A tumor cell detection method based on spectroscopy

Abstract

The present invention proposes a tumor cell detection method based on spectral spectroscopy, which relates to the technical field of cell detection. Monochromatic lights of different wavelengths are transmitted sequentially through a cell sample to be detected, and a spectral spectrum of the cell sample to be detected is collected; a set of absorption peak characteristic values ​​in each monochromatic spectrum curve in the spectral spectrum is obtained; multiple absorption peak characteristic values ​​in the absorption peak characteristic value set in each monochromatic spectrum curve are input into a convolutional neural network, and peak fusion features of the monochromatic spectrum curve are output; the fusion feature values ​​of the multiple monochromatic spectrum curves are combined into a multi-spectral fusion feature value sequence, the multi-spectral fusion feature value sequence is compared with a standard multi-spectral feature value sequence of known tumor cells of different categories, and the category of the tumor cells contained in the cell sample to be detected is judged according to the comparison result, thereby improving the accuracy of tumor detection.

Description

Technical Field

[0001] The invention proposes a tumor cell detection method based on spectral spectroscopy, which is used for non-disease diagnosis and treatment and relates to the technical field of cell detection. Background Art

[0002] Distant metastasis of tumors can be roughly divided into four stages: tumor cells break through the vascular barrier and enter the blood circulation; survive in the blood circulation system; break through the vascular barrier and enter specific tissue sites; proliferate and eventually form metastatic lesions. Among them, tumor cells that survive in the blood circulation system after breaking through the vascular barrier are called circulating tumor cells, which is a heterogeneous group. They may differ from each other in terms of cell size and morphology, molecular phenotype, activity level, metastatic potential, proliferation potential, etc. Circulating tumor cells may exist in the form of single cells, or they may aggregate with each other or with blood-derived cells to form tiny multicellular aggregates (i.e., circulating tumor microemboli). They are the "seeds" for tumor metastasis and an important window for understanding the malignant behavior of tumors. The effective detection of tumor cells is currently a hot topic in oncology research.

[0003] Tumor cell detection technology is an important means in tumor diagnosis and treatment. It can help doctors find tumor cells earlier and thus provide timely treatment. It is particularly suitable for early detection of tumors, metastasis monitoring, efficacy evaluation and prognosis judgment. The currently developed tumor cell detection technology is generally divided into four steps: (1) capture, (2) enrichment, (3) detection, and (4) release. The capture step is the specific interaction between tumor cells and materials (such as physical interaction and antibody / antigen interaction), such as magnetic beads and microfluidic chips; the enrichment step refers to the separation of tumor cells from the blood; after enrichment, tumor cells can be detected by fluorescence (such as fluorescence microscopy, fluorescence spectrophotometer and flow cytometer), surface enhanced Raman scattering (SERS) or electrical impedance; after tumor cells are released, further phenotypic identification and molecular analysis (such as RNA analysis and cell metabolism analysis) can be performed. However, the above tumor cell detection methods require formal labeling of cells during the detection process, which will cause certain damage to the cells.

[0004] Spectrophotometry using a spectrometer has begun to be applied to cell detection technology. This technology is mainly based on the selective absorption of light of a specific wavelength by a substance, and qualitatively and quantitatively analyzes cell samples by measuring absorbance. In tumor cell detection, spectrophotometry can be used to determine tumor cell density, tumor cell type, and monitor tumor cell growth status.

[0005] However, the existing technology of tumor cell detection using a spectrometer for spectroscopic spectrum is susceptible to the influence of stray light and light scattering, resulting in large detection errors. At the same time, because the content of tumor cells in the object to be detected, such as a blood sample, is extremely small, it is difficult to accurately determine whether the blood sample to be detected contains tumor cells based on the spectroscopic spectrum after obtaining the spectroscopic spectrum of the blood sample. Summary of the invention

[0006] In order to solve the above technical problems, the present invention proposes a tumor cell detection method based on spectroscopy for non-disease diagnosis and treatment, comprising the following steps:

[0007] S1. Transmitting monochromatic lights of different wavelengths to the cell samples to be detected in sequence, and collecting the spectroscopic spectrum of the cell samples to be detected;

[0008] S2, obtaining a set of absorption peak characteristic values ​​in each monochromatic spectrum curve in the spectral spectrum diagram;

[0009] S3, inputting a plurality of absorption peak characteristic values ​​in the absorption peak characteristic value set in each monochromatic spectrum curve into a convolutional neural network, and outputting a peak fusion feature of the monochromatic spectrum curve;

[0010] S4. Combining the fused feature values ​​of multiple monochromatic spectral curves into a multi-spectral fusion feature value sequence, comparing the multi-spectral fusion feature value sequence with known standard multi-spectral feature value sequences of different types of tumor cells, and determining the type of tumor cells contained in the cell sample to be detected according to the comparison results.

[0011] Furthermore, in the step S2, each monochromatic spectrum curve in the spectral spectrum diagram is obtained, and each monochromatic spectrum curve is baseline corrected and intensity normalized; and a set of absorption peak characteristic values ​​of each monochromatic spectrum curve after intensity normalization is obtained.

[0012] Furthermore, in step S2, the absorption intensity curves of all cell types in the cell sample to be detected in each monochromatic spectrum curve are obtained, the maximum peak value in the absorption intensity curves of n cell types is saved, and the maximum peak value of the dth cell type at the current input wavelength is set as L d , the maximum peak value L of the d-th cell type at the current input wavelength d The corresponding frequency value of monochromatic light is W d , then the absorption peak characteristic value Z of the d-th cell type d The calculation formula is:

[0013] ;

[0014] Among them, arctan is the inverse tangent function, d = 1,…,n;

[0015] According to the above formula, a set of absorption peak characteristic values ​​of each monochromatic spectrum curve is constructed.

[0016] Furthermore, in step S3, the n absorption peak characteristic values ​​Z of the I-th monochromatic spectrum curve are I1 , Z I2 , …, Z Id , …, Z In Input into the convolutional neural network, the convolutional neural network groups the input n absorption peak eigenvalues ​​into K groups, and calculates the fusion eigenvalue Z of the I-th monochromatic spectrum curve I ,

[0017] .

[0018] in, is the average absorption peak characteristic value of the qth group, W Iq is the weight of the qth group.

[0019] Furthermore, in step S4, standard multi-spectral feature value sequences of different types of tumor cells are constructed, and the standard multi-spectral feature value sequence of the Q-th type of tumor cells is set to [Y Q1 , Y Q2 , …, Y QT , …, Y QM ], where M is the total number of monochromatic spectral curves;

[0020] The fusion feature values ​​of the multiple monochromatic spectrum curves obtained in step S3 are combined into a multi-spectral fusion feature value sequence [Z1, Z2, ..., Z I , …, Z M ], the multispectral fusion feature value sequence is compared with the standard multispectral feature value sequences of different categories of tumor cells respectively, and the category numbers of tumor cells that meet the standard multispectral feature value sequence are obtained. If the category numbers of tumor cells that meet the standard multispectral feature value sequence are not obtained, the cell sample to be detected does not contain tumor cells.

[0021] Furthermore, by calculating the difference value, a multi-spectral fusion feature value sequence [Z1, Z2, ..., Z I , …, Z M ] and the standard multi-spectral feature value sequence of the Q-th category tumor cells [Y Q1 , Y Q2 , …, Y QI , …, Y QM ], where Y QI is the characteristic value of the Ith spectrum of the Qth category of tumor cells, Z I is the fusion eigenvalue of the Ith monochromatic spectrum curve in the multi-spectral fusion eigenvalue sequence;

[0022] The difference value D between the cell sample to be tested and the Qth category of tumor cells Q The formula is as follows:

[0023] .

[0024] Output the difference value D of the comparison Q Less than the threshold D TH Tumor cell type number.

[0025] Furthermore, the intensity of each monochromatic spectral curve is normalized by dividing the absorption intensity value of the maximum peak point of the monochromatic spectral curve corresponding to each monochromatic light source by the absorption intensity value of the maximum peak point in the entire spectral spectrum, so that the absorption intensity value of the peak point of the monochromatic spectral curve is normalized to the range of 0 to 1.

[0026] Compared with the prior art, the present invention has the following beneficial technical effects:

[0027] (1) Transmit monochromatic light of different wavelengths through the cell sample to be tested in turn, collect the spectral spectrum of the cell sample to be tested; obtain the set of absorption peak characteristic values ​​in each monochromatic spectrum curve in the spectral spectrum. By using the spectral spectrum, a detailed spectral analysis of the cell sample can be performed, which is conducive to revealing the composition and structural information inside the cell, and is helpful for subsequent analysis and identification.

[0028] (2) Multiple absorption peak feature values ​​in the absorption peak feature value set in each monochromatic spectral curve are input into a convolutional neural network, and the peak fusion features of the monochromatic spectral curve are output; by using the convolutional neural network to process the absorption peak feature value set, it is possible to effectively learn and extract peak fusion features that are helpful for classification from a large amount of spectral curve data. By calculating the peak fusion features of each monochromatic spectral curve, it is beneficial to combine the peak fusion features of different monochromatic spectral curves, thereby enhancing the subsequent ability to distinguish the types of tumor cells.

[0029] (3) The fused feature values ​​of multiple monochromatic spectral curves are combined into a multi-spectral fusion feature value sequence, and the multi-spectral fusion feature value sequence is compared with the known standard multi-spectral feature value sequences of different types of tumor cells. The type of tumor cells contained in the cell sample to be detected is determined according to the comparison result, so that the type of tumor cells in the cell sample can be accurately determined, thereby improving the accuracy of tumor detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0031] Figure 1 It is a schematic diagram of the process of the tumor cell detection method based on spectroscopy of the present invention;

[0032] Figure 2 is the absorption intensity curve of two cell types under monochromatic light irradiation;

[0033] Figure 3 It is a schematic diagram of eigenvalue clustering;

[0034] Figure 4 Schematic diagram of illuminating a cell sample with two different light sources;

[0035] Figure 5 Schematic diagram of the ultraviolet absorption spectra of various cell samples. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0037] In the drawings of the specific embodiments of the present invention, in order to better and more clearly describe the working principles of the various components in the system, the connection relationship of the various parts in the device is shown, and only the relative wavelength position relationship between the various components is clearly distinguished, and it cannot constitute a limitation on the signal transmission direction, connection sequence and structural size, size and shape of each part within the component or structure.

[0038] The purpose of the tumor cell detection method of the present invention is to detect the type of tumor cells contained in the cell sample in the blood to be detected.

[0039] The cell sample to be detected may be a tumor cell (CTC) in the blood, or may be other cell samples such as an endothelial cell (CEC) or an endothelial progenitor cell (CEP).

[0040] It should be noted that, when directly testing cells in blood, the sample is blood, and the cell sample to be tested may be a state of being taken out from blood, or may be a cell tissue sample.

[0041] Specifically, tumor cells were placed in RPMI1640 culture medium (containing 10% fetal bovine serum, penicillin and streptomycin, 100 U·mL each) -1 ) and cultured in an incubator at 37°C, saturated humidity, and 5% CO2.

[0042] Tumor cell detection based on spectroscopic spectrum is a technology that uses the ability of different substances to absorb or emit light at specific wavelengths to identify tumor cells. Spectroscopic spectrum technology can be divided into many types, including but not limited to ultraviolet / visible light (UV / Vis) spectrum, infrared (IR) spectrum, Raman spectrum, fluorescence spectrum, etc. As a device for collecting spectroscopic spectrum graphs, the spectrophotometer uses a light source that can generate multiple wavelengths, and a series of spectroscopic devices to generate a light source of a specific wavelength. After the light passes through the cell sample to be detected, part of the light is absorbed, and the absorbance value of the cell sample to be detected is calculated.

[0043] The present invention detects tumor cells based on spectral analysis. The flow chart of the detection method is as follows: Figure 1 As shown, the following steps are included:

[0044] S1. Transmit monochromatic lights of different wavelengths sequentially through the cell samples to be detected, and collect the spectral spectrum of the cell samples to be detected.

[0045] Place the cell sample to be tested in the measuring pool of the spectrophotometer, ensure that the cell sample to be tested is evenly distributed and free of bubbles, set the parameters of the spectrophotometer, including wavelength range, scanning speed, etc., transmit monochromatic light of different wavelengths to the cell sample to be tested in turn, measure the absorption degree of monochromatic light of each wavelength, and the spectrophotometer will automatically record the absorption intensity of each wavelength of light, draw a spectral spectrum with wavelength as the horizontal axis and light absorption intensity as the vertical axis, and save the spectral spectrum for subsequent analysis and processing.

[0046] S2. Obtain a set of absorption peak characteristic values ​​in each monochromatic spectrum curve in the spectral spectrum diagram.

[0047] First, the spectroscopic spectrum is imported into the curve processing software to obtain the monochromatic spectrum curves in the spectroscopic spectrum. If there is a baseline shift or drift in each monochromatic spectrum curve, a baseline correction needs to be performed first.

[0048] Here, this can be achieved by selecting data points in the non-peak region and fitting a polynomial, and then subtracting this baseline from the original spectrum.

[0049] Due to the different energy control ratios of light sources with multiple wavelengths, the light intensity ratios of each monochromatic spectrum curve are different. For example, the energy control ratio of the monochromatic light of the corresponding wavelength may be 60% for the monochromatic spectrum curve with a wavelength of 150nm, and the energy control ratio of the monochromatic light of the corresponding wavelength may be 20% for the curve with a wavelength of 320nm, etc. Therefore, it is necessary to normalize the absorption intensity of each monochromatic spectrum curve so that the absorption intensity value of the peak point of the monochromatic spectrum curve is normalized to the range of 0 to 1.

[0050] Specifically, this can be achieved by dividing the absorption intensity value of the maximum peak point of the monochromatic spectrum curve corresponding to each monochromatic light source by the absorption intensity value of the maximum peak point of the entire spectral spectrum.

[0051] After processing according to the above formula, the maximum intensity values ​​of all monochromatic spectral curves will be normalized to the range of 0 to 1, thereby ensuring the comparability between different monochromatic spectral curves.

[0052] Secondly, obtain the absorption peak characteristic value set of each monochromatic spectrum curve after intensity normalization.

[0053] In the monochromatic spectral curve analysis, since there are multiple cell types in the cell sample to be tested, such as red blood cells, white blood cells, tumor cells, etc., each cell type has its own specific chemical composition and molecular structure. These characteristics will appear as different peaks on the monochromatic spectral curve. Each peak represents the characteristic absorption of a specific molecule or chemical component in the cell sample, which helps to identify and distinguish different cell types.

[0054] Detect the main absorption peaks in the monochromatic spectrum curve and record their intensities. These intensity values ​​can be used as the first input parameter to form the characteristic values ​​of the absorption peaks. Determine the frequency positions of the absorption peaks in the spectrum, especially those absorption peaks related to specific components of the cells. These frequency positions can be used as features to distinguish different types of cells to form the second input parameter of the characteristic values ​​of the absorption peaks.

[0055] This step requires obtaining the absorption intensity curves of all cell types in the cell sample to be detected in each monochromatic spectrum curve, such as Figure 2 As shown (only an exemplary graph, only showing the absorption intensity curves of two cell types under monochromatic light irradiation, not representing the actual detection state), the maximum peak value in the absorption intensity curves of all cell types is saved in the data Data, assuming that there are a total of n maximum peak values ​​of cell types in the data Data, and assuming that the maximum peak value of the dth cell type at the current input wavelength is L d , the maximum peak value L of the d-th cell type at the current input wavelength d The corresponding frequency value of monochromatic light is W d , then the absorption peak characteristic value Z of the d-th cell typed The calculation formula is:

[0056] .

[0057] Among them, arctan is the inverse tangent function, d = 1,…,n.

[0058] It should be explained that although the frequency range corresponding to each monochromatic light is the same, when it is absorbed by a certain type of cell, the frequency value corresponding to the maximum peak point is a point in the frequency range corresponding to the above monochromatic light. Therefore, the maximum peak value L of different cell types at the current input wavelength is d The corresponding frequency values ​​of monochromatic light are different.

[0059] According to the above formula, the absorption peak characteristic value set of each monochromatic spectrum curve corresponding to each wavelength is constructed. Let the absorption peak characteristic value set composed of n absorption peaks of the monochromatic spectrum curve corresponding to the I wavelength be [Z I1 , Z I2 , …, Z Id , …, Z In ].

[0060] S3. Input multiple absorption peak characteristic values ​​in the absorption peak characteristic value set in each monochromatic spectrum curve into a convolutional neural network, and output the peak fusion feature of the monochromatic spectrum curve.

[0061] The n absorption peak characteristic values ​​of each monochromatic spectral curve are extracted and input into the convolutional neural network. The n absorption peak characteristic values ​​are processed by the convolutional neural network to obtain the peak fusion characteristics of the monochromatic spectral curve. The fusion characteristic values ​​represent the characteristics of the cell sample to be tested under different monochromatic light irradiation.

[0062] Specifically, the n absorption peak characteristic values ​​Z of the I-th monochromatic spectrum curve are I1 , Z I2 , …, Z Id , …, Z In The input is sent to the convolutional neural network, which groups the input n absorption peak eigenvalues ​​into K groups.

[0063] Here, the grouping method is explained: Since the number of abnormal cell types in the cell sample to be detected is very small, the corresponding absorption peak characteristic values ​​are obviously different from other common cell types. Therefore, grouping can be performed according to the clustering of the absorption peak characteristic values, such as Figure 3For example, the absorption peak characteristic values ​​corresponding to most white blood cells belong to one cluster, that is, they can be divided into one group, and the absorption peak characteristic values ​​corresponding to most red blood cells belong to another cluster, that is, they can be divided into another group, and one or two similar absorption peak characteristic values ​​alone are not enough to form a cluster, that is, they are divided into one group alone, and so on.

[0064] The average absorption peak characteristic value in each group of data is extracted. The average absorption peak characteristic value vector in each group of data represents the most significant or important feature of the group. The average absorption peak characteristic value is used as the peak feature of the group.

[0065] A weight is assigned to each group. The weight can be determined based on the number of absorption peak characteristic values ​​in the group. The larger the number, the smaller the weight. The dot product of the average absorption peak characteristic value of each group and the group weight is performed, and the accumulated result is the fusion characteristic value.

[0066] .

[0067] Among them, Z I is the fusion eigenvalue of the Ith monochromatic spectrum curve, K is the total number of groups, is the average absorption peak characteristic value of the qth group of the Ith monochromatic spectrum curve, W Iq is the weight of the qth group of the Ith monochromatic spectral curve.

[0068] This convolution operation can integrate the local information of multiple absorption peak feature values ​​of different clusters into a global descriptor, which helps to identify and extract key information in complex data sets.

[0069] S4. Combining the fused feature values ​​of multiple monochromatic spectral curves into a multi-spectral fusion feature value sequence, comparing the multi-spectral fusion feature value sequence with known standard multi-spectral feature value sequences of different types of tumor cells, and determining the type of tumor cells contained in the cell sample to be detected based on the comparison results.

[0070] Construct standard multispectral feature value sequences of different types of tumor cells. Suppose the standard multispectral feature value sequence of the Qth type of tumor cells is [Y Q1 , Y Q2 , …, Y QT , …, Y QM ], where M is the total number of monochromatic spectral curves, i.e. the total number of monochromatic light sources in the spectrophotometer,

[0071] Among them, different categories of tumor cells can represent different development stages of tumor cells or different types of tumor cells, and the specific selection method can be adjusted according to actual detection needs.

[0072] The fusion feature values ​​of the multiple monochromatic spectral curves obtained in step S3 are combined into a multi-spectral fusion feature value sequence [Z1, Z2, ..., Z I , …, Z M ], the multispectral fusion feature value sequence is compared with the standard multispectral feature value sequences of different categories of tumor cells, and the category numbers of tumor cells that meet the standard multispectral feature value sequence are obtained. If the category numbers of tumor cells that meet the standard multispectral feature value sequence are not obtained, it is proved that the cell sample to be detected does not contain tumor cells.

[0073] Specifically, the multi-spectral fusion feature value sequence [Z1, Z2, ..., Z I , …, Z M ] and the standard multi-spectral feature value sequence of the Q-th category tumor cells [Y Q1 , Y Q2 , …, Y QI , …, Y QM ], where Y QI is the characteristic value of the Ith spectrum of the Qth category of tumor cells, Z I is the fusion eigenvalue of the Ith monochromatic spectrum curve in the multi-spectral fusion eigenvalue sequence.

[0074] The difference between the cell sample to be tested and the Qth type of tumor cells is D Q The formula is as follows:

[0075] .

[0076] Output the difference value D of the comparison Q Less than the threshold D TH The tumor cell category serial number is output, that is, there are cell samples that meet the category of tumor cells in the cell samples to be tested, and sample identification and warning are performed according to the tumor cell category serial number. If there is no output value, it means that there are no cell samples that meet the tumor cell category in the cell samples to be tested, which proves that the cell samples to be tested do not contain tumor cells.

[0077] Among them, the threshold D TH The method for determining can adopt a distribution-based method, that is, the threshold can be set according to the statistical characteristics of the data (such as mean, median, and quantile). For example, the mean or median of the data can be used as the threshold, or the threshold can be determined by using the percentile based on the data distribution. A model-based method can be used, that is, the threshold is determined by training different models and comparing their performance. For example, the output probability of a logistic regression model can be used as the threshold, or when using a decision tree, a threshold that minimizes the Gini impurity can be selected.

[0078] In a preferred embodiment, the present invention also proposes a tumor cell detection system based on spectral spectroscopy, including a cell collection unit, a spectrophotometer and a spectral processing unit.

[0079] Cell collection units are used to collect cell samples to be tested. Cell collection units such as cell collection and tissue microdissection instruments can effectively cut the target area of ​​tissue sections under a microscope, or collect and deposit cells from any adherent culture grown in standard cell culture plates and culture dishes.

[0080] Among them, the components in the blood sample, such as platelets, plasma, nucleated cells, etc., need to be properly processed and preserved for subsequent detection and analysis.

[0081] As a device for collecting spectroscopic spectra, a spectrophotometer uses a light source that can generate multiple wavelengths, and then generates a light source of a specific wavelength through a series of spectroscopic devices, such as Figure 4 Shown is a schematic diagram of two different light sources irradiating a cell sample.

[0082] After the light passes through the cell sample to be detected, part of the light is absorbed, and the absorbance value of the cell sample to be detected is calculated, thereby converting it into the concentration of the cell sample to be detected.

[0083] The measurement range of the spectrophotometer includes the visible light region with a wavelength range of 400 to 760 nm and the ultraviolet light region with a wavelength range of 200 to 400 nm. Different light sources have their own unique emission spectra. According to the absorption spectrum of the substance and the law of light absorption, the substance can be quantitatively and qualitatively analyzed. Figure 5 As shown, under completely consistent measuring conditions, the ultraviolet absorption spectra of various cell samples were measured, and only the 190-400 nm spectrum is given as an example.

[0084] Among them, the cell sample to be tested is placed in the measuring pool of the spectrophotometer to ensure that the cell sample to be tested is evenly distributed and free of bubbles. The parameters of the spectrophotometer are set, including wavelength range, scanning speed, etc., and monochromatic light of different wavelengths is transmitted to the cell sample to be tested in turn. The absorption degree of monochromatic light of each wavelength is measured. The spectrophotometer will automatically record the absorption intensity of each wavelength of light, draw a spectral spectrum with wavelength as the horizontal axis and light absorption intensity as the vertical axis, and save the spectral spectrum for subsequent analysis and processing.

[0085] The spectral processing unit obtains the absorption peak characteristic value set in each monochromatic spectral curve in the spectral spectrum diagram, inputs multiple absorption peak characteristic values ​​in the absorption peak characteristic value set in each monochromatic spectral curve into a convolutional neural network, outputs the peak fusion feature of the monochromatic spectral curve, combines the fusion characteristic values ​​of multiple monochromatic spectral curves into a multi-spectral fusion characteristic value sequence, compares the multi-spectral fusion characteristic value sequence with the known standard multi-spectral characteristic value sequences of different categories of tumor cells, and determines the category of tumor cells contained in the cell sample to be detected according to the comparison result.

[0086] The tumor cell detection method of the present invention detects tumor cells by acquiring a spectroscopic spectrum. This method is non-contact and non-invasive, so there is no need to label the cells in any form during the detection process, such as attaching marker proteins. The advantage of the present invention is that it reduces damage to cells, allowing fewer damaged cells to be used in subsequent inspections, analyses, and studies. In addition, since it does not rely on the expression of specific marker proteins, a wider range of tumor cells can be detected, including those that may have reduced or no expression of marker proteins during the EMT (epithelial-mesenchymal transition) process. By using deep learning and artificial intelligence technology to feature encode and classify the spectroscopic spectrum, this method can improve the accuracy and efficiency of detection.

[0087] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrations. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.

[0088] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for detecting tumor cells based on spectroscopy for non-disease diagnosis and treatment, characterized in that: The steps include: S1. Transmitting monochromatic lights of different wavelengths to the cell samples to be detected in sequence, and collecting the spectroscopic spectrum of the cell samples to be detected; S2. Obtain a set of absorption peak characteristic values ​​in each monochromatic spectrum curve in the spectroscopic spectrum diagram; obtain the absorption intensity curves of all cell types in the cell sample to be detected in each monochromatic spectrum curve, save the maximum peak value in the absorption intensity curves of n cell types, and set the maximum peak value of the dth cell type at the current input wavelength as L d , the maximum peak value L of the d-th cell type at the current input wavelength d The corresponding frequency value of monochromatic light is W d , then the absorption peak characteristic value Z of the d-th cell type d The calculation formula is: ； Among them, arctan is the inverse tangent function, d = 1,…,n; According to the above formula, a set of absorption peak characteristic values ​​of each monochromatic spectrum curve is constructed; S3, inputting a plurality of absorption peak characteristic values ​​in the absorption peak characteristic value set in each monochromatic spectrum curve into a convolutional neural network, and outputting a peak fusion feature of the monochromatic spectrum curve; The n absorption peak characteristic values ​​Z of the I-th monochromatic spectrum curve I1 , Z I2 , …, Z Id , …, Z In Input into the convolutional neural network, the convolutional neural network groups the input n absorption peak eigenvalues ​​into K groups, and calculates the fusion eigenvalue Z of the I-th monochromatic spectrum curve I , ； The average absorption peak characteristic value in each group of data is extracted. The average absorption peak characteristic value vector in each group of data represents the most significant feature in the group. The average absorption peak characteristic value is used as the peak feature of the group. in, is the average absorption peak characteristic value of the qth group, W Iq is the weight of the qth group; S4. Combining the fused feature values ​​of multiple monochromatic spectral curves into a multi-spectral fusion feature value sequence, comparing the multi-spectral fusion feature value sequence with known standard multi-spectral feature value sequences of different types of tumor cells, and determining the type of tumor cells contained in the cell sample to be detected according to the comparison results.

2. The method for detecting tumor cells based on spectroscopy for non-disease diagnosis and treatment according to claim 1, characterized in that: In the step S2, each monochromatic spectrum curve in the spectroscopic spectrum diagram is obtained, and baseline correction and intensity normalization are performed on each monochromatic spectrum curve; and a set of absorption peak characteristic values ​​of each monochromatic spectrum curve after intensity normalization is obtained.

3. The method for detecting tumor cells based on spectroscopy for non-disease diagnosis and treatment according to claim 1, characterized in that: In step S4, standard multi-spectral feature value sequences of different types of tumor cells are constructed. The standard multi-spectral feature value sequence of the Q-th type of tumor cells is [Y Q1 , Y Q2 , …, Y QT , …, Y QM ], where M is the total number of monochromatic spectral curves; The fusion feature values ​​of the multiple monochromatic spectrum curves obtained in step S3 are combined into a multi-spectral fusion feature value sequence [Z1, Z2, ..., Z I , …, Z M ], the multispectral fusion feature value sequence is compared with the standard multispectral feature value sequences of different categories of tumor cells respectively, and the category numbers of tumor cells that meet the standard multispectral feature value sequence are obtained. If the category numbers of tumor cells that meet the standard multispectral feature value sequence are not obtained, the cell sample to be detected does not contain tumor cells.

4. The method for detecting tumor cells based on spectroscopy for non-disease diagnosis and treatment according to claim 3, characterized in that: By calculating the difference value, the multi-spectral fusion feature value sequence [Z1, Z2, ..., Z I , …, Z M ] and the standard multi-spectral feature value sequence of the Q-th category tumor cells [Y Q1 , Y Q2 , …, Y QI , …, Y QM ], where Y QI is the characteristic value of the Ith spectrum of the Qth category of tumor cells, Z I is the fusion eigenvalue of the Ith monochromatic spectrum curve in the multi-spectral fusion eigenvalue sequence; The difference value D between the cell sample to be tested and the Qth category of tumor cells Q The formula is as follows: ； Output the difference value D of the comparison Q Less than the threshold D TH Tumor cell type number.

5. The method for detecting tumor cells based on spectroscopy for non-disease diagnosis and treatment according to claim 2, characterized in that: The intensity of each monochromatic spectral curve is normalized by dividing the absorption intensity value of the maximum peak point of the monochromatic spectral curve corresponding to each monochromatic light source by the absorption intensity value of the maximum peak point in the entire spectral spectrum, so that the absorption intensity value of the peak point of the monochromatic spectral curve is normalized to the range of 0 to 1.

Images