CN116934776B - FISH microscopic image signal point region extraction method, device, equipment and medium - Google Patents

FISH microscopic image signal point region extraction method, device, equipment and medium Download PDF

Info

Publication number
CN116934776B
CN116934776B CN202310918649.0A CN202310918649A CN116934776B CN 116934776 B CN116934776 B CN 116934776B CN 202310918649 A CN202310918649 A CN 202310918649A CN 116934776 B CN116934776 B CN 116934776B
Authority
CN
China
Prior art keywords
image
signal point
gaussian
scale
fish
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202310918649.0A
Other languages
Chinese (zh)
Other versions
CN116934776A (en
Inventor
王荣荣
蒋和松
吴俊灵
何柯材
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou Micro Shot Technology Co ltd
Original Assignee
Guangzhou Micro Shot Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangzhou Micro Shot Technology Co ltd filed Critical Guangzhou Micro Shot Technology Co ltd
Priority to CN202310918649.0A priority Critical patent/CN116934776B/en
Publication of CN116934776A publication Critical patent/CN116934776A/en
Application granted granted Critical
Publication of CN116934776B publication Critical patent/CN116934776B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/10—Image enhancement or restoration using non-spatial domain filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10056—Microscopic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20048—Transform domain processing
    • G06T2207/20064—Wavelet transform [DWT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20172—Image enhancement details
    • G06T2207/20192—Edge enhancement; Edge preservation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Processing (AREA)

Abstract

The application discloses a method, a device, equipment and a medium for extracting a signal point region of a FISH microscopic image, which comprise the following steps: preprocessing an original image to obtain a first image; carrying out multi-scale wavelet summation denoising on the first image to obtain a second image; performing enhancement processing on a target area of the second image to obtain a third image; suppressing the non-target area of the third image through single-side second-order Gaussian cores in 4 directions to obtain a fourth image; collecting first center point coordinates of a signal point area in a fourth image, and mapping the first center point coordinates to a third image to obtain second center point coordinates; and selecting the signal point area in the third image according to the second center point coordinates as a final signal point area. The technical scheme of the application improves the extraction effect of the signal point region of the FISH microscopic image.

Description

FISH microscopic image signal point region extraction method, device, equipment and medium
Technical Field
The application relates to the technical field of image processing, in particular to a method, a device, equipment and a medium for extracting a signal point area of a FISH microscopic image.
Background
In the tumor canceration cell image processed by the FISH technology, parameters such as the position, the number, the aggregation degree and the like of signal points can effectively reflect the cytopathic degree. However, the data are obtained only manually, so that the method is very complicated, and is influenced by factors such as low imaging quality, complex characteristics of an observed object, strong subjectivity of manual diagnosis, visual fatigue of an observer and the like, so that an analysis result is inaccurate (false positive and false negative are high), a doctor is easy to misjudge the illness state, and the optimal diagnosis time of the patient is missed or medical accidents are caused. Therefore, it is important to extract signal points from images by image processing and analysis techniques to provide objective quantitative data to help quantify and verify the observed life process.
Signal points usually show local bright spots with larger brightness difference from the background in the image, and the existing spot detection methods can be divided into two categories: supervised and unsupervised methods. In the supervised approach, the model is prepared by a training process, where it is required to make predictions and correct when the predictions are wrong. In the unsupervised approach, user-specified values are used for noise suppression, parameter templates or models for speckle enhancement, and thresholds for distinguishing speckle from background.
In addition to a simple threshold method in an unsupervised method, most common spot detection methods construct a suitable function simulation point spread function (point spread function, PSF) by analyzing the formation process of fluorescence microscopic images, and select suitable parameters for the shape and size of spots. Kozubek proposes a watershed technology-based spot detection method, a progressive thresholding method, for detecting spots in 2D and 3D fluorescent stained intermittent nuclear images. Netten proposes a Kozubek-like dot label that enables blob detection by defining a region of interest, top-hat transformation (top-hat transformation), and top-hat transformation with variable thresholds. In addition Gue et al propose a local threshold based seed growth spot detection method, which acquires spot positions in microscopic images by median filtering, top hat transformation and local threshold. Raj proposes image-based multi-thresholding, first filtering the image with laplacian of gaussian (LoG) as defined in the formula to reduce the effects of noise, enhancing the spot structure. The filtered image is then thresholded. The The HDome method is proposed by Smal on the basis of the morphological HDome transformation of Vincent, and finally calculates each spot position through Gaussian filtering, HDome transformation and mean shift algorithm. Rezatofighi proposes a detection method named maximum likelihood HDome (i.e. MPHD), which calculates a suitable threshold h for each spot based on local information and designs an adaptive mask to enhance the spot for locating the spot in the fluorescence microscope image. the top-hat Rotational Morphology Processing (RMP) algorithm is a morphology processing algorithm proposed by Kimori for detecting spots from electron microscope images, and includes three steps of denoising, spot extraction and binarization. The speckle enhancement filtering (Spot ENHANCING FILTER, SE) is a method proposed by Sage for microscopic image speckle detection, which enhances speckle in an image while reducing background structure and suppressing noise. Olivo-Marin proposes a speckle detection method based on wavelet multi-scale product (wavelet multiscale product, WMP) for speckle detection in microscopic images, based on the assumption that: speckle occurs at each scale of the wavelet decomposition, and thus in a multi-scale product. Bo Zhang et al propose a multi-scale VST algorithm (MSVST) in which the image is subjected to multi-scale decomposition after variance-stabilized transformation (variance stabilizing transform, VST), the detail image of each scale is thresholded using FDR, and then HSD iterative reconstruction is performed to obtain a speckle enhancement image, and the target is extracted through gray thresholding. Jaiswal proposes a detection method similar to Sage, called "multiscale spot enhancement filtering" (MSSEF), and uses this method for detection of avian leukosis virus particles in confocal fluorescence microscopy images. Another approach based on optimal scale selection, referred to as adaptive thresholding of laplacian of gaussian images (ATLAS) with automatically selected scales, is proposed by Basset for bubble detection in microscopic imaging, and involves selecting the optimal ratio corresponding to the spot size in the image. Sbalzarini and Koumoutsakos develop feature point detection methods for detecting speckle in microscopic images, in which the detection of the speckle is based on locating local intensity maxima, followed by weighted intensity centroid calculations.
Noise interference exists in fluorescent microscope imaging, and the noise can be mainly classified into 3 types, namely noise based on equipment, noise based on samples, noise caused by insufficient sample preparation and improper microscope operation, and the noise can interfere with the final imaging quality and influence the extraction result of an algorithm; secondly, compared with the background, the contrast of some FISH image signal points is low, the edges are blurred, the visibility is poor, and the detection omission phenomenon is easy to generate in the conventional algorithm; finally, the presence of multiple large highlight regions in some FISH images can cause strong edge interference to the algorithm, resulting in false extraction.
Disclosure of Invention
The application provides a method, a device, equipment and a medium for extracting signal point areas of a FISH microscopic image, which can accurately extract the signal point areas in the FISH microscopic image.
In a first aspect, the present application provides a method for extracting a signal point region of a FISH microscopic image, which adopts the following technical scheme:
A method for extracting a signal point region of a FISH microscopic image comprises the following steps:
preprocessing an original image to obtain a first image;
Carrying out multi-scale wavelet summation denoising on the first image to obtain a second image;
performing enhancement processing on the target area of the second image to obtain a third image;
suppressing the non-target area of the third image through single-side second-order Gaussian cores in 4 directions to obtain a fourth image; collecting first center point coordinates of a signal point area in the fourth image, and mapping the first center point coordinates to the third image to obtain second center point coordinates;
and selecting a signal point area in the third image according to the second center point coordinates as a final signal point area.
Further, the preprocessing of the original image includes graying and normalizing the original image to obtain the first image.
Further, the performing multi-scale wavelet summation denoising on the first image includes the following steps:
Performing multi-scale decomposition on the first image to obtain a plurality of detail images and a plurality of layers of approximate images, wherein the calculation formula is as follows:
Wi(x,y)＝Ii-1(x,y)-Ii(x,y)
Wherein W i is the ith detail image, I i is the ith approximate image, the value range of I is [1, k ], k is the number of layers of the approximate image, I 0 (x, y) is the first image, x, y is the coordinates of the detail image and pixel points in the approximate image;
Calculating the median absolute deviation of a plurality of detail images and performing self-adaptive thresholding to obtain a denoising image, wherein the calculation formula is as follows:
MAD(Wi)＝median(abs(Wi-median(Wi)))
Wherein, For denoising images, abs () is an absolute value function and media () is a median function; MAD (W i) is a function of the median absolute deviation calculation for the detail image;
defining an approximation image I K of a K-th layer as an image background, wherein a first layer detail image W 1 is noise interference, and a calculation formula for denoising the first image is as follows:
wherein I D is the second image.
Further, the enhancing the target area of the second image specifically includes the following steps:
the signal points in the target area of the second image are equivalent to Gaussian spots, and the calculation formula is as follows:
Wherein X is a plane coordinate vector, X= [ X, y ] T,p0 [ E [0,1] represents the intensity of the spot, b 0 [ E [0,1] represents the intensity of the background, and omega 0 represents the scale of the spot;
Convolving the Gaussian spots with the preset radius through a Gaussian Laplace operator, wherein the calculation formula of the Gaussian Laplace operator is as follows:
Wherein, sigma is the standard deviation of Gaussian,
Selecting the scale of the Gaussian Laplace operator as
Determining a scale set S of the Gaussian spots according to the scale, and performing multi-scale filtering on a target area of the second image according to the scale set S to obtain a third image L (x):
Wherein,
Further, the calculation formula of the single-sided second-order gaussian kernel is as follows:
u(X；σ,θ)＝kC(X;σ,θ)+2·kR(X；σ,θ)
Wherein:
θ is the rotation angle of the single-sided second-order gaussian kernel;
and obtaining the single-side second-order Gaussian kernels in 4 directions by adjusting the rotation angle theta of the single-side second-order Gaussian kernels.
Further, the suppressing the non-target area of the third image by the single-sided second-order gaussian matching in 4 directions includes:
The direction D u = { i pi/2|i ∈ {0,1,2,3 };
and calculating the final response U (X) of the single-side second-order Gaussian kernel, wherein the calculation formula is as follows:
wherein S u is a set of scales S.
Further, the mapping the first center point coordinate to the third image specifically includes: let LU (X) =l (X) ·u (X), then perform threshold binarization processing on the fourth image, to obtain center point coordinates of each signal point in the fourth image, and map the center point coordinates to the third image.
In a second aspect, the present application provides a FISH microscopic image signal point area extracting device, which adopts the following technical scheme:
A device for extracting a signal point region of a FISH microscopic image, which applies the method for extracting the signal point region of the FISH microscopic image, comprises the following steps:
the image preprocessing module is used for preprocessing an original image to obtain a first image;
The image denoising module is used for carrying out multi-scale wavelet summation denoising on the first image to obtain a second image;
the image enhancement module is used for carrying out enhancement processing on the target area of the second image to obtain a third image suppression module, and the image enhancement module is used for suppressing the non-target area of the third image through single-side second-order Gaussian cores in 4 directions to obtain a fourth image;
and the output module is used for selecting a final signal point area according to the third image and the fourth image and outputting the final signal point area.
In a third aspect, the present application provides a computer device, which adopts the following technical scheme:
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of the first aspect when executing the program.
In a fourth aspect, the present application provides a computer readable storage medium, which adopts the following technical scheme:
A computer readable storage medium storing a computer program capable of being loaded by a processor and executing any one of the methods of the first aspect.
In summary, the present application includes at least one of the following beneficial technical effects:
According to the method, the device, the equipment and the medium for extracting the signal point region of the FISH microscopic image, the original image, namely the tumor canceration cell image processed by the FISH technology, is preprocessed, generally subjected to graying and normalization processing, so that the subsequent further operation on the image is facilitated; and secondly, carrying out multi-scale wavelet summation denoising on the first image subjected to graying and normalization processing to obtain a second image, wherein noise interference in the fluorescent microscopic image is removed, so that the signal-to-noise ratio can be effectively improved, the image quality and the identification degree of a target are improved, and the main sources of the noise in the fluorescent microscopic image are poisson photon scattering noise and additive Gaussian reading quantization noise. Then, carrying out enhancement processing on a target area of the second image to obtain a third image; after noise reduction, the signal point area needs to be enhanced, so that the spot area in the image can be clearer; meanwhile, as strong response can be generated on the image edge due to image enhancement, suppression is needed to be carried out on the non-target area of the third image through single-side second-order Gaussian cores in 4 directions, and the edge misextraction area and the non-speckle suppression area are removed; finally, in a third image, namely an enhanced image, a signal point area in the third image is selected as a final signal point area according to the second center point coordinate, so that the extraction effect of the signal point area of the FISH microscopic image is greatly improved.
Drawings
Fig. 1 is a schematic flow chart of a method for extracting a signal point region of a FISH microscopic image according to an embodiment of the application.
Fig. 2 is a schematic diagram of a second flow chart of a method for extracting a signal point region of a FISH microscopic image according to an embodiment of the present application.
Fig. 3 is a schematic diagram of the extraction result of FISH microscopic image signal point regions in the embodiment of the present application.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
The embodiment of the application discloses a method for extracting signal point areas of a FISH microscopic image, wherein the FISH microscopic image is a fluorescence in-situ hybridization image, and can be applied to the field of solid oncology, direct evidence of DNA amplification can be observed on interphase cell nuclei, and the quantity and fluorescence intensity of amplified DNA fluorescent signals displayed by the interphase cell nuclei are often related to the level of DNA amplification, so that the method is widely applied to auxiliary diagnosis of solid tumors such as breast cancer, bladder cancer, cervical cancer, lung cancer, lymph cancer and the like. In the tumor canceration cell image processed by the FISH technology, signal points represent spots in the image, and parameters such as the positions, the number, the aggregation degree and the like of the signal points can effectively reflect the cytopathic degree.
Referring to fig. 1, a FISH microscopic image signal point region extraction method includes the steps of:
s101: preprocessing an original image to obtain a first image;
S102: carrying out multi-scale wavelet summation denoising on the first image to obtain a second image;
S103: suppressing the non-target area of the third image through single-side second-order Gaussian cores in 4 directions to obtain a fourth image;
s104: collecting first center point coordinates of a signal point area in the fourth image, and mapping the first center point coordinates to the third image to obtain second center point coordinates;
s105: and selecting a signal point area in the third image according to the second center point coordinates as a final signal point area.
In the embodiment, the original image, namely the tumor canceration cell image processed by the FISH technology is preprocessed, generally, gray-scale processing and normalization processing are carried out, so that the subsequent further operation on the image is facilitated; and secondly, carrying out multi-scale wavelet summation denoising on the first image subjected to graying and normalization processing to obtain a second image, wherein noise interference in the fluorescent microscopic image is removed, so that the signal-to-noise ratio can be effectively improved, the image quality and the identification degree of a target are improved, and the main sources of the noise in the fluorescent microscopic image are poisson photon scattering noise and additive Gaussian reading quantization noise. Then, carrying out enhancement processing on a target area of the second image to obtain a third image; after noise reduction, the signal point area needs to be enhanced, so that the spot area in the image can be clearer; meanwhile, as strong response can be generated on the image edge due to image enhancement, suppression is needed to be carried out on the non-target area of the third image through single-side second-order Gaussian cores in 4 directions, and the edge misextraction area and the non-speckle suppression area are removed; finally, in a third image, namely an enhanced image, a signal point area in the third image is selected as a final signal point area according to the second center point coordinate, so that the extraction effect of the signal point area of the FISH microscopic image is greatly improved.
Referring to fig. 2, in one embodiment of the present application, the preprocessing of the original image in step S101 may include graying and normalizing the original image, that is, converting the original image into a gray image first, which may increase the speed of the subsequent processing and may highlight the signal point area in the original image. And then carrying out normalization processing on the gray level image, namely mapping the gray level value of the pixel point in the gray level image into the range between 0 and 1, and further enhancing the contrast of the image.
Referring to fig. 2, in one embodiment of the present application, step S102 of performing multi-scale wavelet summation denoising on the first image to obtain the second image may include the steps of:
S201: performing multi-scale decomposition on the first image to obtain a plurality of detail images and a plurality of layers of approximate images, wherein the calculation formula is as follows:
Wi(x,y)＝Ii-1(x,y)-Ii(x,y)
Wherein W i is the ith detail image, I i is the ith approximate image, the value range of I is [1, k ], k is the number of layers of the approximate image, I 0 (x, y) is the first image, x, y is the coordinates of the detail image and pixel points in the approximate image;
In this embodiment, removing noise interference in the first image can effectively improve the signal-to-noise ratio, improve the image quality and the recognition degree of the target, and the main sources of noise in the first image are poisson photon scattering noise and additive gaussian type reading quantization noise. In the embodiment of the application, an a trous wavelet based on a B3 spline ([ 1,4,6,4,1 ]/16) is adopted to carry out K-layer multi-scale decomposition on an original image I 0, so as to obtain a detail image W i (i=1, the..K.) and an approximate image I K, wherein I K is the result of convolution of I K-1 on the B3 spline row by row. After each layer of convolution, 2 i-1 -1 0's are inserted between adjacent elements of the B3 spline to adapt to the calculation of the next scale.
S202: calculating the median absolute deviation of a plurality of detail images and performing self-adaptive thresholding to obtain a denoising image, wherein the calculation formula is as follows:
MAD(Wi)＝median(abs(Wi-median(Wi)))
Wherein, For denoising images, abs () is an absolute value function and media () is a median function; MAD (W i) is a function of the median absolute deviation calculation for the detail image;
In the present embodiment, the detail image W i (i=1,., K) is adaptively thresholded by calculating the median absolute deviation (median absolute deviation, MAD) thereof to obtain a denoised detail image
S203: defining an approximation image I K of a K-th layer as an image background, wherein a first layer detail image W 1 is noise interference, and a calculation formula for denoising the first image is as follows:
In one embodiment of the present application, step S103 performs enhancement processing on a target area of the second image to obtain a third image, where I D is the second image, that is, an image obtained by performing wavelet multi-scale summation to obtain a noise interference removed image, referring to fig. 2, and the method may include the following steps:
the signal points in the target area of the second image are equivalent to Gaussian spots, and the calculation formula is as follows:
Wherein X is a plane coordinate vector, X= [ X, y ] T,p0 [ E [0,1] represents the intensity of the spot, b 0 [ E [0,1] represents the intensity of the background, and omega 0 represents the scale of the spot;
S301: convolving the Gaussian spots with the preset radius through a Gaussian Laplace operator, wherein the calculation formula of the Gaussian Laplace operator is as follows:
Wherein, sigma is the standard deviation of Gaussian, In this embodiment, the multi-scale laplacian of gaussian (LAPLACIAN OF GAUSSIAN, loG) is used to enhance the signal point region and reduce image background interference. The two-dimensional gaussian kernel definition formula is as follows:
Wherein x= [ x, y ] T represents a plane coordinate, σ is a gaussian standard deviation, and the normalized laplacian of gaussian formula is as follows:
After the specific radius spot of the image is convolved with the specific scale LoG, a stronger negative response is generated, and the original formula is scaled to obtain the following formula:
S302: selecting the scale of the Gaussian Laplace operator as In this embodiment, in order to obtain the corresponding scale, mathematical modeling needs to be performed on the signal points, and the general signal points may be equivalently gaussian spots, as shown in the following formula:
Where x 0 represents the center, p 0 ε [0,1] represents the intensity of the blob, b 0 ε [0,1] represents the intensity of the background, and ω 0 represents the scale of the blob. For simplicity of calculation, let x 0＝[0,0]T be the above formula becomes:
Let the pixel radius of the gaussian spot be r 0, concentrate within 3 standard deviations of its mean according to 99% gaussian mass, give: r 0≈3ω0
With the mathematical modeling and LoG kernel of the speckle, we can get:
Let x=0, then the response of the spot center coordinates is:
in order to obtain the standard deviation of the LOG kernel and obtain the maximum value of the center coordinate of the Gaussian spot with the standard deviation in the final response plane and meet the target of the enhanced spot, the sigma value needs to be obtained to obtain the maximum response, and the above formula is obtained by deriving:
Let the above equation be 0, when σ=ω 0, i.e. when the LoG scale σ *＝ω0, the gaussian spot center coordinate response is maximum, the response is:
L0(0;σ*)＝p0
thus, the scale selection of LoG may be:
S303: determining a scale set S of the Gaussian spots according to the scale, and performing multi-scale filtering on a target area of the second image according to the scale set S to obtain a third image L (x):
Wherein,
In this embodiment, for a FISH fluorescence image of 500w pixels, the corresponding scale set s=r/3 is obtained by counting the spot pixel radius r e {6,7,8,9,10,11,12 }. After the scale set is determined, the image is subjected to multi-scale Log filtering, and the obtained enhanced image is:
Referring to fig. 2, in one embodiment of the present application, step S104 suppresses the non-target area of the third image by using 4-directional single-sided second-order gaussian matching, that is, four directions, i.e., vertical up, down, left, right, to obtain a fourth image; in this embodiment, the LoG operator also generates a strong response to the image edge, and the application removes the edge misextraction region and suppresses the non-speckle region by single-sided second-order gaussian kernel filtering in 4 directions. The 4 directions of a remarkable spot have positive larger local contrast, the minimum local contrast of the 4 directions is taken, the minimum local contrast of the remarkable spot is larger, but for some non-spot structures, the wavelet+LOG enhanced spot structure can generate strong response on the edge of the large bright spot, so that the extraction of the edge area is misextracted, one remarkable characteristic of the edge is that the edge has larger positive contrast in one direction, the contrast in the other direction is very small and almost 0, and the suppression of the strong edge interference can be completed by taking the minimum response value of the 4 directions.
The second order gaussian kernel formula with directionality is as follows:
Wherein θ is the rotation angle, σ is the gaussian standard deviation, and the R θ rotation matrix is calculated by the following formula:
The second order gaussian kernel is spatially divided into 3 parts, namely a central part k C and two symmetrical lateral parts k R and k L, each of which can be represented as a separate kernel, so that there is:
The second order gaussian kernel measures the local contrast between the central region and its two side regions, so a conventional second order gaussian kernel function typically results in a loss of direction information, since the central portion is always tied to the two side regions. In order to fully utilize the direction information of the local contrast, a single-sided second order Gaussian (USG) kernel is designed as follows:
u(x；σ,θ)＝kC(x;σ,θ)+2·kR(x；σ,θ)
This allows the speckle to be distinguished from the rest of the structure, since the salient speckle has positive local contrast in all directions. The minimum local contrast of the speckle in all directions is still positive compared to the non-speckle structure. At a given scale, the USG kernel is used to measure the local contrast of the region in all directions, and then the minimum is selected as a response, achieving suppression of non-speckle structures. The final response of the USG kernel can be obtained by the following formula:
Wherein S u is consistent with the dimension S of the Log, the directions take 4 directions, namely D u = { i pi/2|i ∈ {0,1,2,3 }.
In one embodiment of the present application, step S104 and step S105 are specifically: and (3) making the fourth image LU (X) =L (X) ·U (X), then performing threshold binarization processing on the fourth image to obtain the center point coordinate of each signal point in the fourth image, mapping the center point coordinate to the third image, and selecting a signal point region image on the third image according to the mapped center point coordinate of the signal point.
The embodiment of the application also discloses a device for extracting the signal point region of the FISH microscopic image, which comprises the following steps of:
the image preprocessing module is used for preprocessing an original image to obtain a first image;
The image denoising module is used for carrying out multi-scale wavelet summation denoising on the first image to obtain a second image;
the image enhancement module is used for carrying out enhancement processing on the target area of the second image to obtain a third image suppression module, and the image enhancement module is used for suppressing the non-target area of the third image through single-side second-order Gaussian cores in 4 directions to obtain a fourth image;
and the output module is used for selecting a final signal point area according to the third image and the fourth image and outputting the final signal point area.
The device for extracting the signal point region of the FISH microscopic image can realize any one of the methods for extracting the signal point region of the FISH microscopic image, and the specific working process of each module in the device for extracting the signal point region of the FISH microscopic image can refer to the corresponding process in the embodiment of the method.
In several embodiments provided by the present application, it should be understood that the methods and systems provided may be implemented in other ways. For example, the system embodiments described above are merely illustrative; for example, a division of a module is merely a logical function division, and there may be another division manner in actual implementation, for example, multiple modules may be combined or may be integrated into another system, or some features may be omitted or not performed.
The embodiment of the application also discloses computer equipment.
Computer apparatus comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing a method xx as described above when executing the computer program.
The embodiment of the application also discloses a computer readable storage medium.
A computer readable storage medium storing a computer program capable of being loaded by a processor and executing any one of the methods xx described above.
Wherein a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device; program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Referring to fig. 3, in one embodiment of the present application, FISH microscopic image signal point region extraction methods are programmed by a c++ language platform and three sets of simulations are performed, where (a) represents a first image, i.e., an original gray scale image; (b) Representing a second image, namely a multi-scale wavelet sum denoising image; (c) Representing a third image, namely a multi-scale LoG filtered enhanced image; (d) Representing a fourth image, namely a single-sided second-order Gaussian kernel filtered image; (e) represents an l×u image after threshold binarization processing; (f) represents the final signal point area image. Through verification, the extraction rate of signal points in the embodiment of the application is more than or equal to 95%, the accuracy rate of extracting the signal points is more than or equal to 90%, the running time of a result algorithm is less than or equal to 2s for 500w pixel images, the extraction rate = the correct extraction number of the algorithm/the total number of gold standards, the accuracy rate = the correct extraction number of the algorithm/the total number of the algorithm extraction, the signal point gold standard of the FISH microscopic fluorescent image refers to the signal point position determined on the image by a professional doctor, the gold standard is the target of the final extraction of the signal points by the algorithm, the better the extracted signal points meet the gold standard representing algorithm, and the total number of the gold standards is the number of all correct signal points on one image.
In summary, according to the method, the device, the equipment and the medium for extracting the signal point region of the FISH microscopic image provided by the embodiment of the application, the signal point region in the original image is highlighted by acquiring the original image and performing graying and normalization processing on the original image, so that the speed of subsequent processing can be improved. For the noise problem of the image, the image is subjected to multi-scale wavelet summation denoising, so that the denoised image is obtained, the signal to noise ratio of the image is effectively improved, and the image quality and the identification degree of the target are improved. And carrying out multi-scale Gaussian Laplace operator filtering on the image after noise reduction, enhancing a signal point area and inhibiting background interference. And (3) for the problem of strong edge interference of the highlight region, carrying out single-side second-order Gaussian kernel filtering in 4 directions on the image after noise reduction, and inhibiting a non-speckle region. The signal point area is obtained by carrying out threshold processing on the plane which is subjected to wavelet denoising and LOG speckle enhancement, the extraction rate can reach more than 95%, but because of the interference of a non-speckle structure and a strong edge structure, more false extraction can be generated, so that the accuracy is lower, and therefore, a single-side second-order Gaussian kernel is adopted to inhibit the non-speckle and the edge area to improve the accuracy. And obtaining a product graph of the Gaussian Laplace plane and the second-order Gaussian kernel plane, binarizing at a threshold value, and analyzing a connected domain to obtain a final spot center coordinate. And obtaining a final signal point area on the Gaussian Laplace plane according to the final coordinates. Aiming at the FISH fluorescent image with poor imaging quality, obvious noise, weak signal point area and strong edge interference, the non-speckle structure is restrained while the signal point area is enhanced, so that the extraction effect and efficiency of the final signal point area are greatly improved.
In the foregoing embodiments, the descriptions of the embodiments are focused on, and for those portions of one embodiment that are not described in detail, reference may be made to the related descriptions of other embodiments.
The foregoing description of the preferred embodiments of the application is not intended to limit the scope of the application in any way, including the abstract and drawings, in which case any feature disclosed in this specification (including abstract and drawings) may be replaced by alternative features serving the same, equivalent purpose, unless expressly stated otherwise. That is, each feature is one example only of a generic series of equivalent or similar features, unless expressly stated otherwise.

Claims (6)

1. The method for extracting the signal point region of the FISH microscopic image is characterized by comprising the following steps of:
preprocessing an original image to obtain a first image;
Carrying out multi-scale wavelet summation denoising on the first image to obtain a second image;
performing enhancement processing on the target area of the second image to obtain a third image;
Suppressing the non-target area of the third image through single-side second-order Gaussian cores in 4 directions to obtain a fourth image;
Collecting first center point coordinates of a signal point area in the fourth image, and mapping the first center point coordinates to the third image to obtain second center point coordinates;
selecting a signal point area in the third image as a final signal point area according to the second center point coordinates;
Said multi-scale wavelet summation denoising of said first image comprises the steps of:
Performing multi-scale decomposition on the first image to obtain a plurality of detail images and a plurality of layers of approximate images, wherein the calculation formula is as follows:
Wi(x,y)＝Ii-1(x,y)-Ii(x,y)
Wherein W i is the ith detail image, I i is the ith approximate image, the value range of I is [1, k ], k is the number of layers of the approximate image, I 0 (x, y) is the first image, x, y is the coordinates of the detail image and pixel points in the approximate image;
Calculating the median absolute deviation of a plurality of detail images and performing self-adaptive thresholding to obtain a denoising image, wherein the calculation formula is as follows:
MAD(Wi)＝median(abs(Wi-median(Wi)))
Wherein, For denoising images, abs () is an absolute value function and media () is a median function; MAD (W i) is a function of the median absolute deviation calculation for the detail image;
defining an approximation image I K of a K-th layer as an image background, wherein a first layer detail image W 1 is noise interference, and a calculation formula for denoising the first image is as follows:
wherein I D is the second image;
the enhancing processing of the target area of the second image specifically includes the following steps:
the signal points in the target area of the second image are equivalent to Gaussian spots, and the calculation formula is as follows:
Wherein X is a plane coordinate vector, X= [ X, y ] T,p0 [ E [0,1] represents the intensity of the spot, b 0 [ E [0,1] represents the intensity of the background, and omega 0 represents the scale of the spot;
Convolving the Gaussian spots with the preset radius through a Gaussian Laplace operator, wherein the calculation formula of the Gaussian Laplace operator is as follows:
Wherein, sigma is the standard deviation of Gaussian, The scale of the Gaussian Laplace operator is selected as/>Determining a scale set S of the Gaussian spots according to the scale, and performing multi-scale filtering on a target area of the second image according to the scale set S to obtain a third image L (x):
Wherein, The calculation formula of the unilateral second-order Gaussian kernel is as follows:
u(X；σ,θ)＝kC(X;σ,θ)+2·kR(X；σ,θ)
Wherein:
θ is the rotation angle of the single-sided second-order gaussian kernel;
Obtaining single-side second-order Gaussian kernels in 4 directions by adjusting the rotation angle theta of the single-side second-order Gaussian kernels;
The suppressing of the non-target area of the third image by the 4-direction single-sided second-order gaussian kernel includes:
The direction D u = { αpi/2|α e {0,1,2,3 };
and calculating the final response U (X) of the single-side second-order Gaussian kernel, wherein the calculation formula is as follows:
wherein S u is a set of scales S.
2. The method according to claim 1, characterized in that: the preprocessing of the original image comprises graying and normalizing the original image to obtain the first image.
3. The method according to claim 1, characterized in that: the mapping the first center point coordinate to the third image specifically includes: let LU (X) =l (X) ·u (X), then perform threshold binarization processing on the fourth image, to obtain center point coordinates of each signal point in the fourth image, and map the center point coordinates to the third image.
4. A FISH microscopic image signal point region extraction apparatus, applying the FISH microscopic image signal point region extraction method according to any one of claims 1 to 3, characterized by comprising:
the image preprocessing module is used for preprocessing an original image to obtain a first image;
The image denoising module is used for carrying out multi-scale wavelet summation denoising on the first image to obtain a second image;
the image enhancement module is used for carrying out enhancement processing on the target area of the second image to obtain a third image suppression module, and the image enhancement module is used for suppressing the non-target area of the third image through single-side second-order Gaussian cores in 4 directions to obtain a fourth image;
and the output module is used for selecting a final signal point area according to the third image and the fourth image and outputting the final signal point area.
5. A computing device, characterized by: a computer program comprising a memory, a processor and stored on said memory and executable on said processor, said processor implementing the method according to any one of claims 1 to 3 when said program is executed.
6. A computer-readable storage medium, characterized by: a computer program being stored which can be loaded by a processor and which performs the method according to any one of claims 1 to 3.
CN202310918649.0A 2023-07-25 2023-07-25 FISH microscopic image signal point region extraction method, device, equipment and medium Active CN116934776B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310918649.0A CN116934776B (en) 2023-07-25 2023-07-25 FISH microscopic image signal point region extraction method, device, equipment and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310918649.0A CN116934776B (en) 2023-07-25 2023-07-25 FISH microscopic image signal point region extraction method, device, equipment and medium

Publications (2)

Publication Number Publication Date
CN116934776A CN116934776A (en) 2023-10-24
CN116934776B true CN116934776B (en) 2024-04-30

Family

ID=88382338

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310918649.0A Active CN116934776B (en) 2023-07-25 2023-07-25 FISH microscopic image signal point region extraction method, device, equipment and medium

Country Status (1)

Country Link
CN (1) CN116934776B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109477759A (en) * 2016-04-07 2019-03-15 堀场(法国)有限公司 The method for obtaining and forming spectroscopic assay image is sampled by adaptive space
CN110866924A (en) * 2019-09-24 2020-03-06 重庆邮电大学 Line structured light center line extraction method and storage medium
CN112215865A (en) * 2019-07-11 2021-01-12 上海交通大学 Automatic detection method for micro-droplets under fluorescence microscopic image

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10598914B2 (en) * 2015-07-14 2020-03-24 Massachusetts Institute Of Technology Enhancement of video-rate fluorescence imagery collected in the second near-infrared optical window

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109477759A (en) * 2016-04-07 2019-03-15 堀场(法国)有限公司 The method for obtaining and forming spectroscopic assay image is sampled by adaptive space
CN112215865A (en) * 2019-07-11 2021-01-12 上海交通大学 Automatic detection method for micro-droplets under fluorescence microscopic image
CN110866924A (en) * 2019-09-24 2020-03-06 重庆邮电大学 Line structured light center line extraction method and storage medium

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
Blob Reconstruction Using Unilateral Second Order Gaussian Kernels with Application to High-ISO Long-Exposure Image Denoising;Gang Wang et al.;《2017 IEEE International Conference on Computer Vision》;4827-4835 *
Noise Estimation for Image Sensor Based on Local Entropy and Median Absolute Deviation;Yongsong Li et al.;《sensors》;1-22 *
图像斑点及边缘检测方法研宄;李藕;《中国博士学位论文全文数据库》;1-124 *
活细胞中运动囊泡的识别和追踪算法及单分子荧光能量共振转移技术方法的研究;张翔;《中国博士学位论文全文数据库》;1-122 *

Also Published As

Publication number Publication date
CN116934776A (en) 2023-10-24

Similar Documents

Publication Publication Date Title
CN110176012B (en) Object segmentation method in image, pooling method, device and storage medium
CN109522908B (en) Image significance detection method based on region label fusion
US10402623B2 (en) Large scale cell image analysis method and system
US8345976B2 (en) Systems and methods for segmenting digital images
JP6660313B2 (en) Detection of nuclear edges using image analysis
CN109685060B (en) Image processing method and device
JP6710135B2 (en) Cell image automatic analysis method and system
US8300938B2 (en) Methods for segmenting objects in images
US9082164B2 (en) Automatic cell migration and proliferation analysis
Wang et al. Automated blob detection using iterative Laplacian of Gaussian filtering and unilateral second-order Gaussian kernels
CN109978848B (en) Method for detecting hard exudation in fundus image based on multi-light-source color constancy model
WO2006024974A1 (en) Feature weighted medical object contouring using distance coordinates
CN110956632B (en) Method and device for automatically detecting pectoralis major region in molybdenum target image
CN107038704B (en) Retina image exudation area segmentation method and device and computing equipment
US20230005140A1 (en) Automated detection of tumors based on image processing
Liu et al. Infrared ship target segmentation through integration of multiple feature maps
Khordehchi et al. Automatic lung nodule detection based on statistical region merging and support vector machines
CN114677525B (en) Edge detection method based on binary image processing
CN112215865A (en) Automatic detection method for micro-droplets under fluorescence microscopic image
CN110490159B (en) Method, device, equipment and storage medium for identifying cells in microscopic image
CN113850792A (en) Cell classification counting method and system based on computer vision
Pelapur et al. Multi-focus image fusion using epifluorescence microscopy for robust vascular segmentation
CN116934776B (en) FISH microscopic image signal point region extraction method, device, equipment and medium
CN116363097A (en) Defect detection method and system for photovoltaic panel
CN113780421B (en) Brain PET image identification method based on artificial intelligence

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant