Cell beating intelligent panoramic dynamic analysis method and system

Abstract

The application relates to a cell beating intelligent panoramic dynamic analysis method and system, relates to the fields of biomedical engineering and cell function analysis technology, realizes automatic division of a single cell beating area through time sequence lossless compression, convolution denoising and unsupervised clustering, calculates a beating direction in combination with a self-developed centric trajectory quantile analysis method, extracts multi-scale deep features by using continuous wavelet transformation and one-side Laplace transformation, and constructs a standardized quality evaluation system containing three biological interpretability indexes. The application can simultaneously complete panoramic heterogeneity analysis of hundreds of single cells, extract more than ten characteristic parameters, and has an analysis efficiency improvement of more than 90%, an accuracy of more than 95%, and is suitable for multiple types of cells and conventional microscopes, and can be applied to multiple scenes such as drug cardiotoxicity screening and stem cell differentiation evaluation.

Description

Technical Field

[0001] This invention relates to the fields of biomedical engineering and cell function analysis technology, and in particular to a method and system for intelligent panoramic dynamic analysis of cell pulsation. Background Technology

[0002] Driven by both life science research and precision medicine development, dynamic monitoring of cell function has become a core element in elucidating physiological mechanisms, promoting disease model construction, and drug development. Cell pulsation, as a key physiological indicator reflecting cell activity and functional coordination, is widely used in cardiovascular disease research, stem cell differentiation assessment, and drug cardiotoxicity screening. The accuracy and efficiency of its analysis directly determine the depth and translational speed of related research. In cell pulsation-related research, universities and research institutions rely on manual observation or traditional analysis software, which suffers from problems such as limited field of view, low signal-to-noise ratio, low analytical efficiency, limited characteristic parameters, and dependence on fluorescent labeling, leading to long research cycles and poor reproducibility of conclusions.

[0003] Existing cell pulsation analysis techniques are mainly divided into two categories: manual annotation and semi-automatic algorithms. Manual annotation relies on manual labeling of cell pulsation peaks frame by frame, requiring 1-3 hours for a single sample analysis. The annotation error rate can reach 15%-20% depending on the person labeling, resulting in low efficiency and strong subjectivity, making it difficult to meet the standardized analysis needs of large-scale samples. Semi-automatic algorithms mostly extract some parameters based on changes in single gray values, with weak ability to capture weak pulsation signals and an analysis accuracy of only 70%-80%. Moreover, the imaging and analysis modules of most systems are independent, requiring image data to be acquired first and then imported into third-party software for offline analysis, making it impossible to achieve real-time feedback and failing to meet the immediate observation needs of dynamic control experiments. For example, the MUSCLEMOTION tool developed by Sala et al., as a mainstream open-source analysis tool, has achieved a certain degree of automated analysis, but it can only extract basic contraction amplitude and frequency parameters and cannot analyze the deep dynamic characteristics of cell pulsation. The Cardio Analyser software proposed by Stummann et al. can only complete basic quantification of pulsation frequency and pulsation region and cannot distinguish the heterogeneity of cell populations.

[0004] In addition, traditional fluorescent labeling detection methods rely on fluorescent dyes such as Cal-520 for staining and detect cell pulsation through fluorescent signals. On the one hand, fluorescent dyes can stimulate cells and affect the cell's own state; on the other hand, they can only detect the overall pulsation signal and cannot accurately distinguish the pulsation state of individual cells, resulting in a mixture of background, real pulsating cells and non-pulsating cells, which greatly reduces the detection capability of real signals. Moreover, only three parameters, amplitude, frequency and period, can be analyzed, which cannot provide a comprehensive feature description of cell pulsation. The review study by Dou et al. [1] also clearly pointed out that the current myocardial contractile function detection platform still has the core bottleneck of limited field of view and insufficient parameter dimensions. The observation field of existing systems is generally limited to local areas of ≤0.5mm², which cannot reflect the heterogeneity and overall coordination of cell populations, resulting in deviations in detection results and failing to meet the needs of long-term dynamic tracking of stem cell differentiation over several days to several weeks.

[0005] Existing related patents and commercial systems also have significant technical limitations. The method for detecting the mechanical pulsation of primary cardiomyocytes disclosed in patent CN116338155A requires dedicated mechanical detection hardware and can only detect single cells, failing to achieve panoramic analysis of the heterogeneity of the entire cell population. The pulsation detection method based on image entropy disclosed in patent CN110706231B can only extract basic frequency and amplitude parameters, unable to analyze the deep dynamic features of cell pulsation. The earlier patent CN200710068077.7 can only analyze morphological parameters such as cell perimeter and area, with limited parameter dimensions and unable to achieve automated real-time analysis. The commercially available RTCA Cardio system relies on customized microplates with micro-gold electrodes, resulting in high hardware costs and the ability to acquire only overall impedance signals, failing to achieve single-cell-level pulsation feature analysis. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention discloses a method and system for intelligent panoramic dynamic analysis of cell pulsation.

[0007] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0008] A method and system for intelligent panoramic dynamic analysis of cell pulsation, comprising the following steps:

[0009] S1. Perform time-series lossless compression on the acquired three-dimensional temporal image data of cells to convert it into two-dimensional planar data that simultaneously retains spatial distribution features and temporal variation features;

[0010] S2. Use a filter convolution kernel to perform convolution noise reduction on the compressed two-dimensional image to eliminate isolated noise points;

[0011] S3. An unsupervised clustering algorithm is used to automatically divide the denoised image into regions, distinguishing between pulsation-positive regions and background regions, and counting the number and area ratio of pulsation regions.

[0012] S4. For each pulsation-positive region, the direction of cell pulsation is calculated using the centroid trajectory quantile analysis method;

[0013] S5. Extract the time-domain pulsation signal of each pulsation positive region, extract the time-frequency features through continuous wavelet transform, and analyze the signal attenuation coefficient through one-sided Laplace transform to quantify the cell repolarization rate.

[0014] S6. Summarize and statistically analyze the characteristic parameters of all pulsation-positive regions, and evaluate the signal quality of the analysis results based on biological interpretability indicators;

[0015] S7. Output a comprehensive analysis report containing multi-dimensional pulsation characteristics and standardized quality assessment results.

[0016] Preferably, the calculation formula for the time-series lossless compression in step S1 is:

[0017]

[0018] in, Let be the image matrix of the t-th frame. For two-dimensional spatial convolution operations, This is an element-wise multiplication operation. This is an element-wise division operation. , For spatial neighborhood parameters, This is the compressed two-dimensional matrix.

[0019] Preferably, the unsupervised clustering algorithm in step S3 is the K-means clustering algorithm, and the number of clusters K ranges from 2 to 200, which is used to adapt to samples with different cell densities.

[0020] Preferably, the specific steps of the centroid trajectory quantile analysis method in step S4 are as follows: calculate the 25%-75% quantile centroid coordinates of each frame image within the pulsating positive region, and obtain the direction slope of cell pulsation by fitting the centroid trajectory through Theil-Sen regression.

[0021] Preferably, the biological interpretability signal quality assessment index in step S6 includes the pulsatility consistency index, the pulsatility signal-to-noise ratio index, and the parameter stability index.

[0022] Preferably, the pulsation consistency index is the standard deviation of the pulsation phase difference between different pulsation-positive regions; the pulsation signal-to-noise ratio index is the ratio of the signal variance of the pulsation-positive region to the signal variance of the background region; and the parameter stability index is the coefficient of variation of the key parameters of pulsation frequency and amplitude within the same continuous monitoring period.

[0023] Preferably, the cells include at least one of cardiomyocytes, nerve cells, tumor cells, and stem cells.

[0024] Preferably, the three-dimensional time-series image data of cells in step S1 is acquired by a bright-field microscope, a phase-contrast microscope, or a high-throughput microscopy imaging device, supporting long-term continuous dynamic monitoring from 1 minute to 30 days.

[0025] Preferably, it can be applied to at least one of drug cardiotoxicity screening, stem cell differentiation assessment, cardiovascular disease model construction, or tumor cell function research, and supports real-time output of analysis results to observe the immediate dynamic response of drug intervention.

[0026] A cell pulsation intelligent panoramic dynamic analysis system, characterized in that it includes a data acquisition module, a time-series compression and noise reduction module, an unsupervised region division module, a multi-scale feature extraction module, a standardized quality assessment module, and a result output module that are connected in sequence.

[0027] The temporal compression and noise reduction module has a built-in temporal lossless compression algorithm unit and a mean filter convolution kernel unit, which are used to perform lossless compression of three-dimensional temporal data and image noise suppression, respectively.

[0028] The unsupervised region segmentation module has a built-in K-means clustering algorithm unit. The number of clusters K can be dynamically adjusted in the range of 2-200. It is used to automatically identify pulsation-positive regions and background regions and count their number and area ratio.

[0029] The multi-scale feature extraction module incorporates a CTQA pulsation direction calculation unit, a continuous wavelet transform unit, and a one-sided Laplace transform unit, which are used to calculate the cell pulsation direction, extract time-frequency features, and analyze the cell repolarization attenuation coefficient, respectively.

[0030] The standardized quality assessment module has a built-in biological interpretability index calculation unit, which is used to calculate the pulsation consistency index, pulsation signal-to-noise ratio index, and parameter stability index.

[0031] The results output module is used to generate a panoramic analysis report that includes single-cell-level multidimensional pulsation characteristics and standardized quality assessment results.

[0032] By employing the technical solution described above, the present invention has the following beneficial effects:

[0033] (1) This invention, through the deep integration of multiple technologies, achieves for the first time a panoramic automated analysis of cell pulsation under bright field conditions, while filling the gap in the field of standardized signal quality assessment. No dedicated mechanical detection hardware is required; analysis can be achieved by adapting to conventional microscopy equipment, and panoramic analysis of hundreds of single cells can be completed simultaneously, overcoming the limitation of this patent which can only detect single cells. Deep dynamic features such as attenuation coefficients can be extracted, with parameter dimensions more than three times that of this patent, providing a more comprehensive description of cell pulsation. Fully automated real-time analysis is achieved without manual intervention, improving analysis efficiency by more than 90%, and distinguishing the heterogeneity of cell populations, overcoming the limitation of this patent which can only analyze overall morphological parameters. No customized microplates are required, reducing hardware costs by more than 90%, and single-cell-level pulsation feature analysis is achieved, far superior to the system's ability to only acquire overall impedance signals.

[0034] (2) The present invention can directly perform cell pulsation analysis based on bright field images, avoiding the stimulation of cells by fluorescent dyes, ensuring that the original state of cells is not affected, reducing experimental costs, solving the cytotoxicity problem of traditional fluorescence methods, and supporting long-term cell dynamic monitoring for several days to several weeks.

[0035] (3) This invention breaks through the limitations of the local field of view of traditional technology, and can realize global cell pulsation analysis under a large field of view, fully reflecting the heterogeneity and overall coordination of the cell population, and avoiding the result deviation caused by local detection.

[0036] (4) This invention achieves fully automated analysis without manual annotation. The analysis time for a single sample can be shortened to minutes, which is far better than the 1-3 hours of traditional manual analysis. The analysis accuracy can reach more than 95%, which is far higher than the 70%-80% of existing semi-automatic algorithms. It eliminates the subjective error of manual annotation and greatly improves the analysis efficiency of large-scale samples.

[0037] (5) The present invention can extract more than ten characteristic parameters such as the number of pulsating cells, the proportion of positive area, pulsation intensity, frequency, direction, duration, and attenuation coefficient, which is far more than the three parameters of the traditional method and is also superior to the parameter dimensions of the existing mainstream tools such as MUSCLEMOTION. It can comprehensively and accurately describe the pulsation state of cells and provide richer quantitative evidence for related research.

[0038] (6) This invention establishes for the first time a standardized signal quality assessment system for biological interpretability, which can quantify and verify the reliability of the analysis results, solves the defect of existing technologies that cannot assess the credibility of analysis results, and provides reliable quality assurance for high-precision research.

[0039] (7) This invention has strong versatility and can be adapted to various cell types such as cardiomyocytes, nerve cells, and tumor cells. It can be adapted to various conventional microscopy equipment such as bright field and phase contrast. It can be applied to multiple scenarios such as drug cardiotoxicity screening, stem cell differentiation assessment, cardiovascular disease model construction, and long-term cell dynamic monitoring. It covers most of the current application needs of cell pulsation analysis. At the same time, the signal quality assessment system is also included in the scope of technical protection, which has a very broad protection space.

[0040] (8) This invention can be directly adapted to conventional microscopic imaging equipment without the need for additional expensive dedicated hardware, and can be quickly promoted and applied to life science research, drug screening and other scenarios. It can significantly improve the analytical efficiency of drug screening, greatly reduce the time and cost of new drug development, and at the same time promote the progress of basic research such as stem cell differentiation and cardiovascular disease mechanisms, and has extremely high industrialization prospects and social value.

[0041] Based on previous validation of cardiomyocyte samples, the analytical results of this invention showed a 96% consistency with the manually labeled results. The LPS intervention experiment also verified that the system can accurately capture the dynamic changes in cell beating under drug intervention, fully demonstrating the feasibility of this invention and the high reliability of the results. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of time-series lossless compression technology;

[0043] Figure 2 This is a schematic diagram of the image convolution process;

[0044] Figure 3 This is a schematic diagram of pulsation signal clustering;

[0045] Figure 4 This is a schematic diagram showing the distribution of the pulsating region;

[0046] Figure 5 A schematic diagram of the division of the pulsating region and the 3D modeling of the pulsating signal;

[0047] Figure 6 This is a schematic diagram illustrating the distribution of panoramic information and the signal analysis of the myocardial cell pulsation region.

[0048] Figure 7 This represents the distribution of signal attenuation and attenuation coefficient in the pulsating region of myocardial cells. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0050] Example 1:

[0051] Combined with appendix Figures 1-7 A method and system for intelligent panoramic dynamic analysis of cell pulsation is presented in this embodiment to verify the application effect of the method in drug cardiotoxicity screening. The experimental subject is a primary cardiomyocyte isolated and cultured for 72 hours. The experimental equipment is a conventional inverted bright field microscope with a high-speed camera, without the need for any custom hardware or fluorescent labeling reagents.

[0052] First, data acquisition was performed. A culture plate containing primary cardiomyocytes was placed on the microscope stage. A 10× objective lens was selected, and the imaging frame rate was set to 30 frames per second. Bright-field video data was continuously acquired for 120 seconds, with a single frame resolution of 1920×1080. The raw data size for a single sample was approximately 1.2GB. The acquired three-dimensional temporal image data underwent temporal lossless compression processing, such as... Figure 1 As shown. The calculation formula for this time-series lossless compression is:

[0053]

[0054] in, Let be the image matrix of the t-th frame. For two-dimensional spatial convolution operations, This is an element-wise multiplication operation. This is an element-wise division operation. In this embodiment, the spatial neighborhood parameter K is set to 3.

[0055] This technology fully leverages the strong correlation between adjacent frames of cell pulsation time-series data. Through a combination of spatial convolution and element-level operations, it compresses the original tensor containing three dimensions—time, X-axis, and Y-axis—into two-dimensional planar data that simultaneously preserves spatial distribution and temporal variation characteristics. Without losing any pulsation signal information, the data volume is compressed to less than 1 / 20 of the original data, reducing data processing complexity by more than 90%. This solves the industry bottleneck of slow processing speed for large-view panoramic data and lays the foundation for subsequent real-time analysis.

[0056] After temporal compression, a 3×3 mean filter convolution kernel is used to perform convolutional noise reduction on the compressed two-dimensional image, such as... Figure 2As shown. This process utilizes the neighborhood correlation of effective image information, replacing the original pixel value with a weighted average of neighboring pixels, effectively eliminating isolated salt-and-pepper noise and camera thermal noise generated during imaging. In this embodiment, the image signal-to-noise ratio is improved to over 12.5 after processing, providing high-quality input data for subsequent unsupervised region segmentation and avoiding false positive pulsation region identification caused by noise.

[0057] Unsupervised K-means clustering algorithm is used to automatically segment the denoised image. The number of clusters K ranges from 2 to 200. In this embodiment, the number of clusters K=5. The algorithm automatically divides the image into different categories based on the gray-level variation characteristics of image pixels (corresponding to the intensity difference of cell pulsation). Without any manual annotation, it can accurately distinguish pulsation-positive areas from background areas. Figure 3 and Figure 4 As shown.

[0058] This method solves the problems of complex cell distribution and crosstalk between adjacent cell pulsation signals in traditional techniques. It can simultaneously identify 127 single-cell pulsation regions within the field of view, automatically counting 127 pulsating cells, with a positive area ratio of 38.2%. Subsequently, a 3D signal map is constructed based on the pulsation signal intensity of each region, such as... Figure 5 As shown, it intuitively displays the differences in pulsation intensity among different cells and quantifies the pulsation heterogeneity of cell populations, which is something that traditional local field-of-view analysis techniques cannot achieve.

[0059] For each identified pulsation-positive region, the pulsation direction of the cell is first calculated using the centroid trajectory quantile analysis (CTQA) method independently developed in this invention. Specifically, pixels with gray values ​​falling within the 25%-75% quantile range in each frame of the image for each pulsation region are extracted, and the centroid coordinates of these pixels are calculated. Then, the centroid trajectory of consecutive frames is fitted using Theil-Sen regression, which has stronger noise resistance, to finally obtain the slope of the pulsation direction of each cell. Compared with the traditional overall centroid calculation method, this method can effectively eliminate the interference of background noise and abnormal edge pixels, and the accuracy of pulsation direction calculation is improved to over 94%.

[0060] The time-domain pulsation signal of each pulsating positive region was extracted and subjected to multi-scale signal decoupling processing: First, the non-stationary cell pulsation signal was decomposed into multiple time-frequency components using continuous wavelet transform to accurately capture the local time-frequency features of the signal, obtaining parameters such as the pulsation frequency, duration, and systolic and diastolic durations of each cell; then, the time-domain signal was transformed to the complex frequency domain using a one-sided Laplace transform, and the signal attenuation coefficient was obtained analytically. This parameter directly quantifies the repolarization rate of the cell and is a key deep feature reflecting the electrophysiological function of cardiomyocytes, such as... Figure 6 and Figure 7 As shown.

[0061] In this embodiment, the system automatically extracted the average beating frequency of all cells to be 62 beats per minute, with an average decay coefficient of 0.12 s. -1 The average pulsation intensity was 0.032.

[0062] After feature extraction, the analysis results were subjected to a standardized biological interpretability signal quality assessment, including three core indicators: pulsation consistency index (BCI), pulsation signal-to-noise ratio (SNR), and parameter stability index. The calculated values ​​were: BCI = 0.08, which represents the standard deviation of the phase difference between different cell pulsation regions; a value closer to 0 indicates better cell pulsation synchronization, and this result indicates good pulsation synchronization in this batch of primary cardiomyocytes; SNR = 12.5, which is the ratio of the variance of the pulsation region to the variance of the background region; a higher value indicates better signal quality; and the average parameter stability index was 2.1%, which is the coefficient of variation of key parameters such as frequency and amplitude during continuous monitoring cycles; a lower value indicates better stability over long periods.

[0063] Finally, the system automatically outputs a panoramic analysis report containing 12 multi-dimensional pulsation characteristic parameters and 3 quality assessment indicators. The total analysis time for a single sample is only 12 minutes, which is far better than the 2-3 hours of traditional manual analysis, and the consistency with the manual annotation results reaches 96%.

[0064] To verify the application of this invention in drug cardiotoxicity screening, lipopolysaccharide (LPS) at a final concentration of 1 μg / mL was added to the culture system for intervention, and the dynamic changes in cell pulsation were continuously monitored over 24 hours using the same procedure. The results showed that after 6 hours of LPS intervention, the average cell pulsation frequency decreased to 48 beats / minute, the pulsation consistency index increased to 0.21, and the attenuation coefficient increased to 0.18 s. -1 The system can capture these subtle functional changes in real time, accurately reflecting the toxic damage of LPS to cardiomyocytes. This application scenario aligns with the field of drug cardiotoxicity screening, and the system supports real-time output of analysis results to observe the immediate dynamic response of drug intervention, demonstrating that this invention can meet the needs of high-throughput screening for drug cardiotoxicity.

[0065] The entire analysis process in this embodiment is executed by the Cell Pulse Intelligent Panoramic Dynamic Analysis System: the data acquisition module completes the acquisition of bright-field video data; the temporal compression and noise reduction module incorporates a temporal lossless compression algorithm unit and a mean filter convolution kernel unit to sequentially complete data compression and noise suppression; the unsupervised region segmentation module incorporates a K-means clustering algorithm unit, with the number of clusters dynamically adjustable from 2 to 200, automatically identifying pulsation-positive regions and counting their quantity and area proportion; the multi-scale feature extraction module incorporates a CTQA pulsation direction calculation unit, a continuous wavelet transform unit, and a one-sided Laplace transform unit, respectively... The module is used to calculate cell pulsation direction, extract time-frequency features, and analyze cell repolarization attenuation coefficients. The standardized quality assessment module has a built-in biological interpretability index calculation unit, which is used to calculate three quality assessment indicators: pulsation consistency index, which is the standard deviation of the pulsation phase difference between different pulsation-positive regions; pulsation signal-to-noise ratio index, which is the ratio of the signal variance of the pulsation-positive region to the signal variance of the background region; and parameter stability index, which is the coefficient of variation of key parameters such as pulsation frequency and amplitude within the same continuous monitoring period. The results output module finally generates a panoramic analysis report containing multi-dimensional pulsation features at the single-cell level and standardized quality assessment results.

[0066] Example 2:

[0067] A method and system for intelligent panoramic dynamic analysis of cell pulsation is disclosed. The difference between this embodiment and Example 1 is that, based on Example 1, this embodiment verifies the long-term label-free monitoring capability of the method in stem cell differentiation assessment. The experimental subjects are cardiomyocytes derived from human induced pluripotent stem cells (hiPSCs) from day 7 to day 21 of induced differentiation. The experimental equipment is a high-throughput bright-field microscope equipped with an automated stage and environmental control box. No fluorescent labeling is required throughout the process, avoiding interference from dyes on the stem cell differentiation process. This experiment supports long-term continuous dynamic monitoring from 1 minute to 30 days.

[0068] Before the experiment, the 96-well plates containing hiPSCs were placed in a microscope environmental control chamber, with the temperature set at 37°C, CO2 concentration at 5%, and humidity at 95% to maintain a normal cell culture environment. Long-term dynamic monitoring was performed for 14 consecutive days using the method of this invention. Bright-field video data was collected from each well for 30 seconds at a frame rate of 20 frames per second at the same time each day, resulting in the acquisition of temporal image data for a total of 96 samples.

[0069] The three-dimensional time-series image data of each sample is subjected to time-series lossless compression. This step significantly reduces the data storage and processing pressure while preserving the spatial morphological changes and temporal pulsation characteristics during cell differentiation. This reduces the batch processing time of the entire 96-well plate data to less than 2 hours, solving the problem that traditional technologies cannot achieve high-throughput long-term monitoring.

[0070] A 5×5 mean filter convolution kernel is used for convolutional noise reduction to eliminate noise caused by illumination fluctuations and environmental vibrations during long-term imaging, ensuring the consistency of signal-to-noise ratio of data at different time points.

[0071] Unsupervised K-means clustering algorithm was used to automatically segment the daily image data. The cluster number K was dynamically adjusted according to changes in cell density (K=3 on day 7 of differentiation, K=15 on day 21 of differentiation). The algorithm automatically identified pulsating positive regions at each time point and statistically analyzed the changing trends of the number of pulsating cells and the percentage of positive region area. Results showed that sporadic pulsating cells began to appear on day 10 of differentiation, with a positive region area percentage of 2.3%; the number of pulsating cells increased significantly on day 14 of differentiation, with the positive region area percentage rising to 21.7%; and on day 21 of differentiation, the positive region area percentage reached 65.4%. The system was able to accurately track the dynamic changes in cell pulsation ability throughout the entire differentiation process.

[0072] For each time point with a positive pulsation region, the pulsation direction was calculated using centroid trajectory quantile analysis. Characteristic parameters such as pulsation frequency, duration, and attenuation coefficient were extracted using continuous wavelet transform and one-sided Laplace transform. Simultaneously, daily signal quality assessment indicators were calculated. Results showed that as differentiation progressed, the average cell pulsation frequency gradually decreased from 112 beats / min on day 10 to 72 beats / min on day 21. The pulsation consistency index decreased from 0.35 to 0.09, and the attenuation coefficient increased from 0.22 s. -1 It dropped to 0.13s -1 The trends in these parameters are highly consistent with the maturation process of cardiomyocytes, and can objectively quantify the differentiation efficiency and maturity of stem cells into cardiomyocytes.

[0073] In this embodiment, the detection method without fluorescent labeling throughout the process avoids the toxic effects of dyes on stem cell proliferation and differentiation, supporting continuous monitoring for up to 30 days. The panoramic analysis capability reflects the differentiation heterogeneity of cells throughout the culture well, avoiding the result bias caused by traditional local sampling. The standardized signal quality assessment system automatically eliminates abnormal data caused by factors such as cell drift and contamination, ensuring the reliability of long-term monitoring results. Compared with traditional immunofluorescence staining endpoint detection methods, this invention can provide continuous dynamic functional data, more comprehensively reflecting the entire process of stem cell differentiation. It provides a novel technical means for optimizing stem cell differentiation processes and controlling quality.

[0074] The long-term batch analysis process in this embodiment is also executed by the Cell Beating Intelligent Panoramic Dynamic Analysis System. The system can automatically complete the batch data acquisition, processing, analysis and report generation of all samples in a 96-well plate without manual intervention, which greatly improves the efficiency of high-throughput experiments.

[0075] The parts of this invention not described in detail are prior art. It will be apparent to those skilled in the art that this invention is not limited to the details of the above exemplary embodiments, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and are intended to encompass all changes falling within the meaning and scope of equivalents within this invention.

Claims

1. A method for intelligent panoramic dynamic analysis of cell pulsation, characterized in that, Includes the following steps: S1. Perform time-series lossless compression on the acquired three-dimensional temporal image data of cells to convert it into two-dimensional planar data that simultaneously retains spatial distribution features and temporal variation features; S2. Use a filter convolution kernel to perform convolution noise reduction on the compressed two-dimensional image to eliminate isolated noise points; S3. An unsupervised clustering algorithm is used to automatically divide the denoised image into regions, distinguishing between pulsation-positive regions and background regions, and counting the number and area ratio of pulsation regions. S4. For each pulsation-positive region, the direction of cell pulsation is calculated using the centroid trajectory quantile analysis method; S5. Extract the time-domain pulsation signal of each pulsation positive region, extract the time-frequency features through continuous wavelet transform, and analyze the signal attenuation coefficient through one-sided Laplace transform to quantify the cell repolarization rate. S6. Summarize and statistically analyze the characteristic parameters of all pulsation-positive regions, and evaluate the signal quality of the analysis results based on biological interpretability indicators; S7. Output a comprehensive analysis report containing multi-dimensional pulsation characteristics and standardized quality assessment results.

2. The intelligent panoramic dynamic analysis method for cell pulsation as described in claim 1, characterized in that, The calculation formula for time-series lossless compression in step S1 is as follows: in, Let be the image matrix of the t-th frame. For two-dimensional spatial convolution operations, This is an element-wise multiplication operation. This is an element-wise division operation. , For spatial neighborhood parameters, This is the compressed two-dimensional matrix.

3. The intelligent panoramic dynamic analysis method for cell pulsation as described in claim 1, characterized in that, The unsupervised clustering algorithm mentioned in step S3 is the K-means clustering algorithm, and the number of clusters K ranges from 2 to 200, which is used to adapt to samples with different cell densities.

4. The intelligent panoramic dynamic analysis method for cell pulsation as described in claim 1, characterized in that, The specific steps of the centroid trajectory quantile analysis method described in step S4 are as follows: calculate the 25%-75% quantile centroid coordinates of each frame image within the pulsating positive region, and obtain the direction slope of cell pulsation by fitting the centroid trajectory through Theil-Sen regression.

5. The intelligent panoramic dynamic analysis method for cell pulsation as described in claim 1, characterized in that, The biological interpretability signal quality assessment metrics mentioned in step S6 include the pulsatility consistency index, the pulsatility signal-to-noise ratio index, and the parameter stability index.

6. The intelligent panoramic dynamic analysis method for cell pulsation as described in claim 5, characterized in that, The pulsation consistency index is the standard deviation of the pulsation phase difference between different pulsation-positive regions; the pulsation signal-to-noise ratio index is the ratio of the signal variance in the pulsation-positive region to the signal variance in the background region; and the parameter stability index is the coefficient of variation of the key parameters of pulsation frequency and amplitude within the same continuous monitoring period.

7. The intelligent panoramic dynamic analysis method for cell pulsation as described in claim 1, characterized in that, The cells include at least one of cardiomyocytes, nerve cells, tumor cells, and stem cells.

8. The intelligent panoramic dynamic analysis method for cell pulsation as described in claim 1, characterized in that, The three-dimensional time-series image data of cells mentioned in step S1 is acquired by a bright-field microscope, phase-contrast microscope or high-throughput microscopy imaging equipment, supporting long-term continuous dynamic monitoring from 1 minute to 30 days.

9. The intelligent panoramic dynamic analysis method for cell pulsation as described in claim 1, characterized in that, It can be applied to at least one of the following: drug cardiotoxicity screening, stem cell differentiation assessment, cardiovascular disease model construction, or tumor cell function research, and supports real-time output of analysis results to observe the immediate dynamic response of drug intervention.

10. A cell pulsation intelligent panoramic dynamic analysis system, characterized in that, It includes a data acquisition module, a temporal compression and noise reduction module, an unsupervised region partitioning module, a multi-scale feature extraction module, a standardized quality assessment module, and a result output module, which are connected in sequence. The temporal compression and noise reduction module has a built-in temporal lossless compression algorithm unit and a mean filter convolution kernel unit, which are used to perform lossless compression of three-dimensional temporal data and image noise suppression, respectively. The unsupervised region segmentation module has a built-in K-means clustering algorithm unit. The number of clusters K can be dynamically adjusted in the range of 2-200. It is used to automatically identify pulsation-positive regions and background regions and count their number and area ratio. The multi-scale feature extraction module incorporates a CTQA pulsation direction calculation unit, a continuous wavelet transform unit, and a one-sided Laplace transform unit, which are used to calculate the cell pulsation direction, extract time-frequency features, and analyze the cell repolarization attenuation coefficient, respectively. The standardized quality assessment module has a built-in biological interpretability index calculation unit, which is used to calculate the pulsation consistency index, pulsation signal-to-noise ratio index, and parameter stability index. The results output module is used to generate a panoramic analysis report that includes single-cell-level multidimensional pulsation characteristics and standardized quality assessment results.

Images