selected single immune cell images; scaling the plurality of selected single immune cell images to a same size; selecting a second number of single immune cell images from the scaled plurality of selected single immune cell images multiple times; and arranging the second number of single immune cell images selected each time into an array to obtain a corresponding multi-immune cell image of the plurality of multi-immune cell images.

10. The method of claim 9, further comprising: rotating foreground regions of a first number of single immune cell images of the plurality of selected single immune cell images to obtain a plurality of rotated single immune cell images, before the scaling the plurality of selected single immune cell images to the same size; and scaling the plurality of rotated single immune cell images to the same size, wherein the selecting the second number of single immune cell images from the scaled plurality of selected single immune cell images multiple times comprises: selecting the second number of single immune cell images from the scaled plurality of rotated single immune cell images and the plurality of selected single immune cell images multiple times.

11. The method of claim 9 or 10, wherein, The second number of single immune cell images are spaced apart from each other in each multi-immune cell image of the plurality of multi-immune cell images.

12. The method of claim 1, wherein, The perturbation condition comprises exposure to a predetermined compound at a predetermined concentration for a predetermined duration.

13. A method for predicting immune cell activation ability of a compound, comprising: subjecting a set of immune cell objects to a perturbation condition, the perturbation condition comprising exposure to the compound at a predetermined concentration for a predetermined duration; capturing the set of immune cell objects by a high-content cell imaging system to obtain a plurality of high-content images; generating a plurality of single immune cell images and a plurality of multi-immune cell images from the plurality of high-content images; processing the plurality of single immune cell images by a first activation state prediction model to obtain a plurality of first activation state prediction values corresponding to the plurality of single immune cell images respectively, which are output by the first activation state prediction model; processing the plurality of multi-immune cell images by a second activation state prediction model to obtain a plurality of second activation state prediction values corresponding to the plurality of multi-immune cell images respectively, which are output by the second activation state prediction model; and determining an activation state of the set of immune cell objects based on the plurality of first activation state prediction values and the plurality of second activation state prediction values, wherein the activation state indicates the immune cell activation ability of the compound; wherein the obtaining the plurality of single immune cell images comprises: obtaining a plurality of high-content images captured by the high-content cell imaging system on the set of immune cell objects, wherein each high-content image of the plurality of high-content images contains a plurality of immune cell objects and has N color channels, each color channel corresponding to a type of organelle or cellular component, N being an integer greater than 3; preprocess each high-content image in the plurality of high-content images to generate a plurality of color images respectively corresponding to the plurality of high-content images; and segment the plurality of single immune cell images from the plurality of color images; wherein the preprocessing each high-content image in the plurality of high-content images comprises: normalizing the high-content image; and merging N color channels of the normalized high-content image into red-green-blue color channels; wherein the first activation state prediction model and the second activation state prediction model are convolutional neural networks, wherein the first activation state prediction model has a depth of 11 layers and the second activation state prediction model has a depth of 20 layers, wherein the second activation state prediction model has an average number of channels that is 1.5 times of the first activation state prediction model.

14. An apparatus for performing the method of any one of claims 1-12, comprising: a first module configured to obtain a plurality of single immune cell images and a plurality of multiple immune cell images, wherein each single immune cell image in the plurality of single immune cell images contains a single immune cell object, each multiple immune cell image in the plurality of multiple immune cell images contains a plurality of immune cell objects, and each immune cell object in the plurality of single immune cell images and the plurality of multiple immune cell images is selected from a set of immune cell objects subjected to a same perturbation condition; a second module configured to process the plurality of single immune cell images using a first activation state prediction model to obtain a plurality of first activation state prediction values respectively corresponding to the plurality of single immune cell images output by the first activation state prediction model; a third module configured to process the plurality of multiple immune cell images using a second activation state prediction model to obtain a plurality of second activation state prediction values respectively corresponding to the plurality of multiple immune cell images output by the second activation state prediction model; and a fourth module configured to determine an activation state of the set of immune cell objects based on the plurality of first activation state prediction values and the plurality of second activation state prediction values.

15. An electronic device, comprising: a processor; and a memory storing instructions executable by the processor, the instructions, when executed by the processor, causing the processor to perform the method according to any one of claims 1-12.

16. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the method according to any one of claims 1-12.

17. A computer program product comprising instructions, wherein the instructions, when executed by a processor, cause the processor to perform the method according to any one of claims 1-12.

Images