Cell intracellular mechanics response visual analysis method based on image feature extraction

Abstract

The present application belongs to the technical field of biological image analysis and cell mechanics research, and particularly relates to a cell intracellular mechanics response visual analysis method based on image feature extraction, comprising the following steps: acquiring original image data containing dynamic change process of cells under mechanical stimulation; structurally segmenting the original image data; extracting morphological features and dynamic behavior features for each data unit respectively to obtain feature elements; classifying the feature elements according to preset biological significance or mechanical response mode to form several feature element groups; and constructing a mapping relationship between local regions or time segments of cells and mechanical responses based on the classified feature element groups. The present application can analyze the mechanical response of local regions and specific time windows of cells under mechanical stimulation, thereby providing analysis basis for cell mechanics mechanism research, differentiation process regulation, drug effect evaluation and the like.

Description

Technical Field

[0001] This invention belongs to the field of biological image analysis and cell mechanics research technology, specifically involving a method for visual analysis of intracellular mechanical responses based on image feature extraction. Background Technology

[0002] Cells are the basic units of life activities, and their morphology and structure are closely related to their mechanical behavior. In the fields of tissue engineering, regenerative medicine, and drug screening, studying the intracellular response mechanisms of cells under external mechanical stimuli (such as stretching, shearing, and compression) is of great significance for understanding cell fate determination and guiding differentiation regulation.

[0003] Currently, mainstream methods for analyzing cell mechanical response mainly rely on global feature extraction from the entire microscopic image. For example, cell outlines can be obtained through image segmentation, and morphological parameters such as nucleocytoplasmic ratio, area, and perimeter can be calculated. Alternatively, the overall mechanical field can be inverted using traction microscopy. Some studies have introduced deep learning models to regress and predict cell elastic modulus or migration ability from images.

[0004] Problems with existing technology:

[0005] Existing methods often treat cells as a whole, extracting global average features. This fails to capture the heterogeneous responses of different regions within the cell (such as the nucleus, pseudopodia, and membrane edge) to mechanical stimulation, resulting in the masking of key mechanically sensitive areas. Furthermore, the extracted features (such as fractal dimension and displacement vector) are mostly mathematical or physical quantities, lacking semantic association with specific biological behaviors or mechanical response patterns. Researchers struggle to intuitively understand "which feature combinations represent which mechanical behaviors." Consequently, existing visualization outputs are mostly force field heatmaps or time trend curves, reflecting only a single dimension of spatial distribution or temporal evolution. They cannot simultaneously present comprehensive information such as "which region," "which time period," and "what type of response," limiting the interpretation of complex dynamic processes. Because they cannot identify the most significant local regions or key time windows of mechanical response, existing methods are insufficient to guide researchers in optimizing mechanical stimulation parameters. In other words, the cellular mechanical behavior features obtained from existing data have low reference value and cannot achieve precise regulation of cell differentiation or functional transformation. Summary of the Invention

[0006] The purpose of this invention is to provide a visualization analysis method for intracellular mechanical response based on image feature extraction, which can realize the analysis of the mechanical response of cells in local areas and specific time windows under mechanical stimulation, thereby providing analytical basis for cell mechanical mechanism research, differentiation process regulation, drug effect evaluation, etc.

[0007] The specific technical solution adopted by this invention is as follows:

[0008] A method for visual analysis of intracellular mechanical response based on image feature extraction includes the following steps:

[0009] Acquire raw image data containing the dynamic changes of cells under mechanical stimulation, wherein the raw image data includes any one of time-series microscopic images, depth image data, or video stream data;

[0010] The original image data is structured and segmented, and a single or multiple frames of images are divided into multiple independent data units according to spatial regions or time segments. Each data unit corresponds to a local region of a cell or a specific time window.

[0011] For each data unit, morphological features and dynamic behavior features are extracted to obtain a set of feature elements that characterize the state of the data unit.

[0012] Based on the preset biological significance or mechanical response mode, the feature elements are classified into several feature element groups, and each feature element group corresponds to a cell mechanical behavior mode or structural response type.

[0013] Based on the categorized feature groups, a mapping relationship between local cell regions or time segments and mechanical responses is constructed, and a visualization atlas is generated to characterize the corresponding intensity and distribution trend of cells in different regions or at different stages under mechanical stimulation.

[0014] According to another aspect of the present invention, the structured segmentation includes at least one of spatial segmentation and temporal segmentation:

[0015] Spatial segmentation refers to dividing a single frame image into regions based on organelles, cell membranes, cell nuclei, or pseudopodia.

[0016] Time segmentation refers to dividing a time series image into multiple time segments based on cell morphological mutation points, mechanical loading cycles, or inflection points of feature changes.

[0017] According to another aspect of the present invention, the morphological features include cell outline complexity, nucleocytoplasmic ratio, local curvature, or fractal dimension;

[0018] The dynamic behavior characteristics include local displacement vector, deformation rate, boundary fluctuation frequency, and regional brightness change gradient.

[0019] According to another aspect of the present invention, the classification criteria include:

[0020] The magnitude of change of characteristic elements before and after mechanical loading;

[0021] The spatial distribution consistency of feature elements across different cellular regions;

[0022] Similarity in the patterns of feature evolution over time;

[0023] By using clustering or rule matching, feature elements with similar variation patterns or spatial distribution rules are grouped into the same feature element group.

[0024] According to another aspect of the present invention, the visualization spectrum is a heat map, a vector field map, or a time-series evolution curve, wherein different colors or arrow directions in the spectrum represent the mechanical response intensity, direction, or dynamic trend corresponding to different feature element groups.

[0025] According to another aspect of the present invention, a visualization and analysis system for intracellular mechanical response based on image feature extraction includes:

[0026] The image input module is used to receive raw image data of cells under mechanical stimulation;

[0027] The data segmentation module is used to perform structured segmentation of the original image data in the spatial or temporal dimensions;

[0028] The feature extraction module is used to extract morphological and dynamic behavior features from each segmented data unit;

[0029] The feature classification module is used to classify the extracted features into feature feature groups according to preset rules or clustering algorithms;

[0030] The visualization generation module is used to construct mechanical response mapping relationships based on the classification results and output visual graphs.

[0031] According to another aspect of the present invention, the element classification module learns the classification pattern automatically based on user-defined classification rules or historical experimental data.

[0032] According to another aspect of the present invention, the visualization generation module selects a specific feature element group or region through interactive operation, dynamically adjusts the visualization parameters, and updates the map display in real time.

[0033] According to another aspect of the present invention, an electronic device is also provided, the electronic device including a memory and a processor; the memory is used to store a program; the processor executes the program to implement the method described in any one of the foregoing.

[0034] According to another aspect of the present invention, a computer-readable storage medium is also provided, the storage medium storing a computer program that, when executed by a processor, implements the method described in any one of the preceding embodiments.

[0035] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the method described in any one of the preceding embodiments.

[0036] The technical effects achieved by this invention are as follows:

[0037] This invention utilizes structured segmentation, unitized feature extraction, semantic classification, and interactive visualization to analyze cellular mechanical responses. First, the image is intelligently segmented into regions such as the nucleus and pseudopodia. Then, features such as deformation and curvature of each unit are extracted independently to capture local antagonistic phenomena such as nuclear shrinkage and pseudopodia expansion. Next, mathematical features are upgraded to biological semantic labels by using rules such as high deformation rate and low curvature to form membrane fluidity groups. Through the interactive superposition of heatmaps, vector fields, and curves, key response windows can be quickly locked.

[0038] In this invention, dynamic behavioral features generate a unique digital fingerprint for each local data unit for subsequent classification and comparison. The principle is unit-based independent computation, that is, the same feature (such as deformation rate) is calculated independently in different spatial or temporal units, thereby preserving and highlighting local differences. Through unit-based independent extraction, the response polarity of different regions inside the cell to the same mechanical stimulus is presented.

[0039] This invention, by being sensitive to instantaneous and local pseudopodia dynamics, enables the capture of the instantaneous germination of tiny pseudopodia and can identify key response signals at a very early stage of cell fate determination. Attached Figure Description

[0040] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0041] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.

[0042] It should be noted that the terms "first," "second," etc., used in this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0043] According to embodiments of the present invention, a method embodiment for visual analysis of intracellular mechanical response based on image feature extraction is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0044] like Figure 1 As shown, the method for visual analysis of intracellular mechanical response based on image feature extraction includes the following steps:

[0045] S1. Acquire raw image data containing the dynamic changes of cells under mechanical stimulation. The raw image data includes any one of time-series microscopic images, depth image data, or video stream data.

[0046] S2. Perform structured segmentation on the original image data, dividing a single or multiple frames of images into multiple independent data units according to spatial regions or time segments. Each data unit corresponds to a local region of a cell or a specific time window.

[0047] S3. Extract morphological features and dynamic behavior features from each data unit to obtain a set of feature elements that characterize the state of the data unit.

[0048] S4. Based on the preset biological meaning or mechanical response mode, classify the feature elements to form several feature element groups. Each feature element group corresponds to a cell mechanical behavior mode or structural response type.

[0049] S5. Based on the categorized feature element groups, construct the mapping relationship between local cell regions or time segments and mechanical responses, and generate a visualization atlas to characterize the corresponding intensity and distribution trend of cells in different regions or at different stages under mechanical stimulation.

[0050] In step S1 above, the original image data is the image of the experimental subject under dynamic state under mechanical stimulation. The mechanical stimulation is the external physical force applied to the cell. The physical force includes, but is not limited to, tensile force (such as uniaxial / biaxial stretching of the base), shear force (such as fluid shear), compressive force (such as micropillar indentation), mechanical tension induced by matrix stiffness gradient, etc. The mechanical stimulation and image acquisition are carried out simultaneously to establish the time correspondence between stimulation and response. Under continuous or periodic mechanical stimulation, the morphology, structure and motion state of the cell undergo observable evolution over time, such as pseudopodia extension and contraction, nuclear deformation, cell migration, skeletal rearrangement, etc. The time span is set according to the speed of cell deformation.

[0051] The methods for acquiring raw image data include the following:

[0052] Time-stamped Z-stack images were acquired using confocal microscopy, STED super-resolution microscopy, and structured light illumination microscopy (SIM).

[0053] Video stream data (frame rate ≥ 1fps) is acquired using a high-speed camera or sCMOS camera in conjunction with a bright field / phase contrast / fluorescence module.

[0054] 4D (x,y,z,t) image data containing depth information are acquired using a light field microscope or a structured light 3D imaging system.

[0055] The substrate deformation field is simultaneously recorded using a traction force microscope (TFM) or digital image correlation (DIC) system as an indirect image characterization of the mechanical response.

[0056] According to step S2, the structured segmentation of the original image data is divided into two dimensions: spatial segmentation and temporal segmentation. The segmentation is based on the biological structure of the cells or the intrinsic rhythm of their mechanical response, and the original image data is divided in a purposeful and non-random manner.

[0057] Structured segmentation includes at least one of spatial segmentation and temporal segmentation:

[0058] Spatial segmentation refers to dividing a single frame image into regions based on organelles, cell membranes, cell nuclei, or pseudopodia.

[0059] Time segmentation refers to dividing a time series image into multiple time segments based on cell morphological mutation points, mechanical loading cycles, or inflection points of feature changes.

[0060] Furthermore, spatial segmentation uses a pre-trained U-Net++ deep learning model to perform pixel semantic segmentation on a single frame image, identifying and outputting binary masks for regions such as the cell nucleus, cytoplasm, pseudopodia, and cell membrane edge. Then, based on these masks, the pseudopodia region is further subdivided into two sub-regions: the leading edge (5μm) and the root. The cell membrane edge is divided into the leading edge, lateral edge, and trailing edge according to the azimuth angle. Each sub-region is an independent spatial data unit.

[0061] Temporal segmentation, whether or not it uses fixed equal time intervals, is based on intelligent segmentation according to the dynamic behavior of the cells themselves. The Global Morphological Entropy (GME) is calculated for each time point in the entire temporal image sequence. The formula for calculating GME is as follows:

[0062]

[0063] in, This is the GME value at time point t, used to quantify the complexity or disorder of cell morphology at that moment. A higher value indicates a more irregular and complex morphology. It is the normalized power of the i-th Fourier descriptor coefficient of the image at time point t, when the GME rate of change is between adjacent time points. When the threshold is exceeded, it is determined to be a morphological change point and the point is used as the starting point of a new time segment. At the same time, the system will also read the mechanical loading log and use the switching points of loading frequency or amplitude as auxiliary segmentation basis.

[0064] According to the formula, if the cell outline is a smooth, regular circle, the energy will be concentrated on a few low-frequency components, resulting in a very low entropy value, close to 0. Conversely, if the cell outline is very complex and irregular, the energy will be dispersed across many high-frequency components, resulting in a very high entropy value. This can be monitored... By analyzing the rate of change of values ​​over time, we can identify inflection points where cell morphology undergoes dramatic changes, thus enabling intelligent time segmentation.

[0065] Based on the above, biologically guided segmentation means that the segmentation is based not on the pixel values ​​or gradients of the image, but on the actual anatomical structure (space) or the intrinsic physiological rhythm (time) of the cells.

[0066] According to step S3, for each independent data unit generated in step S2, a set of morphological features and dynamic behavior features are calculated to form the feature element vector of the unit.

[0067] Among them, morphological features include local curvature, fractal dimension and nucleocytoplasmic ratio. Local curvature is calculated every 3 pixels on the cell boundary and the average of all curvature values ​​in the data unit is taken. Local curvature is obtained by calculating the angle between the tangents of adjacent points on the cell boundary. The larger the value, the more severe the bending.

[0068] Dynamic behavioral characteristics include local displacement vectors, deformation rate, boundary fluctuation frequency, and regional intensity gradient. Deformation rate (DR) reflects the speed of expansion or contraction of the region. Boundary fluctuation frequency (BFF) is calculated by performing a Fast Fourier Transform (FFT) on the X and Y coordinate sequences of the cell boundary and taking the frequency with the largest amplitude in the power spectrum as the BFF, reflecting the speed of boundary vibration. Regional intensity gradient (RIG) is calculated by taking the pixel intensity matrix of the data unit in the fluorescence image and calculating the mean magnitude of its spatial gradient. Gx and Gy are the brightness gradient values ​​in the x and y directions, respectively. Gy is an abbreviation for the brightness gradient of the RIG region. It represents the arithmetic mean of the brightness gradient magnitudes of all pixels in the specified region. The larger the value, the more drastic the brightness change in the region, which usually means that there are obvious edge, texture or protein aggregation / deaggregation phenomena.

[0069] Based on the above, its function is to generate a unique digital fingerprint for each local data unit for subsequent classification and comparison. Its principle is unit-based independent computation, that is, the same feature (such as deformation rate) is calculated independently in different spatial or temporal units, thereby preserving and highlighting local differences. Through unit-based independent extraction, the response polarity of different regions inside the cell to the same mechanical stimulus is presented. For example, when compressive force is applied, the system finds that the deformation rate of the cell nucleus region is negative (-0.8 μm² / h, indicating contraction), while the deformation rate of the adjacent pseudopodia region is positive (+2.1 μm² / h, indicating expansion). The discovery of local mechanical antagonism provides new evidence for understanding how cells maintain structural stability in complex mechanical environments.

[0070] In step S4, the original feature data is given explicit biological semantics. The system classifies the feature vectors extracted in step S3 into different feature groups according to preset feature combination rules. The classification method supports two modes: rule matching and unsupervised clustering.

[0071] The rule matching pattern is preset by domain experts using a series of IF-THEN rules, for example:

[0072] IF (deformation rate greater than 1.5 μm² / h) AND (local curvature less than 0.03 μm⁻¹) AND (regional brightness change gradient greater than 0.5) → classified as membrane fluidity response group;

[0073] IF (change rate of nucleus-mass ratio greater than 4% / h) AND (decrease in fractal dimension greater than 8%) → classified as nuclear strain response group;

[0074] IF (phase difference between pseudopodia extension rate and stretching frequency is less than 30°) → classified as a mechanical-cytoskeleton coupling response group.

[0075] The unsupervised clustering mode is as follows: When there is a lack of prior knowledge, the system uses the K-means++ algorithm to cluster based on the Euclidean distance of the feature vectors. After the clustering is completed, experts assign a biological name to each cluster based on the feature combination of the cluster centers (such as Cluster#1 → "Membrane Fluidity Response Group").

[0076] Based on the above, the principle used to solve the problem of meaningless features is the mapping logic of feature combination → behavioral semantics, which combines multiple low-level mathematical features.

[0077] In step S5, based on the classification results, the system constructs a three-dimensional mapping relationship of spatial region / time segment ↔ feature element group ↔ response intensity, and generates an intuitive visualization map. The specific implementation is as follows:

[0078] Mapping relationship construction: The system records the feature group to which each data unit belongs and the activation strength of that group (usually the normalized feature value or cluster confidence).

[0079] As an alternative embodiment, the visualization spectrum is a heat map, vector field map, or time-series evolution curve. Different colors or arrow directions in the spectrum represent the mechanical response intensity, direction, or dynamic trend corresponding to different feature groups.

[0080] Visualization maps can be generated, such as heatmaps, vector field maps, or time-series evolution curves. Heatmaps use cell outlines as a base map and different colors (e.g., blue → red) to represent the response intensity of specific feature groups in different spatial regions. Users can choose to display a single group (e.g., only the membrane fluidity group) or display multiple groups overlaid. Vector field maps are based on heatmaps, with arrows superimposed to indicate the direction and magnitude of local deformation (arrow direction = displacement vector direction, arrow length = deformation rate). Time-series evolution curves plot the activation curves of different feature groups with time as the X-axis and activation intensity as the Y-axis, for observing the dynamic changes in the response.

[0081] Based on the above, the complex multidimensional analysis results are transformed into intuitive and easy-to-understand visual language, helping researchers to quickly grasp the global situation and local details of cell mechanical response. The principle is multidimensional data fusion visualization, which integrates information from the three dimensions of space, time, and semantics into an interactive map.

[0082] As an optional embodiment, morphological features include not only nucleocytoplasmic ratio, local curvature, or fractal dimension, but also cell contour complexity. Cell contour complexity is a dimensionless index used to quantify the degree of irregularity of cell boundaries. A smooth, round cell has the lowest contour complexity, while a cell covered with pseudopodia, wrinkles, or protrusions has the highest contour complexity. This can be determined using methods such as local curvature to identify cell boundaries, or by combining the fractal dimension method with the Fourier descriptor energy distribution method to comprehensively calculate contour complexity. The specific formula is as follows:

[0083] ;

[0084] Wherein, FD is the fractal dimension of the cell contour, calculated using box counting. FD values ​​range from 1.0 (ideal straight line) to 2.0 (extremely complex plane); the larger the value, the more irregular the contour. The first harmonic energy is the energy of the Fourier descriptor of the cell profile. The Fourier descriptor is the sequence of coefficients obtained by performing a Fourier transform on the cell profile, which is considered as a periodic function. This mainly describes the overall ellipticity of the contour. To outline the total energy of all Fourier descriptors of the subharmonics, and These are weighting coefficients used to balance the contributions of the two algorithms, ensuring sensitivity to high-frequency structural changes such as pseudofoot. As the proportion of high-frequency harmonic energy, when cells extend pseudopodia or form wrinkles, the energy of the high-frequency components of the contour (higher-order Fourier coefficients) increases significantly, leading to... The ratio decreases, thus causing An increase in the value of CCC will ultimately lead to an increase in the CCC value.

[0085] Based on the above, the dynamics of subcellular structures can be captured by cell contour complexity. After spatial segmentation in step S2, calculating the CCC for local units such as pseudopodia or membrane edge regions can capture the dynamic processes of pseudopodia germination, extension, and retraction, and further distinguish mechanical response modes. Different mechanical stimuli will induce different types of contour changes in cells. For example, tensile force mainly induces overall cell elongation (increased FD, but little change in high-frequency energy), while shear force is more likely to induce rapid and irregular extension of pseudopodia (high-frequency energy). (Surge), through the comprehensive calculation of CCC, can distinguish between these two response modes, which also facilitates the classification of feature elements in step S4.

[0086] Furthermore, by being sensitive to instantaneous and local pseudopodia dynamics, it enables the capture of the instantaneous germination of tiny pseudopodia, and can identify key response signals at a very early stage of cell fate determination (such as within minutes after mechanical stimulation).

[0087] Furthermore, it should be noted that fractal dimension is a dimensionless index used to quantify the irregularity of the contours of cells or their local regions and their space-filling ability. The fractal dimension is calculated using box counting, and includes the following steps:

[0088] S101. Place the cell outline (or local region outline) to be analyzed on a two-dimensional grid;

[0089] S102. Cover the outline with a small square box with a side length of ε;

[0090] S103. Count the number of boxes that contain at least one outline pixel, denoted as N(ε);

[0091] S104. Gradually decrease the side length ε of the box (e.g., ε=32,16,8,4,2 pixels), and repeat steps S102 and S103 to obtain a series of (ε,N(ε)) data pairs;

[0092] S105. In a double logarithmic coordinate system, use log(1 / ε) as the horizontal axis and log(N(ε)) as the vertical axis to perform linear fitting on the data points;

[0093] S106. The slope of the fitted line is the fractal dimension FD of the contour.

[0094] Furthermore, the nucleocytoplasmic ratio is an indicator used to quantify the proportion of the cell nucleus relative to the total cell size.

[0095] As an optional implementation, the classification criteria include:

[0096] The magnitude of change of characteristic elements before and after mechanical loading;

[0097] The spatial distribution consistency of feature elements across different cellular regions;

[0098] Similarity in the patterns of feature evolution over time;

[0099] By using clustering or rule matching, feature elements with similar variation patterns or spatial distribution rules are grouped into the same feature element group.

[0100] A visualization and analysis system for intracellular mechanical responses based on image feature extraction includes:

[0101] The image input module is used to receive raw image data of cells under mechanical stimulation;

[0102] The data segmentation module is used to perform structured segmentation of the original image data in the spatial or temporal dimensions;

[0103] The feature extraction module is used to extract morphological and dynamic behavior features from each segmented data unit;

[0104] The feature classification module is used to classify the extracted features into feature feature groups according to preset rules or clustering algorithms;

[0105] The visualization generation module is used to construct mechanical response mapping relationships based on the classification results and output visual graphs.

[0106] As an optional implementation, the element classification module can learn classification patterns automatically based on user-defined classification rules or historical experimental data.

[0107] As an optional embodiment, the visualization generation module allows users to select specific feature groups or regions through interactive operations, dynamically adjust visualization parameters, and update the map display in real time.

[0108] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0109] According to another aspect of the present invention, an electronic device is also provided, the electronic device including a memory and a processor; the memory is used to store a program; the processor executes the program to implement the method of any of the foregoing.

[0110] According to another aspect of the present invention, a computer-readable storage medium is also provided, the storage medium storing a computer program that, when executed by a processor, implements the method of any of the foregoing.

[0111] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the method described in any of the foregoing.

[0112] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A method for visual analysis of intracellular mechanical response based on image feature extraction, characterized in that, Includes the following steps: Acquire raw image data containing the dynamic changes of cells under mechanical stimulation, wherein the raw image data includes any one of time-series microscopic images, depth image data, or video stream data; The original image data is structured and segmented, and a single or multiple frames of images are divided into multiple independent data units according to spatial regions or time segments. Each data unit corresponds to a local region of a cell or a specific time window. For each data unit, morphological features and dynamic behavior features are extracted to obtain a set of feature elements that characterize the state of the data unit. Based on the preset biological significance or mechanical response mode, the feature elements are classified into several feature element groups, and each feature element group corresponds to a cell mechanical behavior mode or structural response type. Based on the categorized feature groups, a mapping relationship between local cell regions or time segments and mechanical responses is constructed, and a visualization atlas is generated to characterize the corresponding intensity and distribution trend of cells in different regions or at different stages under mechanical stimulation.

2. The method for visual analysis of intracellular mechanical response based on image feature extraction according to claim 1, characterized in that: The structured segmentation includes at least one of spatial segmentation and temporal segmentation: Spatial segmentation refers to dividing a single frame image into regions based on organelles, cell membranes, cell nuclei, or pseudopodia. Time segmentation refers to dividing a time series image into multiple time segments based on cell morphological mutation points, mechanical loading cycles, or inflection points of feature changes.

3. The method for visual analysis of intracellular mechanical response based on image feature extraction according to claim 1, characterized in that: The morphological features include cell outline complexity, nucleocytoplasmic ratio, local curvature, or fractal dimension. The dynamic behavior characteristics include local displacement vector, deformation rate, boundary fluctuation frequency, and regional brightness change gradient.

4. The method for visual analysis of intracellular mechanical response based on image feature extraction according to claim 1, characterized in that, The classification criteria include: The magnitude of change of characteristic elements before and after mechanical loading; The spatial distribution consistency of feature elements across different cellular regions; Similarity in the patterns of feature evolution over time; By using clustering or rule matching, feature elements with similar variation patterns or spatial distribution rules are grouped into the same feature element group.

5. The method for visual analysis of intracellular mechanical response based on image feature extraction according to claim 1, characterized in that: The visualization map is a heat map, vector field map or time-series evolution curve. Different colors or arrow directions in the map represent the mechanical response intensity, direction or dynamic trend corresponding to different feature element groups.

6. A visualization and analysis system for intracellular mechanical response based on image feature extraction, used to execute the method as described in any one of claims 1-5, characterized in that, include: The image input module is used to receive raw image data of cells under mechanical stimulation; The data segmentation module is used to perform structured segmentation of the original image data in the spatial or temporal dimensions; The feature extraction module is used to extract morphological and dynamic behavior features from each segmented data unit; The feature classification module is used to classify the extracted features into feature feature groups according to preset rules or clustering algorithms; The visualization generation module is used to construct mechanical response mapping relationships based on the classification results and output visual graphs.

7. The image feature extraction-based visualization analysis system for intracellular mechanical response according to claim 6, characterized in that: The element classification module uses user-defined classification rules or automatically learns classification patterns based on historical experimental data.

8. The image feature extraction-based visualization analysis system for intracellular mechanical response according to claim 6, characterized in that: The visualization generation module allows users to select specific feature groups or regions through interactive operations, dynamically adjust visualization parameters, and update the map display in real time.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

Images