A Sparse Scanning Path Planning Method and System Based on Multimagnification Microscopic Images

Abstract

This invention discloses a sparse scanning path planning method and system based on multi-magnification microscopic images, relating to the field of microscopic microbial quantity statistics. The method involves low-magnification scanning of the observation area of ​​the sample to be tested, selecting high-density areas of bacterial floc distribution from the acquired floc distribution images; calculating the global probability density of each high-density area using a floc-dependent biodistribution model to determine a mid-magnification scanning path; inputting the acquired mid-magnification microscopic images back into the floc-dependent biodistribution model to obtain the total number of the first target organisms; determining a high-magnification scanning path based on the distribution order of the first target organisms in the mid-magnification microscopic images; and inputting the acquired high-magnification microscopic images into a constructed bioassociated distribution model to obtain the total number of the second target organisms. By relying on the distribution association patterns among microorganisms and employing a layered scanning strategy that locks in high-density areas, the scan volume and empty scan rate are significantly reduced.

Description

Technical Field

[0001] This invention relates to the field of microbial quantity statistics technology, specifically to a sparse scanning path planning method and system based on multi-magnification microscopic images. Background Technology

[0002] To address the low switching efficiency of traditional single-magnification scanning devices when monitoring microbial populations, an improved multi-magnification synchronous scanning hardware device was adopted. This device achieves precise slide displacement via dual moving stages (moving vertically) on a base, while a preview camera and multi-path scanning mirrors are mounted on the stand. A "main optical path splitting design" enables multi-magnification scanning. Multiple beam paths are split from the main optical path using a semi-transparent, semi-reflective mirror. Each beam path is equipped with a magnifying glass and imaging camera of different magnifications, allowing simultaneous acquisition of multi-magnification images of the same area without manual objective lens switching. The scanning logic of the multi-magnification synchronous scanning hardware device is "full-area coverage + multi-magnification synchronous imaging." After locating the sample area using the preview camera, the moving stages complete a full-area scan, with each beam path camera simultaneously acquiring images at different magnifications.

[0003] However, multi-magnification scanning devices only optimize operational efficiency, which is a "hardware-level efficiency optimization." Their scanning path remains a "full-area traversal," failing to consider microbial distribution characteristics and thus unable to pinpoint high-value areas. Therefore, while solving the problem of time-consuming magnification switching, the total scan volume remains unchanged, and the empty scan rate is the same as traditional full scan. Furthermore, they lack a model for estimating microbial numbers, only providing image acquisition functionality and unable to directly output quantitative statistical results. Therefore, standalone multi-magnification scanning devices still use a full-area scanning method during scanning, without reducing the scan volume; they offer high accuracy but low efficiency.

[0004] Existing technologies for analyzing the spatial distribution of bacterial communities at the research level employ high-resolution spatial localization techniques, typically using fluorescence in situ hybridization (HiPR-FISH). These techniques focus on "highly accurate identification of complex bacterial communities," with the core principle being "fluorescence encoding + spectral decoding," utilizing the binary combination of 10 fluorescent groups (2... 10 -1=1023 types of codes), designing unique fluorescent probes for different microbial species; the operation process includes probe hybridization (specific probes bind to target bacterial groups, and then bind to fluorescently labeled readout sequences), spectral imaging (five-wavelength laser confocal imaging to obtain unique spectral maps), and machine learning decoding (analyzing the species information corresponding to the spectrum through support vector machines).

[0005] The advantage of high-resolution microbial community spatial localization technology is that it can achieve species localization at single-cell resolution, distinguishing the spatial distribution of hundreds of microorganisms in complex communities. However, firstly, it is extremely expensive, requiring customized fluorescent probes and dedicated spectral imaging equipment, making it unsuitable for low-cost scenarios such as water treatment plants and routine environmental monitoring; secondly, the process is cumbersome, with probe design, hybridization, and spectral decoding taking more than 24 hours, failing to meet the needs of "rapid screening"; and thirdly, the scanning efficiency has not been optimized, still based on full-area fluorescence imaging, and the issues of empty scanning and efficiency have not been resolved. Summary of the Invention

[0006] The purpose of this invention is to provide a sparse scanning path planning method and system based on multi-magnification microscopic images. Different probability models are established according to the distribution association of microorganisms. The method uses low magnification to lock high-density areas and medium and high magnification to focus on the number of target organisms layer by layer to achieve global quantity estimation with "fewer scans and higher accuracy" and improve scanning efficiency.

[0007] To achieve the above objectives, this application proposes the following solution:

[0008] On one hand, the present invention provides a sparse scanning path planning method based on multi-magnification microscopic images, specifically including the following steps:

[0009] S1. Set the low-magnification scanning area of ​​the observation area of ​​the sample to be tested, obtain the complete bacterial floc distribution image, and select the high-density area of ​​bacterial floc distribution from the bacterial floc distribution image according to the bacterial floc distribution probability to establish a bacterial floc probability distribution map.

[0010] S2. Input the probability distribution map of the bacterial floc into the constructed bacterial floc-dependent biological distribution model to obtain the global probability density of each high-density region. Based on the global probability density, determine the medium magnification scanning path of the observation area of ​​the sample to be tested under the microscope and obtain the medium magnification microscopic image.

[0011] S3. Input the medium-magnification microscopic image into the constructed floc-dependent biodistribution model to obtain the total number of the first target organisms in the observation area of ​​the sample to be tested;

[0012] S4. Based on the distribution order of the first target organism in the medium magnification microscopic image, determine the high magnification scanning path of the observation area of ​​the sample to be tested under the microscope, and obtain the high magnification microscopic image.

[0013] S5. Input the high-magnification microscopic image into the constructed biological association distribution model to obtain the total number of the second target organisms in the observation area of ​​the sample to be detected.

[0014] In some specific implementation schemes, the specific process of establishing the probability distribution map of bacterial micelles in step S1 is as follows:

[0015] S11. Perform a full-area scan of the observation area of ​​the sample to be tested to obtain a complete image of the distribution of bacterial flocs;

[0016] S12. Convert the bacterial floc distribution image into a grayscale image, determine the bacterial floc features based on the brightness values ​​of each pixel in the grayscale image, and obtain the probability density of the bacterial flocs based on the bacterial floc features.

[0017] S13. Calculate the mean and standard deviation of the brightness values ​​of all pixels in the grayscale image, and calculate the probability density value of each pixel becoming the center of the fungal floc using the normal distribution based on the mean and standard deviation.

[0018] S14. Sort all pixel probability density values ​​from largest to smallest, select the top few pixels as the center position of the bacterial floc, and obtain the bacterial floc probability distribution map.

[0019] In some specific implementations, the method for calculating the probability density value of each pixel in step S13 is as follows:

[0020]

[0021] in, r ( x , y ) indicates that the coordinates are ( x , y ) represents the probability density value of a pixel, g represents the gray value of the pixel, μ represents the mean, and σ represents the standard deviation.

[0022] In some specific implementation schemes, the method for determining the medium-magnification scanning path is as follows:

[0023] S21. Determine the high-density areas of fungal floc distribution based on the center coordinates of the fungal flocs, and calculate the global distribution probability of fungal floc-dependent organisms appearing in each high-density area based on the fungal floc-dependent organism distribution model.

[0024] S22. Sort the global distribution probability values ​​of all high-density areas from largest to smallest, and select the top-ranked high-density areas as the medium magnification scanning areas of the first target organism in the observation area of ​​the sample to be tested.

[0025] S23. Determine the medium magnification scanning path of the microscope according to the distribution order of the high-density areas corresponding to the medium magnification scanning area, and sequentially acquire images of the medium magnification scanning area according to the medium magnification scanning path to obtain medium magnification microscopic images.

[0026] In some specific implementation schemes, the specific process for obtaining the total number of the first target organisms in step S3 is as follows:

[0027] S31, the coordinate vector of the first target organism in a statistical medium-magnification micrograph (S31, S32 ... x a , y a ) and the first actual number of the first target organism under each coordinate vector. N 实1 ;

[0028] S32. Input the coordinate vector of the first target organism into the constructed floc-dependent biodistribution model to obtain the first biodistribution probability of the first target organism. P ( x a , y a );

[0029] S33. Based on the probability distribution of the first organism and the first actual number, estimate the total number of the first target organisms in the observation area of ​​the sample to be detected. N 总1 :

[0030]

[0031] Where E(·) represents the expectation function.

[0032] In some specific implementation schemes, the distribution model of fungal floc-dependent organisms includes a global probability density function for the fungal floc-dependent organisms, calculated as follows:

[0033]

[0034] in,( x a , y a () represents the coordinate vector of the first target organism in a high-density area. r ( x a , y a ) indicates that at coordinates ( x a , y a The density vector of bacterial flocs near ) is given by ( x f , y f ) represents the coordinate vector of the center position of the fungal floc; s f This represents a constant representing the radius of influence of the bacterial floc on the first target organism. n This indicates the number of pixels in a high-density region. x i , yi ) represents the coordinate value of the i-th pixel, and E(·) represents the calculation of the expectation function.

[0035] In some specific implementation schemes, the process of determining the high-magnification scanning path in step S4 is as follows:

[0036] S41. Using the coordinate vector of the first target organism as a reference, determine the high-power scanning range and obtain several high-power scanning areas.

[0037] S42. Based on the distribution order of the first target organism in the medium magnification microscopic image, determine the high magnification scanning path of the microscope, and sequentially acquire images of the high magnification scanning area according to the high magnification scanning path to obtain a high magnification microscopic image.

[0038] In some specific implementation schemes, the specific process for obtaining the total number of the second target organisms in step S5 is as follows:

[0039] S51, Statistical analysis of the coordinate vector of the second target organism in high-magnification micrographs ( x s , y s ) and the second actual number of the second target organism under each coordinate vector. N 实2 ;

[0040] S52. Input the coordinate vector of the second target organism into the constructed biological association distribution model to obtain the second biological distribution probability of the second target organism. P ( x s , y s );

[0041] S53. Based on the distribution probability of the second organism and the actual number of the second organism, estimate the total number of the second target organisms in the observation area of ​​the sample to be detected. N 总2 :

[0042]

[0043] Where E(·) represents the expectation function.

[0044] In some specific implementation schemes, the biological association distribution model deploys a global probability density function for the biological association distribution, calculated as follows:

[0045]

[0046] in,( x a , y a) represents the coordinate vector of the first target organism, ( x s , y s () represents the coordinate vector of the second target organism; s a This represents the correlation radius between the second target organism and the first target organism. m This indicates the total number of pixels contained in a high-magnification microscope image. x j , y j ) represents the coordinate value of the j-th pixel in a high-magnification microscope image, and E(·) represents the calculation of the expectation function.

[0047] Secondly, this application provides a sparse scanning path planning system based on multi-magnification microscopic images, comprising:

[0048] The low-magnification scanning and recognition module is used to set the low-magnification scanning area of ​​the observation area of ​​the sample to be detected, obtain the complete bacterial floc distribution image, and select the high-density area of ​​bacterial floc distribution from the bacterial floc distribution image according to the bacterial floc distribution probability to establish a bacterial floc probability distribution map.

[0049] The medium magnification scanning control module is used to input the probability distribution map of bacterial micelles into the constructed bacterial micelle-dependent biodistribution model to obtain the global probability density of each high-density region. Based on the global probability density, the medium magnification scanning path of the observation area of ​​the sample to be tested is determined by the microscope to obtain a medium magnification microscopic image.

[0050] The first target organism quantity statistics module is used to input the medium magnification microscopic image into the constructed floc-dependent biological distribution model to obtain the total number of the first target organisms in the observation area of ​​the sample to be detected;

[0051] The high-magnification scanning control module is used to determine the high-magnification scanning path of the observation area of ​​the sample to be tested by the microscope based on the distribution order of the first target organism in the medium-magnification microscopic image, so as to acquire the high-magnification microscopic image.

[0052] The second target organism quantity statistics module inputs high-magnification microscopic images into the constructed biological correlation distribution model to obtain the total number of second target organisms in the observation area of ​​the sample to be detected.

[0053] The advantages of this invention over the prior art are as follows:

[0054] This invention proposes a method combining "biological distribution association modeling + global probability distribution calculation + hierarchical scanning focusing strategy," which specifically overcomes the core shortcomings of existing technologies while achieving a balance between efficiency, accuracy, and cost.

[0055] 1. Biodistribution correlation modeling breaks through the limitations of local probability analysis based on a single coordinate point. Based on the distribution correlation patterns among microorganisms (the dependence between flocs and metazoans / protozoans, and the spatial correlation between small and large protozoans), a global probability density function is constructed. The extraction of floc distribution features is simplified by grayscale image conversion and normal probability conversion, making microbial distribution prediction more in line with actual ecological laws and improving the global accuracy of distribution prediction.

[0056] 2. A layered scanning path is proposed, and a layered focusing strategy is designed: "low magnification full-area scanning to lock high-density areas - medium magnification to select Q% high-density areas and focus on large target organisms - high magnification to select W% high-density areas to capture small target organisms on the basis of medium magnification". Relying on the global probability model, the high probability distribution area of ​​target organisms is accurately locked, which to a certain extent solves the drawback of random selection of scanning areas in existing technologies, reduces the blank scan rate and improves scanning efficiency.

[0057] 3. Flexible scenario adaptation: It establishes a quantitative correspondence between "medium magnification Q% and high magnification W% selection ratio" and "total reduction in scanning volume and estimation accuracy", which overcomes the problem of poor adaptability of existing technical parameters. By adjusting the Q and W parameters, it can flexibly match the needs of different equipment conditions for activated sludge monitoring, allowing users to sacrifice a small amount of accuracy in exchange for controllable project costs and improved efficiency. Attached Figure Description

[0058] Figure 1 A flowchart of a sparse scanning path planning method based on multi-magnification microscopic images provided in an embodiment of the present invention;

[0059] Figure 2 This is a density distribution diagram of fungal flocs provided in an embodiment of the present invention;

[0060] Figure 3 This is a schematic diagram of the scanning area using a medium magnification microscope, provided in an embodiment of the present invention. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention.

[0063] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0064] Furthermore, for clarity and brevity, descriptions of well-known structures, functions, and configurations may have been omitted. Those skilled in the art will recognize that various changes and modifications can be made to the examples described herein without departing from the spirit and scope of this disclosure.

[0065] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0066] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0067] Example 1

[0068] like Figure 1 As shown, this embodiment provides a sparse scanning path planning method based on multi-magnification microscopic images, specifically including the following steps:

[0069] S1. Set the low-magnification scanning area of ​​the observation area of ​​the sample to be tested, obtain the complete bacterial floc distribution image, and select the high-density area of ​​bacterial floc distribution from the bacterial floc distribution image according to the bacterial floc distribution probability to establish a bacterial floc probability distribution map.

[0070] Using low-magnification full scans to identify high-density regions with high-value scans, and subsequent scans based on these high-density regions, the specific process for establishing the bacterial floc probability distribution map in step S1 is as follows:

[0071] S11. Perform a full-area scan of the observation area of ​​the sample to be tested to obtain a complete image of the distribution of bacterial flocs;

[0072] S12. Convert the bacterial floc distribution image into a grayscale image, determine the bacterial floc features based on the brightness values ​​of each pixel in the grayscale image, and obtain the probability density of the bacterial flocs based on the bacterial floc features.

[0073] S13. Calculate the mean and standard deviation of the brightness values ​​of all pixels in the grayscale image, and calculate the probability density value of each pixel becoming the center of the fungal floc using the normal distribution based on the mean and standard deviation.

[0074] The method for calculating the probability density value of each pixel is as follows:

[0075] (1)

[0076] in, r ( x , y ) indicates that the coordinates are ( x , y ) represents the probability density value of a pixel, g represents the gray value of the pixel, μ represents the mean, and σ represents the standard deviation.

[0077] S14. Sort all pixel probability density values ​​from largest to smallest, select the top few pixels as the center position of the bacterial floc, and obtain the bacterial floc probability distribution map.

[0078] S2. Input the probability distribution map of the bacterial floc into the constructed bacterial floc-dependent biological distribution model to obtain the global probability density of each high-density region. Based on the global probability density, determine the medium magnification scanning path of the observation area of ​​the sample to be tested under the microscope and obtain the medium magnification microscopic image.

[0079] Specifically, the method for determining the medium magnification scan path is as follows:

[0080] S21. Determine the high-density areas of fungal floc distribution based on the center coordinates of the fungal flocs, and calculate the global distribution probability of fungal floc-dependent organisms appearing in each high-density area based on the fungal floc-dependent organism distribution model.

[0081] S22. Sort the global distribution probability values ​​of all high-density areas from largest to smallest, and select the top-ranked high-density areas as the medium magnification scanning areas of the first target organism in the observation area of ​​the sample to be tested.

[0082] S23. Determine the medium magnification scanning path of the microscope according to the distribution order of the high-density areas corresponding to the medium magnification scanning area, and sequentially acquire images of the medium magnification scanning area according to the medium magnification scanning path to obtain medium magnification microscopic images.

[0083] S3. Input the medium-magnification microscopic image into the constructed floc-dependent biodistribution model to obtain the total number of the first target organisms in the observation area of ​​the sample to be detected. The first target organisms are large organisms, which belong to the floc-dependent organisms, such as metazoans (e.g., rotifers, nematodes) and large protozoa (e.g., twigworms). The specific process for obtaining the total number of the first target organisms is as follows:

[0084] S31, the coordinate vector of the first target organism in a statistical medium-magnification micrograph (S31, S32 ... x a , y a ) and the first actual number of the first target organism under each coordinate vector. N 实1 ;

[0085] S32. Input the coordinate vector of the first target organism into the constructed floc-dependent biodistribution model to obtain the first biodistribution probability of the first target organism. P ( x a , y a );

[0086] The distribution model of fungal floc-dependent organisms includes the global probability density function of fungal floc-dependent organisms, calculated as follows:

[0087] (2)

[0088] in,( x a , y a () represents the coordinate vector of the first target organism in a high-density area. r ( x a , y a ) indicates that at coordinates ( x a , y a The density vector of bacterial flocs near ) is given by ( x f , y f ) represents the coordinate vector of the center position of the fungal floc; s f This represents a constant representing the radius of influence of the bacterial floc on the first target organism. n This indicates the number of pixels in a high-density region. x i , y i ) represents the coordinate value of the i-th pixel, and E(·) represents the calculation of the expectation function.

[0089] S33. Based on the probability distribution of the first organism and the first actual number, estimate the total number of the first target organisms in the observation area of ​​the sample to be detected. N 总1 :

[0090] (3)

[0091] Where E(·) represents the expectation function.

[0092] S4. Based on the distribution order of the first target organism in the medium magnification microscopic image, determine the high magnification scanning path of the observation area of ​​the sample to be tested under the microscope, and obtain the high magnification microscopic image.

[0093] The specific process for determining the high-magnification scanning path in step S4 is as follows:

[0094] S41. Using the coordinate vector of the first target organism as a reference, determine the high-power scanning range and obtain several high-power scanning areas.

[0095] S42. Based on the distribution order of the first target organism in the medium magnification microscopic image, determine the high magnification scanning path of the microscope, and sequentially acquire images of the high magnification scanning area according to the high magnification scanning path to obtain a high magnification microscopic image.

[0096] S5. Input the high-magnification microscopic image into the constructed biologically related distribution model to obtain the total number of second target organisms in the observation area of ​​the sample to be detected. The second target organisms include small protozoa (such as paramecia and amoebas). The distribution of the second target organisms is closely related to the spatial location of the first target organisms.

[0097] Specifically, the process of obtaining the total number of the second target organisms in step S5 is as follows:

[0098] S51, Statistical analysis of the coordinate vector of the second target organism in high-magnification micrographs ( x s , y s ) and the second actual number of the second target organism under each coordinate vector. N 实2 ;

[0099] S52. Input the coordinate vector of the second target organism into the constructed biological association distribution model to obtain the second biological distribution probability of the second target organism. P ( x s , y s );

[0100] The biological association distribution model deploys a global probability density function for the biological association distribution, calculated as follows: (4)

[0101] in,( x a , y a ) represents the coordinate vector of the first target organism, ( x s , y s () represents the coordinate vector of the second target organism; s a This represents the correlation radius between the second target organism and the first target organism. mThis indicates the total number of pixels contained in a high-magnification microscope image. x j , y j ) represents the coordinates of the j-th pixel in the high-magnification microscope image, and E(·) represents the expectation function calculation. S53. Based on the probability distribution of the second organism and the second actual quantity, estimate the total number of the second target organisms in the observation area of ​​the sample to be detected. N 总2 :

[0102] (5)

[0103] Where E(·) represents the expectation function.

[0104] Understandably, to address the technical problems of traditional microscopic scanning, such as low efficiency due to full-area coverage, poor statistical accuracy due to local random scanning, and high blank scan rate, especially in the scenario of microscopic examination of activated sludge microorganisms, this embodiment relies on the distribution correlation patterns among microorganisms (the dependence relationship between flocs and metazoans / protozoa, and the spatial correlation between small protozoa and large protozoa) to overcome the limitations of single coordinate point probability analysis. A global probability density function is constructed through multi-dimensional sample collection, and the extraction of floc distribution features is simplified through grayscale image conversion and normal probability conversion, quantifying the global distribution characteristics of microorganisms. A layered focusing scanning strategy is adopted: "low magnification full scan to lock high-density areas—medium magnification to select Q% high-density areas to focus on large target organisms—high magnification to select W% areas to capture small target organisms." Locking high-density areas replaces full-area coverage, significantly reducing the scan volume and blank scan rate. Furthermore, a normalized probability model is combined to achieve accurate estimation of the global number of target organisms.

[0105] The method provided in this embodiment is widely used in scenarios involving the statistical analysis of microbial counts under microscopy, such as the microscopic examination of activated sludge in water plants. It optimizes the scanning efficiency and accuracy of existing scanning schemes for estimating microbial counts. Its potential applications include:

[0106] ① Environmental monitoring field: Used for microbial ecological surveys of water and soil environments, such as rapid water quality screening and soil microbial distribution assessment, achieving efficient and accurate statistics of microorganisms in environmental samples through stratified scanning.

[0107] ② Medical testing field: It can be applied to high-precision microbial detection of clinical samples, such as microscopic analysis of pathogenic microorganisms in body fluids and tissue samples, ensuring detection accuracy under the premise of low empty scan rate.

[0108] ③ Scientific research: It provides efficient observation methods for microbial ecology research and is suitable for the investigation and quantitative analysis of the distribution patterns of microorganisms in different habitats.

[0109] To better illustrate the solution concept of this embodiment, the following example uses the statistical analysis of microbial counts in the activated sludge scenario of a water plant to explain the overall solution from model building to application and evaluation. Specifically:

[0110] Step 1: Obtain the original training data for the model and collect basic distribution information of the target organism and related factors.

[0111] Through multi-dimensional sample collection, high-magnification microscopic observation, and quantitative analysis, basic distribution data of the target organism and related factors were obtained. The specific operations are as follows:

[0112] ① Multi-dimensional sample collection: Collect activated sludge samples from different biological treatment units (aerobic tanks, anoxic tanks, etc.), under different water quality conditions (different turbidity, different dissolved oxygen concentrations, etc.) within the same biological treatment unit, and at different time points to construct a diverse sample library to be tested, ensuring that the samples cover the typical living environment of activated sludge.

[0113] ② Multi-magnification microscopic imaging: Each sample was photographed at magnifications of 100x, 400x, and 1000x. During the imaging process, the spatial coordinates and actual numbers of bacterial flocs, metazoans (such as rotifers and nematodes), large protozoa (such as nematodes), and small protozoa (such as paramecia and amoebas) were accurately recorded to form a raw image dataset of microbial distribution.

[0114] ③ Quantitative data processing: Based on the image data captured by microscopy, the quantitative indicators of biological distribution at different magnifications were calculated respectively: the density of bacterial flocs in each image at 100x magnification; the biological distribution density of metazoans and large protozoa at 400x magnification; and the biological distribution density of various small protozoa at 1000x magnification, providing a quantitative basis for the subsequent construction of the global probability density function.

[0115] Step 2: Establish a normalized target organism probability distribution model

[0116] Based on the original training data obtained in step 1, fit the number of target organisms of different ecological types appearing at a certain pixel coordinate (for ease of description, the coordinates in the following text are all pixel coordinates, or simply: coordinates);

[0117] It should be noted that the meaning and construction method of the coordinates in this embodiment are as follows:

[0118] This embodiment involves ( x , y ), ( x a , y a ), ( xf , y f ), ( x s , y s The iso-coordinate system is a spatial coordinate system for microscopic images, encompassing both pixel coordinates and physical coordinates. These two coordinates can be mapped through scale transformation. The specific meaning and construction method are as follows:

[0119] 1) The core meaning of coordinates

[0120] ① Pixel coordinates: Based on the digital image taken by the microscope, with the upper left corner of the image as the origin (0,0), the x-axis is horizontally to the right and the y-axis is vertically downward. The unit is "pixel". It represents the pixel position of the microorganism in the microscopic image and is the basic coordinate form for subsequent probability calculations.

[0121] ② Physical coordinates: derived from pixel coordinates, representing the spatial location of microorganisms in the actual activated sludge sample, in micrometers (μm), reflecting the actual spatial distribution characteristics of the sample.

[0122] 2) Steps for constructing a coordinate system, taking microscopic observation of activated sludge samples from a water plant as an example:

[0123] ① Physical area calibration of the sample: The activated sludge sample is evenly spread on the fixed observation area of ​​the glass slide (e.g., 10). μm ×10 μm (A square region), mark the physical boundary of the observation area on the slide (e.g., the upper left corner is the physical origin O(0,0), and the lower right corner is Oˊ(10). μm 10 μm (), clearly define the actual physical space range of the sample.

[0124] ② When photographing samples using a microscope at the corresponding magnification, record the pixel-to-physical scale conversion coefficient k of the microscope (e.g., k=2 at 100x magnification). μm A pixel, meaning one pixel corresponds to 2 of the actual sample. μm (Length). After the image is captured, a pixel coordinate system is established with the upper left corner of the microscopic image as the pixel origin. At this point, any pixel in the image ( x 像素 , y 像素 ), can be converted to physical coordinates using the following formula: For example, at 100x magnification, the physical coordinates corresponding to pixel coordinates (50, 80) are (100, 80). μm 160 μm ).

[0125] By analyzing the probability density of each coordinate point in the entire region After normalization, a probability density reflecting the global distribution characteristics of microorganisms is obtained. Different biological distribution models are constructed according to different types of target organisms, specifically including:

[0126] 2.1 Floc-dependent biodistribution model (applicable to metazoans and large protozoa)

[0127] The distribution of metazoans (such as rotifers and nematodes) and macroprotozoa (such as *Cyclophorus*) is strongly correlated with the density of fungal flocs. Based on this, a global probability density function for fungal floc-dependent organisms is constructed, expressed as:

[0128] (6)

[0129] Symbol explanation:

[0130] r ( x a , y a ):coordinate( x a , y a The density vector of bacterial flocs near point ) where ( x a , y a ) represents the coordinate vector of the first target organism (metazoans, large protozoa) within a high-density area; x f , y f ): Coordinate vector of the center of the fungal floc;

[0131] s f The radius of influence of the bacterial floc on the target organism is a constant and the only parameter to be estimated for this function. It can be solved by publicly available optimization methods such as maximum likelihood estimation and gradient descent.

[0132] n : Sample size, denominator is the global normalization term, ensuring the function value satisfies P ( x a , y a )∈[0,1];

[0133] E(·): Expectation function, used to quantify the probability expectation at the global level.

[0134] Function meaning: Outputs any location within a high-density region ( x a , y aThe probability value of the presence of a floc-dependent organism is indicated by the presence of a target organism at that location. The higher the probability value, the greater the likelihood of finding the target organism at that location.

[0135] 2.2 Biologically related distribution model (applicable to small protozoa)

[0136] The distribution of small protozoa (such as paramecia and amoebas) is closely related to the spatial location of large protozoa / metazoa. Based on this, a global probability density function for the biologically related distribution is constructed, expressed as follows:

[0137] (7)

[0138] Symbol explanation:

[0139] ( x s , y s ): Represents the coordinate vector of the second target organism;

[0140] ( x a , y a ): Coordinate vectors of large protozoa / metazoa;

[0141] s a The association radius between small protozoa and large protozoa / metazoa can be solved using publicly available optimization methods such as maximum likelihood estimation and gradient descent.

[0142] m : Sample size, denominator is the global normalization term, ensuring the function value satisfies P ( x s , y s )∈[0,1];

[0143] E(·): Expectation function, used to quantify the probability expectation at the global level.

[0144] Function meaning: Outputs any location within a high-density region ( x s , y s The probability value of finding small protozoa is indicated by the probability value. The higher the probability value, the greater the likelihood of finding small protozoa at that location.

[0145] Step 3: Low-magnification scanning and path planning to lock in the high-density scanning area;

[0146] Based on 100x magnification (low magnification), the core area for subsequent scanning is identified through a "full-area scan - high-density area screening" planning path.

[0147] ① Full-area scan: The scanning range covers the entire sample area (e.g., 10). mm ×10 mm ), to obtain a complete image of the distribution of bacterial flocs;

[0148] ② Screening of high-density areas: Extracting the coordinates of the center position of the bacterial flocs ( x f , y f ) and probability density r ( x , y The method for determining the center of the bacterial floc is as follows:

[0149] 1) Grayscale conversion: The color microscopic image is directly converted into a grayscale image. The pixel brightness value (0-255) reflects the characteristics of the fungal flocs. The low grayscale value corresponding to the fungal flocs will be mapped to a higher probability density, while the background corresponds to a low probability density.

[0150] 2) Normal probability transformation: Statistically analyze the brightness value distribution of all pixels in the grayscale image, calculate its mean μ and standard deviation σ, and then convert the grayscale value g of each pixel into its corresponding probability density value using the normal distribution formula. r ( x , y For the specific calculation formula, please refer to formula (1). The higher the probability density value, the more likely the pixel is to become the center of the bacterial floc.

[0151] ③ Establish a probability distribution map of bacterial flocs, such as Figure 2 As shown in the figure, the darker the color, the higher the probability density of that region. r ( x , y The higher the value, the more the red pentagram represents the sampling point of the photo, i.e., the coordinates of the center position (e.g., the first 10 points with the highest probability density).

[0152] Step 4: Medium magnification scanning and path planning, focusing on the largest target organism and estimating its number;

[0153] Based on the high-density areas selected under low magnification, switch to a medium magnification (400x magnification) microscope to focus on floc-dependent organisms (metazoans, large protozoa):

[0154] ① Medium magnification scanning area localization: using the center coordinate vector of the high-density area under low magnification ( x f , yf Based on this, the scanning range of the medium magnification microscope is set to a square area of ​​"center coordinates ± H (e.g., H = 10 pixels)" to cover the potential distribution area of ​​fungal floc-dependent organisms. Then, according to formulas (1) and (6), the areas with high global probability density of the target organism under medium magnification are selected for imaging (e.g., the first 150 areas with high probability density). Figure 3 As shown by the orange dot.

[0155] ② Path generation: Based on the distribution order of high-density areas in the low-power microscope, the scanning areas in the medium-power microscope are located sequentially, and the actual number of metazoans and large protozoa in each scanning area is counted and recorded as: (( x a , y a ), N 实1 ),in( x a , y a () represents the coordinate vector of the first target organism a. N 实1 Indicates in ( x a , y a The quantity vector of the first target organism a under the coordinate vector.

[0156] ③ Global quantity estimation: Substitute the bacterial floc density of the medium magnification scanning area into formula (2) to calculate the distribution probability of the target organism in that area. P ( x a , y a By combining the "local quantity - probability" relationship and substituting it into formula (3), the total number of samples can be calculated. N 总1 .

[0157] Step 5: High-powered scanning and path planning to capture smaller secondary target organisms and estimate their numbers.

[0158] Based on the coordinates of the first target organism identified under medium magnification, switch to 1000x magnification (high magnification), focus on the small protozoa, and repeat step 4:

[0159] ① High-power scanning area localization: Target biological coordinates identified using medium-power microscope ( x a , y aBased on this, the high-power microscope scanning range is set to a square area with "center coordinates ± H (e.g., H = 5 pixels)" to cover the potential distribution area of ​​small protozoa.

[0160] ② Path generation: Based on the distribution order of the target organisms under medium magnification, locate each high magnification scanning area sequentially, count the actual number of small protozoa, and record it as: (( x s , y s ), N 实2 ),in( x s , y s () represents the coordinate vector of the second target organism s. N 实2 Indicates in ( x s , y s The number vector of target organisms under coordinates ).

[0161] ③ Global quantity calculation: The coordinates of the second target organism in the high-power scanning area ( x s , y s Substitute into formula (4) to calculate the distribution probability of small protozoa in this area. P ( x s , y s Based on formula (5), the total number of small protozoa in the entire sample can be estimated. N 总2 .

[0162] Step 6: Display the scanning results and quantitatively analyze the improvement in estimation accuracy and efficiency under different scan volumes.

[0163] The quantitative indicators of scanning effect are defined by the correspondence between "scan volume - estimation accuracy":

[0164] ① Reduce scan rate (X%): refers to the actual number of scans performed in this invention. S 实扫 Compared to traditional full-area scan times S 全扫 The ratio, i.e.

[0165]

[0166] ② Estimation accuracy (P%): refers to the estimated global quantity. N 推算 Compared with the actual full-area count resultsN 实际 The degree of agreement, that is:

[0167]

[0168] The scanning results are shown in Table 1 below:

[0169] Table 1. Comparison of Scanning Results

[0170]

[0171] Note: This invention is based on a low-power full scan. Under medium power, only a Q% region is selected from the high-density area (n%) of the samples screened under low power. Under high power, a W% region is then selected from the medium-power scanning area. The blank scan rate refers to the proportion of the scanned area where no target microorganisms were found out out of the total scanned area. The statistical results in the table were obtained under the following scenarios:

[0172] S 全扫 = Low magnification (100x) scan count (400 scans) + Medium magnification (400x) scan count (6400 scans) + High magnification (1000x) scan count (57600 scans) = 64400 scans

[0173] S 实扫 = Number of scans at low magnification (100x) (400 scans) + Number of scans at medium magnification (400x) (6400 scans × n% × Q%) + Number of scans at high magnification (1000x) (57600 scans × n% × Q% × W%)

[0174] In practical applications, a comparison table of n%, Q%, W% and P% can be established first (as shown in Table 1). Based on the capabilities and accuracy requirements of the imaging equipment, the number of scans can be flexibly selected to improve scanning efficiency and application feasibility.

[0175] Example 2

[0176] To implement the method in Example 1, this example provides a sparse scanning path planning system based on multi-magnification microscopic images, including:

[0177] The low-magnification scanning and recognition module is used to set the low-magnification scanning area of ​​the observation area of ​​the sample to be detected, obtain the complete bacterial floc distribution image, and select the high-density area of ​​bacterial floc distribution from the bacterial floc distribution image according to the bacterial floc distribution probability to establish a bacterial floc probability distribution map.

[0178] The medium magnification scanning control module is used to input the probability distribution map of bacterial micelles into the constructed bacterial micelle-dependent biodistribution model to obtain the global probability density of each high-density region. Based on the global probability density, the medium magnification scanning path of the observation area of ​​the sample to be tested is determined by the microscope to obtain a medium magnification microscopic image.

[0179] The first target organism quantity statistics module is used to input the medium magnification microscopic image into the constructed floc-dependent biological distribution model to obtain the total number of the first target organisms in the observation area of ​​the sample to be detected;

[0180] The high-magnification scanning control module is used to determine the high-magnification scanning path of the observation area of ​​the sample to be tested by the microscope based on the distribution order of the first target organism in the medium-magnification microscopic image, so as to acquire the high-magnification microscopic image.

[0181] The second target organism quantity statistics module inputs high-magnification microscopic images into the constructed biological correlation distribution model to obtain the total number of second target organisms in the observation area of ​​the sample to be detected.

[0182] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Based on the technical essence of the present invention, any simple modifications, equivalent substitutions, and improvements made to the above embodiments within the spirit and principles of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A sparse scanning path planning method based on multi-magnification microscopic images, characterized in that, Specifically, the following steps are included: S1. Set the low-magnification scanning area of ​​the observation area of ​​the sample to be tested, obtain the complete bacterial floc distribution image, and select the high-density area of ​​bacterial floc distribution from the bacterial floc distribution image according to the bacterial floc distribution probability to establish a bacterial floc probability distribution map. S2. Input the probability distribution map of the bacterial floc into the constructed bacterial floc-dependent biological distribution model to obtain the global probability density of each high-density region. Based on the global probability density, determine the medium magnification scanning path of the observation area of ​​the sample to be tested under the microscope and obtain the medium magnification microscopic image. S3. Input the medium-magnification microscopic image into the constructed floc-dependent biodistribution model to obtain the total number of the first target organisms in the observation area of ​​the sample to be tested; S4. Based on the distribution order of the first target organism in the medium magnification microscopic image, determine the high magnification scanning path of the observation area of ​​the sample to be tested by the microscope, and obtain the high magnification microscopic image. S5. Input the high-magnification microscopic image into the constructed biological association distribution model to obtain the total number of the second target organisms in the observation area of ​​the sample to be detected.

2. The sparse scanning path planning method based on multi-magnification microscopic images according to claim 1, characterized in that, The specific process of establishing the probability distribution map of bacterial flocs in step S1 is as follows: S11. Perform a full-area scan of the observation area of ​​the sample to be tested to obtain a complete image of the distribution of bacterial flocs; S12. Convert the bacterial floc distribution image into a grayscale image, determine the bacterial floc features based on the brightness values ​​of each pixel in the grayscale image, and obtain the probability density of the bacterial flocs based on the bacterial floc features. S13. Calculate the mean and standard deviation of the brightness values ​​of all pixels in the grayscale image, and calculate the probability density value of each pixel becoming the center of the fungal floc using the normal distribution based on the mean and standard deviation. S14. Sort all pixel probability density values ​​from largest to smallest, select the top few pixels as the center position of the bacterial floc, and obtain the bacterial floc probability distribution map.

3. The sparse scanning path planning method based on multi-magnification microscopic images according to claim 2, characterized in that, The method for calculating the probability density value of each pixel becoming the center of the fungal floc in step S13 is as follows: in, ρ ( x , y ) indicates that the coordinates are ( x , y ) represents the probability density value of a pixel, g represents the gray value of the pixel, μ represents the mean, and σ represents the standard deviation.

4. The sparse scanning path planning method based on multi-magnification microscopic images according to claim 2, characterized in that, The method for determining the medium magnification scan path is as follows: S21. Determine the high-density areas of fungal floc distribution based on the center coordinates of the fungal flocs, and calculate the global distribution probability of fungal floc-dependent organisms appearing in each high-density area based on the fungal floc-dependent organism distribution model. S22. Sort the global distribution probability values ​​of all high-density areas from largest to smallest, and select the top-ranked high-density areas as the medium magnification scanning areas of the first target organism in the observation area of ​​the sample to be tested. S23. Determine the medium magnification scanning path of the microscope according to the distribution order of the high-density areas corresponding to the medium magnification scanning area, and acquire the images of the medium magnification scanning area in sequence according to the medium magnification scanning path to obtain medium magnification microscopic images.

5. The sparse scanning path planning method based on multi-magnification microscopic images according to claim 2, characterized in that, The specific process for obtaining the total number of the first target organisms in the observation area of ​​the sample to be tested in step S3 is as follows: S31, the coordinate vector of the first target organism in a statistical medium-magnification micrograph (S31, S32 ... x a , y a ) and the first actual number of the first target organism under each coordinate vector. N 实1 ; S32. Input the coordinate vector of the first target organism into the constructed floc-dependent biodistribution model to obtain the first biodistribution probability of the first target organism. P ( x a , y a ); S33. Based on the probability distribution of the first organism and the first actual number, estimate the total number of the first target organisms in the observation area of ​​the sample to be detected. N 总1 : Where E(·) represents the expectation function.

6. The sparse scanning path planning method based on multi-magnification microscopic images according to claim 1, characterized in that, The distribution model of fungal floc-dependent organisms includes the global probability density function of fungal floc-dependent organisms, calculated as follows: in,( x a , y a () represents the coordinate vector of the first target organism in a high-density area. ρ ( x a , y a ) indicates that at coordinates ( x a , y a The density vector of bacterial flocs near ) is given by ( x f , y f () represents the coordinate vector of the center position of the fungal floc; σ f This represents a constant representing the radius of influence of the bacterial floc on the first target organism. n This indicates the number of pixels in a high-density region. x i , y i ) represents the coordinate value of the i-th pixel, and E(·) represents the calculation of the expectation function.

7. The sparse scanning path planning method based on multi-magnification microscopic images according to claim 5, characterized in that, The specific process of determining the high-magnification scanning path of the observation area of ​​the sample to be tested under the microscope in step S4 is as follows: S41. Using the coordinate vector of the first target organism as a reference, determine the high-power scanning range and obtain several high-power scanning areas. S42. Based on the distribution order of the first target organism in the medium magnification microscopic image, determine the high magnification scanning path of the microscope, and sequentially acquire images of the high magnification scanning area according to the high magnification scanning path to obtain a high magnification microscopic image.

8. The sparse scanning path planning method based on multi-magnification microscopic images according to claim 7, characterized in that, The specific process for obtaining the total number of the second target organisms in step S5 is as follows: S51, Statistical analysis of the coordinate vector of the second target organism in high-magnification micrographs ( x s , y s ) and the second actual number of the second target organism under each coordinate vector. N 实2 ; S52. Input the coordinate vector of the second target organism into the constructed biological association distribution model to obtain the second biological distribution probability of the second target organism. P ( x s , y s ); S53. Based on the distribution probability of the second organism and the actual number of the second organism, estimate the total number of the second target organisms in the observation area of ​​the sample to be detected. N 总2 : Where E(·) represents the expectation function.

9. A sparse scanning path planning method based on multi-magnification microscopic images according to claim 8, characterized in that, The biological association distribution model deploys a global probability density function for the biological association distribution, calculated as follows: in,( x a , y a ) represents the coordinate vector of the first target organism, ( x s , y s () represents the coordinate vector of the second target organism; σ a This represents the correlation radius between the second target organism and the first target organism. m This indicates the total number of pixels contained in a high-magnification microscope image. x j , y j ) represents the coordinate value of the j-th pixel in a high-magnification microscope image, and E(·) represents the calculation of the expectation function.

10. A sparse scanning path planning system based on multi-magnification microscopic images, characterized in that, include: The low-magnification scanning and recognition module is used to set the low-magnification scanning area of ​​the observation area of ​​the sample to be detected, obtain the complete bacterial floc distribution image, and select the high-density area of ​​bacterial floc distribution from the bacterial floc distribution image according to the bacterial floc distribution probability to establish a bacterial floc probability distribution map. The medium magnification scanning control module is used to input the probability distribution map of bacterial micelles into the constructed bacterial micelle-dependent biodistribution model to obtain the global probability density of each high-density region. Based on the global probability density, the medium magnification scanning path of the observation area of ​​the sample to be tested is determined by the microscope to obtain a medium magnification microscopic image. The first target organism quantity statistics module is used to input the medium magnification microscopic image into the constructed floc-dependent biological distribution model to obtain the total number of the first target organisms in the observation area of ​​the sample to be detected; The high-magnification scanning control module is used to determine the high-magnification scanning path of the observation area of ​​the sample to be tested by the microscope based on the distribution order of the first target organism in the medium-magnification microscopic image, so as to acquire the high-magnification microscopic image. The second target organism quantity statistics module inputs high-magnification microscopic images into the constructed biological correlation distribution model to obtain the total number of second target organisms in the observation area of ​​the sample to be detected.

Images