A method, device, equipment and storage medium for identifying minimum inhibitory concentration

Abstract

The present invention discloses a method, device, equipment and computer-readable storage medium for identifying minimum inhibitory concentration, and relates to the field of image processing and calculation. The method comprises: obtaining a to-be-tested image of a to-be-tested bacterial species; obtaining a test control image corresponding to the to-be-tested bacterial species; calculating the similarity between each to-be-tested image and the test control image corresponding to each to-be-tested image, and obtaining a similarity test result corresponding to each to-be-tested image; determining the minimum inhibitory concentration of an antibiotic to be tested corresponding to the to-be-tested bacterial species according to the similarity test result and a preset gradient concentration; the present invention can utilize the similarity calculation between the to-be-tested image of an antibiotic-bacterial species with different concentration gradients and the test control image corresponding to each to-be-tested image to determine the similarity or dissimilarity corresponding to each to-be-tested image, thereby obtaining the minimum inhibitory concentration of the antibiotic to be tested corresponding to the to-be-tested bacterial species, and can quickly and accurately perform automatic detection and identification on the minimum inhibitory concentration of the antibiotic, thereby reducing the detection cost.

Description

Technical Field

[0001] The present invention relates to the field of image processing and computing, and in particular to a method, device, equipment and computer-readable storage medium for identifying a minimum inhibitory concentration. Background Art

[0002] The abuse of antibiotics is a major problem facing mankind in the 21st century. The appropriate use of antibiotics has reached a point where it is urgent, so it is increasingly important to accurately and quickly detect the minimum antibiotic dose that works on bacteria. The minimum inhibitory concentration refers to the minimum antibiotic concentration that can effectively kill or block the growth of bacteria, which is the minimum inhibitory concentration (MIC) of antibiotics.

[0003] In the prior art, there are two main methods for determining the minimum inhibitory concentration of drugs: the dilution method and the automated instrument detection method. Both of these methods have some problems. For example, the entire operation of the broth dilution method requires at least 24 to 48 hours, which is very unfavorable for clinical use. Although the reporting time of the automated instrument has been shortened, it still takes about 10 to 12 hours, which is difficult to meet the rapid diagnosis needs of clinical emergencies and is very expensive.

[0004] Therefore, how to quickly and accurately automate the detection of the minimum inhibitory concentration of antibiotics and reduce the detection cost is an urgent problem that needs to be solved today. Summary of the invention

[0005] The purpose of the present invention is to provide a method, device, equipment and computer-readable storage medium for identifying the minimum inhibitory concentration, so as to quickly and accurately automatically detect and identify the minimum inhibitory concentration of antibiotics and reduce the detection cost.

[0006] In order to solve the above technical problems, the present invention provides a method for identifying a minimum inhibitory concentration, comprising:

[0007] Acquire the image to be tested of the bacterial species to be tested; wherein the image to be tested includes images corresponding to the mixed solution of the antibiotic to be tested and the bacterial species to be tested at each preset gradient concentration at the moment when the inhibition is completed;

[0008] Acquire a test control image corresponding to the bacterial species to be tested;

[0009] Calculating the similarity between each of the images to be tested and the corresponding test control images, and obtaining the similarity test results corresponding to each of the images to be tested; wherein the similarity test results are similar or dissimilar;

[0010] According to the similarity test result and the preset gradient concentration, the minimum inhibitory concentration of the antibiotic to be tested corresponding to the bacterial species to be tested is determined.

[0011] Optionally, the test control image includes a preset control image of the test bacterial species and / or images corresponding to the mixed solution of the test antibiotic and the test bacterial species at each of the preset gradient concentrations at the time when bacteriostasis is not completed; wherein the time when bacteriostasis is not completed is before the time when bacteriostasis is completed.

[0012] Optionally, when the test control image includes images corresponding to the mixed solution of the test antibiotic of each preset gradient concentration and the test bacterial species at the time when the inhibition is not completed, the similarity between each of the images to be tested and the corresponding test control image is calculated to obtain the similarity test results corresponding to each of the images to be tested, including:

[0013] The similarity between the image to be tested and the test control image corresponding to each of the preset gradient concentrations is calculated to obtain a similarity test result corresponding to each of the images to be tested.

[0014] Optionally, obtaining the image to be tested of the strain to be tested includes:

[0015] After the bacterial suspension of the strain to be tested is added to each microwell of the drug sensitivity plate for a preset incubation time, each microwell is imaged without a lens to obtain the original image of the strain to be tested; wherein each microwell is provided with a preset gradient concentration of the antibiotic to be tested corresponding to the strain to be tested;

[0016] Identifying a micropore area in the original image to be tested;

[0017] An object region of interest of a preset size is extracted from each of the micropore regions, and an image of the object region of interest is used as the image to be tested.

[0018] Optionally, the minimum inhibitory concentration is any of the preset gradient concentrations.

[0019] Optionally, determining the minimum inhibitory concentration of the antibiotic to be tested corresponding to the tested bacterial species according to the similarity test result and the preset gradient concentration includes:

[0020] When the similarity test results corresponding to all the images to be tested include similarities and dissimilarities, the minimum preset gradient concentration corresponding to the images to be tested whose similarity test results are dissimilarities is determined as the minimum inhibitory concentration.

[0021] Optionally, the calculating the similarity between each of the images to be tested and the corresponding test control image to obtain the similarity test result corresponding to each of the images to be tested includes:

[0022] The preset Siamese neural network model is used to calculate the similarity between each of the images to be tested and the corresponding test control images, and obtain the similarity test results corresponding to each of the images to be tested.

[0023] Optionally, the sister network in the preset Siamese neural network model adopts the feature extraction layer of MobileNet-V2.

[0024] Optionally, when the test control image includes a preset control image of the strain to be tested, the preset Siamese neural network model is used to calculate the similarity between each of the images to be tested and the test control image corresponding to each of the images to be tested, and the similarity test results corresponding to each of the images to be tested are obtained, including:

[0025] Utilizing the preset Siamese neural network model, extracting the dimensionality reduction features corresponding to each of the images to be tested and the preset control images;

[0026] Calculating the Euclidean distance between each of the reduced-dimensionality features of the image to be tested and the reduced-dimensionality features of the preset control image;

[0027] According to the Euclidean distance and a preset distance threshold, a similarity test result corresponding to each of the images to be tested is determined.

[0028] Optionally, the method further comprises: a training process of the preset Siamese neural network model;

[0029] The training process of the preset Siamese neural network model includes:

[0030] Obtain a training sample set and a training loss function; wherein the training sample set includes the preset number of pairs of training images and the image similarity labels corresponding to each pair of the training images, the image similarity labels are similar or dissimilar, and the training loss function is L(x0,x1,y)=y||f(x0)-f(x1)||+(1-y)max(0,(m-||f(x0)-f(x1)||)), x0 and x1 are two training images in any pair of training images, f(x0) and f(x1) are the dimensionality reduction features extracted by the model corresponding to x0 and x1, y is the image similarity label corresponding to x0 and x1, m is the preset distance threshold, and ||f(x0)-f(x1)|| is the square of the Euclidean distance;

[0031] The initial Siamese neural network model is iteratively trained using the training sample set and the training loss function to obtain the preset Siamese neural network model.

[0032] The present invention also provides a minimum inhibitory concentration identification device, comprising:

[0033] The first acquisition module is used to acquire the image to be tested of the bacterial species to be tested; wherein the image to be tested includes images corresponding to the mixed solution of the antibiotic to be tested and the bacterial species to be tested at each preset gradient concentration at the moment when the inhibition is completed;

[0034] A second acquisition module is used to acquire a test control image corresponding to the bacterial species to be tested;

[0035] A similarity calculation module is used to calculate the similarity between each of the images to be tested and the corresponding test control images, and obtain the similarity test results corresponding to each of the images to be tested; wherein the similarity test results are similar or dissimilar;

[0036] The concentration determination module is used to determine the minimum inhibitory concentration of the antibiotic to be tested corresponding to the bacterial species to be tested according to the similarity test result and the preset gradient concentration.

[0037] The present invention also provides a minimum inhibitory concentration identification device, comprising:

[0038] Memory for storing computer programs;

[0039] A processor is used to implement the steps of the method for identifying the minimum inhibitory concentration as described above when executing the computer program.

[0040] In addition, the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method for identifying the minimum inhibitory concentration as described above are implemented.

[0041] A method for identifying a minimum inhibitory concentration provided by the present invention comprises: obtaining a test image of a test bacterial species; wherein the test image comprises images corresponding to the mixed solution of the test antibiotic and the test bacterial species at each preset gradient concentration at the moment of completion of bacteriostasis; obtaining a test control image corresponding to the test bacterial species; calculating the similarity between each test image and the corresponding test control image, and obtaining a similarity test result corresponding to each test image; wherein the similarity test result is similar or dissimilar; determining the minimum inhibitory concentration of the test antibiotic corresponding to the test bacterial species according to the similarity test result and the preset gradient concentration;

[0042] It can be seen that the present invention obtains the similarity test results corresponding to each image to be tested by calculating the similarity between each image to be tested and the corresponding test control image, and can use the similarity calculation between the images to be tested of antibiotic-bacteria species with different concentration gradients and the corresponding test control images to determine the similarity or dissimilarity between the images to be tested of each preset gradient concentration and the corresponding test control image, so that the minimum inhibitory concentration of the antibiotic to be tested corresponding to the tested bacteria species can be obtained according to the similarity or dissimilarity corresponding to each preset gradient concentration, and the minimum inhibitory concentration of the antibiotic can be automatically detected and identified quickly and accurately, reducing the detection cost of the minimum inhibitory concentration of the antibiotic. In addition, the present invention also provides a minimum inhibitory concentration identification device, equipment and computer-readable storage medium, which also have the above-mentioned beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0044] Figure 1 A flowchart of a method for identifying a minimum inhibitory concentration provided by an embodiment of the present invention;

[0045] Figure 2 A schematic flow chart of another method for identifying the minimum inhibitory concentration provided by an embodiment of the present invention;

[0046] Figure 3 It is a structural diagram of a Siamese neural network model in the prior art;

[0047] Figure 4 A schematic diagram of the structure of a preset Siamese neural network model for another minimum inhibitory concentration identification method provided by an embodiment of the present invention;

[0048] Figure 5 A structural block diagram of a minimum inhibitory concentration identification device provided by an embodiment of the present invention;

[0049] Figure 6 A schematic diagram of the structure of a minimum inhibitory concentration identification device provided by an embodiment of the present invention;

[0050] Figure 7 A schematic diagram of the specific structure of a minimum inhibitory concentration identification device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] Please refer to Figure 1 , Figure 1 A flowchart of a method for identifying a minimum inhibitory concentration provided by an embodiment of the present invention. The method may include:

[0053] Step 101: obtaining images of the tested bacteria species; wherein the images include images corresponding to the mixed solutions of the tested antibiotics and the tested bacteria species at preset gradient concentrations at the moment when the inhibition is completed.

[0054] Among them, the strain to be tested in this embodiment may be a strain that needs to be detected and identified by the minimum inhibitory concentration of the antibiotic to be tested; the antibiotic to be tested in this embodiment may be an antibiotic that needs to be detected and identified by the minimum inhibitory concentration of the strain to be tested. The image to be tested in this embodiment may be an image of a mixed solution of the antibiotic to be tested and the strain to be tested that is required for detecting and identifying the minimum inhibitory concentration of the antibiotic to be tested corresponding to the strain to be tested, that is, an image corresponding to each mixed solution at the moment when the inhibition is completed after the mixed solution of the antibiotic to be tested and the strain to be tested of each preset gradient concentration is configured and waits for the preset incubation time.

[0055] Correspondingly, the specific concentration values ​​of the above-mentioned preset gradient concentrations can be confidently set by the designer according to practical scenarios and user needs. For example, the concentration gradient of the antibiotic to be tested with the preset gradient concentration can be designed according to the rules of the Clinical Laboratory Standard Institute (CLSI). As long as all the preset gradient concentrations can form a corresponding concentration gradient to facilitate the detection and identification of the minimum inhibitory concentration, this embodiment does not impose any restrictions on this.

[0056] Specifically, the specific content of the image to be tested of the test bacterial species in this embodiment can be set by the designer according to practical scenarios and user needs. For example, the image to be tested can be the original image of the mixed solution of the test antibiotic and the test bacterial species at each preset gradient concentration collected by the photographing device (i.e., the original image to be tested). For example, the processor can control the photographing platform to perform lensless imaging (or microscope imaging) on ​​each microwell at a preset incubation time (i.e., the moment when the inhibition is completed) after the bacterial suspension of the test bacterial species is added to each microwell of the drug-sensitive plate, so as to obtain the original image to be tested of the test bacterial species (i.e., the holographic image of the microwell). The image to be tested may also be an image obtained after image processing is performed on each original image to be tested; for example, the processor may use image processing technology to identify the micropore area on the disk for the original image to be tested corresponding to each micropore, and then extract the target region of interest (ROI region) from the micropore area, and input the ROI region as the image to be tested into the corresponding model (such as a preset Siamese neural network model) for minimum inhibitory concentration detection; that is, in this step, the processor may perform lensless imaging on each micropore at a preset incubation time after the bacterial suspension of the test strain is added to each micropore of the drug sensitivity plate, and obtain the original image to be tested of the test strain; wherein each micropore is provided with a corresponding preset gradient concentration of the test antibiotic; identify the micropore area in the original image to be tested; extract the target region of interest of a preset size from each micropore area, and use the image of the target region of interest as the image to be tested.

[0057] Among them, this embodiment does not limit the specific size of the ROI area, that is, the specific specifications of the preset size. For example, the preset size can be 1600 pixels * 1600 pixels, that is, the size of the ROI area extracted in this embodiment can be 1600 * 1600, because the ROI area of ​​1600 * 1600 pixels can cover most of the characteristic areas of the micropore image (that is, the original image to be tested); during processing, the corresponding relationship between each ROI area and the micropore image can be matched at the same time, so as to correspond to the type and concentration of the antibiotic to be tested and the category of the bacterial species to be tested recorded in each micropore image.

[0058] Step 102: Obtain a test control image corresponding to the strain to be tested.

[0059] The test control image in this embodiment may be a control image required for detecting and identifying the minimum inhibitory concentration of the tested antibiotic corresponding to the tested bacterial species, that is, a control image for performing a similarity test with the corresponding tested image.

[0060] Specifically, the specific content of the test control image in this embodiment can be set by the designer according to practical scenarios and user needs. For example, the test control image can include a preset control image of the test bacterial species, such as the preset control image can be a preset non-inhibitory image, that is, the image corresponding to the bacterial suspension of the test bacterial species without the addition of the test antibiotic at the moment of completion of the inhibition, that is, the images of the moment of completion of the inhibition corresponding to the micropores of the antibiotic-species combination with different concentration gradients on the drug-sensitive plate can be respectively tested for similarity with the images of the moment of completion of the inhibition corresponding to the micropores of the test bacterial species without the addition of the antibiotic (control group); the preset control image can also be a preset antibacterial success image, that is, the image corresponding to the bacterial suspension of the test bacterial species with a preset high concentration of the test antibiotic added at the moment of completion of the inhibition, that is, the images of the moment of completion of the inhibition corresponding to the micropores of the antibiotic-species combination with different concentration gradients on the drug-sensitive plate can be respectively tested for similarity with the images of the moment of completion of the inhibition corresponding to the micropores of the test bacterial species (control group) that have completed the inhibition. The test control image liquid may also include images corresponding to the mixed solution of the tested antibiotic and the tested bacterial species at each preset gradient concentration at the time when the inhibition is not completed before the inhibition is completed, such as when the bacterial suspension of the tested bacterial species is added to each microwell of the drug-sensitive plate, each microwell is imaged (such as lens-free imaging or microscope imaging) to obtain the test control image corresponding to the tested bacterial species. The test control image may also include a combination of any two or more of the above images, such as the test control image includes a preset non-inhibitory image of the tested bacterial species and images corresponding to the mixed solution of the tested antibiotic and the tested bacterial species at each preset gradient concentration at the time when the inhibition is not completed, so as to improve the accuracy of the minimum inhibitory concentration detection by testing the similarity of each image to be tested with the corresponding multiple test control images. This embodiment does not impose any restrictions on this.

[0061] Specifically, the specific method for the processor to obtain the test control image corresponding to the tested strain in this embodiment can be set by the designer according to practical scenarios and user needs. For example, the processor can directly receive or search for the stored test control image corresponding to the tested strain; the processor can also control the acquisition of the test control image corresponding to the tested strain. Figure 2As shown, when the test control image includes images corresponding to the mixed solutions of the tested antibiotics and the tested bacterial species at each preset gradient concentration at the time when the inhibition is not completed (T0 moment), the processor can perform lensless imaging (or microscope imaging) on ​​each micropore at the initial moment (T0 moment, i.e., the time when the inhibition is not completed) after the bacterial suspension of the tested bacterial species is added to each micropore of the drug-sensitive plate to obtain the control original image of the tested bacterial species; identify the micropore area in the control original image; extract a control target area of ​​interest of a preset size from each micropore area in the original image, and use the image of the control target area of ​​interest as the image to be tested; perform lensless imaging (or microscope imaging) on ​​each micropore again at the time when the inhibition is completed (T1 moment) after the preset incubation time at the initial moment to obtain the original image to be tested of the tested bacterial species, so as to obtain the image to be tested of the tested bacterial species. When the test control image includes a preset non-inhibitory image of the test bacterial species, after the bacterial suspension of the test bacterial species is added to the control microwells (i.e., positive control wells) of the drug sensitivity plate for a preset incubation time, lensless imaging (or microscope imaging) is performed on each control microwell to obtain a control original image of the test bacterial species; wherein the control microwells are not provided with the test antibiotics; the microwell area in the control original image is identified; a control target area of ​​interest of a preset size is extracted from the microwell area of ​​the control original image, and the image of the control target area of ​​interest is used as the test control image.

[0062] Among them, the drug sensitivity plate is a commonly used experimental device in drug sensitivity test, which consists of several regularly arranged micropores. In actual use, there are drug sensitivity plates of different specifications, and the common ones are 96-well plates and 128-well plates. In this implementation, the operator first configures the bacterial suspension of the sample (i.e., the strain to be tested), opens the drug sensitivity plate coated with antibiotics, and the micropores of the drug sensitivity plate are pre-coated with different gradient concentrations (i.e., preset gradient concentrations) of the antibiotics to be tested. The antibiotic concentration gradient can be designed according to the rules of the American Clinical Trial Standards Organization, and the bacterial suspension is added to each micropore through the sample addition hole. After the sample addition is completed, the drug sensitivity plate is placed on the camera platform, and the processor can control the camera platform to perform lensless imaging and save data for each micropore at the initial moment after the sample addition is completed (i.e., the moment when the inhibition is not completed) and after the predetermined incubation time (i.e., the moment when the inhibition is completed), so as to obtain the image to be tested and the test control image corresponding to the strain to be tested; the processor can also perform lensless imaging and save data for each micropore and the control micropore after the predetermined incubation time.

[0063] It should be noted that there is no necessary logical sequence between step 101 and step 102 in this embodiment. Step 101 may be performed first and then step 102, or step 102 may be performed first and then step 101, or both steps may be performed simultaneously. This embodiment does not impose any limitation on this.

[0064] Step 103: Calculate the similarity between each image to be tested and its corresponding test control image, and obtain the similarity test result corresponding to each image to be tested; wherein the similarity test result is similarity or dissimilarity.

[0065] It can be understood that in this step, the processor calculates the similarity between the test images of each preset gradient concentration and the corresponding test control images to obtain a similarity test result of whether each test image is similar to the corresponding test control image, thereby using the similarity test results corresponding to each preset gradient concentration to determine the minimum inhibitory concentration of the test antibiotic corresponding to the test bacterial species, thereby realizing the detection and identification of the minimum inhibitory concentration.

[0066] Correspondingly, the specific method for the processor to calculate the similarity between each image to be tested and the corresponding test control image in this step to obtain the similarity test result corresponding to each image to be tested can be set by the designer according to practical scenarios and user needs. For example, when the test control image includes images corresponding to the mixed solution of the tested antibiotic and the tested bacteria at each preset gradient concentration at the moment when the inhibition is not completed, the processor can calculate the similarity between the image to be tested and the test control image corresponding to each preset gradient concentration, and obtain the similarity test result corresponding to each image to be tested, that is, the image to be tested at each preset gradient concentration and the same The similarity test result of the similarity between the test control images of different concentrations is used to judge the minimum inhibitory concentration of the antibiotic from the time dimension; when the test control image includes a preset non-inhibited image or a preset successful inhibition image of the tested bacteria, the processor can calculate the similarity between the tested image of each preset gradient concentration and the preset non-inhibited image or the preset successful inhibition image, and obtain the similarity test result corresponding to each image to be tested, that is, the similarity test result of the similarity between the tested image of each preset gradient concentration and the same preset non-inhibited image or the preset successful inhibition image, so as to judge the minimum inhibitory concentration of the antibiotic from the antibiotic dimension. This embodiment does not impose any restrictions on this.

[0067] Among them, Figure 2 As shown, in this step, the processor can first calculate the similarity between each image to be tested and its corresponding test control image, and then use the similarity threshold method to determine the similarity test result corresponding to each image to be tested, that is, according to the similarity corresponding to each image to be tested and the preset distance threshold, determine the similarity test result corresponding to each image to be tested.

[0068] Specifically, the specific process of the processor calculating the similarity between each image to be tested and the corresponding test control image in this step to obtain the similarity test result corresponding to each image to be tested can be set by the designer according to the practical scenario and user needs. For example, the processor can use the preset Siamese neural network model to calculate the similarity between each image to be tested and the corresponding test control image to obtain the similarity test result corresponding to each image to be tested; that is, in this step, the processor can use the preset Siamese neural network (Siamese Net, also known as dual neural network) model to calculate the similarity between each image to be tested and the corresponding test control image to obtain the similarity test result corresponding to each image to be tested. The processor can also use other algorithm models to calculate the similarity between each image to be tested and the corresponding test control image, and this embodiment does not impose any restrictions on this.

[0069] Among them, the Siamese neural network is an artificial neural network that uses the same weights when working in series on two different input vectors to calculate comparable output vectors. The structure of the Siamese neural network can be as follows Figure 3 As shown, the sister networks (Sister Network #1 and Sister Network #2) in the Siamese neural network are identical and share weights; each subnetwork of the Siamese neural network receives a different input image and uses contrastive loss or triplet loss to train the neural network.

[0070] Specifically, the sister network in the preset Siamese neural network model in this embodiment is a feature extraction layer using MobileNet-V2, which is a lightweight neural network composed of inverted residual and deep separable convolution. Based on the idea of ​​transfer learning, the MobileNet-V2 classification model trained by Google on the ImageNet (a large visual database for visual object recognition software research) data set is used. The last fully connected layer of the model outputs 1000 nodes, and then after the Softmax (classification network) layer, the probability of each class in the 1000 class samples in the ImageNet data set is output; in this embodiment, the MobileNet-V2 model is migrated to the image similarity measurement problem, and the global pooling layer and the fully connected layer of the MobileNet-V2 model are removed to only retain the full convolution feature extraction layer of the model, and then a 1*1 convolution layer and a global average pooling layer are spliced ​​to map the image features to 128 dimensions. The entire network is implemented using full convolution, which can be applied to any image size, and the fully connected layer is abandoned and replaced with a 1*1 convolution layer, which reduces the model parameters. Among them, the process of the classic residual block is: 1x1 (dimensionality reduction) -> 3x3 (convolution) -> 1x1 (dimensionality increase), while mobileNet-V2 first expands the channels of the feature map through a 1x1 pointwise (PW) convolution operation to enrich the number of features and improve accuracy, and then extracts features through a deep convolution layer (DP) to limit the input feature dimension. This process is just the opposite of the order of the residual block, so it is called inverted residual: 1x1 (dimensionality increase) -> 3x3 (dw-conv+relu) -> 1x1 (dimensionality reduction+linear transformation). Depthwise separable convolution is an operation that combines depthwise (DW) convolution and pointwise (PW) convolution to extract feature maps. Compared with conventional convolution, the number of parameters and computational cost are lower.

[0071] Correspondingly, the sister network in the preset Siamese neural network model can also be a feature extraction layer of other convolutional neural networks (CNN networks), such as MobileNet (deep separable convolution) network and VGG (a classic convolutional neural network) network, etc. This embodiment does not impose any restrictions on this.

[0072] Correspondingly, in this embodiment, in order to measure the similarity between samples, metric learning can be used; metric learning can be considered as similarity learning or distance metric learning. There are many learning methods for metric learning, such as contrastive loss, margin loss, hinge loss, and triplet loss. Among them, the purpose of contrastive loss is to increase the "inter-class" difference of the classifier, and triplet loss is an improvement on contrastive loss, which not only makes the "inter-class" difference larger but also ensures that the "intra-class" difference is smaller. Among them, the formula of the contrastive loss function can be L(x0,x1,y)=y||f(x0)-f(x1)||+(1-y)max(0,(m-||f(x0)-f(x1)||)), where x0 and x1 are two samples, respectively, and f(x0) and f(x1) are sample features extracted by the model. y is the label between the two samples. When y=1, the samples are similar, and when y=0, the samples are not similar. ||f(x0)-f(x1)|| is the square of the Euclidean distance; m (margin) is a constant hyperparameter (i.e., the preset distance threshold), which is equivalent to setting an upper bound (margin square) for the loss. If the distance is greater than or equal to margin, it means that it has been optimized very well, and loss = 0, which can well express the matching degree of paired samples.

[0073] Among them, the formula of Euclidean distance can be Where x and y are two points in n-dimensional space: x(x1,x2,...,x n ) and y(y1,y2,...,y n ).

[0074] Correspondingly, the formula of the triple loss function can be L(x a ,x p ,x n )=max(0,q+||f(x a )-f(x p )||-||f(x a )-f(x n )||); where x a 、x p and x n They are anchor samples, positive samples, and negative samples respectively. q is a constant hyperparameter like m in the contrast loss function, which can force the model to learn hard, making the distance between heterogeneous samples larger and the distance between similar samples smaller.

[0075] Specifically, the method provided in this embodiment may further include a training process of a preset Siamese neural network model. For example, the loss function used in the training process of the preset Siamese neural network model in this embodiment may be a contrast loss, that is, the processor may obtain a training sample set and a training loss function; wherein the training sample set includes a preset number of pairs of training images and image similarity labels corresponding to each pair of training images, the image similarity labels are similar or dissimilar, and the training loss function is L(x0,x1,y)=y||f(x0)-f(x1)||+(1-y)max(0,(m-||f(x0)-f(x1)||)), x0 and x1 are two training images in any pair of training images, f(x0) and f(x1) are respectively the dimensionality reduction features extracted by the model corresponding to x0 and x1, y is the image similarity label corresponding to x0 and x1, m is the preset distance threshold, and ||f(x0)-f(x1) is the square of the Euclidean distance; using the training sample set and the training loss function, the initial Siamese neural network model is iteratively trained to obtain the preset Siamese neural network model.

[0076] For example, the processor can randomly select two ROI images (i.e., images of the target region of interest) in the training database and add corresponding image similarity labels as a pair of training images. When the image similarity labels of the two ROI images are 1, they are similar, and when the image similarity labels of the two ROI images are 0, they are dissimilar. A preset number of pairs of ROI images are selected to construct a training sample set for training a preset Siamese neural network model; wherein, the image similarity labels of the two ROI images can be pre-comprehensively annotated by a method combining broth dilution method and image vision. The processor can train the initial Siamese neural network model based on the constructed training sample set to obtain the preset Siamese neural network model. The training loss function can use contrast loss. The hyperparameter (margin, i.e., preset distance threshold) in this paper can be 5. The optimizer selects Adam (Adaptive Moment Estimation), the initial learning rate is 0.001, the input image of the training is resized (changed) to 256, the batch size is 64, the model is iterated for 600 rounds, and l2 regularization (a common regularization type) is used during the training process.

[0077] Correspondingly, when the test control image includes a preset control image of the strain to be tested (such as a preset non-inhibitory image), Figure 4As shown, in this embodiment, the processor can use the preset Siamese neural network model to extract the corresponding dimensionality reduction features (such as dimensionality reduction feature Embedding#1 and dimensionality reduction feature Embedding#2) of each image to be tested (antibiotic bacteria image #1) and the preset control image (positive control image #2), so as to avoid extracting the dimensionality reduction features of the preset control image once for each Euclidean distance calculation; calculate the Euclidean distance (Euclidean Distance) between the dimensionality reduction features of each image to be tested and the dimensionality reduction features of the preset control image; and determine the similarity test results corresponding to each image to be tested according to the Euclidean distance and the preset distance threshold. That is, in this embodiment, the Euclidean distance between the dimensionality reduction features of each image to be tested and the dimensionality reduction features of the preset control image can be used as the similarity to perform the similarity threshold method, that is, if the Euclidean distance is greater than the preset distance threshold (margin), the similarity test result is determined to be dissimilar, and if the Euclidean distance is not greater than the preset distance threshold (margin), the similarity test result is determined to be similar.

[0078] Step 104: Determine the minimum inhibitory concentration of the antibiotic to be tested corresponding to the tested bacterial species according to the similarity test result and the preset gradient concentration.

[0079] It is understandable that in this step, the processor can determine the minimum inhibitory concentration of the tested antibiotic corresponding to the tested bacterial species according to the similarity test results corresponding to each tested image and the preset gradient concentration, so as to realize the detection and identification of the minimum inhibitory concentration of the antibiotic.

[0080] Specifically, the specific manner in which the processor determines the minimum inhibitory concentration of the tested antibiotic corresponding to the tested bacterial species based on the similarity test results and the preset gradient concentration in this step can be set by the designer according to practical scenarios and user needs. For example, the minimum inhibitory concentration of the tested antibiotic corresponding to the tested bacterial species determined by the processor can be any preset gradient concentration; for example, when the test control image includes preset non-inhibitory images of the tested bacterial species and / or images corresponding to the mixed solutions of the tested antibiotic and the tested bacterial species at each preset gradient concentration at the moment when inhibition is not completed, the processor can, when the similarity test results corresponding to all the images to be tested include similarities and dissimilarities, determine the minimum preset gradient concentration corresponding to the image to be tested with a dissimilar similarity test result as the minimum inhibitory concentration, or determine the maximum preset gradient concentration corresponding to the image to be tested with a similarity test result as the minimum inhibitory concentration. When the test control image includes a preset antibacterial completion image of the test bacterial species, the processor can determine the minimum preset gradient concentration corresponding to the test image with similar test results as the minimum inhibitory concentration when the similarity test results corresponding to all the test images include similarities and dissimilarities, or determine the maximum preset gradient concentration corresponding to the test image with dissimilar similarity test results as the minimum inhibitory concentration.

[0081] Correspondingly, the minimum inhibitory concentration of the tested antibiotic corresponding to the tested bacterial species determined by the processor may not be the preset gradient concentration. When the test control image includes the preset non-inhibited image of the tested bacterial species and / or the images corresponding to the mixed solution of the tested antibiotic and the tested bacterial species at each preset gradient concentration at the time when the inhibition is not completed, the processor may determine the minimum inhibitory concentration as the intermediate value (such as the average value of the two) between the minimum preset gradient concentration corresponding to the tested image with a dissimilar similarity test result and the maximum preset gradient concentration corresponding to the tested image with a similarity test result as the minimum inhibitory concentration when the similarity test results corresponding to all the tested images include similarity and dissimilarity. This embodiment does not impose any limitation on this.

[0082] For example, the minimum inhibitory concentration identification results of two common clinical quality control strains (ATCC25922 Escherichia coli and ATCC25923 Staphylococcus aureus) using the method provided in this embodiment can be used to prove the feasibility of this method. Preparation of strains: The strains are transferred to the second to fourth generations for standby use, and the colonies are picked to make a 0.5 McFarland bacterial suspension. The antibiotics are pre-coated in the microwells of the drug sensitivity plate (the antibiotic concentration is prepared according to the CLSI standard). The bacterial liquid is mixed evenly with agarose with a concentration of 0.6%-1.0% and a temperature of 37-40°C at a dilution ratio of 300-500 times, and the sample is added to each microwell by centrifugation, and then the drug sensitivity plate card is placed in an incubator (37±2°C) and incubated for 2h~4h. Data acquisition: After the incubation is completed, the image of each well (i.e., microwell) is taken on a lens-free imaging platform, and the images under different antibiotic concentrations are recorded, so as to use these images to obtain the minimum inhibitory concentration (MIC) identified by the method provided in this embodiment. As shown in Tables 1 and 2, the MIC results identified by the method are compared with the drug sensitivity results of the clinical gold standard broth microdilution test (BMD), and the MIC results identified by the method provided in this embodiment are more accurate and reliable, and the MIC results can be read in 2h to 4h, which is faster.

[0083] Table 1 Drug sensitivity results of this method and BMD method for ATCC25922 Escherichia coli

[0084] antibiotic name This technology (ug / ml) BMD (ug / ml) Remark Amikacin 4 1-4 Within the standard range Ampicillin 4 4 Consistency Cefoxitin 1 2-4 One gradient difference is acceptable Gentamycin 0.25 0.25-0.5 Within the standard range Imipenem 0.12 0.06-0.25 Within the standard range

[0085] Table 2 Drug sensitivity results of this method and BMD method for ATCC25923 Staphylococcus aureus

[0086] antibiotic name This technology (ug / ml) BMD (ug / ml) Remark Erythromycin 0.5 0.25-0.5 Within the standard range Cefoxitin 4 1-4 Within the standard range Ampicillin 1 1 Consistency

[0087] In the present embodiment, the embodiment of the present invention calculates the similarity between each image to be tested and its corresponding test control image to obtain the similarity test result corresponding to each image to be tested. The similarity calculation between the images to be tested of antibiotic-bacterial species with different concentration gradients and their corresponding test control images can be used to determine the similarity or dissimilarity between the images to be tested of each preset gradient concentration and the corresponding test control image. Therefore, the minimum inhibitory concentration of the antibiotic to be tested corresponding to the test bacterial species can be obtained based on the similarity or dissimilarity corresponding to each preset gradient concentration. The minimum inhibitory concentration of the antibiotic can be automatically detected and identified quickly and accurately, thereby reducing the detection cost of the minimum inhibitory concentration of the antibiotic.

[0088] Corresponding to the above method embodiment, the embodiment of the present invention further provides a minimum inhibitory concentration identification device. The minimum inhibitory concentration identification device described below and the minimum inhibitory concentration identification method described above can be referred to each other.

[0089] Please refer to Figure 5 , Figure 5 This is a structural block diagram of a minimum inhibitory concentration identification device provided by an embodiment of the present invention. The device may include:

[0090] The first acquisition module 10 is used to acquire the image to be tested of the strain to be tested; wherein the image to be tested includes images corresponding to the mixed solution of the antibiotic to be tested and the strain to be tested at each preset gradient concentration at the moment when the inhibition is completed;

[0091] The second acquisition module 20 is used to acquire the test control image corresponding to the strain to be tested;

[0092] The similarity calculation module 30 is used to calculate the similarity between each image to be tested and the corresponding test control image, and obtain the similarity test result corresponding to each image to be tested; wherein the similarity test result is similarity or dissimilarity;

[0093] The concentration determination module 40 is used to determine the minimum inhibitory concentration of the antibiotic to be tested corresponding to the tested bacterial species according to the similarity test result and the preset gradient concentration.

[0094] Optionally, the test control image includes a preset control image of the tested bacterial species and / or images corresponding to the mixed solution of the tested antibiotic and the tested bacterial species at each preset gradient concentration at the time when the inhibition is not completed; wherein the time when the inhibition is not completed is before the time when the inhibition is completed.

[0095] Optionally, when the test control image includes images corresponding to the mixed solution of the test antibiotic and the test bacterial species at each preset gradient concentration at the moment when antibacterial inhibition is not completed, the similarity calculation module 30 can be specifically used to calculate the similarity between the test image corresponding to each preset gradient concentration and the test control image, and obtain the similarity test results corresponding to each image to be tested.

[0096] Optionally, the first acquisition module 10 may include:

[0097] The acquisition submodule is used to perform lensless imaging on each microwell after the bacterial suspension of the tested bacterial species is added to each microwell of the drug sensitivity plate for a preset incubation time to obtain the original image of the tested bacterial species; wherein each microwell is provided with a corresponding preset gradient concentration of the tested antibiotic;

[0098] The recognition submodule is used to recognize the micropore area in the original image to be tested;

[0099] The extraction submodule is used to extract a target region of interest of a preset size from each micropore region, and use the image of the target region of interest as the image to be tested.

[0100] Optionally, the minimum inhibitory concentration is any preset gradient concentration.

[0101] Optionally, the concentration determination module 40 may be specifically configured to determine the minimum preset gradient concentration corresponding to the image to be tested whose similarity test result is dissimilar as the minimum inhibitory concentration when the similarity test results corresponding to all the images to be tested include similarity and dissimilarity.

[0102] Optionally, the similarity calculation module 30 may include:

[0103] The model calculation submodule is used to calculate the similarity between each image to be tested and its corresponding test control image using a preset Siamese neural network model, and obtain the similarity test results corresponding to each image to be tested.

[0104] Optionally, the sister network in the preset Siamese neural network model adopts the feature extraction layer of MobileNet-V2.

[0105] Optionally, when the test control image includes a preset control image of the strain to be tested, the model calculation submodule may include:

[0106] A feature extraction unit, used to extract the dimensionality reduction features corresponding to each of the images to be tested and the preset control images by using a preset Siamese neural network model;

[0107] A distance calculation unit, used to calculate the Euclidean distance between the dimension reduction feature of each image to be tested and the dimension reduction feature of a preset reference image;

[0108] The result determination unit is used to determine the similarity test results corresponding to each image to be tested according to the Euclidean distance and a preset distance threshold.

[0109] Optionally, the device may further include:

[0110] Model training module, used to train and obtain a preset Siamese neural network model;

[0111] Among them, the model training module may include:

[0112] an acquisition submodule, for acquiring a training sample set and a training loss function; wherein the training sample set includes a preset number of pairs of training images and image similarity labels corresponding to each pair of training images, the image similarity labels are similar or dissimilar, and the training loss function is L(x0,x1,y)=y||f(x0)-f(x1)||+(1-y)max(0,(m-||f(x0)-f(x1)||)), x0 and x1 are two training images in any pair of training images, f(x0) and f(x1) are dimensionality reduction features extracted by the model corresponding to x0 and x1, y is the image similarity label corresponding to x0 and x1, m is a preset distance threshold, and ||f(x0)-f(x1) is the square of the Euclidean distance;

[0113] The training submodule is used to iteratively train the initial Siamese neural network model using the training sample set and the training loss function to obtain a preset Siamese neural network model.

[0114] In the embodiment, the embodiment of the present invention calculates the similarity between each image to be tested and the corresponding test control image through the similarity calculation module 30, and obtains the similarity test results corresponding to each image to be tested. The similarity calculation between the images to be tested of antibiotic-bacterial species with different concentration gradients and the corresponding test control images can be used to determine the similarity or dissimilarity between the images to be tested of each preset gradient concentration and the corresponding test control image, so that the minimum inhibitory concentration of the antibiotic to be tested corresponding to the test bacterial species can be obtained according to the similarity or dissimilarity corresponding to each preset gradient concentration, and the minimum inhibitory concentration of the antibiotic to be tested can be automatically detected and identified quickly and accurately, thereby reducing the detection cost of the minimum inhibitory concentration of the antibiotic.

[0115] Corresponding to the above method embodiment, the embodiment of the present invention further provides a minimum inhibitory concentration identification device. The minimum inhibitory concentration identification device described below and the minimum inhibitory concentration identification method described above can be referred to each other.

[0116] Please refer to Figure 6 , Figure 6 A schematic diagram of the structure of a minimum inhibitory concentration identification device provided by an embodiment of the present invention. The identification device may include:

[0117] A memory D1, for storing computer programs;

[0118] The processor D2 is used to implement the steps of the method for identifying the minimum inhibitory concentration provided by the above method embodiment when executing the computer program.

[0119] For details, please refer to Figure 7 , Figure 7A schematic diagram of the specific structure of a minimum inhibitory concentration identification device provided in an embodiment of the present invention. The identification device 310 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 322 (for example, one or more processors) and a memory 332, and one or more storage media 330 (for example, one or more mass storage devices) storing application programs 342 or data 344. Among them, the memory 332 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the data processing device. Furthermore, the central processing unit 322 can be configured to communicate with the storage medium 330 to execute a series of instruction operations in the storage medium 330 on the identification device 310.

[0120] The minimum inhibitory concentration identification device 310 may also include one or more power supplies 326, one or more wired or wireless network interfaces 350, one or more input and output interfaces 358, and / or one or more operating systems 341. For example, Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0121] The steps in the method for identifying the minimum inhibitory concentration described above can be implemented by the structure of the device for identifying the minimum inhibitory concentration.

[0122] Corresponding to the above method embodiment, the embodiment of the present invention further provides a computer-readable storage medium. The computer-readable storage medium described below and the minimum inhibitory concentration identification method described above can be referred to each other.

[0123] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for identifying the minimum inhibitory concentration provided by the above method embodiment are implemented.

[0124] The computer-readable storage medium may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or any other readable storage medium that can store program codes.

[0125] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. For the devices, equipment, and computer-readable storage media disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0126] The above is a detailed introduction to the minimum inhibitory concentration identification method, device, equipment and computer-readable storage medium provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, the present invention can also be improved and modified in a number of ways, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A method for identifying minimum inhibitory concentration, characterized in that: include: Acquire the image to be tested of the bacterial species to be tested; wherein the image to be tested includes images corresponding to the mixed solution of the antibiotic to be tested and the bacterial species to be tested at each preset gradient concentration at the moment when the inhibition is completed; Acquire a test control image corresponding to the bacterial species to be tested; Calculating the similarity between each of the images to be tested and the corresponding test control images, and obtaining the similarity test results corresponding to each of the images to be tested; wherein the similarity test results are similar or dissimilar; According to the similarity test result and the preset gradient concentration, the minimum inhibitory concentration of the antibiotic to be tested corresponding to the bacterial species to be tested is determined.

2. The method for identifying the minimum inhibitory concentration according to claim 1, characterized in that: The test control image includes a preset control image of the tested bacterial species and / or images corresponding to the mixed solution of the tested antibiotic and the tested bacterial species at each of the preset gradient concentrations at the time when bacteriostasis is not completed; wherein the time when bacteriostasis is not completed is before the time when bacteriostasis is completed.

3. The method for identifying the minimum inhibitory concentration according to claim 2, characterized in that: When the test control image includes images corresponding to the mixed solution of the test antibiotic of each preset gradient concentration and the test bacterial species at the time when the inhibition is not completed, the similarity between each of the test images and the corresponding test control image is calculated to obtain the similarity test results corresponding to each of the test images, including: The similarity between the image to be tested and the test control image corresponding to each of the preset gradient concentrations is calculated to obtain a similarity test result corresponding to each of the images to be tested.

4. The method for identifying the minimum inhibitory concentration according to claim 1, characterized in that: The step of obtaining the image to be tested of the strain to be tested comprises: After the bacterial suspension of the strain to be tested is added to each microwell of the drug sensitivity plate for a preset incubation time, each microwell is imaged without a lens to obtain the original image of the strain to be tested; wherein each microwell is provided with a preset gradient concentration of the antibiotic to be tested corresponding to the strain to be tested; Identifying a micropore area in the original image to be tested; An object region of interest of a preset size is extracted from each of the micropore regions, and an image of the object region of interest is used as the image to be tested.

5. The method for identifying the minimum inhibitory concentration according to claim 1, characterized in that: The minimum inhibitory concentration is any of the preset gradient concentrations.

6. The method for identifying the minimum inhibitory concentration according to claim 5, characterized in that: In the case where the test control image includes a preset control image of the tested bacterial species and / or images corresponding to the images of the mixed solution of the tested antibiotic at each preset gradient concentration and the tested bacterial species at the time when the inhibition is not completed, determining the minimum inhibitory concentration of the tested antibiotic corresponding to the tested bacterial species according to the similarity test result and the preset gradient concentration includes: When the similarity test results corresponding to all the images to be tested include similarities and dissimilarities, the minimum preset gradient concentration corresponding to the images to be tested whose similarity test results are dissimilarities is determined as the minimum inhibitory concentration.

7. The method for identifying the minimum inhibitory concentration according to any one of claims 1 to 6, characterized in that: The calculating the similarity between each of the images to be tested and the corresponding test control images to obtain the similarity test results corresponding to each of the images to be tested includes: The preset Siamese neural network model is used to calculate the similarity between each of the images to be tested and the corresponding test control images, and obtain the similarity test results corresponding to each of the images to be tested.

8. The method for identifying the minimum inhibitory concentration according to claim 7, characterized in that: The sister network in the preset Siamese neural network model adopts the feature extraction layer of MobileNet-V2.

9. The method for identifying the minimum inhibitory concentration according to claim 7, characterized in that: When the test control image includes a preset control image of the strain to be tested, the preset Siamese neural network model is used to calculate the similarity between each of the images to be tested and the test control image corresponding to each of the images to be tested, and the similarity test results corresponding to each of the images to be tested are obtained, including: Utilizing the preset Siamese neural network model, extracting the dimensionality reduction features corresponding to each of the images to be tested and the preset control images; Calculating the Euclidean distance between each of the reduced-dimensionality features of the image to be tested and the reduced-dimensionality features of the preset control image; According to the Euclidean distance and a preset distance threshold, a similarity test result corresponding to each of the images to be tested is determined.

10. The method for identifying the minimum inhibitory concentration according to claim 9, characterized in that: Also includes: The training process of the preset Siamese neural network model; The training process of the preset Siamese neural network model includes: Obtain a training sample set and a training loss function; wherein the training sample set includes a preset number of pairs of training images and image similarity labels corresponding to each pair of the training images, the image similarity labels are similar or dissimilar, and the training loss function is L(x0,x1,y)=y||f(x0)-f(x1)||+(1-y)max(0,(m-||f(x0)-f(x1)||)), x0 and x1 are two training images in any pair of training images, f(x0) and f(x1) are respectively the dimensionality reduction features extracted by the model corresponding to x0 and x1, y is the image similarity label corresponding to x0 and x1, m is the preset distance threshold, and ||f(x0)-f(x1)|| is the square of the Euclidean distance; The initial Siamese neural network model is iteratively trained using the training sample set and the training loss function to obtain the preset Siamese neural network model.

11. A minimum inhibitory concentration identification device, characterized in that: include: The first acquisition module is used to acquire the image to be tested of the bacterial species to be tested; wherein the image to be tested includes images corresponding to the mixed solution of the antibiotic to be tested and the bacterial species to be tested at each preset gradient concentration at the moment when the inhibition is completed; A second acquisition module is used to acquire a test control image corresponding to the bacterial species to be tested; A similarity calculation module is used to calculate the similarity between each of the images to be tested and the corresponding test control images, and obtain the similarity test results corresponding to each of the images to be tested; wherein the similarity test results are similar or dissimilar; The concentration determination module is used to determine the minimum inhibitory concentration of the antibiotic to be tested corresponding to the bacterial species to be tested according to the similarity test result and the preset gradient concentration.

12. A minimum inhibitory concentration identification device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the method for identifying the minimum inhibitory concentration as claimed in any one of claims 1 to 10 when executing the computer program.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for identifying the minimum inhibitory concentration according to any one of claims 1 to 10 are implemented.

Images