Cell DNA quantitative analysis method and system based on microscope imaging technology

Abstract

The present invention proposes a cell DNA quantitative analysis method and system based on microscope imaging technology; the method co-incubates a cell sample with a fluorescent DNA probe and a PAFP-labeled NLS protein; uses SIM imaging to obtain DNA and PAFP-NLS signals; accurately locates a single cell nucleus through two-photon excitation; performs image segmentation on the fluorescence signal, and corrects the DNA signal based on the PAFP-NLS signal intensity to eliminate the nuclear thickness difference error; uses a multi-scale convolutional neural network for quantitative analysis, and obtains the DNA content based on an absolute quantitative model; the present invention improves the accuracy, efficiency and reliability of DNA quantitative analysis, and provides a powerful new tool for biomedical research and clinical applications.

Description

Technical Field

[0001] The invention belongs to the technical field of cell DNA quantitative analysis, and in particular relates to a cell DNA quantitative analysis method and system based on microscope imaging technology. Background Art

[0002] Accurate quantitative analysis of cellular DNA content is of great significance for fields such as cell cycle research, tumor diagnosis and drug screening. Traditional DNA quantitative analysis methods mainly include flow cytometry and image cytometry. Flow cytometry has the advantages of high throughput and rapid analysis, but it cannot provide cell morphological information and has high requirements for cell suspension, making it difficult to apply to samples such as tissue sections. Image cytometry can obtain cell morphology and DNA content information at the same time, but its spatial resolution and quantitative accuracy are limited by the diffraction limit of traditional optical microscope imaging systems.

[0003] In recent years, the development of super-resolution microscopy technology has brought new opportunities for quantitative analysis of cellular DNA. Structured illumination microscopy (SIM), as a wide-field super-resolution imaging technology, can increase the spatial resolution to about 100nm while maintaining a high imaging speed and low phototoxicity. However, SIM technology still faces many challenges in quantitative analysis of DNA: the difference in thickness of the cell nucleus in the z-axis direction will lead to changes in the intensity of the fluorescence signal, affecting the quantitative accuracy of the DNA content. Cell autofluorescence and nonspecific staining will introduce background noise and reduce the signal-to-noise ratio. The distribution of DNA in the cell nucleus has complex multi-scale characteristics, and traditional image analysis methods are difficult to fully capture these characteristics. Absolute quantitative standards: The lack of reliable absolute quantitative standards makes it difficult to directly compare different experimental results.

[0004] To solve the above problems, researchers have proposed a variety of improved methods. For example, using fluorescent proteins labeled with nuclear localization sequences (NLS) as internal references can partially compensate for the influence of sample thickness heterogeneity. However, this method still has nonlinear effects and needs further optimization.

[0005] In terms of background signal removal, image segmentation algorithms based on machine learning show good performance, but their generalization ability and robustness still need to be improved. In terms of multi-scale feature extraction, multi-scale analysis methods such as wavelet transform have been introduced, but how to effectively integrate these features remains a challenge; in terms of establishing absolute quantitative standards, researchers have tried to cross-validate flow cytometry and image analysis results, but the sample processing processes of the two methods are quite different and difficult to compare directly; in addition, with the rapid development of deep learning technology in the field of image analysis, models such as convolutional neural networks (CNN) have shown great potential in cell image segmentation and feature extraction. However, how to design a network structure suitable for DNA quantitative analysis and how to solve the interpretability problem of deep learning models still require further research.

[0006] Although super-resolution microscopy technology provides new possibilities for quantitative analysis of cellular DNA, there are still key issues that need to be addressed, such as sample thickness compensation, background signal removal, multi-scale feature extraction, and absolute quantitative standards. Summary of the invention

[0007] The present invention provides a method and system for quantitative analysis of cell DNA based on microscope imaging technology, aiming to combine super-resolution microscope technology, fluorescent labeling strategy and deep learning algorithm, apply them to DNA quantitative analysis, and solve the problems faced in the prior art.

[0008] The scheme of the present invention is as follows:

[0009] In a first aspect, the present invention provides a method for quantitative analysis of cell DNA based on microscope imaging technology, the method comprising the following steps:

[0010] Step a), the cell sample to be analyzed is incubated with a fluorescent DNA probe and a nuclear localization sequence NLS protein labeled with a photo-activated fluorescent protein PAFP.

[0011] Step b), using structured light illumination microscopy (SIM) to image the sample and obtain DNA fluorescence signals and PAFP-NLS fluorescence signals.

[0012] Step c) selectively activate PAFP-NLS in the target area by two-photon excitation light activation to achieve precise positioning of single cell nuclei.

[0013] The activation function is defined as: ;

[0014] in, Indicates that the excitation light is at the coordinate The strength of is the reference intensity of the excitation light, The target activation position center coordinates, and They are the activation range parameters of the target area in the xy plane and z axis direction respectively.

[0015] Step d) performing image segmentation on the acquired fluorescence signal to obtain a cell region, and correcting the DNA fluorescence signal based on the PAFP-NLS signal intensity in the cell region to eliminate the error caused by the difference in cell nucleus thickness.

[0016] The correction function for the DNA fluorescence signal is: ;in After correction The signal value at is the maximum value of the FP-NLS signal, is the nonlinear correction factor, , Before correction The signal value at FP-NLS signal The value at .

[0017] Step e) quantitatively analyzes the corrected DNA fluorescence signal using a trained multi-scale convolutional neural network, and obtains the cellular DNA content based on an absolute quantitative model.

[0018] Furthermore, the structured light illumination microscope SIM in step b) adopts a three-beam interference mode, and the distribution of illumination intensity is expressed as: ;in, express The light intensity at is the average light intensity, is the contrast, and Respectively Direction and The spatial frequency coordinates of the direction, , is the size of the spatial frequency, is the phase, is the modulation depth, ranging from 0 to 1. The larger it is, the more obvious the stripe pattern is. The smaller it is, the less noticeable the stripes will be.

[0019] By rotating the illumination pattern and phase shifting, 9 original images are obtained, and then the super-resolution image is obtained using the Wiener filter reconstruction algorithm: ;in, is the representation of the reconstructed super-resolution image in the frequency domain, is the Fourier transform of the mth original image, is the value of the optical transfer function OTF, for The complex conjugate of is the regularization parameter.

[0020] Furthermore, the two-photon excitation technique in step c) utilizes femtosecond pulse laser to excite the With laser intensity Proportional to the square of: ,in is the two-photon absorption cross section, is the pulse duration, is the repetition frequency; by adjusting the laser power and scanning speed, precise activation of cell nuclei at a specific depth can be achieved.

[0021] Furthermore, in step d), performing image segmentation on the acquired fluorescence signal to obtain the cell region includes: segmenting the DNA fluorescence signal and the PAFP-NLS fluorescence signal to obtain the precise boundary of the cell nucleus and cytoplasm.

[0022] The segmentation results are post-processed through morphological operations, including small area removal and boundary smoothing, to finally obtain the accurate cell area.

[0023] Furthermore, in the step e), the corrected DNA fluorescence signal is quantitatively analyzed by a trained multi-scale convolutional neural network, wherein the encoder path in the multi-scale convolutional neural network consists of multiple convolutional layers, including parallel convolutional layers of different scales, and the decoder path consists of upsampling and convolution operations, and an attention gating mechanism is added between the encoder and the decoder; an adaptive learning rate optimization strategy is adopted, with an initial learning rate of 0.001, which decays by 10% every 50 epochs.

[0024] Furthermore, in step e), the cell DNA content is obtained based on the absolute quantitative model, and the specific method is: using flow cytometry to measure the DNA content of some samples as calibration data; comparing and calibrating the flow cytometry results with the output of the deep learning network, and establishing an absolute quantitative model for DNA content:

[0025] ;in is the fitting parameter, The absolute quantification model obtains the DNA content. For the DNA content output of the deep learning network, the fitting parameters were determined by minimizing the mean square error: ; is the DNA content measured by flow cytometry, is the number of samples, when the mean square error When it is the smallest, the corresponding fitting parameters are the optimal parameters of the absolute quantitative model.

[0026] Furthermore, the parallel convolutional layers of different scales of the multi-scale convolutional neural network are expressed as: ;in, Represents the feature map obtained after multi-scale convolution, They represent convolution operations with kernel sizes of 1, 3, 5, and 7 respectively, and Concat represents feature concatenation operations. The multi-scale features are weightedly fused using the attention mechanism, and the cell DNA content is obtained through the fully connected layer. ; ;in, represents the attention weight vector, and is a learnable parameter matrix, which is used for feature transformation and attention score calculation respectively. For the The attention weight of each scale, For the Feature maps of different scales, is the final feature after weighted fusion.

[0027] in, and are the weights and biases of the fully connected layer, DNA content of the output of the deep learning network.

[0028] In a second aspect, the present invention provides a cell DNA quantitative analysis system based on microscope imaging technology, which is used to execute the method described in the first aspect of the present invention. The system includes: a sample processing module, a structured light illumination microscope SIM, a two-photon excitation module, an image processing module, a multi-scale convolutional neural network module and an absolute quantification module connected in sequence.

[0029] Furthermore, the sample processing module is used to incubate the cell sample to be analyzed with the fluorescent DNA probe and the nuclear localization sequence NLS protein labeled with the photoactivated fluorescent protein PAFP.

[0030] Furthermore, the structured light illumination microscope SIM is used to image the sample and obtain DNA fluorescence signals and PAFP-NLS fluorescence signals.

[0031] Furthermore, the two-photon excitation module is used to selectively activate PAFP-NLS in the target area by two-photon excitation light activation to achieve precise positioning of single cell nuclei.

[0032] Furthermore, the image processing module is used to perform image segmentation on the acquired fluorescence signal to obtain a cell region, and to correct the DNA fluorescence signal based on the PAFP-NLS signal intensity in the cell region.

[0033] Furthermore, the multi-scale convolutional neural network module is used to perform quantitative analysis on the corrected DNA fluorescence signal, and the absolute quantification module is used to derive the cellular DNA content based on the absolute quantification model.

[0034] Furthermore, the structured light illumination microscope SIM adopts a three-beam interference mode, including: an illumination light source, a phase modulator, a rotating platform, an image acquisition device and an image reconstruction module.

[0035] Furthermore, the illumination light source is used to generate an interference fringe pattern; the phase modulator is used to adjust the phase of the illumination pattern; the rotating platform is used to rotate the illumination pattern; the image acquisition device is used to obtain 9 original images; and the image reconstruction module is used to obtain a super-resolution image using a Wiener filter reconstruction algorithm.

[0036] Furthermore, the system also includes: a flow cytometer, which is used to measure the DNA content of some samples as calibration data; a calibration module, which is used to compare and calibrate the flow cytometry results with the output of the deep learning network to establish an absolute quantitative model of DNA content; and a parameter optimization module, which is used to determine the optimal parameters of the absolute quantitative model by minimizing the mean square error.

[0037] Compared with the prior art, the present invention has the following beneficial effects: the present invention utilizes PAFP-NLS as an internal reference and proposes a nonlinear correction method, which effectively compensates for the change in fluorescence signal intensity caused by uneven cell nucleus thickness. This correction mechanism significantly improves the reliability of DNA quantification; by innovatively combining super-resolution microscopy technology, fluorescent labeling strategy and deep learning algorithm, the accuracy, efficiency and reliability of cell DNA quantitative analysis are comprehensively improved, providing a powerful new tool for biomedical research and clinical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flow chart of the method for quantitative analysis of cell DNA based on microscope imaging technology of the present invention;

[0039] Figure 2 The figure is a schematic diagram of the composition of the cell DNA quantitative analysis system based on microscope imaging technology of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Embodiment 1, as Figure 1 The figure is a flow chart of the cell DNA quantitative analysis method based on microscope imaging technology of the present invention, which comprises the following steps: Step a), incubating the cell sample to be analyzed with a fluorescent DNA probe and a nuclear localization sequence NLS protein labeled with a photoactivated fluorescent protein PAFP.

[0042] The incubation method here is not limited, and only a specific example is given as an illustration: human HeLa cell line is selected as the sample to be analyzed.

[0043] The cells were cultured in DMEM medium containing 10% fetal bovine serum and 1% penicillin / streptomycin. When the cells grew to 70-80% confluence, they were digested with 0.25% trypsin-EDTA; the cells were collected and washed twice with PBS.

[0044] SYTOX Orange nucleic acid dye was selected as the fluorescent DNA probe, and a 1 mM SYTOX Orange stock solution (in DMSO) was prepared; mEos3.2-NLS fusion protein was used as PAFP-NLS, and the mEos3.2-NLS protein was expressed and purified in Escherichia coli in advance, and the purified protein was diluted with PBS to a concentration of 1 mg / mL.

[0045] Resuspend the collected HeLa cells in serum-free DMEM medium and adjust the cell concentration to 1×10^6 cells / mL. Add 5 μL SYTOX Orange stock solution (final concentration 5 μM) and 10 μL mEos3.2-NLS protein solution (final concentration 10 μg / mL) to 1 mL of cell suspension; Mix gently to avoid bubbles; Incubate the mixture in an incubator at 37°C and 5% CO2 for 30 minutes.

[0046] After incubation, wash the cells twice with pre-warmed PBS to remove unbound probes and proteins, and resuspend the cells in fresh serum-free DMEM medium; drop the incubated cell suspension onto a glass slide pre-coated with poly-L-lysine and let it stand for 15 minutes to allow the cells to adhere.

[0047] In this specific example, the cell sample was labeled with SYTOX Orange (for DNA staining) and mEos3.2-NLS (for nuclear localization and thickness correction). SYTOX Orange can efficiently penetrate the cell membrane and bind to DNA, providing a strong fluorescent signal; while mEos3.2-NLS will be localized in the cell nucleus, providing a reference for subsequent DNA signal correction.

[0048] Step b), using structured light illumination microscopy (SIM) to image the sample and obtain DNA fluorescence signals and PAFP-NLS fluorescence signals.

[0049] Structured light illumination microscopy SIM uses a three-beam interference mode, and the distribution of illumination intensity is expressed as: ,in, express The light intensity at is the average light intensity, is the contrast, and Respectively Direction and The spatial frequency coordinates of the direction, , is the size of the spatial frequency, is the phase, is the modulation depth, ranging from 0 to 1. The larger it is, the more obvious the stripe pattern is. The smaller it is, the less obvious the stripes are. By rotating the illumination pattern and shifting the phase, 9 original images are obtained, and then the super-resolution image is obtained using the Wiener filter reconstruction algorithm: ,in, is the representation of the reconstructed super-resolution image in the frequency domain, is the Fourier transform of the mth original image, is the value of the optical transfer function OTF, for The complex conjugate of is the regularization parameter.

[0050] It should be noted that in terms of light source selection, an appropriate excitation light source is selected for the DNA fluorescence signal (SYTOX Orange), usually a 532nm or 561nm laser; a 405nm laser is selected for photoactivation for the PAFP-NLS fluorescence signal (mEos3.2), and a 488nm or 514nm laser is used for fluorescence excitation; appropriate dichroic mirrors and emission filters are used to separate the fluorescence signals of DNA and PAFP-NLS, ensuring that the filter combination can effectively separate the two fluorescences to avoid signal crosstalk; a high-quality grating or spatial light modulator (SLM) is used to generate a structured light illumination pattern to ensure the contrast and uniformity of the fringe pattern.

[0051] For each plane, at least 3 phases (0°, 120°, 240°) and 3 directions (0°, 60°, 120°) are collected, for a total of 9 original images, to ensure the accuracy and repeatability of phase and direction switching; for PAFP-NLS signals, special attention should be paid to controlling the power of the 405nm activation laser to avoid over-activation; the Z-axis step size is usually 100-200nm to meet the Nyquist sampling theorem.

[0052] Step c) selectively activate the PAFP-NLS in the target area by two-photon excitation light activation to achieve precise positioning of single cell nuclei; the activation function is defined as:

[0053] ,in, Indicates that the excitation light is at the coordinate The strength of is the reference intensity of the excitation light, The target activation position center coordinates, and They are the activation range parameters of the target area in the xy plane and z axis direction respectively.

[0054] Two-photon excitation technology uses femtosecond pulse laser to excite the probability With laser intensity Proportional to the square of: ,in is the two-photon absorption cross section, is the pulse duration, is the repetition frequency; by adjusting the laser power and scanning speed, precise activation of cell nuclei at a specific depth can be achieved.

[0055] Typically, a titanium sapphire laser is used, with a wavelength tunable range of about 700-1000nm; for mEos3.2-NLS, the typical excitation wavelength is about 800nm, the pulse duration is usually around 100fs, and the repetition frequency is generally 80MHz.

[0056] Step d), performing image segmentation on the acquired fluorescence signal to obtain the cell region, and based on the PAFP-NLS signal intensity in the cell region, correcting the DNA fluorescence signal to eliminate the error caused by the difference in cell nucleus thickness; performing image segmentation on the acquired fluorescence signal to obtain the cell region includes: segmenting the DNA fluorescence signal and the PAFP-NLS fluorescence signal to obtain the precise boundary of the cell nucleus and cytoplasm; post-processing the segmentation result through morphological operations, including small area removal and boundary smoothing, to finally obtain the precise cell region.

[0057] It should be noted that before image segmentation, the acquired fluorescence images were first preprocessed, including the application of a non-local mean filtering algorithm to remove noise and the use of a rolling ball algorithm to correct the non-uniform background. For the DNA and PAFP-NLS dual-channel images, a mutual information-based registration method was used to ensure spatial correspondence.

[0058] Image segmentation uses a multi-scale 3D watershed algorithm combined with the complementary information of DNA and PAFP-NLS signals. To deal with the problem of cell adhesion, a deep learning-based instance segmentation network (such as Mask R-CNN) can be introduced, trained on manually annotated datasets. After segmentation, morphological operations are applied for post-processing: an area threshold (usually set to 1 / 10 of the average cell size) is used to remove small areas, and the level set method is used to smooth the boundaries. To evaluate the segmentation quality, the Dice coefficient and Hausdorff distance are calculated and compared with the manual annotation results. If the accuracy is lower than the preset threshold (such as Dice coefficient <0.9), the user interactive correction mechanism is triggered.

[0059] DNA fluorescence signal correction The PAFP-NLS signal intensity was used to estimate nuclear thickness. The correction function used a piecewise linear model and applied different correction coefficients to thin regions (<2 μm) and thick regions. The correction parameters were optimized by minimizing the mean square error and combining cross-validation.

[0060] The correction function for the DNA fluorescence signal is: ,in After correction The signal value at is the maximum value of the FP-NLS signal, is the nonlinear correction factor, , Before correction The signal value at FP-NLS signal The value at .

[0061] Step e) quantitatively analyzes the corrected DNA fluorescence signal using a trained multi-scale convolutional neural network, and obtains the cellular DNA content based on an absolute quantitative model.

[0062] In the step e), the corrected DNA fluorescence signal is quantitatively analyzed by a trained multi-scale convolutional neural network, wherein the encoder path in the multi-scale convolutional neural network consists of multiple convolutional layers, including parallel convolutional layers of different scales, and the decoder path consists of upsampling and convolution operations, and an attention gating mechanism is added between the encoder and the decoder; an adaptive learning rate optimization strategy is adopted, with an initial learning rate of 0.001 and a decay of 10% every 50 epochs.

[0063] The parallel convolutional layers of different scales of the multi-scale convolutional neural network are expressed as: ,in, Represents the feature map obtained after multi-scale convolution, They represent convolution operations with kernel sizes of 1, 3, 5, and 7 respectively, and Concat represents feature concatenation operations. The multi-scale features are weightedly fused using the attention mechanism, and the cell DNA content is obtained through the fully connected layer. ; ;in, represents the attention weight vector, and is a learnable parameter matrix, which is used for feature transformation and attention score calculation respectively. For the The attention weight of each scale, For the Feature maps of different scales, is the final feature after weighted fusion; ;in and are the weights and biases of the fully connected layer, DNA content of the output of the deep learning network.

[0064] Flow cytometry was used to measure the DNA content of some samples as calibration data. The flow cytometry results were compared and calibrated with the output of the deep learning network to establish an absolute quantitative model for DNA content: ;in is the fitting parameter, The absolute quantification model obtains the DNA content. For the DNA content output of the deep learning network, the fitting parameters were determined by minimizing the mean square error: , is the DNA content measured by flow cytometry, is the number of samples, when the mean square error When it is the smallest, the corresponding fitting parameters are the optimal parameters of the absolute quantitative model.

[0065] Embodiment 2, as Figure 2As shown, the cell DNA quantitative analysis system based on microscope imaging technology of the present invention is used to execute the method described in Example 1. The system includes: a sample processing module, a structured light illumination microscope SIM, a two-photon excitation module, an image processing module, a multi-scale convolutional neural network module and an absolute quantification module connected in sequence.

[0066] The sample processing module is used to incubate the cell sample to be analyzed with the fluorescent DNA probe and the nuclear localization sequence NLS protein labeled with the photoactivated fluorescent protein PAFP; the module also includes a temperature control unit and a precision liquid handling device to ensure the consistency and repeatability of the incubation process. The temperature control unit can be accurately adjusted within the range of 37±0.1°C, and the liquid handling device has a liquid separation accuracy of ±0.1μL.

[0067] The structured light illumination microscope SIM is used to image the sample and obtain DNA fluorescence signals and PAFP-NLS fluorescence signals. SIM is equipped with 488nm and 561nm dual-wavelength lasers with powers of 100mW and 150mW respectively; in addition, SIM also includes a piezoelectric autofocus system that can achieve an axial resolution of 50nm within the range of ±10μm to maintain the stability of long-term imaging.

[0068] The two-photon excitation module is used to selectively activate the PAFP-NLS in the target area by two-photon excitation light activation to achieve precise positioning of single cell nuclei. The module uses a wavelength-tunable titanium sapphire laser (700-1000nm), a pulse width of 100fs, and a repetition frequency of 80MHz. The module also includes a spatial light modulator that can achieve submicron activation accuracy within a range of 10×10μm.

[0069] The image processing module is used to perform image segmentation on the acquired fluorescence signal to obtain the cell area, and to correct the DNA fluorescence signal based on the PAFP-NLS signal intensity in the cell area. The module is equipped with a high-performance GPU (such as NVIDIA Tesla V100), supports real-time 3D image processing, and the processing speed can reach 1GB / s.

[0070] The multi-scale convolutional neural network module is used to quantitatively analyze the corrected DNA fluorescence signal. The network uses ResNet50 as the backbone and combines the feature pyramid network (FPN) to achieve multi-scale feature extraction. The network training uses the Adam optimizer, with an initial learning rate of 0.001 and a batch size of 16.

[0071] The absolute quantification module is used to obtain the cell DNA content based on the absolute quantification model. The module includes a calibration curve management system that can store and manage calibration models of up to 100 different cell types.

[0072] The structured light illumination microscope SIM adopts a three-beam interference mode, and includes: an illumination light source, a phase modulator, a rotating platform, an image acquisition device and an image reconstruction module.

[0073] The illumination light source is used to generate an interference fringe pattern; the phase modulator is used to adjust the phase of the illumination pattern, and the phase modulation accuracy reaches λ / 100; the rotating platform is used to rotate the illumination pattern, and the angle accuracy is 0.1°; the image acquisition device is an sCMOS camera with a pixel number of 2048×2048 and a frame rate of up to 100fps; the image reconstruction module is used to obtain a super-resolution image using a Wiener filter reconstruction algorithm, and the reconstruction time is less than 1s / frame.

[0074] The system also includes: a flow cytometer, which is used to measure the DNA content of some samples as calibration data; a calibration module, which is used to compare and calibrate the flow cytometry results with the output of the deep learning network to establish an absolute quantitative model of DNA content; and a parameter optimization module, which is used to determine the optimal parameters of the absolute quantitative model by minimizing the mean square error.

[0075] In addition, the system also includes the following functional modules: Central control unit: uses a real-time operating system to coordinate the work of each module, optimize the workflow, and has a response time of <10ms.

[0076] Data management system: Equipped with a 20TB RAID5 storage array for storing raw images and analysis results. Contains a PostgreSQL-based relational database to support efficient data retrieval and analysis.

[0077] Quality Control Module: Automatically runs a standard sample test every 24 hours and generates a system performance report. If key indicators deviate from the preset range, the system will issue an alarm and recommend calibration actions.

[0078] User Interface: An interactive web-based interface that supports real-time monitoring and parameter adjustment. It includes visualization tools to display analysis results and system status.

[0079] Through the collaborative work of these functional modules, this system can achieve high-precision, high-throughput quantitative analysis of cell DNA, providing reliable technical support for biomedical research and clinical diagnosis.

[0080] It should be noted that those skilled in the art should understand that various changes and equivalent substitutions may be made to the present invention without departing from the scope of the present invention. In addition, various modifications may be made to the present invention for specific situations or materials without departing from the scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed, but should include all embodiments falling within the scope of the claims of the present invention.

Claims

1. A method for quantitative analysis of cell DNA based on microscope imaging technology, characterized in that: The method comprises the following steps: step a), incubating the cell sample to be analyzed with a fluorescent DNA probe and a nuclear localization sequence NLS protein labeled with a photoactivated fluorescent protein PAFP; step b), imaging the sample using a structured light illumination microscope SIM to obtain a DNA fluorescent signal and a PAFP-NLS fluorescent signal; step c), selectively activating the PAFP-NLS in the target area by two-photon excitation light activation to achieve precise positioning of a single cell nucleus; the activation function is defined as: ;in, Indicates that the excitation light is at the coordinate The strength of is the reference intensity of the excitation light, The target activation position center coordinates, and are activation range parameters of the target area in the xy plane and z axis direction respectively; step d), performing image segmentation on the acquired fluorescence signal to obtain the cell area, and correcting the DNA fluorescence signal based on the PAFP-NLS signal intensity in the cell area to eliminate the error caused by the difference in cell nucleus thickness; the correction function of the DNA fluorescence signal is: ;in After correction The signal value at is the maximum value of the FP-NLS signal, is the nonlinear correction factor, , Before correction The signal value at FP-NLS signal In step e), the corrected DNA fluorescence signal is quantitatively analyzed by the trained multi-scale convolutional neural network, and the cellular DNA content is obtained based on the absolute quantitative model.

2. The method for quantitative analysis of cell DNA based on microscope imaging technology according to claim 1, characterized in that: The structured light illumination microscope SIM in step b) adopts a three-beam interference mode, and the distribution of illumination intensity is expressed as: ;in, express The light intensity at is the average light intensity, is the contrast, and Respectively Direction and The spatial frequency coordinates of the direction, , is the size of the spatial frequency, is the phase, is the modulation depth, ranging from 0 to 1. The larger it is, the more obvious the stripe pattern is. The smaller it is, the less obvious the stripes are. By rotating the illumination pattern and shifting the phase, 9 original images are obtained, and then the super-resolution image is obtained using the Wiener filter reconstruction algorithm: ;in, is the representation of the reconstructed super-resolution image in the frequency domain, is the Fourier transform of the mth original image, is the value of the optical transfer function OTF, for The complex conjugate of is the regularization parameter.

3. The method for quantitative analysis of cell DNA based on microscope imaging technology according to claim 2, characterized in that: The two-photon excitation technique in step c) uses a femtosecond pulse laser to excite the probability With laser intensity Proportional to the square of: ,in is the two-photon absorption cross section, is the pulse duration, is the repetition frequency; by adjusting the laser power and scanning speed, precise activation of cell nuclei at a specific depth can be achieved.

4. The method for quantitative analysis of cell DNA based on microscope imaging technology according to claim 3, characterized in that: In the step d), performing image segmentation on the acquired fluorescence signal to obtain the cell region includes: segmenting the DNA fluorescence signal and the PAFP-NLS fluorescence signal to obtain the precise boundary of the cell nucleus and cytoplasm; post-processing the segmentation result by morphological operation, including small area removal and boundary smoothing, to finally obtain the precise cell region.

5. The method for quantitative analysis of cell DNA based on microscope imaging technology according to claim 4, characterized in that: In the step e), the corrected DNA fluorescence signal is quantitatively analyzed by a trained multi-scale convolutional neural network, wherein the encoder path in the multi-scale convolutional neural network consists of multiple convolutional layers, including parallel convolutional layers of different scales, and the decoder path consists of upsampling and convolution operations, and an attention gating mechanism is added between the encoder and the decoder; an adaptive learning rate optimization strategy is adopted, with an initial learning rate of 0.001 and a decay of 10% every 50 epochs.

6. The method for quantitative analysis of cell DNA based on microscope imaging technology according to claim 5, characterized in that: In step e), the cell DNA content is obtained based on the absolute quantitative model. The specific method is: using flow cytometry to measure the DNA content of some samples as calibration data; comparing and calibrating the flow cytometry results with the output of the deep learning network to establish an absolute quantitative model for DNA content: ;in is the fitting parameter, The absolute quantification model obtains the DNA content. For the DNA content output of the deep learning network, the fitting parameters were determined by minimizing the mean square error: ; is the DNA content measured by flow cytometry, is the number of samples, when the mean square error When it is the smallest, the corresponding fitting parameters are the optimal parameters of the absolute quantitative model.

7. The method for quantitative analysis of cell DNA based on microscope imaging technology according to claim 4, characterized in that: The parallel convolutional layers of different scales of the multi-scale convolutional neural network are expressed as: ;in, Represents the feature map obtained after multi-scale convolution, They represent convolution operations with kernel sizes of 1, 3, 5, and 7 respectively, and Concat represents feature concatenation operations. The multi-scale features are weightedly fused using the attention mechanism, and the cell DNA content is obtained through the fully connected layer. ; ;in, represents the attention weight vector, and is a learnable parameter matrix, which is used for feature transformation and attention score calculation respectively. For the The attention weight of each scale, For the Feature maps of different scales, is the final feature after weighted fusion; ;in and are the weights and biases of the fully connected layer, DNA content of the output of the deep learning network.

8. A cell DNA quantitative analysis system based on microscope imaging technology, used to perform the method according to any one of claims 1 to 7, characterized in that: The system includes: a sample processing module, a structured light illumination microscope SIM, a two-photon excitation module, an image processing module, a multi-scale convolutional neural network module and an absolute quantification module connected in sequence; the sample processing module is used to co-incubate the cell sample to be analyzed with a fluorescent DNA probe and a nuclear localization sequence NLS protein labeled with a light-activated fluorescent protein PAFP; the structured light illumination microscope SIM is used to image the sample to obtain a DNA fluorescent signal and a PAFP-NLS fluorescent signal; the two-photon excitation module is used to selectively activate the PAFP-NLS in the target area by two-photon excitation light activation to achieve accurate positioning of a single cell nucleus; the image processing module is used to perform image segmentation on the acquired fluorescent signal to obtain a cell area, and correct the DNA fluorescent signal based on the PAFP-NLS signal intensity in the cell area; the multi-scale convolutional neural network module is used to quantitatively analyze the corrected DNA fluorescent signal, and the absolute quantification module is used to derive the cell DNA content based on the absolute quantification model.

9. The cell DNA quantitative analysis system based on microscope imaging technology according to claim 8, characterized in that: The structured light illumination microscope SIM adopts a three-beam interference mode, including: an illumination light source, a phase modulator, a rotating platform, an image acquisition device and an image reconstruction module; the illumination light source is used to generate an interference fringe pattern; the phase modulator is used to adjust the phase of the illumination pattern; the rotating platform is used to rotate the illumination pattern; the image acquisition device is used to obtain 9 original images; the image reconstruction module is used to obtain a super-resolution image using a Wiener filter reconstruction algorithm.

10. The cell DNA quantitative analysis system based on microscope imaging technology according to claim 8, characterized in that: The system also includes: a flow cytometer, which is used to measure the DNA content of some samples as calibration data; a calibration module, which is used to compare and calibrate the flow cytometry results with the output of the deep learning network to establish an absolute quantitative model of DNA content; and a parameter optimization module, which is used to determine the optimal parameters of the absolute quantitative model by minimizing the mean square error.

Images