Candida category identification method, device and equipment based on morphological characteristics and medium

Abstract

The invention discloses a candida category identification method, device and equipment based on morphological characteristics and a medium. The method comprises the following steps: firstly, acquiring sample data of a leucorrhea sample microscopic image containing candida spores; obtaining morphological characteristics of each candida spore in the sample data according to the sample data and the semantic segmentation model; determining the category of each candida spore according to a spore distinguishing rule and the morphological characteristics of each candida spore; according to the category of each candida spore, the hyphae in the sample data and a sample distinguishing rule, the candida category in the sample data is determined, and the sample distinguishing rule is used for matching the candida category corresponding to the category of the candida spores. In the process, after morphological feature data are extracted from the leucorrhea sample microscopic image containing the candida through the model, candida category identification can be automatically carried out according to rules, the efficiency of candida category identification can be improved by utilizing the model, and the accuracy can also be improved.

Description

Technical Field

[0001] The present application relates to the technical field of Candida species identification, and in particular to a method, device, equipment and medium for Candida species identification based on morphological characteristics. Background Art

[0002] In recent years, research and development on the classification of Candida species have found that Candida identification can help clinicians develop targeted treatment plans and provide a reference for the clinical prescription of targeted antifungal drugs. In addition, due to Candida identification, patients can be given targeted medications to reduce difficult-to-heal relapses and reduce patient pain and economic burden.

[0003] At present, the identification method of Candida albicans mainly adopts qualitative test under conventional microscope, biochemical identification, color plate identification, and molecular biology method. However, the microscopic examination relies on manual judgment, is prone to errors, and has low accuracy; biochemical identification and color plate identification are time-consuming and labor-intensive, and inefficient. Moreover, the use of molecular biological means is costly and not universal. Therefore, the present application needs a Candida albicans identification method that can overcome the above technical defects. Summary of the invention

[0004] The present application provides a method, device, equipment and medium for identifying the type of Candida based on morphological features, which uses models and judgment rules to identify and distinguish the morphological features in Candida image samples, directly determine the type of Candida, thereby achieving improved efficiency and accuracy at a low cost.

[0005] In a first aspect, the present application provides a method for identifying the type of Candida species based on morphological characteristics, the method comprising:

[0006] Acquiring sample data of a leucorrhea sample microscopic image, wherein the leucorrhea sample microscopic image includes an image of Candida spores;

[0007] Obtaining morphological features of each Candida spore in the sample data according to the sample data and a semantic segmentation model, wherein the semantic segmentation model is obtained by training a neural network using training data of the Candida spores;

[0008] Determining the category of each Candida spore according to a spore differentiation rule and the morphological characteristics of each Candida spore, wherein the spore differentiation rule is used to match the category of the Candida spore corresponding to the morphological characteristics of the Candida spore;

[0009] The Candida category in the sample data is determined according to the category of each Candida spore, hyphae in the sample data and a sample differentiation rule, wherein the hyphae are obtained by identifying the sample data, and the sample differentiation rule is used to match the Candida category corresponding to the category of the Candida spore.

[0010] Optionally, according to the sample data and the semantic segmentation model, the morphological features of each Candida spore in the sample data are obtained, including:

[0011] Input the sample data into the semantic segmentation model to obtain the mask of each Candida spore in the sample data;

[0012] Extracting the contour of each Candida spore according to the mask and the contour recognition algorithm of Candida spore;

[0013] The morphological characteristics of each Candida spore are calculated using the contour, and the morphological characteristics of each Candida spore include: the perimeter, area, roundness and flatness of the contour.

[0014] Optionally, before obtaining the morphological features of each Candida spore in the sample data according to the sample data and the semantic segmentation model, the method further includes:

[0015] Inputting the sample data into the target detection model, identifying the formed components in the sample data, wherein the formed components include Candida monospores, blastospores and hyphae, and the target detection model is obtained by training a neural network with the formed component training data of the leucorrhea sample;

[0016] The step of inputting the sample data into the semantic segmentation model to obtain a mask of each Candida spore in the sample data includes:

[0017] The Candida monospores and blastospores are input into the semantic segmentation model to obtain the mask of each Candida spore in the sample data.

[0018] Optionally, the category of each Candida spore is determined based on spore differentiation rules and morphological characteristics of each Candida spore, including:

[0019] According to the perimeter and area of ​​the outline, obtaining morphological size indication data of each Candida spore;

[0020] According to the roundness and flatness of the outline, shape characteristic indication data of each Candida spore is obtained, wherein the roundness is used to describe the degree to which the outline of the Candida spore is close to a theoretical circle, and the flatness is used to describe the degree to which the outline of the Candida spore is flat;

[0021] The morphological size indication data and the shape characteristic indication data are input into the spore differentiation rule to determine the category of each Candida spore. The spore differentiation rule includes the corresponding relationship between the morphological size indication data, the shape characteristic indication data and the Candida spore category.

[0022] Optionally, determining the Candida category in the sample data according to the category of each Candida spore, hyphae in the sample data, and a sample differentiation rule includes:

[0023] Calculate the total number of Candida spores in the sample data;

[0024] Count the number of different types of Candida spores in the sample data;

[0025] Calculating the number of hyphae in the sample data;

[0026] The Candida category in the sample data is determined based on the total number of Candida spores, the number of different types of Candida spores, the number of hyphae and the sample differentiation rules.

[0027] Optionally, determining the Candida category in the sample data according to the total number of Candida spores, the number of different types of Candida spores and a sample differentiation rule includes:

[0028] The number of different types of Candida spores was divided by the total number of Candida spores to obtain the Candida spore type ratio data;

[0029] The Candida spore category proportion data and the number of hyphae are matched with the sample differentiation rule to determine the Candida category in the sample data. The sample differentiation rule includes the correspondence between the Candida spore category proportion data and the number of hyphae and the Candida category.

[0030] In a second aspect, the present application provides a device for identifying Candida species based on morphological characteristics, the device comprising:

[0031] An acquisition unit, used for acquiring sample data of a leucorrhea sample microscopic image, wherein the leucorrhea sample microscopic image includes an image of Candida spores;

[0032] An obtaining unit, used for obtaining the morphological features of each Candida spore in the sample data according to the sample data and a semantic segmentation model, wherein the semantic segmentation model is obtained by training a neural network with the training data of the Candida spores;

[0033] a determination unit, configured to determine the category of each Candida spore according to a spore differentiation rule and the morphological characteristics of each Candida spore, wherein the spore differentiation rule is configured to match the category of the Candida spore corresponding to the morphological characteristics of the Candida spore;

[0034] The determination unit is further used to determine the Candida category in the sample data according to the category of each Candida spore, hyphae in the sample data and a sample differentiation rule, wherein the hyphae are obtained by identifying the sample data, and the sample differentiation rule is used to match the Candida category corresponding to the category of the Candida spore.

[0035] Optionally, the obtaining unit is specifically used for:

[0036] Inputting the sample data into the semantic segmentation model to obtain a mask of each Candida spore in the sample data;

[0037] Extracting the contour of each Candida spore according to the mask and the contour recognition algorithm of Candida spore;

[0038] The morphological characteristics of each Candida spore are calculated using the contour, and the morphological characteristics of each Candida spore include: perimeter, area, roundness and flatness of the contour.

[0039] Optionally, the device further comprises:

[0040] An identification unit is used to input the sample data into a target detection model to identify the formed components in the sample data, wherein the formed components include Candida monospores, blastospores and hyphae, and the target detection model is obtained by training a neural network with formed component training data of leucorrhea samples;

[0041] The acquisition unit is specifically used for:

[0042] The Candida monospores and blastospores are input into the semantic segmentation model to obtain the mask of each Candida spore in the sample data.

[0043] Optionally, the determining unit is used to determine the category of each Candida spore according to the spore differentiation rule and the morphological characteristics of each Candida spore, specifically for:

[0044] According to the perimeter and area of ​​the outline, obtaining morphological size indication data of each Candida spore;

[0045] According to the roundness and flatness of the outline, shape characteristic indication data of each Candida spore is obtained, wherein the roundness is used to describe the degree to which the outline of the Candida spore is close to a theoretical circle, and the flatness is used to describe the degree to which the outline of the Candida spore is flat;

[0046] The morphological size indication data and the shape characteristic indication data are input into the spore differentiation rule to determine the category of each Candida spore. The spore differentiation rule includes the corresponding relationship between the morphological size indication data, the shape characteristic indication data and the Candida spore category.

[0047] Optionally, the determining unit is used to determine the category of Candida in the sample data according to the category of each Candida spore, hyphae in the sample data, and the sample differentiation rule, including:

[0048] A calculation subunit, used to calculate the total number of Candida spores in the sample data;

[0049] The subunits are also used to calculate the number of different types of Candida spores in the sample data;

[0050] The calculation subunit is further used to calculate the number of hyphae in the sample data;

[0051] The subunit is determined, and is used to determine the Candida category in the sample data according to the total number of Candida spores, the number of Candida spores of different categories, the number of hyphae, and the sample differentiation rule.

[0052] Optionally, the determining subunit has a function for:

[0053] The number of different types of Candida spores was divided by the total number of Candida spores to obtain the Candida spore type ratio data;

[0054] The Candida spore category proportion data and the number of hyphae are matched with the sample differentiation rule to determine the Candida category in the sample data. The sample differentiation rule includes the correspondence between the Candida spore category proportion data and the number of hyphae and the Candida category.

[0055] In a third aspect, the present application provides an electronic device, the electronic device comprising a memory and a processor:

[0056] The memory is used to store computer programs;

[0057] The processor is used to execute the method provided by the first aspect according to the computer program.

[0058] In a fourth aspect, the present application further provides a computer-readable storage medium, which is used to store a computer program, and the computer program is used to execute the method provided in the first aspect above.

[0059] It can be seen that this application has the following beneficial effects:

[0060] The present application provides a method for identifying the classification of Candida albicans based on morphological features. First, sample data of a microscopic image of a leucorrhea sample is obtained, wherein the microscopic image of the leucorrhea sample contains an image of Candida albicans spores; according to the sample data and a semantic segmentation model, the morphological features of each Candida albicans spore in the sample data are obtained, wherein the semantic segmentation model is obtained by training a neural network with the training data of Candida albicans spores; according to the spore differentiation rule and the morphological features of each Candida albicans spore, the category of each Candida albicans spore is determined, wherein the spore differentiation rule is used to match the category of Candida albicans spores corresponding to the morphological features of the Candida albicans spores; according to the category of each Candida albicans spore, the hyphae in the sample data and the sample differentiation rule, the category of Candida albicans in the sample data is determined, wherein the hyphae is obtained by identifying the sample data, and the sample differentiation rule is used to match the Candida albicans category corresponding to the category of the Candida albicans spores. In this process, after the morphological feature data of the microscopic image of the leucorrhea sample containing Candida albicans is extracted through the model, the Candida albicans category can be automatically identified according to the rule, and the efficiency of Candida albicans category identification can be improved by using the model, and the accuracy can also be improved. In addition, the model identification method can also achieve low-cost Candida species identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0062] Figure 1 This is a flow chart of an embodiment of a method for identifying Candida species based on morphological characteristics in the embodiments of the present application;

[0063] Figure 2 This is an architecture diagram of the semantic segmentation model in the embodiment of the present application;

[0064] Figure 3 This is a diagram of the architecture of the target detection model in the embodiment of the present application;

[0065] Figure 4 This is a flow chart of another embodiment of a method for identifying Candida species based on morphological characteristics in the embodiments of the present application;

[0066] Figure 5 This is a schematic diagram of the structure of a Candida species identification device based on morphological characteristics in an embodiment of the present application;

[0067] Figure 6 This is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0069] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0070] At present, the method for identifying Candida albicans mainly adopts qualitative inspection, biochemical identification and color plate identification under conventional microscope, as well as molecular biology. For qualitative inspection under conventional microscope, it is mainly limited to the magnification of visual field and the experience of operators. Specifically, the operation of distinguishing Candida albicans according to morphology requires high professional ability of personnel, and untrained inspectors have basically no ability to distinguish. If trained inspectors want to judge high accuracy, they also require high ability; and the visual field under the microscope is small, and the human eye has weak ability to distinguish morphological features such as size, length, thickness, and roundness, and cannot be durable and easy to fatigue, so the qualitative inspection under conventional microscope relies on manual judgment and is easy to make mistakes, and the accuracy is not high and unstable. In addition, the two methods of biochemical identification and color plate identification are time-consuming, with a cycle of 2 to 3 days, which is difficult to meet clinical needs; the method of molecular biology is used for fast and accurate diagnosis, but the cost is high, the experimental requirements are high, the operation is cumbersome, and fewer hospitals can meet the detection conditions. Therefore, the application needs to provide a method for identifying Candida albicans that can overcome the above technical defects.

[0071] In the examples of the present application, reports and studies were found that different Candida morphological differences can be used for identification, that is, by gradually classifying, labeling, and segmenting the morphological characteristics of different Candida types in different Candida microscopic images, training an artificial intelligence model, and subsequently using the artificial intelligence model to automatically identify Candida, thereby improving efficiency and accuracy at a low cost.

[0072] In specific implementation, the method may, for example, include: acquiring sample data of a leucorrhea sample microscopic image, wherein the leucorrhea sample microscopic image includes an image of Candida spores; obtaining the morphological features of each Candida spore in the sample data according to the sample data and a semantic segmentation model, wherein the semantic segmentation model is obtained by training a neural network through training data of Candida spores; determining the category of each Candida spore according to a spore distinction rule and feature indication data of each Candida spore, wherein the spore distinction rule is used to match the category of Candida spores corresponding to the morphological features of the Candida spores; determining the category of Candida in the sample data according to the category of each Candida spore, hyphae in the sample data, and the sample distinction rule, wherein the hyphae are obtained by identifying the sample data, and the sample distinction rule is used to match the category of Candida corresponding to the category of Candida spores.

[0073] It can be seen that the method provided by the implementation of the present application can obtain the morphological characteristics of each Candida spore in the sample data by inputting the sample data of the image of Candida spores into the semantic segmentation model, and then determine the Candida category in the sample data based on the morphological characteristics of the Candida spores and the established distinction rules, thereby realizing automatic distinction of Candida categories, which not only improves the efficiency of Candida category identification, but also improves the accuracy.

[0074] To facilitate understanding of the specific implementation of a method for identifying Candida species based on morphological characteristics provided in an embodiment of the present application, it will be described below with reference to the accompanying drawings.

[0075] It should be noted that the subject implementing the method for identifying the Candida species based on morphological features can be a device for identifying the Candida species based on morphological features provided in the embodiment of the present application, and the device for identifying the Candida species based on morphological features can be carried in an electronic device or a functional module of an electronic device. The electronic device in the embodiment of the present application can be any device capable of implementing the method for identifying the Candida species based on morphological features in the embodiment of the present application, for example, an Internet of Things (IoT) device.

[0076] Figure 1 The flowchart of a method for identifying the type of Candida albicans based on morphological features provided in an embodiment of the present application is shown in FIG. The method can also be applied to a device for identifying the type of Candida albicans based on morphological features, which can be, for example, Figure 5 The Candida species identification device 500 based on morphological characteristics is shown, or the Candida species identification device based on morphological characteristics can also be integrated into Figure 6 Functional modules in the electronic device 600 are shown.

[0077] like Figure 1 As shown, the method includes the following S101 to S104:

[0078] S101: Acquire sample data of a leucorrhea sample microscopic image, wherein the leucorrhea sample microscopic image includes an image of Candida spores.

[0079] In order to realize the identification of Candida category in sample data, first obtain sample data of leucorrhea sample microscopic images, wherein the leucorrhea sample microscopic images contain images of Candida spores; obtain the morphological characteristics of each Candida spore in the sample data according to the sample data and the semantic segmentation model, wherein the semantic segmentation model is obtained by training the neural network with the training data of Candida spores; determine the category of each Candida spore according to the spore distinction rule and the morphological characteristics of each Candida spore, wherein the spore distinction rule is used to match the category of Candida spores corresponding to the morphological characteristics of Candida spores; determine the category of Candida in the sample data according to the category of each Candida spore, the hyphae of the sample data and the sample distinction rule, wherein the hyphae are obtained by identifying the sample data, and the sample distinction rule is used to match the category of Candida spores corresponding to the category of Candida spores. Therefore, the embodiment of the present application obtains the sample data of the leucorrhea sample microscopic images through S101, and makes a pre-preparation for the subsequent acquisition of the morphological characteristics of each Candida spore in the sample data.

[0080] S102: Obtaining morphological features of each Candida spore in the sample data according to the sample data and a semantic segmentation model, wherein the semantic segmentation model is obtained by training a neural network using training data of the Candida spores.

[0081] As an example, S102 may include: inputting the sample data into the semantic segmentation model to obtain the mask of each Candida spore in the sample data; extracting the contour of each Candida spore according to the mask and the contour recognition algorithm of the Candida spore; using the contour to calculate the morphological characteristics of each Candida spore, wherein the morphological characteristics of each Candida spore include: the perimeter, area, roundness and flatness of the contour. Among them, the roundness describes the degree to which the contour of the Candida spore is close to the theoretical circle, and the calculation formula is, roundness = contour area / contour minimum circumscribed circle area, when the contour is circular, the roundness is 1; the flatness describes the degree to which the Candida contour is flat, and the calculation formula is, flatness = contour fitting ellipse major axis / contour fitting ellipse minor axis, when the contour is circular, the flatness is 1.

[0082] Among them, the above-mentioned Candida spore contour recognition algorithm and the method of using the contour to calculate the morphological characteristics of each Candida spore both use image processing algorithms, and the specific engineering implementation is to use OpenCV to call the corresponding algorithm library.

[0083] It should be noted that the key points of morphological identification of different Candida species are mainly the morphology and size of the megaspores in the Candida blastospores, followed by the size and morphology of the single spores, and finally the presence or absence of hyphae. Based on this, the size and morphological characteristics of spores are crucial for the identification of Candida species, so accurate segmentation of spores is the basis for morphological differentiation. For this purpose, it is necessary to establish a semantic segmentation model that can obtain the morphological characteristics of Candida spores.

[0084] The specific process of establishing the semantic segmentation model may include: firstly manually intercepting the typical morphology of Candida monospores and blastospores of different Candida using image extraction software, then annotating their precise contours using segmentation and annotation software, and using the acquired contours as training samples for training to obtain the semantic segmentation model. Among them, the segmentation and annotation software in the embodiment of the present application can use Labelme.

[0085] like Figure 2As shown, the semantic segmentation model of the embodiment of the present application is based on a fully convolutional neural network and adopts a symmetrical architecture of an encoder-decoder, wherein the left side is a feature extraction network (encoder) that uses ten convolutional layers and four downsamplings to extract multi-scale features while reducing the amount of calculation; the right side is a feature fusion network (decoder) that uses nine convolutional layers and four upsamplings to restore the feature map and output the segmentation result, while adding jump connections to improve positioning accuracy; finally, a weighted loss function is used to improve the segmentation effect of adhesion cells.

[0086] It should be noted that the embodiment of the present application can directly use the semantic segmentation model to segment the sample data to obtain the outline of the Candida spores. However, since conventional leucorrhea detection also focuses on other cells, such as Trichomonas, leukocytes, etc., the embodiment of the present application can also perform target detection on the sample data before using the semantic segmentation model to obtain Candida monospores and blastospores, that is, to improve the accuracy of the subsequent semantic segmentation model.

[0087] To this end, before obtaining the morphological features of each Candida spore in the sample data based on the sample data and the semantic segmentation model, the method also includes: inputting the sample data into a target detection model to identify formed components in the sample data, wherein the formed components include Candida monospores, blastospores and hyphae; when performing semantic segmentation subsequently, the Candida monospores and blastospores are also input into the semantic segmentation model to obtain a mask of each Candida spore in the sample data.

[0088] It should be noted that the target detection model in the embodiment of the present application is used to detect the visible components in the vaginal discharge sample, that is, the microscopic image of the vaginal discharge sample is input, and the location, category, and confidence of the visible components are output. The detectable targets include but are not limited to: Candida monospores, blastospores, and hyphae. Figure 3 As shown, the target detection model in the embodiment of the present application is composed of multiple neural network modules. Taking darknet-53 as an example, each module includes a convolutional layer and an activation layer, and the connection methods between modules include direct connection and jump connection; using a multi-detection head output method, 3 scale target prediction results are provided; the loss function calculation in the target detection model includes position loss calculation, confidence loss calculation, and classification loss calculation.

[0089] The specific process of training the target detection model includes: collecting a large number of microscopic images of real Candida samples, using LabelImg to annotate the positions of Candida monospores, blastospores and hyphae to form a training data set; then using Mosaic, MixUp and random scaling, cropping, flipping and other methods to enhance the data of the original microscopic images; finally, setting the number of training iterations to 300, the mini-batch size to 64, the L2 regularization coefficient to 0.0001, the initial learning rate to 0.001, the momentum parameter to 0.9, and the gradient threshold to 0.05.

[0090] In this process, the semantic segmentation model is trained in advance with the annotated contours of the Candida spores. When the Candida category is subsequently identified, the sample data can be directly input into the semantic segmentation model to automatically obtain the contours of the Candida spores in the sample data and the morphological characteristics of the Candida spores. The process of automating the acquisition of the morphological characteristics of the Candida spores through the semantic segmentation model not only improves the efficiency of Candida category identification, but also improves the accuracy of Candida category identification. And it can be completed using only the software model, which also completes the Candida category identification at a low cost.

[0091] S103: Determine the category of each Candida spore according to a spore differentiation rule and the morphological characteristics of each Candida spore, wherein the spore differentiation rule is used to match the category of the Candida spore corresponding to the morphological characteristics of the Candida spore.

[0092] As an example, S103 may include: obtaining morphological size indication data of each Candida spore according to the perimeter and area of ​​the contour; obtaining shape feature indication data of each Candida spore according to the roundness and flatness of the contour; and finally inputting the morphological size indication data and the shape feature indication data into a spore differentiation rule to determine the category of each Candida spore, wherein the spore differentiation rule includes a correspondence between the morphological size indication data, the shape feature indication data and the Candida spore category.

[0093] Among them, the perimeter or area data of the contour can be used as the spore morphology size indication data, and the roundness and flatness can be used as the spore morphology feature indication data, which can specifically include: for example, when the optical magnification is 40 and the camera pixel size is 3.3 microns, the perimeter of Candida albicans is 55-64, the area is 240-320, the roundness is 0.75-0.95, and the flatness feature is 1.0-1.2; the perimeter of Candida glabrata and Candida portugalensis is 47-53, the area is 170-22 0, roundness is 0.75-0.95, and flatness is 1.0-1.2; the perimeter of Candida parapsilosis is 53-55, the area is 220-240, the roundness is 0.75-0.95, and the flatness is 2-3; the perimeter of Candida krusei is 55-67, the area is 240-340, the roundness is 0.5-0.7, and the flatness is 2-4; the perimeter of Candida tropicalis is >67, the area is >340, the roundness is 0.5-0.7, and the flatness is 1.0-1.2. The above spore morphology size indication data and spore morphology characteristic indication data can be continuously optimized according to the increase in the amount of sample data input into the training model.

[0094] The spore differentiation rule is: when any morphological size indication data and any shape characteristic indication data meet the characteristic data requirements of a certain type of spore, the spore is determined to be this type of Candida: the correspondence between the morphological size indication data, the shape characteristic indication data and the Candida spore category in the spore differentiation rule may include: if a spore area is 260 and the flatness is 1, it is determined to be Candida albicans, a spore area is 200 and the flatness is 1, it is determined to be Candida glabrata, a spore area is 300 and the flatness is 3, it is determined to be Candida krusei, and other types of Candida are similar, and examples are not given here one by one.

[0095] It should be noted that the sample data used as an example in the embodiments of the present application is image acquisition under a 40x microscope system, and a 100x oil objective lens image can also be used.

[0096] In this process, the present application summarizes the method of Candida identification based on the microscopic morphological characteristics of different Candida by observing the microscopic morphological characteristics of a large number of clinical Candida. Therefore, the corresponding Candida category can be determined based on the morphological state of Candida according to the spore differentiation rules.

[0097] S104: Determine the Candida category in the sample data according to the category of each Candida spore, hyphae in the sample data, and a sample differentiation rule, wherein the hyphae are obtained by identifying the sample data, and the sample differentiation rule is used to match the Candida category corresponding to the Candida spore category.

[0098] As an example, S104 may include: calculating the total number of Candida spores in the sample data; calculating the number of different types of Candida spores in the sample data; calculating the number of hyphae in the sample data; and determining the Candida category in the sample data based on the total number of Candida spores, the number of different types of Candida spores and the sample differentiation rule.

[0099] Among them, the above-mentioned determination of the Candida category in the sample data based on the total number of Candida spores, the number of different types of Candida spores and the sample differentiation rule specifically includes: dividing the number of different types of Candida spores by the total number of Candida spores to obtain the Candida spore category proportion data; matching the Candida spore category proportion data and the number of hyphae with the sample differentiation rule to determine the Candida category in the sample data, and the sample differentiation rule includes the correspondence between the Candida spore category proportion data and the number of hyphae and the Candida category.

[0100] The number of different types of Candida spores is divided by the total number of Candida spores to obtain the Candida spore category proportion data D as follows:

[0101] D albicans = the number of spores judged as Candida albicans / the total number of Candida spores × %;

[0102] D smooth = number of spores judged as Candida glabrata / total number of Candida spores × %;

[0103] D krusei = number of spores of Candida krusei / total number of Candida spores × %;

[0104] D parapsilosis = the number of spores judged as Candida parapsilosis / the total number of Candida spores × %;

[0105] D tropical = number of spores judged as tropical Candida spores / total number of Candida spores × %;

[0106] D Portugal = the number of Candida spores judged as Portugal / the total number of Candida spores × %;

[0107] And so on.

[0108] The sample differentiation rules include the data of the proportion of Candida spore categories and the corresponding relationship between the number of hyphae and the Candida categories, and may include:

[0109] (1) If D albicans = 100%, then Candida albicans is output; if D smooth = 100%, then Candida smooth is output; and so on.

[0110] (2) D white + D smooth = 100%, when D white > 0%, Candida albicans is output;

[0111] (3) When D krusei > 0%, Candida krusei is exported;

[0112] (4) When D parasmoothing > 0%, Candida parasmoothing is output;

[0113] (5) When D tropical > 0%, Candida tropicalis is exported;

[0114] (6) When D krusei, D near smooth, and D tropical conflict, Candida krusei is exported;

[0115] (7) When D is nearly smooth and D is tropical, Candida tropicalis is exported;

[0116] (8) D smooth = 100%, and the number of hyphae identified is greater than 0, then Candida albicans is output;

[0117] Except for the above 8 cases, the rest of the outputs are other Candida species.

[0118] It should be noted that the conflict in the above correspondence relationship is specifically illustrated by (6) as an example. When the sample data contains both Candida krusei, Candida parapsilosis, or Candida tropicalis, that is, D krusei>0% and / or D parapsilosis>0% and / or D tropicalis>0%, Candida krusei is output.

[0119] It can be seen that the embodiment of the present application can obtain the morphological characteristics of each Candida spore in the sample data by inputting the sample data of the image of Candida spores into the semantic segmentation model, and then determine the Candida category in the sample data based on the morphological characteristics of the Candida spores and the established distinction rules, thereby realizing automatic distinction of Candida categories, which not only improves the efficiency of Candida category identification, but also improves the accuracy.

[0120] In order to make the method provided in the embodiment of the present application clearer and easier to understand, Figure 4 A specific example of this method is described.

[0121] like Figure 4 As shown, the implementation may include:

[0122] S401: Acquire sample data of a leucorrhea sample microscopic image, wherein the leucorrhea sample microscopic image includes an image of Candida spores.

[0123] In the embodiment of the present application, clinical Candida samples were collected using sample data of microscopic images of leucorrhea samples from Antu Biotechnology AW500.

[0124] S402: Inputting the sample data into the target detection model, identifying the formed components in the sample data, wherein the formed components include Candida monospores, blastospores and hyphae.

[0125] Similarly, in order to improve the accuracy of subsequent identification, the embodiment of the present application first inputs the sample data into the target detection model to identify the tangible components in the sample data so that subsequent identification can be directly based on the tangible components. The target detection model can be the target detection model obtained by training multiple neural network modules as exemplified in the above embodiment.

[0126] S403: Input the Candida spores and blastospores into the semantic segmentation model to obtain the mask of each Candida spore in the sample data.

[0127] Since the identification of Candida albicans in the embodiment of the present application is determined by morphological features, it is first necessary to obtain the mask of Candida albicans spores according to the semantic segmentation model, wherein the semantic segmentation model can be the semantic segmentation model obtained by training a fully convolutional neural network as exemplified in the above embodiment.

[0128] S404: Using an image processing algorithm to process the mask of each Candida spore to obtain the morphological characteristics of each Candida spore.

[0129] After obtaining the mask of each Candida spore in the sample data, it is necessary to use an image processing algorithm to process the mask of each Candida spore, specifically including: extracting the contour of each Candida spore according to the mask and the contour recognition algorithm of the Candida spore; using the contour to calculate the morphological characteristics of each Candida spore, the morphological characteristics of each Candida spore include: the perimeter, area, roundness and flatness of the contour. That is, the morphological characteristics of each Candida spore are obtained using the image processing algorithm.

[0130] S405: Determine the category of each Candida spore according to a spore differentiation rule and the morphological characteristics of each Candida spore, wherein the spore differentiation rule includes a correspondence between the morphological characteristics of the Candida spore and Candida albicans, Candida glabrata, and Candida krusei.

[0131] In the examples of the present application, only the spore differentiation rules of Candida albicans, Candida glabrata and Candida krusei are used as an illustration, that is, other types of Candida are classified into other Candida. Among them, the corresponding relationship between the morphological characteristics of Candida spores and Candida albicans, Candida glabrata and Candida krusei is shown in the above examples, and will not be repeated here.

[0132] S406: Divide the number of different types of Candida spores by the total number of Candida spores to obtain Candida spore type ratio data.

[0133] Because the determination result of the Candida category in the sample data depends on the proportion of different types of Candida spores in the total number of Candida spores, it is necessary to divide the number of different types of Candida spores by the total number of Candida spores to obtain the Candida spore category proportion data. The specific calculation formula is shown in the above embodiment and will not be repeated here.

[0134] S407: Matching the Candida spore category proportion data and the number of hyphae with the sample differentiation rule to determine the Candida category in the sample data, wherein the sample differentiation rule includes the correspondence between the Candida spore category proportion data and the number of hyphae and Candida albicans, Candida glabrata and Candida krusei.

[0135] Since only the identification of Candida albicans, Candida glabrata and Candida krusei is listed in the examples of the present application, the sample differentiation rules in the examples of the present application only include the correspondence between Candida albicans, Candida glabrata and Candida krusei, and the rest belong to other Candida species. The specific correspondence is shown in the above examples and will not be repeated here.

[0136] S408: Use the chromogenic plate mass spectrometry identification method to compare and test the Candida species in the sample data.

[0137] After obtaining the Candida species in the sample data, in order to test the accuracy of the method, the embodiment of the present application used a chromogenic plate mass spectrometry identification method (chromogenic plate + mass spectrometry identification) for a comparison, and the results are shown in Table 1.

[0138] Table 1 Comparison test table

[0139]

[0140]

[0141] In the above Table 1, according to the method of morphological feature identification of the present application, in the sample data identified as Candida albicans, there are 174 Candida albicans, 1 Candida glabrata and 4 other Candida spp., with a total of 179 Candida spp.; in the sample data identified as Candida glabrata, there are 2 Candida albicans, 53 Candida glabrata and 3 other Candida spp., with a total of 57 Candida spp.; in the sample data identified as Candida krusei, there are 1 Candida albicans, 8 Candida krusei and 1 other Candida spp., with a total of 10 Candida spp.; in the sample data identified as other Candida spp., there are 1 Candida krusei and 4 other Candida spp., with a total of 5 Candida spp.; the total number of Candida spp. is 251.

[0142] In the above Table 1, using the method of chromogenic plate mass spectrometry, in the sample data identified as Candida albicans, there are 174 Candida albicans, 2 Candida glabrata and 1 Candida krusei, with a total of 177 Candida species; in the sample data identified as Candida glabrata, there are 1 Candida albicans and 52 Candida glabrata, with a total of 53 Candida species; in the sample data identified as Candida krusei, there are 8 Candida krusei and 1 other Candida species, with a total of 9 Candida species; in the sample data identified as other Candida species, there are 4 Candida albicans, 3 Candida glabrata, 1 Candida krusei and 4 other Candida species, with a total of 12 Candida species; the total number of Candida species is 251.

[0143] In this regard, the accuracy of the morphological feature identification method of the present application can be tested by using several indicator data, including overall compliance rate, white-reading sensitivity, white-reading positive expected value, smooth sensitivity, smooth positive expected value, Kerro sensitivity and Kerro positive expected value.

[0144] (1) Calculate the overall compliance rate. The specific formula is: overall compliance rate = (a0 + b0 + c0 + d0) / E ×%, that is, overall compliance rate = (174 + 52 + 8 + 4) / 251 × % = 94.82%.

[0145] (2) Calculate the white-light sensitivity, the specific formula is white-light sensitivity = a0 / A1×%, that is, white-light sensitivity = 174 / 177×% = 98.31%; calculate the white-light positive expected value, the specific formula is white-light positive expected value = a0 / A2×%, that is, white-light positive expected value = 174 / 179×% = 97.21%.

[0146] (3) Calculate the smooth sensitivity. The specific formula is smooth sensitivity = b0 / B1×%, that is, smooth sensitivity = 52 / 53×% = 98.11%; calculate the smooth positive expected value. The specific formula is smooth positive expected value = b0 / B2×%, that is, smooth positive expected value = 52 / 57×% = 91.23%.

[0147] (4) Calculate the Kru sensitivity, the specific formula is Kru sensitivity = c0 / C1×%, that is, Kru sensitivity = 8 / 9×% = 88.89%; calculate the Kru positive expected value, the specific formula is Kru positive expected value = c0 / B2×% = 8 / 10×% = 80.00%.

[0148] That is, it can be seen from the above-calculated multiple index values ​​that the overall compliance rate of the morphological feature identification method in the embodiment of the present application can reach more than 94%, and it only takes 2 minutes to complete the Candida differentiation at the fastest, which is fast and convenient.

[0149] This embodiment provides a method for identifying the Candida species based on morphological features. First, sample data of images of Candida spores are input into a target detection model to identify the formed components therein, and the identified Candida spores are input into a semantic segmentation model for segmentation, so that a mask of each Candida spore in the sample data can be obtained, and then the morphological features of each Candida spore are obtained based on the mask and the image processing algorithm, and then the Candida species in the sample data are determined based on the morphological features of the Candida spores and the established distinction rules, thereby realizing automatic distinction of Candida species, which not only improves the efficiency of Candida species identification, but also improves the accuracy. In addition, the method of model identification can also realize low-cost Candida species identification.

[0150] See also Figure 5 The present application embodiment provides a Candida species identification device 500 based on morphological characteristics, the device comprising:

[0151] An acquisition unit 501 is used to acquire sample data of a leucorrhea sample microscopic image, wherein the leucorrhea sample microscopic image includes an image of Candida spores;

[0152] An obtaining unit 502 is used to obtain the morphological features of each Candida spore in the sample data according to the sample data and a semantic segmentation model, wherein the semantic segmentation model is obtained by training a neural network using the training data of the Candida spores;

[0153] A determination unit 503 is used to determine the category of each Candida spore according to a spore differentiation rule and a morphological feature of each Candida spore, wherein the spore differentiation rule is used to match the category of the Candida spore corresponding to the morphological feature of the Candida spore;

[0154] The determination unit 503 is further used to determine the Candida category in the sample data according to the category of each Candida spore, hyphae in the sample data and a sample differentiation rule, wherein the hyphae are obtained by identifying the sample data, and the sample differentiation rule is used to match the Candida category corresponding to the Candida spore category.

[0155] Optionally, the obtaining unit 502 is specifically configured to:

[0156] Inputting the sample data into the semantic segmentation model to obtain a mask of each Candida spore in the sample data;

[0157] Extracting the contour of each Candida spore according to the mask and the contour recognition algorithm of Candida spore;

[0158] The morphological characteristics of each Candida spore are calculated using the contour, and the morphological characteristics of each Candida spore include: perimeter, area, roundness and flatness of the contour.

[0159] Optionally, the device 500 further includes:

[0160] An identification unit is used to input the sample data into a target detection model to identify the formed components in the sample data, wherein the formed components include Candida monospores, blastospores and hyphae, and the target detection model is obtained by training a neural network with formed component training data of leucorrhea samples;

[0161] The obtaining unit 502 is specifically used for:

[0162] The Candida monospores and blastospores are input into the semantic segmentation model to obtain the mask of each Candida spore in the sample data.

[0163] Optionally, the determining unit 503 is used to determine the category of each Candida spore according to the spore differentiation rule and the morphological characteristics of each Candida spore, specifically for:

[0164] According to the perimeter and area of ​​the outline, obtaining morphological size indication data of each Candida spore;

[0165] According to the roundness and flatness of the outline, shape characteristic indication data of each Candida spore is obtained, wherein the roundness is used to describe the degree to which the outline of the Candida spore is close to a theoretical circle, and the flatness is used to describe the degree to which the outline of the Candida spore is flat;

[0166] The morphological size indication data and the shape characteristic indication data are input into the spore differentiation rule to determine the category of each Candida spore. The spore differentiation rule includes the corresponding relationship between the morphological size indication data, the shape characteristic indication data and the Candida spore category.

[0167] Optionally, the determining unit 503 is configured to determine the type of Candida in the sample data according to the type of each Candida spore, hyphae in the sample data, and a sample differentiation rule, including:

[0168] A calculation subunit, used to calculate the total number of Candida spores in the sample data;

[0169] The subunits are also used to calculate the number of different types of Candida spores in the sample data;

[0170] The calculation subunit is further used to calculate the number of hyphae in the sample data;

[0171] The subunit is determined, and is used to determine the Candida category in the sample data according to the total number of Candida spores, the number of Candida spores of different categories, the number of hyphae, and the sample differentiation rule.

[0172] Optionally, the determining subunit has a function for:

[0173] The number of different types of Candida spores was divided by the total number of Candida spores to obtain the Candida spore type ratio data;

[0174] The Candida spore category proportion data and the number of hyphae are matched with the sample differentiation rule to determine the Candida category in the sample data. The sample differentiation rule includes the correspondence between the Candida spore category proportion data and the number of hyphae and the Candida category.

[0175] It should be noted that the specific implementation method and effect of the Candida species identification device 500 based on morphological characteristics can be found in the above Figure 1 or Figure 4 The relevant descriptions in the provided methods will not be repeated here.

[0176] The present application embodiment also provides an electronic device 600, such as Figure 6 As shown, the device 600 includes a memory 601 and a processor 602:

[0177] The memory 601 is used to store computer programs;

[0178] The processor 602 is used to execute the above Figure 1 or Figure 4 The method provided.

[0179] In addition, the present application also provides a computer-readable storage medium, the computer-readable storage medium is used to store a computer program, and the computer program is used to execute Figure 1 or Figure 4 The method provided.

[0180] Through the description of the above implementation methods, it can be known that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a storage medium, such as a read-only memory (ROM) / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network communication device such as a router) to execute the methods described in each embodiment of the present application or some parts of the embodiments.

[0181] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, in which the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the goals of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.

[0182] The above description is merely an exemplary embodiment of the present application and is not intended to limit the protection scope of the present application.

Claims

1. A method for identifying Candida species based on morphological characteristics. It is characterized in that The method comprises: Acquiring sample data of a leucorrhea sample microscopic image, wherein the leucorrhea sample microscopic image includes an image of Candida spores; Obtaining the morphological features of each Candida spore in the sample data according to the sample data and the semantic segmentation model, wherein the semantic segmentation model is obtained by training a neural network using the training data of the Candida spores; Determining the category of each Candida spore according to a spore differentiation rule and the morphological characteristics of each Candida spore, wherein the spore differentiation rule is used to match the category of the Candida spore corresponding to the morphological characteristics of the Candida spore; The Candida category in the sample data is determined according to the category of each Candida spore, hyphae in the sample data and a sample differentiation rule, wherein the hyphae are obtained by identifying the sample data, and the sample differentiation rule is used to match the Candida category corresponding to the category of the Candida spore.

2. The method according to claim 1, It is characterized in that The step of obtaining the morphological features of each Candida spore in the sample data according to the sample data and the semantic segmentation model includes: Inputting the sample data into the semantic segmentation model to obtain a mask of each Candida spore in the sample data; extracting the contour of each Candida spore according to the mask and the contour recognition algorithm of Candida spores; The morphological characteristics of each Candida spore are calculated using the contour, and the morphological characteristics of each Candida spore include: the perimeter, area, roundness and flatness of the contour.

3. The method according to claim 2, It is characterized in that Before obtaining the morphological features of each Candida spore in the sample data according to the sample data and the semantic segmentation model, the method further includes: Inputting the sample data into a target detection model to identify formed components in the sample data, wherein the formed components include Candida monospores, blastospores and hyphae, and the target detection model is obtained by training a neural network using formed component training data of leucorrhea samples; The step of inputting the sample data into the semantic segmentation model to obtain a mask of each Candida spore in the sample data includes: The Candida monospore and blastospore are input into the semantic segmentation model to obtain a mask of each Candida spore in the sample data.

4. The method according to claim 2, It is characterized in that Determining the category of each Candida spore according to the spore differentiation rule and the morphological characteristics of each Candida spore comprises: According to the perimeter and area of ​​the outline, obtaining morphological size indication data of each Candida spore; According to the roundness and flatness of the outline, shape characteristic indication data of each Candida spore is obtained, wherein the roundness is used to describe the degree to which the outline of the Candida spore is close to a theoretical circle, and the flatness is used to describe the degree to which the outline of the Candida spore is flat; Obtaining flatness indication data of each Candida spore according to the ratio of the major and minor axes of the fitted ellipse of the contour; The morphological size indication data and the shape feature indication data are input into the spore differentiation rule to determine the category of each Candida spore. The spore differentiation rule includes the corresponding relationship between the morphological size indication data, the shape feature indication data and the Candida spore category.

5. The method according to claim 1, It is characterized in that The determining the Candida category in the sample data according to the category of each Candida spore, the number of hyphae in the sample data and the sample differentiation rule comprises: Calculating the total number of Candida spores in the sample data; Calculate the number of different types of Candida spores in the sample data; Calculating the number of hyphae in the sample data; The Candida category in the sample data is determined according to the total number of the Candida spores, the number of the Candida spores of different categories, the number of hyphae and the sample differentiation rule.

6. The method according to claim 5, It is characterized in that The determining the Candida category in the sample data according to the total number of the Candida spores, the number of the different types of Candida spores and the sample differentiation rule comprises: Dividing the number of the different types of Candida spores by the total number of the Candida spores to obtain the Candida spore type ratio data; The Candida spore category proportion data and the number of hyphae are matched with the sample differentiation rule to determine the Candida category in the sample data. The sample differentiation rule includes the correspondence between the Candida spore category proportion data and the number of hyphae and the Candida category.

7. A device for identifying Candida species based on morphological characteristics, It is characterized in that The device comprises: An acquisition unit, used for acquiring sample data of a leucorrhea sample microscopic image, wherein the leucorrhea sample microscopic image includes an image of Candida spores; an obtaining unit, configured to obtain the morphological features of each Candida spore in the sample data according to the sample data and a semantic segmentation model, wherein the semantic segmentation model is obtained by training a neural network using training data of Candida spores; a determination unit, configured to determine the category of each Candida spore according to a spore differentiation rule and the morphological characteristics of each Candida spore, wherein the spore differentiation rule is used to match the category of the Candida spore corresponding to the morphological characteristics of the Candida spore; The determination unit is further used to determine the Candida category in the sample data according to the category of each Candida spore, hyphae in the sample data and a sample differentiation rule, wherein the hyphae are obtained by identifying the sample data, and the sample differentiation rule is used to match the Candida category corresponding to the Candida spore category.

8. The device according to claim 7, It is characterized in that The obtaining unit is specifically used for: Inputting the sample data into the semantic segmentation model to obtain a mask of each Candida spore in the sample data; extracting the contour of each Candida spore according to the mask and the contour recognition algorithm of Candida spores; The morphological characteristics of each Candida spore are calculated using the contour, and the morphological characteristics of each Candida spore include: the perimeter, area, roundness and flatness of the contour.

9. An electronic device, It is characterized in that The electronic device comprises a memory and a processor, wherein the processor is used to execute a program stored in the memory to run the method according to any one of claims 1 to 6.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium is used to store a computer program, and the computer program is used to execute the method according to any one of claims 1 to 6.

Images