Fungus detection method, device and equipment based on image recognition assistance and medium

Abstract

The invention discloses a fungus detection method, device and equipment based on image recognition assistance and a medium. The method comprises the following steps: acquiring a to-be-detected image of to-be-detected sample liquid based on image acquisition equipment; fusing the to-be-detected image according to the acquisition angle to obtain a target fused image; performing feature extraction on the fused image to obtain a feature extraction result; performing matching in a first fungus feature database based on the fungus morphological features and the color features to obtain first target fungus features, and performing matching in a second fungus feature database based on texture features formed by feature metabolite dyeing to obtain second target fungus features; determining the type of fungi in the to-be-detected sample liquid based on the first target fungi characteristics, and determining the survival state of the fungi in the to-be-detected sample liquid based on the second target fungi characteristics. According to the invention, fungi can be accurately and rapidly detected and identified.

Description

Technical Field

[0001] The present invention relates to the field of biotechnology, and in particular, to a fungal detection method, device, equipment and medium assisted by image recognition. Background Art

[0002] In the current fields of medical and biological research, fungal infection has become an increasingly serious problem. With the extensive development of immunosuppressive therapy, organ transplantation surgery, and the abuse of antibiotics, the incidence of fungal infection shows a significant upward trend. Traditional fungal detection methods include microscopic observation and manual observation methods.

[0003] Microscopic observation methods include direct smear microscopy and histopathological section examination. Although they can provide information quickly to a certain extent, their sensitivity is relatively low, and it is easy to miss the diagnosis of samples with a small number of fungi or atypical morphology. In addition, these methods often have difficulty in distinguishing live bacteria from dead bacteria and cannot provide an accurate basis for evaluating the treatment effect. The manual observation method is to collect a fungal distribution map and detect and identify fungi by manually comparing with known normal and abnormal sample images. However, the subjectivity and misjudgment caused by fatigue in manual film reading may lead to deviation of the detection results. Summary of the Invention

[0004] In view of the above problems, the present invention provides a fungal detection method, device, equipment and medium assisted by image recognition to achieve accurate and rapid detection and identification of fungi.

[0005] In a first aspect, the present invention provides a fungal detection method assisted by image recognition, including:

[0006] Collecting a to-be-detected image of a to-be-detected sample solution based on an image acquisition device; the to-be-detected image is an image collected after the to-be-detected sample solution is stained with a fungal triple fluorescence staining solution;

[0007] Fusing the to-be-detected image according to the acquisition angle to obtain a target fused image;

[0008] Performing feature extraction on the fused image to obtain a feature extraction result; the feature extraction result includes fungal morphological features, color features, and texture features formed by staining with characteristic metabolites;

[0009] Match in the first fungal feature database based on the morphological and color features of the fungi to obtain the first target fungal feature, and match in the second fungal feature database based on the texture features formed by the staining of characteristic metabolites to obtain the second target fungal feature; the first fungal feature database contains the standard features of various fungi under specific culture and staining conditions, and the second fungal feature database contains the texture features formed by the staining of the metabolites of live fungi and the texture features formed by the staining of the metabolites of dead fungi;

[0010] Determine the type of fungi in the sample liquid to be detected based on the first target fungal feature, and determine the survival status of the fungi in the sample liquid to be detected based on the second target fungal feature.

[0011] In a second aspect, the present invention provides a fungal detection device assisted by image recognition, including:

[0012] An image acquisition module, configured to acquire a to-be-detected image of the sample liquid to be detected based on an image acquisition device; the to-be-detected image is an image acquired after the sample liquid to be detected is stained with a triple fluorescence staining solution for fungi;

[0013] An image fusion module, configured to fuse the to-be-detected images according to the acquisition angles to obtain a target fused image;

[0014] A feature extraction module, configured to extract features from the fused image to obtain a feature extraction result; the feature extraction result includes fungal morphological features, color features, and texture features formed by the staining of characteristic metabolites;

[0015] A feature matching module, configured to match based on the fungal morphological features and color features in the first fungal feature database to obtain a first target fungal feature, and match based on the texture features formed by the staining of characteristic metabolites in the second fungal feature database to obtain a second target fungal feature; the first fungal feature database contains the standard features of various fungi under specific culture and staining conditions, and the second fungal feature database contains the texture features formed by the staining of the metabolites of live fungi and the texture features formed by the staining of the metabolites of dead fungi;

[0016] A fungal detection module, configured to determine the type of fungi in the sample liquid to be detected based on the first target fungal feature, and determine the survival status of the fungi in the sample liquid to be detected based on the second target fungal feature.

[0017] In a third aspect, the present invention further provides an electronic device, including: a memory, configured to store a computer software program; a processor, configured to read and execute the computer software program, so as to implement the fungal detection method assisted by image recognition as described in the first aspect above.

[0018] Fourthly, the present invention also provides a non-transitory computer-readable storage medium, in which a computer software program is stored, and when the computer software program is executed by a processor, the fungal detection method assisted by image recognition as described in the first aspect above is implemented.

[0019] Fifthly, the present invention provides a computer program product, including a computer program, and when the computer program is executed by a processor, the fungal detection method assisted by image recognition as described in the first aspect above is implemented.

[0020] The fungal detection method assisted by image recognition provided by the present invention can obtain more comprehensive fungal sample information through multi-angle high-resolution image acquisition, and avoid missed diagnosis due to a small number of fungi or atypical morphology in the sample. During the image preprocessing, segmentation and feature extraction, through feature extraction, the morphological features, color features of fungi and the texture features formed by staining with characteristic metabolites can be comprehensively judged, and the fungal features can be accurately highlighted, unlike the microscopic observation method which is difficult to distinguish between live bacteria and dead bacteria. Therefore, compared with the manual observation method, feature extraction and matching based on an objective image recognition algorithm avoid the subjectivity of manual film reading and misjudgment caused by fatigue, greatly improving the accuracy of the detection result and achieving accurate and rapid detection and identification of fungi. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only one embodiment of the present invention, and those skilled in the art can obtain other drawings without creative efforts based on these drawings.

[0022] Figure 1 is a schematic flow chart of the fungal detection method assisted by image recognition provided by the present invention;

[0023] Figure 2 is a schematic structural diagram of the fungal detection device assisted by image recognition provided by the present invention;

[0024] Figure 3 is a schematic structural diagram of the electronic device provided by the present invention;

[0025] Figure 4 is a schematic diagram of an embodiment of the computer-readable storage medium provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0027] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, the meaning of "a plurality" is two or more, unless otherwise specifically defined. In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or having more advantages than other embodiments. In order for any person skilled in the art to implement and use the present invention, the following description is given. In the following description, details are set forth for the purpose of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other instances, well-known structures and processes will not be elaborated in detail to avoid unnecessary details from obscuring the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope that conforms to the principles and features disclosed in the present invention.

[0028] Optionally, referring to Figure 1 , Figure 1 is a schematic flowchart of the fungal detection method assisted by image recognition provided by the present invention. The execution subject of the fungal detection method assisted by image recognition provided by the present invention can be a fungal detection device, such as Figure 1 shown, the fungal detection method assisted by image recognition includes the following:

[0029] Step 10, acquiring a to-be-detected image of a to-be-detected sample solution based on an image acquisition device. The to-be-detected image is an image acquired after the to-be-detected sample solution is stained with a fungal triple fluorescence staining solution.

[0030] Optionally, the image acquisition device in the fungal detection device is used to obtain an image of the to-be-detected sample solution. Before this, the to-be-detected sample solution needs to be stained with a fungal triple fluorescence staining solution. This staining solution can produce a specific fluorescence reaction with the fungi in the sample, making the fungi easier to identify and analyze in the image. The image acquisition device needs to have appropriate parameters such as resolution and sensitivity to ensure that the acquired image is clear and contains sufficient detailed information for subsequent processing and analysis.

[0031] In one embodiment, the fungus detection device is equipped with a high-resolution fluorescence microscope camera as an image acquisition device. First, prepare the sample liquid to be detected, place it on a glass slide, and evenly drip the fungal triple fluorescent staining solution. After an appropriate staining time, place the glass slide on the stage of the fluorescence microscope. The fluorescent substance in the sample is stimulated to emit light through the optical path system of the microscope. The camera is set to a high-resolution mode (e.g., 5000*5000 pixels), and the sensitivity is adjusted to obtain a clear fluorescent image. At this time, the fungus detection device controls the camera to capture an image of the sample liquid to be detected, and obtains an image to be detected containing the fungi to be detected. The fungi in the image appear brightly colored due to fluorescent staining, which forms a sharp contrast with the background.

[0032] Step 20, fuse the images to be detected according to the acquisition angle to obtain a target fused image.

[0033] Furthermore, since the images to be detected acquired from different acquisition angles may contain information on different aspects of fungi, in order to more comprehensively analyze the characteristics of fungi, it is necessary to fuse these images. The fungus detection device processes multiple images to be detected according to the acquisition angles, and merges the useful information in these images through a specific image fusion algorithm to generate a target fused image. The fused image can comprehensively reflect the morphology, color and other characteristics of fungi at various angles, and provide richer data for subsequent feature extraction, as described in steps 201 to 204.

[0034] Continuing with the above example, the fungus detection device collects 4 images to be detected from multiple angles (such as 0°, 30°, 60°, 90°, etc.) by controlling the rotation of the stage of the fluorescence microscope. The device uses a weighted average fusion algorithm, which assigns different weights to each image based on factors such as the clarity and contrast of the image at each angle. For example, the image clarity at 0° is the highest, and the weight is set to 0.4; the images at 30° and 60° are second, and the weights are set to 0.25 respectively; the image at 90° is relatively poor, and the weight is set to 0.1. Through algorithm calculation, the color values of the corresponding pixels of the four images are weighted averaged according to the weights, and finally a target fused image is generated.

[0035] Step 30, extracting features from the fused image to obtain feature extraction results. The feature extraction results include fungal morphological features, color features, and texture features formed by staining with characteristic metabolites.

[0036] Further, according to the characteristics of fungi, the features extracted by the fungal detection device mainly include fungal morphological features (such as shape, size, thickness and branching of hyphae, etc.), color features (due to fluorescence staining, fungi exhibit specific colors, and information such as hue and saturation can be used as features), and texture features formed by staining with characteristic metabolites (unique texture patterns are formed on the surface or around the fungi after the metabolites react with the staining solution). In one embodiment, for morphological feature extraction, the fungal detection device uses an edge detection algorithm to identify the contour of the fungi, calculates the size of the fungi through parameters such as the perimeter and area of the contour, and analyzes the shape of the contour to determine whether it is circular, oval or irregular, etc. For the hyphae part, a thinning algorithm is used to thin the hyphae to a single-pixel width. In terms of color feature extraction, the fused image is converted from the RGB color space to the HSV color space, and the hue, saturation and value of the fungal region are extracted as color features. For texture feature extraction, the gray-level co-occurrence matrix algorithm is used to calculate parameters such as contrast, correlation, energy and entropy of the texture of the fungal region. For example, it is analyzed that the target fungus has an oval shape, the major axis length is 20 μm, and the minor axis length is 15 μm; in the color features, the hue value is 120, the saturation is 0.8, and the value is 0.6; in the texture features, the contrast is 0.5, the correlation is 0.6, the energy is 0.7, and the entropy is 0.4.

[0037] Step 40, based on the fungal morphological features and color features, perform matching in the first fungal feature database to obtain the first target fungal feature, and based on the texture features formed by staining with characteristic metabolites, perform matching in the second fungal feature database to obtain the second target fungal feature.

[0038] Further, the fungal detection device performs feature matching in two different fungal feature databases respectively. The first fungal feature database stores the standard features of various fungi under specific culture and staining conditions. The device compares the extracted fungal morphological features and color features with the features in this database to find the most matching feature and determine the first target fungal feature, as specifically described in steps 401 to 404.

[0039] Further, the second fungal feature database contains the texture features formed by staining the metabolites of live and dead fungi. The device performs matching of the extracted texture features in this database to determine the second target fungal feature, as specifically described in steps 405 to 408.

[0040] Step 50, determine the type of fungi in the sample liquid to be detected based on the first target fungal feature, and determine the survival status of the fungi in the sample liquid to be detected based on the second target fungal feature.

[0041] Further, the fungal detection device determines the types of fungi in the sample liquid to be detected based on the first target fungal features, as specifically described in Steps 501 to 503. Further, the fungal detection device can judge the survival state of the fungi in the sample liquid to be detected based on the second target fungal features. Since the second fungal feature database differentiates the staining texture features of the metabolites of live and dead fungi, a conclusion on whether the fungi are alive or dead can be drawn according to the matching results, as specifically described in Steps 504 to 506.

[0042] In the embodiment of the present invention, through multi-angle high-resolution image acquisition, more comprehensive fungal sample information can be obtained, avoiding missed diagnosis due to a small number of fungi or atypical morphology in the sample. In the processes of image preprocessing, segmentation, and feature extraction, through feature extraction, the fungal morphological features, color features, and texture features formed by the staining of characteristic metabolites of fungi can be comprehensively judged, which can accurately highlight the fungal features. Different from the microscopic observation method that is difficult to distinguish between live and dead fungi, therefore, compared with the manual observation method, based on the objective image recognition algorithm for feature extraction and matching, the subjectivity of manual film reading and misjudgment caused by fatigue are avoided, greatly improving the accuracy of the detection results and achieving accurate and rapid detection and identification of fungi.

[0043] In one embodiment, the descriptions of Steps 201 to 204 are as follows:

[0044] Step 201: Group each image to be detected into multiple initial detection image groups according to a preset angle interval based on the acquisition angle.

[0045] Optionally, the fungal detection device classifies each acquired image to be detected according to a preset angle interval based on the acquisition angle. The preset angle interval is set in advance according to the detection requirements and the necessity of multi-angle observation of fungi. Through this grouping method, the images to be detected with similar acquisition angles can be grouped together. Continuing with the previous example where the fungal detection device uses a fluorescence microscope camera to collect images of the sample liquid to be detected from multiple angles. For example, the preset angle interval set by the device is 30°, that is, 0° - 30° is a group, 30° - 60° is a group, 60° - 90° is a group, and so on. The camera collects 6 images to be detected from angles such as 0°, 15°, 35°, 45°, 65°, and 75°. Then, the images collected at 0° and 15° will be assigned to the initial detection image group of 0° - 30°; the images collected at 35° and 45° will be assigned to the initial detection image group of 30° - 60°; the images collected at 65° and 75° will be assigned to the initial detection image group of 60° - 90°, thus obtaining multiple initial detection image groups.

[0046] Step 202: Arrange the images to be detected within each initial detection image group according to the size of the acquisition angle to obtain an updated detection image group.

[0047] Further, for each initial detection image group, the fungal detection device rearranges the images to be detected within the group in ascending order of the acquisition angle. This can make the images present an orderly angular change, facilitating better utilization of the correlation between images at different angles during subsequent feature extraction and fusion processes, so that the fusion result can more accurately reflect the true situation of the fungus under continuous angular changes. Taking the initial detection image group of 0° - 30° as an example, there are images acquired at 0° and 15° within the group. The fungal detection device identifies the acquisition angles of these two images, and then arranges the image acquired at 0° in the front and the image acquired at 15° in the back in ascending order of the angle to form an updated detection image group. Similarly, for the initial detection image group of 30° - 60°, if there are images acquired at 35° and 45° within the group, the device will arrange the image acquired at 35° in the front and the image acquired at 45° in the back, completing the arrangement of the images within each initial detection image group to obtain each updated detection image group.

[0048] Step 203: For each image within each updated detection image group, extract the key feature points of each image to be detected based on the feature extraction algorithm.

[0049] Further, within each updated detection image group, the fungal detection device uses a specific feature extraction algorithm to extract key feature points for each image to be detected within the group. These key feature points can accurately represent the important structures and texture information of the fungus in the image and are the key basis for subsequent image fusion. Different feature extraction algorithms (such as SIFT, SURF, etc.) will find those points with uniqueness and stability in the image as key feature points according to information such as the grayscale and gradient of the image.

[0050] For example, the fungal detection device uses the SIFT (Scale-Invariant Feature Transform) algorithm to extract key feature points. Taking the images acquired at 35° and 45° within the updated detection image group of 30° - 60° as an example. The device runs the SIFT algorithm on the image acquired at 35°. The algorithm first constructs an image pyramid and detects extreme points in different scale spaces. After a series of steps such as Gaussian difference operations, key point localization, and direction assignment, key feature points of the fungal part in this image are extracted. For example, multiple key feature points are determined at positions such as the edge of the fungus and the bifurcation of the hyphae. Similarly, the image acquired at 45° is also processed using the SIFT algorithm to extract the corresponding key feature points. These key feature points can reflect the unique structures and texture features of the fungus at different angles and provide key data for subsequent fusion operations.

[0051] Step 204: Based on the key feature points of each image to be detected in each updated group of detected images, perform fusion to obtain a target fused image.

[0052] Furthermore, the fungal detection device performs fusion based on the key feature points of each image to be detected in each updated group of detected images to obtain a target fused image, as specifically described in Steps 2041 to 2044.

[0053] The embodiments of the present invention can reasonably group and orderly arrange the images to be detected obtained from different acquisition angles, and achieve precise fusion based on key feature points. The finally obtained target fused image synthesizes important structure, texture and other information of the fungus from multiple angles. Compared with the images from a single angle, the target fused image can more comprehensively and accurately present the true form and characteristics of the fungus, providing a data basis for subsequent determination of the fungus species and analysis of the survival status, and improving the accuracy and reliability of fungal detection.

[0054] In one embodiment, the descriptions of Steps 2041 to 2044 are as follows:

[0055] Step 2041: For the images in each updated group of detected images, determine the correspondence between different images based on the positions and descriptors of the key feature points of each image to be detected, and obtain an image matching result. The image matching result indicates that the feature points represent the same object or region in different images.

[0056] Optionally, for the images in each updated group of detected images, the fungal detection device uses the positions and descriptors of the key feature points of each image to be detected to explore the correspondence between different images. The position information of the key feature points indicates their coordinates in the image, and the descriptor is a quantitative description of the features of the surrounding area of the feature point, such as the gray change pattern around the feature point. By comparing the descriptors of the key feature points in different images, find the feature point pairs with high similarity. These feature point pairs represent the same object or region in different images, and then obtain the image matching result.

[0057] In one embodiment, taking the updated detection image group at 30° - 60° as an example, there are images collected at 35° and 45° in this group, and key feature points have been extracted. For example, there is a feature point A in the image collected at 35°, its position coordinates are (x1, y1), and the descriptor is a set of quantization values D1, including information such as the gray-scale change around this point; there is a feature point B in the image collected at 45°, the position coordinates are (x2, y2), and the descriptor is D2. The fungal detection device calculates the similarity between D1 and D2, for example, uses the Euclidean distance to calculate the difference between the two. If the distance is less than the set threshold, it is determined that the feature points A and B match. By performing such comparisons on all key feature points in the two groups of images, multiple groups of matching feature point pairs are found, such as (A, B), (C, D), etc. These matching pairs constitute the image matching result, indicating that the matching feature points in different images represent the same part of the fungus.

[0058] Step 2042, based on the image matching result, determine the overlapping region between the images within each updated detection image group. The overlapping region represents the part that is commonly covered in different images.

[0059] Furthermore, the fungal detection device determines the overlapping region between the images within each updated detection image group. Since the matching feature points represent the same object or region, by analyzing the position distribution of these matching feature points in different images, the part that is commonly covered in different images, that is, the overlapping region, can be inferred. Usually, the coordinate range of the matching feature points can be used to define the boundary of the overlapping region. Continuing with the example of the updated detection image group at 30° - 60°, it is known that there are matching feature point pairs such as (A, B), (C, D), etc. The coordinate of feature point A is (x1, y1), B is (x2, y2), C is (x3, y3), D is (x4, y4), etc. The fungal detection device finds the minimum and maximum coordinate values of all matching feature points in the horizontal and vertical directions. For example, the minimum coordinate in the horizontal direction is min_x, the maximum coordinate is max_x, the minimum coordinate in the vertical direction is min_y, and the maximum coordinate is max_y. Then, in the images collected at 35° and 45°, the rectangular region with (min_x, min_y) as the upper left vertex and (max_x, max_y) as the lower right vertex is the overlapping region. Within this overlapping region, the same part of the fungus is captured in both images, only with a slightly different perspective.

[0060] Step 2043, for the images within the same updated detection image group, use an image fusion algorithm to fuse the images to be detected within the overlapping region to generate a fused image within the group.

[0061] Further, for the images within the same group of updated detected images, within the determined overlapping region, the fungal detection device performs a fusion operation on the images to be detected using an image fusion algorithm to generate a fused image within the group. The image fusion algorithm comprehensively considers the pixel information of different images within the overlapping region and generates an image that contains more information and can more accurately reflect the characteristics of this part of the fungus. Common fusion algorithms, such as the fusion method based on multi-resolution analysis, will process the images at different resolutions and combine the processing results.

[0062] Continuing with the example of the group of updated detected images from 30° to 60°, within the determined overlapping region, a multi-resolution analysis fusion algorithm based on wavelet transform is adopted. The parts of the images collected at 35° and 45° within the overlapping region are wavelet decomposed to obtain sub-band images of different frequencies. For the low-frequency sub-bands, the average value of the corresponding pixels of the two is taken; for the high-frequency sub-bands, the pixel corresponding to the larger value is selected according to the absolute value of the coefficient. Then, the processed sub-band images are subjected to inverse wavelet transform to obtain the fused image part of the overlapping region. For the non-overlapping region, the corresponding part of the original image is directly retained. The fused result of the overlapping region and the non-overlapping region are combined to generate the fused image within the group of updated detected images from 30° to 60°.

[0063] Step 2044, for the images between different groups of updated detected images, the fused images within each group of updated detected images are fused again to obtain the target fused image.

[0064] Further, the fungal detection device fuses the fused images within each group of different updated detected images again to obtain the target fused image. This step will comprehensively integrate the information of different angular intervals represented by each group of updated detected images and further generate an image that comprehensively reflects the overall characteristics of the fungus. The fusion method is similar to the fusion of images within the group, but the images involved come from different groups of updated detected images in different angular intervals. For example, the fused images within the groups of updated detected images from 0° to 30°, 30° to 60°, 60° to 90°, etc. have been generated. The fungal detection device determines the overlapping regions and matching relationships between these fused images within the groups again, for example, by detecting the edge features of the images. Then, a similar algorithm as the fusion of images within the group (such as the multi-resolution analysis fusion algorithm based on wavelet transform) is used for fusion within the overlapping region. For the non-overlapping region, reasonable splicing and integration are also carried out. Finally, the target fused image is generated, which covers the information of the fungus in multiple angular intervals from 0° to 90° and can comprehensively and meticulously display the morphology, structure, texture and other characteristics of the fungus.

[0065] In the embodiments of the present invention, starting from determining the corresponding relationships of key feature points of different images, the overlapping regions are gradually determined and intra-group and inter-group image fusion is performed. Finally, the obtained target fused image integrates the detailed information of the fungus in multiple angular intervals. Therefore, it can present the whole picture of the fungus more comprehensively and accurately, including its complex structure and texture. This enables more abundant and accurate data support for subsequent fungus species recognition and survival status judgment based on the image, improving the accuracy and reliability of fungus detection.

[0066] In one embodiment, the descriptions of steps 401 to 404 are as follows:

[0067] Step 401: Decompose the morphological features and color features of the fungus according to different dimensions to obtain the decomposed morphological feature and color feature components.

[0068] Optionally, the fungus detection device performs a decomposition operation on the extracted morphological features and color features of the fungus. For the morphological features, they are split according to different dimensions of their composition. For example, the shape feature is separated from the size feature, hypha feature, etc. For the color features, they are decomposed from different component dimensions such as hue, saturation, and lightness to obtain more targeted color feature components.

[0069] In one embodiment, among the morphological features of the fungus extracted by the fungus detection device before, the shape is oval, the size is a major axis of 20 μm and a minor axis of 15 μm, and there are branched hyphae. In the HSV color space, the hue value of the color feature is 120, the saturation is 0.8, and the lightness is 0.6. The device decomposes the morphological features into: the oval feature in the shape dimension; the major axis of 20 μm and minor axis of 15 μm features in the size dimension; the branched hypha feature in the hypha dimension. The color feature is decomposed into a hue component of 120, a saturation component of 0.8, and a lightness component of 0.6.

[0070] Step 402: Based on the decomposed morphological features, perform a preliminary comparison with the standard fungal morphological features at the same classification level in the first fungal feature database according to the types and combination modes of basic geometric elements, to obtain a morphological feature matching result. The classification levels include phylum, class, order, family, genus, and species.

[0071] Furthermore, the fungus detection device, according to the decomposed morphological features, performs a preliminary comparison with the standard fungal morphological features at the same classification level in the first fungal feature database according to the types (such as basic elements representing different fungal morphologies like circles, ovals, straight lines, etc.) and combination modes (such as the branching patterns of hyphae) of basic geometric elements. The classification levels from large to small include phylum, class, order, family, genus, and species.

[0072] Continuing with the above example, the fungal detection device compares the decomposed elliptical shape characteristics, the major axis size of 20 μm and the minor axis size of 15 μm, and the characteristics of branched hyphae with the standard fungal morphological characteristics under the "Phylum Fungi" level in the first fungal feature database. In the database, for fungi of different classes under the "Phylum Fungi", there are corresponding standard morphological descriptions. The device finds that the basic geometric element types and combination methods are relatively similar to those of some fungi under the "Ascomycetes". These fungi have an elliptical cell morphology and may have branched hyphae. Thus, the morphological feature matching result is some fungal categories related to the "Ascomycetes", and the subsequent matching range is narrowed down to within the "Ascomycetes".

[0073] Step 403: Based on the color feature components, compare with the standard fungal color features corresponding to the index position in the first fungal feature database according to the morphological feature matching result, and obtain the color feature matching result.

[0074] Furthermore, the fungal detection device finds the standard fungal color features at the corresponding index position in the first fungal feature database, and then compares the previously decomposed color feature components with them. Since the morphological feature matching result has limited a certain range of fungal categories, performing color feature comparison within this range at this time can more accurately determine the target fungal features. By comparing the differences between color feature components such as hue, saturation, and lightness and the standard values in the database, the matching degree of the color features is judged.

[0075] Continuing with the above embodiment, since the morphological feature matching result in step 402 points to some fungi under the "Ascomycetes", the fungal detection device finds the standard color features corresponding to these fungi within the "Ascomycetes" in the first fungal feature database. For example, the standard color features of a possible matching fungus in the database are hue 115 - 125, saturation 0.7 - 0.9, and lightness 0.5 - 0.7. Comparing the previously decomposed hue component of 120, saturation component of 0.8, and lightness component of 0.6 with this standard color feature, it is found that each component is within the standard range, indicating a relatively high color feature matching degree. Thus, the color feature matching result is obtained, further confirming that the fungal category corresponding to this standard color feature is within the matching range.

[0076] Step 404: Determine the first target fungal feature based on the color feature matching result.

[0077] Furthermore, based on the color feature matching result obtained in step 403 and in combination with the previous morphological feature matching situation, the fungal detection device finally determines the first target fungal feature. If the color feature matching result and the morphological feature matching result point to the same type of fungus in the database and both have a high degree of matching, then the features of this type of fungus can be determined as the first target fungal feature. This feature represents the standard fungal feature that is closest to the fungal features in the sample to be detected under specific culture and staining conditions. After the previous steps, the morphological feature matching points to some fungi under "Ascomycetes", and the color feature matching also highly conforms to the standard features of a specific fungus species under "Ascomycetes". For example, through comprehensive judgment, it is determined that the features of the fungus to be detected are most consistent with the standard features of "Ascomycetes - Saccharomycetales - Candida - Candida albicans". Therefore, the fungal detection device determines that the first target fungal feature is the standard feature of Candida albicans in the first fungal feature database, including its specific morphological and color feature descriptions.

[0078] In the embodiment of the present invention, the fungal features are first decomposed and then compared with the first fungal feature database in stages. Finally, the first target fungal feature that best matches the morphological and color features of the fungus to be detected can be accurately selected from the database. This provides an accurate reference when determining the type of fungus in the sample to be detected. Therefore, the matching object can be found more efficiently and accurately from the large database, improving the accuracy and efficiency of fungus type determination.

[0079] In one embodiment, the descriptions of steps 405 to 408 are as follows:

[0080] Step 405, extract the primary and secondary direction information and structural information in the texture features formed by the staining of characteristic metabolites based on the image structure tensor algorithm.

[0081] Optionally, the fungal detection device uses the image structure tensor algorithm to process the texture features formed by staining characteristic metabolites. This algorithm calculates the gradient information of the local area of the image to construct a structure tensor matrix. From this matrix, the primary and secondary direction information in the texture features can be extracted, that is, the main and secondary directions of the texture in the image, which can reflect the directionality of the distribution of fungal metabolites. At the same time, structural information can also be obtained, such as the complexity and regularity of the texture. In an embodiment, for example, the fungal detection device has obtained the texture image formed by staining the characteristic metabolites in the sample to be detected. The device applies the image structure tensor algorithm to this image. In a small local area of the image, a structure tensor matrix is constructed by calculating the gradient values of the pixel points. For example, in a 3*3 pixel neighborhood, the structure tensor matrix is obtained through calculation. The eigenvalue decomposition is performed on this matrix. The direction of the eigenvector corresponding to the larger eigenvalue is the main direction of the texture, and the direction of the eigenvector corresponding to the smaller eigenvalue is the secondary direction. At the same time, according to some attributes of the matrix, such as the ratio of the eigenvalues, the structural information of the texture in this area can be evaluated to determine whether it is a regular texture or an irregular texture, and the complexity of the texture. After processing the entire texture image, the primary and secondary direction information and structural information in the texture features are obtained.

[0082] Step 406, traverse the hierarchical storage structure in the second fungal feature database to determine the relevant information in each storage layer. The second fungal feature database stores the texture features of live and dead fungal metabolites separately, and within each storage area, it is further stratified according to different fungal species and metabolite types.

[0083] Furthermore, the second fungal feature database has a specific hierarchical storage structure. The fungal detection device starts to traverse this database. First, the database stores the texture features of live and dead fungal metabolites separately. Within each storage area, it is further stratified according to different fungal species and metabolite types. During the traversal process, the device sequentially determines the relevant information in each storage layer, including the standard descriptions of the texture features formed by staining the metabolites of different fungal species in the live or dead state, and these descriptions include the primary and secondary direction information and structural information extracted in a similar manner to step 405, etc. In an embodiment, start traversing the second fungal feature database, first enter the storage area of the texture features of live fungal metabolites, and this area is further stratified according to fungal species, such as the Candida albicans layer, the Aspergillus layer, etc. In the Candida albicans layer, it is further subdivided according to different metabolite types. Enter the Candida albicans - a specific metabolite type layer, and read the standard information about the texture features formed by staining with this metabolite in the live state of Candida albicans stored in this layer, including the standard value range of the primary and secondary directions of the texture, the description of the structural information, etc., and traverse each layer under the storage area of the texture features of dead fungal metabolites to obtain the corresponding information.

[0084] Step 407: Determine the similarity evaluation results between the primary and secondary direction information and the structural information in the texture features and the relevant information in each storage layer based on the similarity measurement algorithm.

[0085] Furthermore, the fungal detection device uses the similarity measurement algorithm to compare the primary and secondary direction information and the structural information of the texture features to be detected extracted in Step 405 with the relevant information obtained from each storage layer of the second fungal feature database in Step 406. The similarity measurement algorithm will calculate the similarity degree between the two according to the set rules to obtain the similarity evaluation results. For example, for the primary and secondary direction information, the difference in direction angles can be calculated; for the structural information, the quantified values of texture complexity can be compared, etc. By synthesizing these comparison results, an overall similarity evaluation score or grade can be obtained. For example, the main direction angle of the texture features to be detected obtained from Step 405 is 45°, the secondary direction angle is 135°, and the structural information shows that the texture complexity is 0.6 (quantified value). In the second fungal feature database, the standard main direction angle of the texture features of a certain metabolite in the viable state of Candida albicans is 40° - 50°, the secondary direction angle is 130° - 140°, and the structural information is described as a complexity of 0.5 - 0.7. The fungal detection device uses the similarity measurement algorithm to calculate that the difference in the main direction angle is within a reasonable range, the difference in the secondary direction angle is also within a reasonable range, and the quantified value of the texture complexity is also similar. Through the comprehensive calculation of the algorithm, the similarity evaluation result between this storage layer and the texture features to be detected is a high matching degree, which is recorded in a certain form (such as a score of 80 points, with a full score of 100 points). The device performs such similarity evaluation operations on all storage layers in the database.

[0086] Step 408: Determine the second target fungal feature based on the similarity evaluation results.

[0087] Furthermore, the fungal detection device determines the second target fungal feature based on the similarity evaluation results obtained in Step 407. It will find the fungal feature corresponding to the storage layer with the highest similarity among the similarity evaluation results of all storage layers and determine it as the second target fungal feature. This feature represents the standard feature in the second fungal feature database that most closely matches the texture features formed by the staining of characteristic metabolites in the sample to be detected, thereby enabling the determination of the viability status and related species information of the fungi in the sample to be detected.

[0088] For example, after the similarity evaluation of all storage layers of the second fungal feature database, it is found that the similarity evaluation score of the storage layer where the texture feature of a certain metabolite is located in the viable state of Candida albicans is the highest, which is 85 points (for example, the scores of other storage layers are all lower than this). Then, the fungal detection device determines that the second target fungal feature is the texture feature formed by staining with this specific metabolite in the viable state of Candida albicans, and this feature will be used to judge information such as the viability of the fungi in the sample liquid to be detected.

[0089] The embodiment of the present invention can accurately screen out the second target fungal feature that best matches the stained texture feature of the characteristic metabolite in the sample to be detected from the second fungal feature database, which provides an accurate basis for judging the viability and related species of the fungi in the sample liquid to be detected. Compared with the method of randomly searching or not performing detailed feature extraction and hierarchical comparison, the accuracy and efficiency of judging the viability and related species of the fungi are improved.

[0090] In one embodiment, the descriptions of steps 501 to 503 are as follows:

[0091] Step 501, map the first target fungal feature to the corresponding network node of the phylogenetic tree. The phylogenetic tree is a feature relationship network constructed based on the first fungal feature database. The network nodes in the feature relationship network are different fungal species, and the edges are the similarity associations of features between different species.

[0092] Optionally, the phylogenetic tree used by the fungal detection device is a feature relationship network constructed based on the first fungal feature database. The network nodes of this phylogenetic tree represent different fungal species, and the edges between the nodes reflect the similarity associations of the features of different fungal species. The device matches and maps the previously obtained first target fungal feature with the nodes in the phylogenetic tree. By analyzing the various parameters of the first target fungal feature, such as the specific numerical values or descriptions of features such as morphology and color, the most suitable network node is found in the phylogenetic tree, and this node preliminarily corresponds to the possible fungal species.

[0093] For example, after the previous matching, the first target fungal feature is determined to be highly similar to the standard feature of Candida albicans in the first fungal feature database. The fungal detection device searches for network nodes with similar Candida albicans feature descriptions in the phylogenetic tree constructed based on this database. In the phylogenetic tree, each node stores detailed feature information of the corresponding fungal species. By comparing the features, the device finds that the fungal species represented by a specific node in the phylogenetic tree has morphological features (cells are oval and within a certain size range) and color features (hue, saturation, etc. presented under specific staining) that are almost the same as the first target fungal feature. Then, the first target fungal feature is mapped to this network node, and this node corresponds to the Candida albicans species.

[0094] Step 502, traverse the network nodes based on the first target fungal feature along the edge of the evolutionary tree, and determine the matching degree between the first target fungal feature and the features of each node by comparing the position of the first target fungal feature on the evolutionary tree with the distribution of known fungal species, combined with the topological structure and branch evolution law of the evolutionary tree.

[0095] Furthermore, the fungus detection device starts from the network node of the mapped first target fungus feature and traverses along the edge of the evolutionary tree. During the traversal process, the position of the first target fungus feature on the evolutionary tree is compared with the distribution of known fungal species in combination with the topological structure of the evolutionary tree (such as the hierarchical relationship of the nodes, the number and direction of the branches, etc.) and the law of branch evolution (such as the trend of feature changes from simple to complex, from primitive to evolutionary). By analyzing the degree of difference between the features of different nodes and the first target fungus feature, the degree of match between them is determined. For example, the nodes that are close to the mapping node and are closely connected in evolutionary relationship may have a higher degree of match between their features and the first target fungus feature; while the nodes that are far away and on different branches may have a lower degree of match.

[0096] Let's continue with the example of mapping the first target fungal feature to the node where Candida albicans is located. The fungal detection device starts from this node and traverses to the adjacent nodes along the edge of the evolutionary tree. In the evolutionary tree, the nodes adjacent to Candida albicans may represent other Candida species that are evolutionarily close to it. The device compares the features of these adjacent nodes with the first target fungal features and finds that the fungal species represented by a certain adjacent node is similar to Candida albicans in morphology, but differs in some subtle structures (such as slight differences in cell wall thickness); in terms of color features, the hue is slightly deviated. According to the topological structure of the evolutionary tree, these adjacent nodes are in the same branch as the Candida albicans node and are close to each other. Through comprehensive analysis, the device determines that the features of these adjacent nodes have a certain degree of match with the first target fungal features, but not as good as the Candida albicans node itself. Continue to traverse to more distant nodes and find that the fungal species in other branches have large differences in morphology and color features from the first target fungal features, and the degree of match is low.

[0097] Step 503: Determine the type of fungus in the sample liquid to be detected based on the matching degree between the first target fungus feature and each node feature.

[0098] Furthermore, the fungus detection device finally determines the type of fungi in the sample liquid to be detected based on the matching degree between the first target fungus feature and the features of each node. Among all the traversed nodes, the fungus species represented by the node with the highest matching degree is the type of fungi in the sample liquid to be detected. This is because the feature of the node is most similar to the first target fungus feature, and the first target fungus feature is obtained by matching a large number of standard features in the first fungus feature database, and has a high reliability.

[0099] For example, after the traversal and comparison in step 502, the fungal detection device finds that the node where Candida albicans is located has the highest degree of matching with the first target fungal feature, and the matching degrees of other nodes are all lower than it. Therefore, the device determines that the type of fungus in the sample liquid to be detected is Candida albicans. This result is based on a comprehensive analysis and comparison of the characteristics of numerous nodes on the evolutionary tree, ensuring the accuracy of the determination of the fungal species.

[0100] The embodiment of the present invention can accurately determine the type of fungus in the sample liquid to be detected by means of the feature relationship network of the evolutionary tree. Therefore, by using the evolutionary tree structure and evolution law, the connections and differences between fungal features can be analyzed more comprehensively and deeply. This makes the determination of the fungal species no longer limited to the simple matching of surface features, but comprehensively considered from the perspective of evolution, improving the accuracy and scientificity of the determination of the fungal species.

[0101] In one embodiment, the descriptions of steps 504 to 506 are as follows:

[0102] Step 504, determine the periodicity and repeatability of the texture, and the arrangement pattern of texture units based on the second target fungal feature.

[0103] Optionally, the fungal detection device focuses on analyzing the periodicity and repeatability of the texture formed by the staining of characteristic metabolites and the arrangement pattern of texture units for the second target fungal feature. Whether the periodic fingerprint texture appears regularly repeated within a certain area; repeatability is a quantitative description of periodicity, reflecting information such as the repetition frequency. Texture units are the basic elements that make up the texture, and their arrangement patterns include ordered arrangement, disordered arrangement, etc. By analyzing these texture characteristics, key information reflecting the metabolic state of the fungus can be obtained, because when the fungus is in the live and dead states, the production and distribution of its metabolites are different, resulting in differences in texture characteristics.

[0104] For example, the second target fungal feature corresponds to the texture feature of the metabolite of Candida albicans. The fungal detection device analyzes this texture image. It is observed that within a certain image area, there is a texture unit similar to a "ring", which repeats every certain distance, showing obvious periodicity. Further measurement reveals that the "ring" texture unit reappears approximately every 5 pixels, which determines its repeatability. At the same time, the device analyzes the arrangement pattern of the "ring" texture units and finds that they are arranged approximately in a certain direction and interval in an orderly manner, rather than randomly. These analysis results regarding the texture periodicity, repeatability, and arrangement pattern.

[0105] Step 505: Based on the periodicity and repeatability of the texture, as well as the arrangement pattern of texture units, perform matching in the survival state feature knowledge base to obtain the state feature matching result. The survival state feature knowledge base is a knowledge base about the survival state features of fungi constructed with the second fungal feature database as the core and combined with relevant microbiological research results. The survival state feature knowledge base contains the typical manifestations and variation rules of metabolite texture features of different fungi in the live and dead states.

[0106] Furthermore, the fungal detection device performs matching of feature information such as the periodicity, repeatability of the texture, and the arrangement pattern of texture units in the survival state feature knowledge base. The survival state feature knowledge base takes the second fungal feature database as the core and combines a large amount of relevant microbiological research results, and details the typical manifestations and variation rules of metabolite texture features of different fungi in the live and dead states. The device finds the most matching result by comparing the input texture features with the texture feature standards of various fungi in different survival states stored in the knowledge base to obtain the state feature matching result.

[0107] Continuing with Candida albicans as an example, the fungal detection device searches for a match in the survival state feature knowledge base for the features such as the periodicity (repeating every 5 pixels), repeatability, and ordered arrangement pattern of the Candida albicans texture obtained from the previous analysis. The description of the texture features of Candida albicans in the live state in the knowledge base is: there are specific texture units, with obvious periodicity, the repeat interval is between 4 - 6 pixels, and the arrangement pattern is in an ordered state; while in the dead state, the texture units usually become irregular, the periodicity weakens or even disappears, and the arrangement pattern is chaotic. After comparison, it is found that the texture features of the Candida albicans to be detected are closer to the description in the live state in the knowledge base.

[0108] Step 506: Determine the survival state of the fungi in the sample liquid to be detected based on the state feature matching result.

[0109] Furthermore, the fungal detection device finally determines the survival state of the fungi in the sample liquid to be detected according to the state feature matching result. If the state feature matching result is highly consistent with the feature description of a certain survival state (live or dead) in the knowledge base, then it can be determined that the fungi in the sample to be detected are in this survival state. This determination result is based on the accurate analysis of the texture features before and the precise matching with the standard features in the knowledge base. For example, since the state feature matching result in Step 505 shows that the texture features of the Candida albicans to be detected are more consistent with the description in the live state, the fungal detection device determines that the Candida albicans in the sample liquid to be detected is in the survival state.

[0110] The embodiments of the present invention can accurately determine the survival status of fungi in a sample liquid to be detected. By conducting a detailed analysis of texture features and matching them with a professional knowledge base, the judgment result no longer relies solely on simple observation, but is based on scientific feature analysis and the support of a large number of research results, improving the accuracy and reliability of judging the survival status of fungi.

[0111] The following describes the fungal detection device assisted by image recognition provided by the present invention. The fungal detection device assisted by image recognition described below can be correspondingly referred to the fungal detection method assisted by image recognition described above. Figure 2 It is a schematic structural diagram of the fungal detection device assisted by image recognition provided by the present invention. The fungal detection device assisted by image recognition includes:

[0112] An image acquisition module 210, configured to acquire a to-be-detected image of the sample liquid to be detected based on an image acquisition device; the to-be-detected image is an image acquired after the sample liquid to be detected is stained with a triple fluorescence staining solution for fungi.

[0113] An image fusion module 220, configured to fuse the to-be-detected images according to the acquisition angles to obtain a target fused image.

[0114] A feature extraction module 230, configured to extract features from the fused image to obtain a feature extraction result; the feature extraction result includes fungal morphological features, color features, and texture features formed by staining with characteristic metabolites.

[0115] A feature matching module 240, configured to perform matching in a first fungal feature database based on the fungal morphological features and color features to obtain a first target fungal feature, and perform matching in a second fungal feature database based on the texture features formed by staining with characteristic metabolites to obtain a second target fungal feature; the first fungal feature database contains standard features of various fungi under specific culture and staining conditions, and the second fungal feature database contains texture features formed by staining with metabolites of live fungi and texture features formed by staining with metabolites of dead fungi.

[0116] A fungal detection module 250, configured to determine the type of fungi in the sample liquid to be detected based on the first target fungal feature, and determine the survival status of fungi in the sample liquid to be detected based on the second target fungal feature.

[0117] In the embodiments of the present invention, through multi-angle high-resolution image acquisition, more comprehensive information of fungal samples can be obtained, avoiding missed diagnosis due to a small number of fungi or atypical morphology in the samples. During the processes of image preprocessing, segmentation, and feature extraction, through feature extraction, the morphological features, color features, and texture features formed by the staining of characteristic metabolites of fungi can be comprehensively judged, which can accurately highlight the fungal features. Different from the microscopic observation method that is difficult to distinguish between live bacteria and dead bacteria, therefore, compared with the manual observation method, based on the objective image recognition algorithm for feature extraction and matching, the subjectivity of manual film reading and misjudgment caused by fatigue are avoided, greatly improving the accuracy of the detection results and achieving accurate and rapid detection and identification of fungi.

[0118] Please refer to Figure 3 , Figure 3 which is a schematic diagram of an embodiment of an electronic device provided by an embodiment of the present invention. As Figure 3 shown, an embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored on the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, the following steps are implemented:

[0119] Collect a to-be-detected image of a to-be-detected sample solution based on an image acquisition device; the to-be-detected image is an image collected after the to-be-detected sample solution is stained with a triple fluorescent staining solution for fungi;

[0120] Fuse the to-be-detected images according to the acquisition angles to obtain a target fused image;

[0121] Extract features from the fused image to obtain a feature extraction result; the feature extraction result includes the morphological features, color features, and texture features formed by the staining of characteristic metabolites of fungi;

[0122] Match based on the morphological features and color features of fungi in a first fungal feature database to obtain a first target fungal feature, and match based on the texture features formed by the staining of characteristic metabolites in a second fungal feature database to obtain a second target fungal feature; the first fungal feature database contains standard features of various fungi under specific culture and staining conditions, and the second fungal feature database contains the texture features formed by the staining of the metabolites of live bacteria and the texture features formed by the staining of the metabolites of dead bacteria;

[0123] Determine the type of fungi in the to-be-detected sample solution based on the first target fungal feature, and determine the survival state of the fungi in the to-be-detected sample solution based on the second target fungal feature.

[0124] Please refer to Figure 4 , Figure 4 which is a schematic diagram of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. AsFigure 4 As shown in the figure, this embodiment provides a computer-readable storage medium 400, on which a computer program 311 is stored. When the computer program 311 is executed by a processor, the following steps are implemented:

[0125] Collect a to-be-detected image of a to-be-detected sample liquid based on an image acquisition device; the to-be-detected image is an image collected after the to-be-detected sample liquid is stained with a triple fungal fluorescence staining solution;

[0126] Fuse the to-be-detected image according to the acquisition angle to obtain a target fused image;

[0127] Extract features from the fused image to obtain a feature extraction result; the feature extraction result includes fungal morphological features, color features, and texture features formed by staining with characteristic metabolites;

[0128] Match based on the fungal morphological features and color features in a first fungal feature database to obtain a first target fungal feature, and match based on the texture features formed by staining with characteristic metabolites in a second fungal feature database to obtain a second target fungal feature; the first fungal feature database contains standard features of various fungi under specific culture and staining conditions, and the second fungal feature database contains texture features formed by staining with metabolites of live bacteria and texture features formed by staining with metabolites of dead bacteria;

[0129] Determine the type of fungus in the to-be-detected sample liquid based on the first target fungal feature, and determine the survival state of the fungus in the to-be-detected sample liquid based on the second target fungal feature.

[0130] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the fungus detection method assisted by image recognition provided by the above various methods. The fungus detection method assisted by image recognition includes:

[0131] Collect a to-be-detected image of a to-be-detected sample liquid based on an image acquisition device; the to-be-detected image is an image collected after the to-be-detected sample liquid is stained with a triple fungal fluorescence staining solution;

[0132] Fuse the to-be-detected image according to the acquisition angle to obtain a target fused image;

[0133] Extract features from the fused image to obtain a feature extraction result; the feature extraction result includes fungal morphological features, color features, and texture features formed by staining with characteristic metabolites;

[0134] Match in the first fungal feature database based on the morphological and color features of fungi to obtain the first target fungal feature, and match in the second fungal feature database based on the texture features formed by the staining of characteristic metabolites to obtain the second target fungal feature; the first fungal feature database contains the standard features of various fungi under specific culture and staining conditions, and the second fungal feature database contains the texture features formed by the staining of the metabolites of live bacteria and the texture features formed by the staining of the metabolites of dead bacteria;

[0135] Determine the types of fungi in the sample liquid to be detected based on the first target fungal feature, and determine the survival status of the fungi in the sample liquid to be detected based on the second target fungal feature.

[0136] It should be noted that in the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0137] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0138] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A fungal detection method assisted by image recognition, characterized in that Including: Collecting a to-be-detected image of a to-be-detected sample liquid by an image acquisition device; The to-be-detected image is an image collected after the to-be-detected sample liquid is stained with a triple-fluorescence staining solution for fungi; Fusing the to-be-detected images according to the acquisition angle to obtain a target fused image; Performing feature extraction on the fused image to obtain a feature extraction result; the feature extraction result includes fungal morphological features, color features, and texture features formed by staining with characteristic metabolites; Matching based on the fungal morphological features and color features in a first fungal feature database to obtain a first target fungal feature, and matching based on the texture features formed by staining with characteristic metabolites in a second fungal feature database to obtain a second target fungal feature; the first fungal feature database contains standard features of various fungi under specific culture and staining conditions, and the second fungal feature database contains texture features formed by staining with metabolites of live bacteria and texture features formed by staining with metabolites of dead bacteria; Determining the type of fungi in the to-be-detected sample liquid based on the first target fungal feature, and determining the survival state of the fungi in the to-be-detected sample liquid based on the second target fungal feature.

2. The fungal detection method assisted by image recognition according to claim 1, wherein The determining the type of fungi in the to-be-detected sample liquid based on the first target fungal feature includes: Mapping the first target fungal feature to a corresponding network node of an evolutionary tree; the evolutionary tree is a feature relationship network constructed based on the first fungal feature database; the network nodes in the feature relationship network are different fungal species, and the edges are the similarity associations of features between different species; Traversing along the edges of the evolutionary tree based on the network node of the first target fungal feature, and determining the matching degree between the first target fungal feature and the features of each node by comparing the position of the first target fungal feature on the evolutionary tree with the distribution of known fungal species, and combining the topological structure and branch evolution law of the evolutionary tree; Determining the type of fungi in the to-be-detected sample liquid based on the matching degree between the first target fungal feature and the features of each node.

3. The fungal detection method based on image recognition assistance according to claim 1, characterized in that The determining the survival state of the fungi in the to-be-detected sample liquid based on the second target fungal feature includes: Determining the periodicity and repeatability of the texture, and the arrangement pattern of texture units based on the second target fungal feature; Matching in a survival state feature knowledge base based on the periodicity and repeatability of the texture, and the arrangement pattern of texture units to obtain a state feature matching result; the survival state feature knowledge base is a knowledge base about fungal survival state features constructed with the second fungal feature database as the core and combined with relevant microbiological research results; the survival state feature knowledge base contains the typical manifestations and change rules of metabolite texture features of different fungi in the live and dead states; Determining the survival state of the fungi in the to-be-detected sample liquid based on the state feature matching result.

4. The fungal detection method assisted by image recognition according to claim 1, wherein The matching based on the fungal morphological features and color features in a first fungal feature database to obtain a first target fungal feature includes: Decompose the fungal morphological characteristics and the color characteristics according to different dimensions to obtain the decomposed morphological characteristic and color characteristic components; Based on the decomposed morphological characteristics, conduct a preliminary comparison with the standard fungal morphological characteristics at the same classification level in the first fungal characteristic database according to the types and combination methods of basic geometric elements; the classification level includes phylum, class, order, family, genus, and species; Based on the color characteristic components, compare with the standard fungal color characteristics corresponding to the indexing position of the morphological characteristic matching result in the first fungal characteristic database to obtain the color characteristic matching result; Determine the first target fungal characteristic based on the color characteristic matching result.

5. The fungal detection method based on image recognition assistance according to claim 1, wherein, Match the texture characteristics formed by staining with characteristic metabolites in the second fungal characteristic database to obtain the second target fungal characteristic, including: Extract the primary and secondary direction information and structural information in the texture characteristics formed by staining with characteristic metabolites based on the image structure tensor algorithm; Traverse the hierarchical storage structure in the second fungal characteristic database to determine the relevant information in each storage layer; the second fungal characteristic database stores the texture characteristics of live and dead bacteria metabolites separately, and is stratified according to different fungal species and metabolite types within each storage area; Based on the similarity measurement algorithm, determine the similarity evaluation results between the primary and secondary direction information and structural information in the texture characteristics and the relevant information in each storage layer; Determine the second target fungal characteristic based on the similarity evaluation results.

6. The fungal detection method based on image recognition assistance according to any one of claims 1 to 5, characterized in that, The step of fusing the images to be detected according to the acquisition angle to obtain the target fused image includes: Group each image to be detected according to a preset angle interval based on the acquisition angle to obtain multiple initial detection image groups; Arrange the images to be detected within each initial detection image group according to the size of the acquisition angle to obtain the updated detection image group; For each image in the updated detection image group, extract the key feature points of each image to be detected based on the feature extraction algorithm; the key feature points represent the important structure and texture information of the image; Fuse the key feature points of each image to be detected in each updated detection image group to obtain the target fused image.

7. The method for detecting fungi assisted by image recognition according to claim 6, wherein, The step of fusing the key feature points of each image to be detected in each updated detection image group to obtain the target fused image includes: For each image in the updated detection image group, determine the corresponding relationship between different images based on the positions and descriptors of the key feature points of each image to be detected to obtain the image matching result; the image matching result indicates that the feature points represent the same object or area in different images; Determine the overlapping area between the images within each updated detection image group based on the image matching result; the overlapping area represents the common covered part in different images; For the images within the same updated detection image group, fuse the images to be detected within the overlapping area using the image fusion algorithm to generate the intra-group fused image; For the images between different groups of updated detected images, the images obtained by intra-group fusion of each group of updated detected images are fused again to obtain the target fused image.

8. A fungal detection device assisted by image recognition, characterized in that, Including: An image acquisition module, configured to acquire a to-be-detected image of a to-be-detected sample liquid based on an image acquisition device; The to-be-detected image is an image acquired after the to-be-detected sample liquid is stained with a triple fluorescence staining solution for fungi; An image fusion module, configured to fuse the to-be-detected images according to the acquisition angle to obtain a target fused image; A feature extraction module, configured to extract features from the fused image to obtain a feature extraction result; the feature extraction result includes fungal morphological features, color features, and texture features formed by staining with characteristic metabolites; A feature matching module, configured to perform matching in a first fungal feature database based on fungal morphological features and color features to obtain a first target fungal feature, and perform matching in a second fungal feature database based on the texture features formed by staining with characteristic metabolites of fungi to obtain a second target fungal feature; the first fungal feature database contains standard features of various fungi under specific culture and staining conditions, and the second fungal feature database contains texture features formed by staining with metabolites of live fungi and texture features formed by staining with metabolites of dead fungi; A fungal detection module, configured to determine the type of fungi in the to-be-detected sample liquid based on the first target fungal feature, and determine the survival state of the fungi in the to-be-detected sample liquid based on the second target fungal feature.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the fungal detection method assisted by image recognition according to any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the fungal detection method assisted by image recognition according to any one of claims 1-7.

Images