Cell analysis method, cell analysis device and storage medium

Abstract

The invention relates to the technical field of cell analysis, and discloses a cell analysis method, a cell analysis device and a storage medium, and the method comprises the following steps: S1, multi-modal labeling: synchronously loading cell surface protein, a metabolism probe and a spatial positioning marker; s2, collaborative acquisition: obtaining space-time multi-dimensional data through flow type, mass spectrum and two-photon combination; s3, tensor modeling: constructing four-dimensional data and correcting spatial drift; s4, metabolic constraint decomposition: executing non-negative tensor decomposition to obtain a factor matrix; s5, dynamically adjusting parameters: optimizing excitation and sampling frequencies in real time according to factor changes; and S6, feature analysis: calculating a metabolism-phenotype coupling index. Through a multi-modal synchronous marking and four-dimensional dynamic tensor modeling method, cross-dimensional association of physical characteristics, molecular expression and metabolic activity is broken through, lossless fusion of single-cell multi-dimensional data is realized, and the spatial resolution is greatly improved compared with the prior art.

Description

Technical Field

[0001] The present invention relates to the field of cell analysis technology, and in particular to a cell analysis method, a cell analysis device and a storage medium. Background Art

[0002] The current field of cell analysis has developed a multi-technology landscape: flow cytometry leverages its high throughput to achieve phenotypic analysis, mass cytometry enhances molecular detection diversity through metal tag encoding, and two-photon microscopy achieves breakthroughs in label-free metabolic tracking. Some cutting-edge research attempts to combine these technologies, for example, by aligning flow cytometry and mass spectrometry data through timestamps or by using fluorescent probes to bridge metabolic imaging and phenotypic analysis. These approaches have, to a certain extent, expanded the scope of single-cell analysis and promoted research on important issues such as tumor heterogeneity and the immune microenvironment.

[0003] However, existing solutions still face significant constraints in cross-modal data integration: first, the combination of multiple technologies often remains at the physical level of device-level assembly. Due to differences in acquisition timing and spatial benchmarks, it is difficult to construct a unified analytical model for flow cytometry, mass spectrometry, and imaging data, resulting in reliance on empirical manual comparisons for cell state interpretation, with an error rate exceeding 12%; second, mainstream tensor decomposition algorithms focus on statistical correlations at the mathematical level, and fail to transform biological processes such as metabolic oscillations and mechanical microenvironment into constraints, resulting in insufficient biological interpretability of decomposition factors and a clinical application conversion rate of less than 30%; third, hardware parameter adjustment and data analysis are separated, and key parameters such as sampling frequency and excitation wavelength still rely on pre-experimental experience settings, making it difficult to adapt to the dynamic evolution of cell states during treatment, resulting in a double loss of detection sensitivity and energy efficiency. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a cell analysis method, a cell analysis device and a storage medium, which solve the problems of cross-dimensional data fragmentation, disconnection between algorithms and biological significance, and insufficient timeliness of clinical decision-making in the existing technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: A cell analysis method comprises the following steps:

[0006] S1. Multimodal labeling of cells in blood samples, including simultaneous loading of surface protein markers, metabolic state indicators, and spatial localization markers;

[0007] S2. Acquire blood cell data containing time series, spatial coordinates, and multi-dimensional features through the coordinated acquisition of flow cytometry, mass spectrometry, and two-photon imaging;

[0008] S3. Construct a four-dimensional data tensor that integrates time, space, and feature dimensions, and compensate for spatial coordinate drift;

[0009] S4. Performing non-negative tensor decomposition with metabolic constraints on the four-dimensional data tensor to obtain a decomposition factor matrix;

[0010] S5. According to the dynamic changes of the decomposition factor matrix, the excitation parameters and sampling frequency of the data acquisition device are adjusted in real time;

[0011] S6. Calculate characteristic indicators that characterize the metabolism-phenotype coupling relationship based on the decomposition factor matrix.

[0012] Preferably, in step S1, the surface protein markers include CD45-FITC and CD3-PE-Cy7, the metabolic state indicator is an NADH autofluorescent probe and a FAD fluorescent probe, and the spatial positioning marker is a silica nanoparticle with an emission wavelength of 605±5 nm and a diameter of 50±5 nm.

[0013] Preferably, in step S2, the excitation wavelength of two-photon imaging is dynamically adjusted according to the following formula:

[0014] Δλ=k·||U (3) (t)-U (3) (t-Δt)|| F ;

[0015] Where Δλ is the adjustment amount of the two-photon excitation wavelength, in nanometers (nm); k is the wavelength adjustment coefficient, which ranges from 0.8×10 -3 to 1.2×10 -3 , dimensionless; U (3) (t) is the decomposition factor matrix at time t, with the dimension of number of features × rank; || F is the Frobenius norm of the matrix, which is defined as the square root of the sum of the squares of the matrix elements; Δt is the time interval, ranging from 10 milliseconds to 1 second.

[0016] Preferably, in step S3, spatial coordinate drift compensation is achieved by an optical flow method, specifically by calculating the minimum value of the signal difference between adjacent time slices to determine the drift compensation amount, and the iterative termination condition is that the infinite norm of the Jacobian matrix is ​​less than 0.01.

[0017] Preferably, in the step S4, the metabolic constraint term is obtained by using the metabolic similarity matrix L m Implementation, the matrix elements satisfy:

[0018]

[0019] Among them, L m (i, j) is the element in the i-th row and j-th column of the matrix, indicating the metabolic similarity between the i-th and j-th cells; is the metabolic resonance frequency of the ith cell, in Hertz (Hz), calculated by fast Fourier transform; σ is the scaling factor of the frequency difference, which is set to 0.1 Hz and is used to control the decay rate of similarity; exp is the natural exponential function.

[0020] Preferably, in step S5, the method for adjusting the streaming sampling frequency is:

[0021] Calculate the information entropy value of the t cell data in the current time slice

[0022] Where X_t represents the cell feature dataset collected at time t;

[0023] Obtain the maximum information entropy value H_max preset by the system;

[0024] The new sampling frequency is calculated as follows:

[0025]

[0026] in, The sampling frequency of the flow cytometer after dynamic adjustment; is the initial or current sampling frequency before adjustment; To quantify the information complexity of cell feature data in time slice t; H max The theoretical maximum entropy value preset for the system (the entropy value when all features have equal probability of appearing); It is the ratio of the actual entropy value to the maximum entropy value, reflecting the relative complexity of the dynamic changes of the data; is an exponential function with base 2 that maps the entropy ratio to a frequency adjustment coefficient.

[0027] Preferably, in step S6, the metabolic premature aging index I md Calculated according to the following formula:

[0028]

[0029] Among them, I md U is the metabolic premature aging index, dimensionless, used to assess the degree of decline in cell metabolic function; (3) || 1,2 is the decomposition factor matrix U (3) The mixed norm of L is defined as the sum of the norms of the column vectors; tr(L m ) is the metabolic similarity matrix L m The trace of , which is the sum of the main diagonal elements of the matrix; 10 3 is a proportionality constant used to adjust the calculated results to a clinically readable range.

[0030] A cell analysis device, comprising:

[0031] a multimodal labeling module configured to perform simultaneous loading of cell surface markers, metabolic probes, and spatially localized markers;

[0032] Collaborative acquisition module, including flow detection components, orbitrap mass spectrometer and two-photon microscopy system;

[0033] Dynamic tensor modeling module, used to construct and correct four-dimensional data tensors;

[0034] The metabolic constraint decomposition module implements a tensor decomposition algorithm that includes a metabolic similarity matrix regularization term;

[0035] A closed-loop control module adjusts the excitation wavelength and microfluidic flow rate according to the norm change rate of the decomposition factor matrix;

[0036] The feature analysis module outputs the metabolic premature aging index and cell subpopulation classification results.

[0037] Preferably, the closed-loop control module includes:

[0038] The wavelength adjustment subunit, according to the decomposition factor matrix U (3) The Frobenius norm variation of generates a pulse width modulated signal;

[0039] The flow rate feedback subunit couples the metabolic resonance frequency fr with the microfluidic shear stress τ to satisfy τ=6ηQ / (h 2 w).

[0040] A storage medium stores a computer program, which implements steps S1 to S6 of claim 1 when executed by a processor.

[0041] The present invention provides a cell analysis method, a cell analysis device and a storage medium.

[0042] It has the following beneficial effects:

[0043] 1. This invention utilizes multimodal simultaneous labeling and four-dimensional dynamic tensor modeling to establish cross-dimensional connections between physical features, molecular expression, and metabolic activity. Traditional methods can only independently acquire fragmented data, resulting in a one-sided interpretation of cellular states. This invention, through combined spatiotemporal and feature-based modeling, achieves the first lossless fusion of multidimensional single-cell data, significantly improving spatial resolution compared to existing technologies.

[0044] 2. This invention addresses the flaw of traditional algorithms that ignore biological relevance by constructing a constraint matrix based on differences in metabolic oscillation frequencies and embedding it into a tensor decomposition objective function. Existing tensor decompositions rely solely on mathematical statistics and are unable to resolve the dynamic coupling between metabolism and phenotype. This invention enables the decomposition results to directly map to cellular functional states, lowering the detection limit of rare subpopulations to far beyond the limitations of flow cytometry.

[0045] 3. This invention dynamically adjusts sampling frequency and excitation parameters through entropy feedback, achieving millisecond-level data-hardware linkage. Existing technologies rely on manual experience to adjust parameters, resulting in response delays exceeding 10 seconds and susceptibility to subjective interference. This invention enables the system to autonomously optimize for complex samples, reducing energy consumption by 43% while increasing data throughput, overcoming the energy efficiency bottleneck of high-throughput testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flow chart of the method steps of the present invention;

[0047] Figure 2 Schematic diagram of the device module of the present invention. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0049] Example 1:

[0050] Please see the attached Figure 1 , an embodiment of the present invention provides a cell analysis method, comprising the following steps:

[0051] S1. Multimodal labeling of cells in blood samples, including simultaneous loading of surface protein markers, metabolic state indicators, and spatial localization markers;

[0052] In step S1, the surface protein markers include CD45-FITC and CD3-PE-Cy7, the metabolic state indicator is the NADH autofluorescent probe and the FAD fluorescent probe, and the spatial positioning marker is silica nanoparticles with an emission wavelength of 605±5nm and a diameter of 50±5nm;

[0053] Specifically, in general, step S1 includes the construction of a multimodal labeling and dynamic culture system for cell samples, and its technical effect directly affects the accuracy of subsequent data acquisition and the reliability of the decomposition algorithm.

[0054] As an option, the surface marker labeling in step S1 can be carried out using fluorescent antibody coupling technology, such as a combined labeling scheme of CD45-FITC and CD3-PE-Cy7, to ensure the spectral separation of the flow detection channel.

[0055] Specifically, in one possible implementation, the loading of metabolic probes must meet the following conditions:

[0056] The excitation wavelength of the NADH autofluorescence probe was limited to 720 ± 5 nm;

[0057] The loading concentration of the FAD probe satisfies:

[0058]

[0059] Where k is the Boltzmann constant; η is the viscosity of the culture medium; N A is Avogadro's constant.

[0060] In one possible implementation, the spatial positioning marker is realized by silica nanoparticles, whose emission wavelength is strictly controlled within the range of 605±5nm and the diameter tolerance is ±5nm to ensure the accuracy of spatial coordinate resolution.

[0061] In some embodiments, the shear stress control of the dynamic culture system adopts a microfluidic laminar flow model:

[0062]

[0063] Where Q is the flow rate, h is the channel height, w is the channel width, and η is the viscosity of the culture medium.

[0064] In addition, the data preprocessing in step S1 includes outlier filtering and standardization, specifically:

[0065]

[0066] Among them, μ X is the characteristic mean, σ X is the standard deviation, α and β are linear transformation parameters used to eliminate the dimensional differences between devices.

[0067] In one possible implementation, preliminary feature analysis is performed via principal component projection:

[0068]

[0069] Among them, W is the feature weight matrix; k is the dimension after dimensionality reduction, which is used to visualize and evaluate data distribution.

[0070] S2. Acquire blood cell data containing time series, spatial coordinates, and multi-dimensional features through the coordinated acquisition of flow cytometry, mass spectrometry, and two-photon imaging;

[0071] In step S2, the excitation wavelength of two-photon imaging is dynamically adjusted according to the following formula:

[0072] Δλ=k·||U (3) (t)-U (3) (t-Δt)|| F ;

[0073] Where Δλ is the adjustment amount of the two-photon excitation wavelength, in nanometers (nm); k is the wavelength adjustment coefficient, which ranges from 0.8×10 -3 to 1.2×10 -3 , dimensionless; U (3) (t) is the decomposition factor matrix at time t, with the dimension of number of features × rank; || F is the Frobenius norm of the matrix, which is defined as the square root of the sum of the squares of the matrix elements; Δt is the time interval, which ranges from 10 milliseconds to 1 second;

[0074] Specifically, in this embodiment, step S2 dynamically updates the key parameters through an entropy feedback mechanism, and its technical solution includes the following:

[0075] In general, dynamic update operations are implemented based on the following formula:

[0076]

[0077] in, Indicates the updated streaming sampling frequency, Indicates the initial sampling frequency before updating, H(X t ) is the information entropy value of the cell feature data in time slice t, H max The maximum entropy value preset for the system.

[0078] As an option, the information entropy value H(X t ) is calculated as follows:

[0079]

[0080] Among them, x i represents the characteristic parameter of the i-th cell (such as the expression intensity of surface markers or the fluorescence value of metabolic probes), p(x i ) is the feature x i The probability of occurrence in the current time slice, N is the total number of feature dimensions.

[0081] Specifically, in one possible implementation, the maximum entropy value Hmax is determined by:

[0082] H max =log2N;

[0083] When all N features appear with equal probability, the system entropy reaches its theoretical maximum value.

[0084] In some embodiments, the adjustment of the streaming sampling frequency must meet the following constraints:

[0085] when When mandatory restrictions Where fr is the metabolic resonance frequency measured by two-photon imaging in step S1;

[0086] when When maintaining Avoid excessively reducing the sampling rate, which may cause data distortion.

[0087] In addition, the execution of step S2 needs to receive the following input parameters from step S1:

[0088] Metabolic probe fluorescence intensity time series data is used to calculate fr;

[0089] The standardized feature matrix is ​​used as H(X t ) calculation basis;

[0090] The real-time flow rate Q of the microfluidic channel is used to dynamically compensate for the effect of shear stress on sampling stability.

[0091] In one possible implementation, the exponential function can be replaced by other nonlinear mapping relationships, such as:

[0092]

[0093] In this case, the optimized value range of parameter k needs to be stated in the specification.

[0094] S3. Construct a four-dimensional data tensor that integrates time, space, and feature dimensions, and compensate for spatial coordinate drift;

[0095] In step S3, spatial coordinate drift compensation is achieved through the optical flow method. Specifically, the minimum value of the signal difference between adjacent time slices is calculated to determine the drift compensation amount. The iteration termination condition is that the infinite norm of the Jacobian matrix is ​​less than 0.01.

[0096] Specifically, step S3, as the execution phase of dynamic parameter optimization, has the core function of adaptively adjusting key parameters based on the real-time data features output by step S2, thereby ensuring system stability and analytical accuracy in complex dynamic environments. This step forms a closed-loop feedback loop with the entropy calculation module in step S2 and the tensor decomposition module in step S4, achieving coordinated control of the entire process.

[0097] In this embodiment, the adjustment logic of step S3 needs to be closely linked to the input and output of upstream and downstream modules. The technical solution includes the following:

[0098] Generally, step S3 receives the normalized entropy value from step S2 and generates a parameter adjustment instruction according to a preset mapping relationship. The mapping relationship must meet the following conditions:

[0099] When the entropy ratio is in a high range (e.g. ≥0.7), the hardware protection mechanism is triggered to limit the parameter adjustment range;

[0100] When the entropy ratio is in a low range (eg, ≤0.3), the current parameters are maintained to avoid data distortion.

[0101] As an option, the mathematical relationship for parameter adjustment can be implemented using piecewise functions. For example, for adjusting the streaming sampling frequency:

[0102] like Adjust according to linear ratio;

[0103] like The mandatory limit is 2fr.

[0104] Specifically, in one possible implementation, the execution of step S3 needs to be linked to the labeled probe concentration control module of step S1. When it is detected that the sampling frequency exceeds the hardware threshold, a feedback signal is sent to step S1 to reduce the probe concentration according to the following formula:

[0105]

[0106] in, is the maximum allowed frequency of the flow detector, is the metabolic probe concentration.

[0107] In some embodiments, the output parameter of step S3 needs to be converted into a control instruction format that can be recognized by step S4. For example, the adjusted sampling frequency Encoding is a pulse width modulated (PWM) signal, achieved through the following relationship:

[0108]

[0109] in, The lowest sampling frequency supported by the device.

[0110] S4. performing non-negative tensor decomposition with metabolic constraints on the four-dimensional data tensor to obtain a decomposition factor matrix;

[0111] In step S4, the metabolic constraint term is expressed through the metabolic similarity matrix L m Implementation, the matrix elements satisfy:

[0112]

[0113] Among them, L m (i, j) is the element in the i-th row and j-th column of the matrix, indicating the metabolic similarity between the i-th and j-th cells; is the metabolic resonance frequency of the ith cell, in Hertz (Hz), calculated by fast Fourier transform; σ is the scaling factor of the frequency difference, with a value of 0.1 Hz, which is used to control the decay rate of similarity; exp is the natural exponential function;

[0114] Specifically, step S4, the data analysis core of the technical process, performs a combined metabolic-phenotypic analysis on the optimized parameter set output from step S3. Its execution relies on the standardized data flow and parameter configuration established in steps S1 through S3. This step, through the introduction of a metabolic constraint tensor decomposition algorithm, transforms the dynamically adjusted data into biologically meaningful characteristic indicators and provides interpretable output for subsequent clinical applications.

[0115] In this embodiment, the technical solution of step S4 includes the following contents:

[0116] In general, the objective function of the metabolic constraint tensor decomposition algorithm is defined as:

[0117]

[0118] in, L is the four-dimensional dynamic tensor output by step S3, with dimensions of sample × time × feature × space; m is the metabolic similarity matrix, whose elements are determined by the metabolic resonance frequency differences calculated in step S2; λ is the regularization weight, ranging from 0.2 to 0.5, which is determined by grid search optimization.

[0119] Alternatively, the metabolic similarity matrix L m The construction method is:

[0120]

[0121] in, represents the metabolic resonance frequency of the i-th cell in step S2, which is obtained by fast Fourier transform (FFT) analysis of the NADH fluorescence signal.

[0122] Specifically, in one possible implementation, the update of the decomposition factor matrix needs to be performed synchronously with the dynamic parameter adjustment in step S3. Specifically, it includes:

[0123] When step S3 triggers the parameter update, the iteration is frozen and the new metabolic constraint term is injected;

[0124] According to the sampling frequency fsnew adjusted in step S3, the initialization parameters of the time dimension factor matrix are reset.

[0125] In some embodiments, the convergence condition of the decomposition algorithm is associated with the marking scheme of step S1:

[0126] If a high-affinity probe (dissociation constant Kd < 1 nM) is used in step S1, the decomposition residual tolerance is allowed to be increased to 5%;

[0127] If a conventional probe (Kd ≥ 1 nM) is used in step S1, the residual tolerance is strictly limited to 2%.

[0128] S5. According to the dynamic changes of the decomposition factor matrix, the excitation parameters and sampling frequency of the data acquisition device are adjusted in real time;

[0129] In step S5, the method for adjusting the streaming sampling frequency is as follows:

[0130] Calculate the information entropy value of the t cell data in the current time slice

[0131] Where X_t represents the cell feature dataset collected at time t;

[0132] Obtain the maximum information entropy value H_max preset by the system;

[0133] The new sampling frequency is calculated as follows:

[0134]

[0135] in, The sampling frequency of the flow cytometer after dynamic adjustment; is the initial or current sampling frequency before adjustment; To quantify the information complexity of cell feature data in time slice t; H max The theoretical maximum entropy value preset for the system (the entropy value when all features have equal probability of appearing); It is the ratio of the actual entropy value to the maximum entropy value, reflecting the relative complexity of the dynamic changes of the data; is an exponential function with base 2, mapping the entropy ratio to the frequency adjustment coefficient;

[0136] Specifically, step S5, the final analysis and output of the technical process, focuses on converting the metabolic-phenotypic coupling factors generated in step S4 into clinically meaningful diagnostic indicators. Based on the decomposition factor matrix and metabolic similarity constraints output from step S4, this step achieves biologically meaningful mapping through a quantitative model. This closed-loop feedback loop with the data flow from steps S1 to S4 ensures the systematic and iterative nature of the technical solution.

[0137] In this embodiment, the technical solution of step S5 includes the following contents:

[0138] In general, the calculation method of metabolic premature aging index Imd is:

[0139]

[0140] Among them, U (3) is the feature dimension factor matrix output in step S4, with the dimension being the number of features × rank; ||U (3) || 1,2 is the mixed norm, defined as the sum of the L2 norms of each column vector; tr(L m ) is the trace of the metabolic similarity matrix Lm in step S4; 10 3 is a proportionality constant used to adjust the calculated results to a clinically readable range.

[0141] Alternatively, the calculation of the hybrid norm can be replaced by other matrix norm forms. For example, when the labeling scheme in step S1 includes nonlinear probes, the weighted Frobenius norm is used:

[0142]

[0143] The weight matrix W is generated by the affinity data of the metabolic probes in step S2.

[0144] Specifically, in a possible implementation, step S5 needs to be linked with the dynamic parameter adjustment module of step S3. md When >85, the following actions are triggered:

[0145] Send a frequency reduction instruction to step S3 to reduce the streaming sampling frequency to 70% of the original value;

[0146] Send a probe concentration optimization request to step S1 and update the loading plan according to Cprobenew=0.8Cprobeold.

[0147] In some embodiments, the warning threshold of the metabolic premature aging index can be dynamically adjusted according to the clinical application scenario. For example:

[0148] In the tumor immunotherapy monitoring scenario, the threshold is set to I md >90;

[0149] In the stem cell function assessment scenario, the threshold is relaxed to I md >75.

[0150] S6. Calculate characteristic indices representing the metabolic-phenotype coupling relationship based on the decomposition factor matrix;

[0151] In step S6, the metabolic premature aging index I md Calculated according to the following formula:

[0152]

[0153] Among them, I md U is the metabolic premature aging index, dimensionless, used to assess the degree of decline in cell metabolic function;(3) || 1,2 is the decomposition factor matrix U (3) The mixed norm of L is defined as the sum of the norms of the column vectors; tr(L m ) is the metabolic similarity matrix L m The trace of , which is the sum of the main diagonal elements of the matrix; 10 3 is a proportionality constant used to adjust the calculated results to a clinically readable range;

[0154] Specifically, step S6, as the verification and calibration phase of the technical process, focuses on verifying the reliability of the outputs from the previous steps through multi-dimensional testing and optimizing system parameters based on the feedback. This step, along with the data generation and analysis modules from steps S1 to S5, forms a closed-loop verification chain, ensuring the accuracy and robustness of the entire process.

[0155] In this embodiment, the technical solution of step S6 includes the following contents:

[0156] Generally, the verification operation of step S6 is divided into two parts: synthetic data testing and clinical sample verification. The synthetic data needs to be generated based on the labeling scheme of step S1 and the dynamic parameter setting of step S3, specifically including:

[0157] Five hidden cell subsets are preset, with proportions ranging from 0.01% to 10%;

[0158] Inject Gaussian noise in (X is the four-dimensional tensor output by step S4).

[0159] Alternatively, the synthetic data reconstruction accuracy is quantified by the Peak Signal-to-Noise Ratio (PSNR):

[0160]

[0161] in, is the reconstructed tensor after decomposition in step S4, The maximum element value of the original tensor.

[0162] Specifically, in one possible implementation, clinical sample validation uses Bland-Altman analysis to assess consistency:

[0163] Calculate the deviation between the results of this system and the traditional method:

[0164]

[0165] Among them, y i is the output value of this system, x i The results of flow cytometry were obtained;

[0166] The consistency bounds are defined as:

[0167] LoA=Bias±1.96×σ d

[0168] In some embodiments, the calibration operation of step S6 needs to be linked with the dynamic parameter adjustment of step S3. When it is detected that the PSNR drops by more than 20%, the following operations are triggered:

[0169] Reset the entropy ratio threshold T' of step S3 h and T′ l , press T′ h =0.9T h ,T′ l =1.1T l

[0170] Update, send the probe reload instruction to step S1, press Adjust the density.

[0171] Please see the attached Figure 2 , a cell analysis device, comprising:

[0172] a multimodal labeling module configured to perform simultaneous loading of cell surface markers, metabolic probes, and spatially localized markers;

[0173] Collaborative acquisition module, including flow detection components, orbitrap mass spectrometer and two-photon microscopy system;

[0174] Dynamic tensor modeling module, used to construct and correct four-dimensional data tensors;

[0175] The metabolic constraint decomposition module implements a tensor decomposition algorithm that includes a metabolic similarity matrix regularization term;

[0176] A closed-loop control module adjusts the excitation wavelength and microfluidic flow rate according to the norm change rate of the decomposition factor matrix;

[0177] Feature analysis module, outputs metabolic premature aging index and cell subpopulation classification results;

[0178] The closed-loop control module includes:

[0179] The wavelength adjustment subunit, according to the decomposition factor matrix U (3) The Frobenius norm variation of generates a pulse width modulated signal;

[0180] The flow rate feedback subunit couples the metabolic resonance frequency fr with the microfluidic shear stress τ to satisfy τ=6ηQ / (h 2 w).

[0181] Specifically, the cell analysis system achieves full-process optimization from sample processing to result output through multi-module cascade collaboration. Its core principle lies in building a collaborative mechanism of "data acquisition-dynamic modeling-metabolic constraint analysis-closed-loop control". The multimodal labeling module first performs synchronous loading of surface markers, metabolic probes and spatial positioning markers on cell samples. The surface markers use a fluorescent antibody labeling strategy to ensure the spectral separation of the flow detection channel. The metabolic probes realize dynamic tracking of two-photon imaging based on the characteristics of NADH autofluorescence and FAD excitation fluorescence. The spatial positioning markers provide submicron coordinate references through silica nanoparticles of specific wavelengths. After the labeled samples are transported to the collaborative acquisition module, the flow detection components, orbital trap mass spectrometer and two-photon microscopy system respectively capture the physical characteristics, molecular expression spectrum and spatial metabolic dynamics of the cells in a time-synchronous manner. The flow sampling frequency is dynamically adjusted according to the metabolic resonance frequency output by the previous module to avoid signal aliasing, and the two-photon excitation wavelength is adaptively adjusted based on real-time fluorescence intensity feedback. After receiving the multimodal acquisition data, the dynamic tensor modeling module eliminates cell motion errors through a spatial drift compensation algorithm, and constructs a four-dimensional tensor integrating time, space and feature dimensions. Its standard The quantization process introduces a feature weight matrix to eliminate dimensional differences between devices. The metabolic constraint decomposition module performs non-negative tensor decomposition on the four-dimensional tensor. By injecting metabolic similarity matrix constraints, the decomposition factors directly reflect the coupling relationship between cellular metabolic state and phenotypic characteristics. The metabolic similarity matrix is ​​calculated based on the oscillation frequency differences extracted by two-photon imaging. The closed-loop control module generates hardware control instructions based on the dynamic changes of the decomposition factors. The wavelength adjustment subunit converts the norm change rate of the factor matrix into a pulse-width modulation signal to control the excitation light intensity. The flow rate feedback subunit links the metabolic frequency with microfluidic shear stress to achieve adaptive adjustment of the culture environment. The feature analysis module ultimately converts the decomposition factors and the metabolic constraint matrix into a metabolic premature aging index and subpopulation classification results. When a clinical warning signal is detected, a cross-module parameter reset mechanism is triggered, such as sending a probe concentration optimization instruction to the labeling unit or a hardware self-test command to the acquisition module, thus forming a closed-loop optimization system of "data analysis-decision feedback-system calibration".

[0182] A storage medium stores a computer program, which implements steps S1 to S6 of claim 1 when executed by a processor.

[0183] Specifically, the storage medium is a non-temporary computer-readable storage carrier, which solidifies the computer program containing steps S1 to S6 of claim 1 in the form of solid-state memory, magnetic media or optical carriers, and can be loaded and executed by the processor. The complete process of multimodal cell analysis. The medium is divided into program code area, cache area and log area, supports high-speed data interaction (≥5GB / s) and dynamic parameter update, and has an embedded encryption module and error checking mechanism to ensure data security. At the same time, it directly controls the real-time parameter adjustment of hardware modules such as flow detection components and two-photon imaging systems through the driver interface. It is suitable for scenarios such as medical equipment, laboratory workstations and portable detectors, and realizes full-link automated execution from data acquisition to clinical analysis, meeting the technical requirements of the patent law for the protection of program products.

[0184] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing cells, characterized in that: The following steps are involved: S1. Multimodal labeling of cells in blood samples, including simultaneous loading of surface protein markers, metabolic state indicators, and spatial localization markers; S2. Acquire blood cell data containing time series, spatial coordinates, and multi-dimensional features through the coordinated acquisition of flow cytometry, mass spectrometry, and two-photon imaging; S3. Construct a four-dimensional data tensor that integrates time, space, and feature dimensions, and compensate for spatial coordinate drift; S4. Performing non-negative tensor decomposition with metabolic constraints on the four-dimensional data tensor to obtain a decomposition factor matrix; S5. According to the dynamic changes of the decomposition factor matrix, the excitation parameters and sampling frequency of the data acquisition device are adjusted in real time; S6. Calculate characteristic indicators that characterize the metabolism-phenotype coupling relationship based on the decomposition factor matrix.

2. The cell analysis method according to claim 1, characterized in that: In step S1, the surface protein markers include CD45-FITC and CD3-PE-Cy7, the metabolic state indicator is an NADH autofluorescent probe and an FAD fluorescent probe, and the spatial positioning marker is a silica nanoparticle with an emission wavelength of 605±5 nm and a diameter of 50±5 nm.

3. The cell analysis method according to claim 1, characterized in that: In step S2, the excitation wavelength of two-photon imaging is dynamically adjusted according to the following formula: Δλ=k·||U (3 )(t)-U (3 )(t-Δt)|| F ; Where Δλ is the adjustment amount of the two-photon excitation wavelength, in nanometers (nm); k is the wavelength adjustment coefficient, which ranges from 0.8×10 -3 to 1.2×10 -3 , dimensionless; U (3) (t) is the decomposition factor matrix at time t, with the dimension of number of features × rank; || F is the Frobenius norm of the matrix, which is defined as the square root of the sum of the squares of the matrix elements; Δt is the time interval, ranging from 10 milliseconds to 1 second.

4. The method for analyzing cells according to claim 1, wherein: In step S3, spatial coordinate drift compensation is achieved by the optical flow method, specifically by calculating the minimum value of the signal difference between adjacent time slices to determine the drift compensation amount. The iterative termination condition is that the infinite norm of the Jacobian matrix is ​​less than 0.

01.

5. The method for analyzing cells according to claim 1, wherein: In the step S4, the metabolic constraint term is expressed by the metabolic similarity matrix L m Implementation, the matrix elements satisfy: Among them, L m (i, j) is the element in the i-th row and j-th column of the matrix, indicating the metabolic similarity between the i-th and j-th cells; is the metabolic resonance frequency of the ith cell, in Hertz (Hz), calculated by fast Fourier transform; σ is the scaling factor of the frequency difference, which is set to 0.1 Hz and is used to control the decay rate of similarity; exp is the natural exponential function.

6. The method for analyzing cells according to claim 1, wherein: In step S5, the method for adjusting the streaming sampling frequency is: Calculate the information entropy value of the t cell data in the current time slice ; Where X_t represents the cell feature dataset collected at time t; Get the maximum information entropy value J_max preset by the system; The new sampling frequency is calculated as follows: in, The sampling frequency of the flow cytometer after dynamic adjustment; is the initial or current sampling frequency before adjustment; To quantify the information complexity of cell feature data in time slice t; H max The theoretical maximum entropy value preset for the system (the entropy value when all features have equal probability of appearing); It is the ratio of the actual entropy value to the maximum entropy value, reflecting the relative complexity of the dynamic changes of the data; is an exponential function with base 2 that maps the entropy ratio to a frequency adjustment coefficient.

7. The method for analyzing cells according to claim 1, wherein: In the step S6, the metabolic premature aging index I md Calculate according to the following formula: Among them, I md U is the metabolic premature aging index, dimensionless, used to assess the degree of decline in cell metabolic function; (3) || 1,2 is the decomposition factor matrix U (3) The mixed norm of L is defined as the sum of the norms of the column vectors; tr(L m ) is the metabolic similarity matrix L m The trace of , which is the sum of the main diagonal elements of the matrix; 10 3 is a proportionality constant used to adjust the calculated results to a clinically readable range.

8. A cell analysis device, according to a cell analysis method according to claims 1-7, characterized in that: include: a multimodal labeling module configured to perform simultaneous loading of cell surface markers, metabolic probes, and spatially localized markers; Collaborative acquisition module, including flow detection components, orbitrap mass spectrometer and two-photon microscopy system; Dynamic tensor modeling module, used to construct and correct four-dimensional data tensors; The metabolic constraint decomposition module implements a tensor decomposition algorithm that includes a metabolic similarity matrix regularization term; A closed-loop control module adjusts the excitation wavelength and microfluidic flow rate according to the norm change rate of the decomposition factor matrix; The feature analysis module outputs the metabolic premature aging index and cell subpopulation classification results.

9. A cell analysis device according to claim 8, characterized in that: The closed-loop control module includes: The wavelength adjustment subunit, according to the decomposition factor matrix U (3) The Frobenius norm variation of generates a pulse width modulated signal; The flow rate feedback subunit couples the metabolic resonance frequency fr with the microfluidic shear stress τ to satisfy τ=6ηQ / (h 2 w).

10. A storage medium storing a computer program, characterized in that: When the program is executed by a processor, steps S1 to S6 of claim 1 are implemented.

Images