CN113065403A - Hyperspectral imaging-based machine learning cell classification method and device - Google Patents
Hyperspectral imaging-based machine learning cell classification method and device Download PDFInfo
- Publication number
- CN113065403A CN113065403A CN202110246544.6A CN202110246544A CN113065403A CN 113065403 A CN113065403 A CN 113065403A CN 202110246544 A CN202110246544 A CN 202110246544A CN 113065403 A CN113065403 A CN 113065403A
- Authority
- CN
- China
- Prior art keywords
- cell
- image
- hyperspectral
- dimensional
- hyperspectral imaging
- 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.)
- Pending
Links
- 238000000701 chemical imaging Methods 0.000 title claims abstract description 53
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000010801 machine learning Methods 0.000 title claims abstract description 44
- 239000000126 substance Substances 0.000 claims abstract description 28
- 238000007781 pre-processing Methods 0.000 claims abstract description 18
- 238000012545 processing Methods 0.000 claims description 10
- 238000012706 support-vector machine Methods 0.000 claims description 9
- 238000001237 Raman spectrum Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 6
- 230000001427 coherent effect Effects 0.000 claims description 4
- 238000003384 imaging method Methods 0.000 claims description 4
- 238000002329 infrared spectrum Methods 0.000 claims description 4
- 230000001052 transient effect Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- 238000001157 Fourier transform infrared spectrum Methods 0.000 claims description 2
- 238000000862 absorption spectrum Methods 0.000 claims description 2
- 238000000295 emission spectrum Methods 0.000 claims description 2
- 210000004027 cell Anatomy 0.000 description 137
- 230000000877 morphologic effect Effects 0.000 description 12
- 238000012549 training Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 8
- 238000001069 Raman spectroscopy Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 230000010339 dilation Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000003628 erosive effect Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 210000002798 bone marrow cell Anatomy 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 238000004611 spectroscopical analysis Methods 0.000 description 2
- 238000005033 Fourier transform infrared spectroscopy Methods 0.000 description 1
- 238000004566 IR spectroscopy Methods 0.000 description 1
- 238000004497 NIR spectroscopy Methods 0.000 description 1
- 238000004847 absorption spectroscopy Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000001724 coherent Stokes Raman spectroscopy Methods 0.000 description 1
- 238000002082 coherent anti-Stokes Raman spectroscopy Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004993 emission spectroscopy Methods 0.000 description 1
- 238000011223 gene expression profiling Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000001774 stimulated Raman spectroscopy Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/698—Matching; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/693—Acquisition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/695—Preprocessing, e.g. image segmentation
Abstract
The invention discloses a hyperspectral imaging-based machine learning cell classification method and device, which comprise the following steps: performing hyperspectral imaging on cells to obtain three-dimensional cell hyperspectral images with the size of S multiplied by M multiplied by N, wherein S represents different wave numbers, the size of each wave number corresponds to a two-dimensional image with the size of M multiplied by N, the intensity of each pixel in the two-dimensional images represents the signal intensity under the corresponding wave number, and the signal intensity combination of each pixel under the different wave numbers can reflect the chemical characteristics of substances corresponding to the pixels; preprocessing a cell hyperspectral image and segmenting cells to obtain a cell image block; and classifying the cell image blocks by using a machine learning model to obtain a cell classification result. A cell hyperspectral image containing chemical information and two-dimensional image information is obtained through hyperspectral imaging, cell classification is carried out based on the cell hyperspectral image, and automatic high-speed and high-precision classification of living body single cells is achieved.
Description
Technical Field
The invention belongs to the field of cell analysis and medical image processing, and particularly relates to a hyperspectral imaging-based machine learning cell classification method and device.
Background
Cells can be classified into different types according to differences in genes and phenotypes. The method can accurately distinguish different types of cells, particularly classify the cells at the single cell level, and has important application in many fields such as life science research, medical diagnosis and the like.
Existing single cell classification techniques are typically based on gene expression profiling, such as various single cell sequencing techniques. However, many studies and medical diagnostics actually require not only a focus on cell types, but also a classification of the more dynamic concept of cell phenotype, and it becomes more important to capture the transient molecular state of cells. Therefore, various new technologies have been developed to measure the characteristics of single cells at the genomic, transcriptomic, proteomic or metabolome level, providing targeted molecular changes of the cells under interaction with the microenvironment. However, in such a scenario, finding a high-speed and high-precision living single cell classification technique is a problem to be solved urgently.
In the prior art, cell classification is mostly based on two-dimensional image information of cells, such as an automatic bone marrow cell classification method disclosed in the patent application with application publication No. CN108376400A, and a deep learning-based bone marrow cell classification method and classification device disclosed in the patent application with application publication No. CN 109360198A. The cell classification method based on two-dimensional image information is not accurate enough in classification accuracy.
Disclosure of Invention
In view of the above, an object of the present invention is to provide a method and an apparatus for machine learning cell classification based on hyperspectral imaging, in which a hyperspectral image of a cell containing chemical information and two-dimensional image information is obtained by hyperspectral imaging, and cell classification is performed based on the hyperspectral image of the cell, so as to realize automatic high-speed and high-precision classification of living single cells.
In order to achieve the purpose, the invention provides the following technical scheme:
in a first aspect, a hyperspectral imaging-based machine learning cell classification method includes the following steps:
performing hyperspectral imaging on cells to obtain three-dimensional cell hyperspectral images with the size of S multiplied by M multiplied by N, wherein S represents different wave numbers, the size of each wave number corresponds to a two-dimensional image with the size of M multiplied by N, the intensity of each pixel in the two-dimensional images represents the signal intensity under the corresponding wave number, and the signal intensity combination of each pixel under the different wave numbers can reflect the chemical characteristics of substances corresponding to the pixels;
preprocessing a cell hyperspectral image and segmenting cells to obtain a cell image block;
and classifying the cell image blocks by using a machine learning model to obtain a cell classification result.
Preferably, the hyperspectral imaging of the cell is performed using visible light spectroscopy, near infrared spectroscopy, raman spectroscopy, stimulated raman spectroscopy, coherent anti-stokes raman spectroscopy, coherent stokes raman spectroscopy, transient absorption spectroscopy, stimulated emission spectroscopy, infrared spectroscopy or fourier transform infrared spectroscopy.
Preferably, when hyperspectral imaging is performed on the cells, the following method is adopted:
obtaining a two-dimensional image through a single-point photoelectric detector and laser scanning or sample scanning; or the like, or, alternatively,
obtaining a two-dimensional image through a one-dimensional array detector and laser scanning or sample scanning; or, a two-dimensional image is obtained through exposure of the two-dimensional array detector and the imaging system;
and forming a three-dimensional cell hyperspectral image by the two-dimensional images under different wave numbers.
Preferably, the preprocessing of the cell hyperspectral image comprises region of interest extraction and noise filtering, binarization processing and image morphology processing.
Preferably, the cell segmentation is performed on the preprocessed image by using a watershed algorithm to obtain a cell image block.
Preferably, the machine learning model employs a support vector machine.
Preferably, before the cell image block classification is performed by using the machine learning model, the machine learning model needs to be subjected to parameter optimization.
In a second aspect, a hyperspectral imaging-based machine learning cell classification device includes:
the image acquisition module is used for performing hyperspectral imaging on the cells to obtain a three-dimensional cell hyperspectral image with the size of S multiplied by M multiplied by N, wherein S represents different wave numbers, each wave number corresponds to a two-dimensional image with the size of M multiplied by N, the intensity of each pixel in the two-dimensional image represents the signal intensity under the corresponding wave number, and the signal intensity combination of each pixel under the different wave numbers can reflect the chemical characteristics of substances corresponding to the pixel;
the image preprocessing module is used for preprocessing the hyperspectral image of the cell and segmenting the cell to obtain a cell image block;
and the cell classification module is used for classifying the cell image blocks by utilizing a machine learning model to obtain a cell classification result.
In a third aspect, a hyperspectral imaging-based machine-learned cell classification apparatus includes a computer memory, a computer processor, and a computer program stored in the computer memory and executable on the computer processor, wherein a parameter-optimized machine-learned model is stored in the computer memory, and the computer processor executes the computer program to implement the following steps:
preprocessing and cell segmenting the collected hyperspectral image of the cell with the size of S multiplied by M multiplied by N to obtain a cell image block;
and classifying the cell image blocks by using a machine learning model to obtain a cell classification result.
Compared with the prior art, the invention has the beneficial effects that at least:
the hyperspectral imaging-based machine learning cell classification method and device provided by the embodiment of the invention can be used for hyperspectral imaging of cells, so that the obtained hyperspectral image of the cells simultaneously contains two-dimensional morphological information and chemical components of the cells, and then, the cell classification is carried out by combining the chemical components and two-dimensional heart states of the cells, thereby improving the speed and accuracy of cell classification.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a flowchart of a hyperspectral imaging-based machine learning cell classification method according to an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the detailed description and specific examples, while indicating the scope of the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention.
In order to solve the problem of automatic high-speed, high-precision and living body single cell classification which is difficult to realize by using the existing cell sequencing technology, the embodiment of the invention provides a hyperspectral imaging-based machine learning cell classification method and a hyperspectral imaging-based machine learning cell classification device. The method is different from the method for classifying cells by utilizing the morphological characteristics of the traditional single-frame cells, simultaneously combines the chemical components provided by hyperspectral imaging and the multidimensional information of the distribution of the chemical components, realizes quick and high-accuracy single-cell classification by identifying the spatial information and the hyperspectral information in the three-dimensional image, and further obviously improves the accuracy and the efficiency of cell classification.
Fig. 1 is a flowchart of a hyperspectral imaging-based machine learning cell classification method according to an embodiment of the invention. As shown in fig. 1, the hyperspectral imaging-based machine learning cell classification method provided by the embodiment includes the following steps:
step 1, performing hyperspectral imaging on cells to obtain a three-dimensional hyperspectral cell image.
In the embodiment, the prepared cell sample is subjected to hyperspectral imaging to obtain a cell hyperspectral image containing high-dimensional spectral information. As a novel imaging technology with chemical bond selectivity, hyperspectral imaging provides a brand new idea for developing capture of cell phenotypes. Hyperspectral imaging utilizes molecular vibration spectroscopy to superimpose chemical information on traditional two-dimensional image information with another layer of dimensionality-spectral dimensionality-to form hyperspectral image data.
Visible light spectrum, near infrared spectrum, Raman spectrum, stimulated Raman spectrum, coherent anti-Stokes Raman spectrum, coherent Stokes Raman spectrum, transient absorption spectrum, stimulated emission spectrum, infrared spectrum and Fourier transform infrared spectrum can be adopted in hyperspectral imaging.
When hyperspectral imaging is performed on cells, the following method can be adopted: (a) obtaining a two-dimensional image through a single-point photoelectric detector and laser scanning or sample scanning; (b) obtaining a two-dimensional image through a one-dimensional array detector and laser scanning or sample scanning; (c) exposing through a two-dimensional array detector and an imaging system to obtain a two-dimensional image; and forming a three-dimensional cell hyperspectral image by the two-dimensional images under different wave numbers.
The term (a) is to be understood as: detection is performed by a single-point photodetector (e.g., a photodiode) having only one pixel, and in order to achieve two-dimensional detection, either the light source is scanned, i.e., the illumination beam moves in the XY plane to have an opportunity to cover the entire sample, or the sample is scanned, i.e., the sample to be observed moves in the XY plane.
In an embodiment, a stimulated Raman scattering hyperspectral imaging device is used to obtain a hyperspectral image of a cell to be measured, the device comprises a frequency omegapPump laser of and a frequency of omegasThe stokes laser of (2). When the frequency difference between the two lasers matches the vibration frequency of the Raman vibration band, i.e., ωp-ωs＝νvibThe intensity of the pump laser will decrease (stimulated raman loss) while the stokes laser increases (stimulated raman gain). By detecting the modulation of one of the narrow band lasers, while the other modulated laser is adjusted in the spectral range, or using a broadband laser, to produce a frequency range covering the region of interest, and detecting the Raman gain (or loss) from that laser by usingThe two narrow-band lasers scan at the interested frequency, or the full spectrum is recorded point by using a broadband laser method, so that the three-dimensional cell hyperspectral image can be obtained.
And obtaining a three-dimensional cell hyperspectral image with the size of S multiplied by M multiplied by N, wherein S represents different wave numbers, the size of the two-dimensional image corresponding to each wave number is M multiplied by N, the intensity of each pixel in the two-dimensional image represents the signal intensity under the corresponding wave number, the combination of the signal intensities of each pixel under different wave numbers can reflect the chemical characteristics of substances corresponding to the pixels, and the chemical characteristics are key factors for cell classification.
And 2, preprocessing the cell hyperspectral image and segmenting the cell to obtain a cell image block.
The three-dimensional cell hyperspectral image contains chemical information of cells, and may also contain complex information components which are not beneficial to cell classification information acquisition, such as system noise and the like. Therefore, how to deal with the increased dimension of chemical information to serve the task of single-cell classification is a major challenge of current technologies. In order to solve the problem, cell hyperspectral image preprocessing is required, the preprocessing comprises at least one of region of interest (ROI) extraction and noise filtering, binarization processing, image morphology processing and the like, and the preprocessing sequence can be adjusted.
Wherein, the average filter can be adopted to filter noise to improve the image quality. Specifically, the image is filtered using a filter (filter) of size m x n,
wherein x is 0,1,2, M-1, y is 0,1,2, N-1, a is (M-1)/2, and b is (N-1)/2.
The binarization process is to convert the grayscale map into a binary image, and in one embodiment, an adaptive threshold method must be used to obtain sufficient information due to the different grayscale intensities of different regions in the sample, rather than using a fixed threshold, the binarization method of Otsu is preferably selected.
Morphological image processing involves binary erosion, dilation, opening, closing, and reconstruction (bounding and reconstruction). The effect of erosion is to "shrink" and "thin" objects in the image, while dilation is used to "grow" and "thicken" objects in the image. Second, morphological opening (morphological opening) is a combined process of erosion and dilation, while morphological closing (morphological closing) continues erosion by using the concept of dilation. In other words, the function of the form opening is to remove, break, shrink connections or objects that do not contain structural elements. Instead, the function of morphological closure is to connect, fill, and establish connections and objects in the image. Morphological processing will continuously apply morphological open fields and morphological closed to achieve ideal images.
The single hyperspectral image obtained contains a plurality of cells which may be from different types and needs to be identified first. Therefore, after the cell hyperspectral image is preprocessed, the foreground and the background of the image need to be distinguished, namely, whether each pixel in the image belongs to the foreground (cell) or the background is distinguished, cell segmentation is realized, cell image blocks are obtained, and each cell image block is used as a sample. In the embodiment, a watershed algorithm (watershed algorithm) is used for cell segmentation, and the specific process is as follows:
(a) a local minimum in the image is found. Each minimum value is assigned a unique label; (b) the priority queues are scanned in sequence from a small value of the variable h to a large value. Then, an element is selected from the first non-empty queue. If all queues of the priority queue are empty, the algorithm terminates. (c) The selected element is removed from the queue and its tag is passed to all unmarked neighbors. (d) All neighbors marked in the previous step are placed in the priority queue and then go back to step (a).
And 3, classifying the cell image blocks by using a machine learning model to obtain a cell classification result.
After the treatment is completed, cell sorting is performed using signals in each cell. In the traditional method for analyzing the single cell type, components in an image are judged mainly through a phasor analysis or clustering method, and then the single cell type is judged according to the position characteristics of different substances through people. Methods for cell classification using machine learning methods include training and applying predictions of machine learning models.
In this embodiment, the machine learning model adopts a Support Vector Machine (SVM), and the support vector machine after parameter optimization is used to classify the cells of the cell image block.
When the parameters of the support vector machine are optimized, an already labeled data set is required to be given as a training set. (x)1,y1),...,(xn,yn),xi∈Rd and yiE (-1, +1), where xiIs a feature vector representation, yiAre known labels of the training components i. The aim of training the support vector machine is to find an optimal hyperplane, wxT+ b is 0, w is the weight vector, x is the input feature vector, and b is the offset. w and b satisfy the following inequality: wxi T+b≥+1if yi＝1，wxi T+b≤-1if yiThe training process is to find the most suitable w and b so that the hyperplane can distinguish the boundaries, maximizing the boundary 1/| w | | survival2。
In a non-linear problem, the kernel function can be used to add an extra dimension to the raw data, making it a linear problem in the resulting high dimensional space. Its definition is: k (x, y) ═ K<f(x),f(y)>. Where K is a kernel function, x, y are n-dimensional inputs, and f is used to map the inputs from an n-dimensional space to an m-dimensional space.<x，y>The dot product is represented. By means of the kernel function, a scalar product between two data points in the high-dimensional space can be calculated without explicitly calculating a mapping from the input space to the high-dimensional space. In one embodiment, RBF is used as the kernel function, defined as follows: kRBF(x,y)＝exp(-γ||x-y||2)。
During training, a pre-labeled training sample is input into an algorithm for learning, and the training sample can be from a sample obtained in a previous experiment or from part of previous data obtained in a single experiment process, for example, the hyperspectral image obtained in step 1, and part but not all cells in the pre-labeled image. The labeling operation refers to the region and cell type given the segmentation of the cells and the background.
In embodiments, a support vector machine is selected, not requiring too many labeled samples. The typical neural network requires a large amount of sample training to obtain reasonable results, and the support vector machine is free from the problem.
During prediction, cell image blocks obtained by segmentation are input into a machine learning model with optimized parameters, and cell classification results are obtained through calculation.
The hyperspectral imaging-based machine learning cell classification method provided by the embodiment performs hyperspectral imaging on cells, enables the obtained hyperspectral images of the cells to simultaneously contain two-dimensional morphological information and chemical components of the cells, then performs cell classification by combining the chemical components and two-dimensional heart states of the cells, and improves the speed and accuracy of cell classification.
An embodiment also provides a machine learning cell classification device based on hyperspectral imaging, including:
the image acquisition module is used for performing hyperspectral imaging on the cells to obtain a three-dimensional cell hyperspectral image with the size of S multiplied by M multiplied by N, wherein S represents different wave numbers, each wave number corresponds to a two-dimensional image with the size of M multiplied by N, the intensity of each pixel in the two-dimensional image represents the signal intensity under the corresponding wave number, and the signal intensity combination of each pixel under the different wave numbers can reflect the chemical characteristics of substances corresponding to the pixel;
the image preprocessing module is used for preprocessing the hyperspectral image of the cell and segmenting the cell to obtain a cell image block;
and the cell classification module is used for classifying the cell image blocks by utilizing a machine learning model to obtain a cell classification result.
It should be noted that, when the hyperspectral imaging-based machine learning cell classification device provided in the above embodiment performs cell classification, the above division of each functional module is taken as an example, and the above function distribution may be completed by different functional modules according to needs, that is, the internal structure of the terminal or the server is divided into different functional modules to complete all or part of the above described functions. In addition, the hyperspectral imaging-based machine learning cell classification device and the hyperspectral imaging-based machine learning cell classification method provided by the embodiment belong to the same concept, and specific implementation processes thereof are detailed in the hyperspectral imaging-based machine learning cell classification method embodiment, and are not repeated here.
Embodiments also provide a hyperspectral imaging based machine-learned cell classification apparatus comprising a computer memory, a computer processor, and a computer program stored in the computer memory and executable on the computer processor, wherein the computer memory has a parameter-optimized machine learning model stored therein, and the computer processor executes the computer program to perform the steps of:
preprocessing and cell segmenting the collected hyperspectral image of the cell with the size of S multiplied by M multiplied by N to obtain a cell image block;
and classifying the cell image blocks by using a machine learning model to obtain a cell classification result.
In practical applications, the computer memory may be volatile memory at the near end, such as RAM, or may be non-volatile memory, such as ROM, FLASH, floppy disk, mechanical hard disk, etc., or may be a remote storage cloud. The computer processor can be a Central Processing Unit (CPU), a microprocessor unit (MPU), a Digital Signal Processor (DSP) or a Field Programmable Gate Array (FPGA), namely steps of the hyperspectral imaging-based machine learning cell classification method can be realized through the processors.
The above-mentioned embodiments are intended to illustrate the technical solutions and advantages of the present invention, and it should be understood that the above-mentioned embodiments are only the most preferred embodiments of the present invention, and are not intended to limit the present invention, and any modifications, additions, equivalents, etc. made within the scope of the principles of the present invention should be included in the scope of the present invention.
Claims (9)
1. A hyperspectral imaging-based machine learning cell classification method is characterized by comprising the following steps:
performing hyperspectral imaging on cells to obtain three-dimensional cell hyperspectral images with the size of S multiplied by M multiplied by N, wherein S represents different wave numbers, the size of each wave number corresponds to a two-dimensional image with the size of M multiplied by N, the intensity of each pixel in the two-dimensional images represents the signal intensity under the corresponding wave number, and the signal intensity combination of each pixel under the different wave numbers can reflect the chemical characteristics of substances corresponding to the pixels;
preprocessing a cell hyperspectral image and segmenting cells to obtain a cell image block;
and classifying the cell image blocks by using a machine learning model to obtain a cell classification result.
2. The hyperspectral imaging-based machine-learned cell classification method according to claim 1, wherein hyperspectral imaging is performed on the cell by using at least one of visible light spectrum, near infrared spectrum, raman spectrum, stimulated raman spectrum, coherent anti-stokes raman spectrum, coherent stokes raman spectrum, transient absorption spectrum, stimulated emission spectrum, infrared spectrum, and fourier transform infrared spectrum.
3. The hyperspectral imaging-based machine learning cell classification method according to claim 1 or 2, wherein the following method is adopted when hyperspectral imaging is performed on the cell:
obtaining a two-dimensional image through a single-point photoelectric detector and laser scanning or sample scanning; or the like, or, alternatively,
obtaining a two-dimensional image through a one-dimensional array detector and laser scanning or sample scanning; or the like, or, alternatively,
exposing through a two-dimensional array detector and an imaging system to obtain a two-dimensional image;
and forming a three-dimensional cell hyperspectral image by the two-dimensional images under different wave numbers.
4. The hyperspectral imaging-based machine learning cell classification method of claim 1, wherein the preprocessing of the hyperspectral image of the cell comprises region of interest extraction and noise filtering, binarization processing and image morphology processing.
5. The hyperspectral imaging-based machine-learned cell classification method of claim 1, wherein a watershed algorithm is used to perform cell segmentation on the preprocessed image to obtain a cell image block.
6. The hyperspectral imaging-based machine learning cell classification method of claim 1, wherein the machine learning model employs a support vector machine.
7. The hyperspectral imaging-based machine-learned cell classification method according to claim 1 or 6, wherein before cell image block classification using the machine-learned model, parameter optimization needs to be performed on the machine-learned model.
8. The utility model provides a machine learning cell sorter based on hyperspectral imaging which characterized in that includes:
the image acquisition module is used for performing hyperspectral imaging on the cells to obtain a three-dimensional cell hyperspectral image with the size of S multiplied by M multiplied by N, wherein S represents different wave numbers, each wave number corresponds to a two-dimensional image with the size of M multiplied by N, the intensity of each pixel in the two-dimensional image represents the signal intensity under the corresponding wave number, and the signal intensity combination of each pixel under the different wave numbers can reflect the chemical characteristics of substances corresponding to the pixel;
the image preprocessing module is used for preprocessing the hyperspectral image of the cell and segmenting the cell to obtain a cell image block;
and the cell classification module is used for classifying the cell image blocks by utilizing a machine learning model to obtain a cell classification result.
9. A hyperspectral imaging based machine-learned cell classification apparatus comprising a computer memory, a computer processor and a computer program stored in the computer memory and executable on the computer processor, wherein a parametrically optimized machine-learned model is stored in the computer memory and the computer processor executes the computer program to perform the steps of:
preprocessing and cell segmenting the collected hyperspectral image of the cell with the size of S multiplied by M multiplied by N to obtain a cell image block;
and classifying the cell image blocks by using a machine learning model to obtain a cell classification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110246544.6A CN113065403A (en) | 2021-03-05 | 2021-03-05 | Hyperspectral imaging-based machine learning cell classification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110246544.6A CN113065403A (en) | 2021-03-05 | 2021-03-05 | Hyperspectral imaging-based machine learning cell classification method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113065403A true CN113065403A (en) | 2021-07-02 |
Family
ID=76559793
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110246544.6A Pending CN113065403A (en) | 2021-03-05 | 2021-03-05 | Hyperspectral imaging-based machine learning cell classification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113065403A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114018789A (en) * | 2021-10-08 | 2022-02-08 | 武汉大学 | Acute leukemia typing method based on imaging flow cytometry detection and machine learning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102385751A (en) * | 2011-07-19 | 2012-03-21 | 中国科学院自动化研究所 | Liver tumor region segmentation method based on watershed transform and classification through support vector machine |
| CN103604737A (en) * | 2013-12-10 | 2014-02-26 | 山东大学 | Automatic blood cell recognition device and operating method thereof |
| CN107064019A (en) * | 2017-05-18 | 2017-08-18 | 西安交通大学 | The device and method for gathering and splitting for dye-free pathological section high spectrum image |
| CN108460400A (en) * | 2018-01-02 | 2018-08-28 | 南京师范大学 | A kind of hyperspectral image classification method of combination various features information |
| CN110852227A (en) * | 2019-11-04 | 2020-02-28 | 中国科学院遥感与数字地球研究所 | Hyperspectral image deep learning classification method, device, equipment and storage medium |
| CN111259913A (en) * | 2020-01-14 | 2020-06-09 | 哈尔滨工业大学 | Cell spectral image classification method based on bag-of-word model and textural features |
-
2021
- 2021-03-05 CN CN202110246544.6A patent/CN113065403A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102385751A (en) * | 2011-07-19 | 2012-03-21 | 中国科学院自动化研究所 | Liver tumor region segmentation method based on watershed transform and classification through support vector machine |
| CN103604737A (en) * | 2013-12-10 | 2014-02-26 | 山东大学 | Automatic blood cell recognition device and operating method thereof |
| CN107064019A (en) * | 2017-05-18 | 2017-08-18 | 西安交通大学 | The device and method for gathering and splitting for dye-free pathological section high spectrum image |
| CN108460400A (en) * | 2018-01-02 | 2018-08-28 | 南京师范大学 | A kind of hyperspectral image classification method of combination various features information |
| CN110852227A (en) * | 2019-11-04 | 2020-02-28 | 中国科学院遥感与数字地球研究所 | Hyperspectral image deep learning classification method, device, equipment and storage medium |
| CN111259913A (en) * | 2020-01-14 | 2020-06-09 | 哈尔滨工业大学 | Cell spectral image classification method based on bag-of-word model and textural features |
Non-Patent Citations (2)
| Title |
|---|
| 贾高杰: "基于显微高光谱成像的血液细胞识别研究与应用", 《中国优秀硕士学位论文全文数据库》 * |
| 郑欣: "基于显微高光谱成像的乳腺肿瘤组织微阵列识别分析方法研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114018789A (en) * | 2021-10-08 | 2022-02-08 | 武汉大学 | Acute leukemia typing method based on imaging flow cytometry detection and machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wagner et al. | SPHIRE-crYOLO is a fast and accurate fully automated particle picker for cryo-EM | |
| Meng et al. | Large-scale multi-class image-based cell classification with deep learning | |
| Salvi et al. | The impact of pre-and post-image processing techniques on deep learning frameworks: A comprehensive review for digital pathology image analysis | |
| US10671833B2 (en) | Analyzing digital holographic microscopy data for hematology applications | |
| Li et al. | Apple quality identification and classification by image processing based on convolutional neural networks | |
| Wang et al. | A spectral and morphologic method for white blood cell classification | |
| CN111915596A (en) | Method and device for predicting benign and malignant pulmonary nodules | |
| Moradi et al. | New features for automatic classification of human chromosomes: A feasibility study | |
| CN110245657B (en) | Pathological image similarity detection method and detection device | |
| Mishra et al. | Histopathological diagnosis for viable and non-viable tumor prediction for osteosarcoma using convolutional neural network | |
| Safdar et al. | Intelligent microscopic approach for identification and recognition of citrus deformities | |
| Ngugi et al. | A new approach to learning and recognizing leaf diseases from individual lesions using convolutional neural networks | |
| CN109815974A (en) | A kind of cell pathology slide classification method, system, equipment, storage medium | |
| Nissen et al. | Convolutional neural networks for segmentation and object detection of human semen | |
| Vidyarthi et al. | Classification of breast microscopic imaging using hybrid CLAHE-CNN deep architecture | |
| CN113065403A (en) | Hyperspectral imaging-based machine learning cell classification method and device | |
| Admass et al. | Convolutional neural networks and histogram-oriented gradients: a hybrid approach for automatic mango disease detection and classification | |
| Khan et al. | GLNET: global–local CNN's-based informed model for detection of breast cancer categories from histopathological slides | |
| Zahir et al. | Applying the deep learning model on an IoT board for breast cancer detection based on histopathological images | |
| Pati et al. | Deep positive-unlabeled learning for region of interest localization in breast tissue images | |
| Sema et al. | Automatic Detection and Classification of Mango Disease Using Convolutional Neural Network and Histogram Oriented Gradients | |
| CN112669288B (en) | Cell target expression prediction method, system and device based on digital pathological image | |
| Fang et al. | Recent progress and applications of Raman spectrum denoising algorithms in chemical and biological analyses: A review | |
| Kanojia et al. | Image processing techniques for breast cancer detection: A review | |
| JP6329651B1 (en) | Image processing apparatus and image processing method |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210702 |
|
| RJ01 | Rejection of invention patent application after publication |