Accurate feature extraction system and method based on AFM multi-source data analysis

Abstract

The invention discloses a feature accurate extraction system and method based on AFM multi-source data analysis, and relates to the field of big data processing, and the method comprises the steps: obtaining historical morphology and mechanical data; identifying a morphology gradient salient region, extracting morphology and mechanical gradient values, establishing a reference mapping relation, and determining a gradient threshold according to a prediction residual error; collecting real-time morphology and mechanical data, and identifying a real-time gradient salient region; performing consistency judgment by using a gradient threshold value, identifying and removing invalid data, and generating a compliance data set; the system comprises a historical data acquisition module, a gradient threshold determination module, a real-time data acquisition and preprocessing module and a mechanical verification and compliance data generation module. According to the method, the quantitative correlation model of the morphology and mechanical data is established, so that the automatic validity verification of the multi-source measurement data is realized, and the accuracy and reliability of sample feature extraction are improved.

Description

Technical Field

[0001] This invention relates to the field of big data processing technology, specifically to a feature extraction system and method based on AFM multi-source data analysis. Background Technology

[0002] Atomic force microscopy (AFM), a key tool for nanoscale characterization, has been widely used in the study of the physical properties of biological samples. While traditional single-probe AFM techniques can acquire sample surface morphology, they have significant limitations in the simultaneous, in-situ measurement of multiple physical parameters. Typically, single-probe systems require sequential scanning of different modes, resulting in data acquisition that is not fully synchronized in time and space. This makes it difficult to establish precise correspondences between different physical fields, and this asynchrony introduces significant errors, especially for dynamically changing biological samples.

[0003] Existing techniques face numerous challenges in characterizing the mechanical properties of biological samples. A complex correlation exists between gradient changes in sample surface morphology and local mechanical responses; however, due to the complexity of probe-sample interactions during measurement and environmental interference, inconsistencies often arise between the acquired morphological and mechanical data. This inconsistency leads to insufficient reliability of analytical results based on single-modal data, making it difficult to accurately extract sample features.

[0004] Furthermore, existing technologies lack effective mechanisms for multi-source data fusion and verification. Although some studies have attempted to acquire synchronous data through multi-probe systems, establishing a quantitative correlation model between morphological and mechanical data, and using this model to automatically verify the validity of measurement data, remains a pressing technical challenge. Particularly during real-time measurement, the lack of an automatic identification and removal mechanism for abnormal data directly impacts the accuracy and reliability of subsequent feature analysis.

[0005] Therefore, there is an urgent need in this field for a technical solution that can realize collaborative analysis and automatic verification of multi-source data, in order to solve the problems of inconsistency between morphological and mechanical data and insufficient measurement reliability in the existing technology. Summary of the Invention

[0006] The purpose of this invention is to provide a system and method for accurate feature extraction based on AFM multi-source data analysis, so as to solve the problems raised in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a system and method for accurate feature extraction based on AFM multi-source data analysis. The methods include: Step S1: Obtain the historical dataset of the physical feature quantitative characterization experiment based on three-probe atomic force nanomanipulation. The historical dataset contains several sets of synchronously acquired historical morphological data and historical mechanical data. The historical morphological data includes historical sample surface height information obtained by probe scanning. By establishing historical datasets, sufficient training samples are provided for subsequent analysis, ensuring the reliability and generalization ability of the model. Step S2: Traverse the historical morphology data to identify regions with significant historical morphology gradients; based on these regions, simultaneously extract the morphology gradient values ​​and mechanical gradient values ​​to form several sets of morphology and mechanical gradient correlation data; based on this correlation data, establish a baseline mapping relationship between the morphology gradient and the mechanical gradient through regression analysis, and determine the gradient threshold based on the predicted residual distribution of the baseline mapping relationship. By establishing a quantitative mapping relationship between morphology and mechanical gradient, a theoretical basis is provided for the validity verification of real-time data; a threshold is set based on the predicted residual distribution to ensure the statistical significance of the anomaly judgment criteria. Step S3: Simultaneously acquire real-time morphological data and real-time mechanical data of the biological sample to be tested; perform gradient calculation on the real-time morphological data to identify regions with significant real-time morphological gradients; Using the same threshold standard as historical data processing ensures consistency in data collection and processing, providing a comparable basis for subsequent verification. Step S4: Using the gradient threshold, perform consistency judgment on real-time topography data and real-time mechanical data. If the data exceeds the gradient threshold, it is marked as invalid data of real-time topography data and real-time mechanical data. Remove the invalid data marked as invalid data of real-time topography data and real-time mechanical data from the real-time data to generate a compliant dataset. An automated threshold judgment mechanism is used to verify data validity, eliminate abnormal data, and improve the accuracy and reliability of subsequent feature analysis.

[0008] Furthermore, step S2 includes: Step S2-1: Define the plane of the sample-supporting substrate as the xOy reference plane, define the scanning direction of the physical characteristic quantitative characterization experiment based on three-probe atomic force nanomanipulation as the reference direction as the positive x-axis, the direction perpendicular to the x-axis as the positive y-axis, and the direction perpendicular to the xOy reference plane upward as the positive z-axis to establish the sample reference coordinate system. Establish a unified coordinate system to ensure accurate spatial registration of multi-source data and provide a benchmark for gradient calculation; Step S2-2: Divide the sample surface into several grid regions; calculate the average gradient magnitude in all grid regions, and identify those whose average gradient magnitude is greater than a preset gradient magnitude threshold. The grid region is considered as a region with significant historical morphological gradient; among which, , This represents the median gradient magnitude across all grid regions in the historical topographic data. Scaling factor indicating the determination of significant regions in historical morphological gradients; A gridded analysis method is used to systematically identify regions with significant morphological changes, ensuring the comprehensiveness and representativeness of feature extraction. Step S2-3: Select m sampling locations within the historical morphological gradient significant region, and simultaneously extract the morphological gradient values ​​and mechanical gradient values ​​at the m sampling locations. The morphological gradient value specifically refers to the rate of change of the morphological data in the reference direction, and the mechanical gradient value specifically refers to the rate of change of the mechanical data in the reference direction; form a data pair set composed of morphological gradient values ​​and mechanical gradient values. ,in, , , ..., , ..., These represent data pairs consisting of the topographic gradient value at the 1st, 2nd, ..., ith, ..., mth sampling positions and the mechanical gradient value at the 1st, 2nd, ..., ith, ..., mth sampling positions, respectively. By synchronously collecting morphological and mechanical gradient data, spatially correlated data pairs are established, providing a data foundation for establishing mapping relationships. Step S2-4: Based on the data, adjust the set The least squares method is used to establish a baseline mapping relationship between the topographic gradient and the mechanical gradient. This baseline mapping relationship is represented by the following linear equation: ,in, Represents the morphological gradient. This represents the predicted mechanical gradient. The regression intercept coefficient, The regression slope coefficient; A linear regression model is used to establish quantitative relationships. The model is simple, reliable, and easy to implement and interpret. Step S2-5: Calculate the data pair set The i-th data pair Predicted residuals under the reference mapping relationship : The predicted residual set is obtained. ,in, , , ..., , ..., Let represent the prediction residuals of the 1st, 2nd, ..., i-th, ..., m-th data pairs, respectively; Residual analysis is used to quantify the deviation between actual measurements and theoretical predictions, providing a quantitative basis for anomaly identification. Step S2-6: Set the predicted residuals All elements are sorted numerically, and the values ​​at preset quantile positions among the sorted elements are set as the gradient threshold. ; Thresholds are set based on statistical quantiles to avoid the influence of extreme values ​​and ensure the robustness of the threshold setting.

[0009] Furthermore, step S3 includes: Step S3-1: Simultaneously acquire real-time morphological data and real-time mechanical data of the biological sample to be tested; Ensure the synchronization of real-time data acquisition and maintain consistency with historical data acquisition conditions; Step S3-2: Perform gradient calculation on the real-time topography data, and identify gradient magnitudes that are greater than a preset gradient magnitude. The region is designated as the region with significant real-time topographic gradient; The same identification criteria as historical data analysis are used to ensure consistency in data processing logic.

[0010] Furthermore, step S4 includes: Step S4-1: Select q sampling locations within the significant region of the real-time topography gradient, and extract the real-time topography gradient value at the q-th sampling location. Simultaneously extract the real-time mechanical gradient value at the q-th sampling position. This forms a set of real-time topographic and mechanical gradient data. ,in, , , ..., These represent real-time data pairs consisting of the real-time topographic gradient value at the 1st, 2nd, ..., qth sampling positions and the real-time mechanical gradient value at the 1st, 2nd, ..., qth sampling positions, respectively. Intensive sampling was conducted in key areas to ensure the representativeness and reliability of the validation data; Step S4-2: Calculate the set of real-time topography and mechanical gradient data pairs. The q-th data pair Prediction residuals under the baseline mapping relationship : To obtain the real-time predicted residual set ,in, , , ..., Let represent the real-time prediction residuals of the 1st, 2nd, ..., qth real-time data pairs, respectively; The mapping relationship established using historical data is used to ensure the theoretical consistency of the verification process; Step S4-3: Set the real-time prediction residuals All predicted residuals and gradient thresholds Compare, if greater than If so, it is marked as invalid data for real-time topography and real-time mechanical data; Automated anomaly identification is achieved through threshold comparison, improving processing efficiency; Step S4-4: Remove invalid data identified as real-time topographic data and real-time mechanical data from the real-time data; use the data obtained after removing invalid data as compliant data to form a compliant dataset; Removing outlier data ensures dataset quality and provides a reliable data foundation for subsequent analysis.

[0011] The system includes: a historical data acquisition module, a gradient threshold determination module, a real-time data acquisition and preprocessing module, and a mechanical verification and compliance data generation module; The historical data acquisition module is used to acquire multimodal historical data of biological samples collected simultaneously by three probes, providing a data foundation for subsequent analysis; The gradient threshold determination module is used to establish a quantitative mapping relationship based on the spatial correlation between historical morphology and mechanical data, and to determine the judgment criteria for mechanical anomalies. The real-time data acquisition and preprocessing module is used to simultaneously acquire real-time topographic data and real-time mechanical data, and to perform gradient calculation on the real-time topographic data to identify regions with significant gradients in the real-time topographic data. The mechanical verification and compliance data generation module is used to verify the consistency of morphology and mechanics based on gradient thresholds and generate high-quality compliance datasets.

[0012] The gradient threshold determination module includes: The coordinate system establishment unit is used to establish a unified coordinate reference datum to ensure spatial alignment of multi-source data; Historical gradient significant region identification unit is used to identify regions with significant historical morphological changes and locate key analysis sites; The gradient data acquisition unit is used to simultaneously acquire topographic and mechanical gradient data and establish spatially correlated data pairs. The benchmark mapping establishment unit is used to establish the quantitative mapping relationship between the topographic gradient and the mechanical gradient; The prediction residual calculation unit is used to calculate the deviation between the actual measured value and the predicted value, and to quantify the degree of mechanical response anomaly. The threshold setting unit is used to set gradient thresholds based on the residual distribution to determine the anomaly judgment criteria.

[0013] The real-time data acquisition and preprocessing module includes: The real-time data acquisition unit is used to synchronously acquire real-time multimodal data to ensure data spatiotemporal consistency. The real-time gradient saliency region identification unit is used to identify regions with significant real-time topographic gradients.

[0014] The real-time data acquisition and mechanical verification module includes: The real-time gradient data acquisition unit is used to acquire real-time topographic and mechanical gradient data pairs. The real-time prediction residual calculation unit is used to calculate the prediction residual of real-time data; The invalid mechanical data identification unit is used to identify abnormal mechanical data based on a gradient threshold. The compliance dataset generation unit is used to remove all invalid data and generate high-quality compliance datasets.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. By analyzing the spatial distribution characteristics of historical morphology and mechanical data, a benchmark mapping relationship between morphological gradient and mechanical gradient was established, realizing the quantitative correlation analysis between morphological features and mechanical response, and solving the technical problem of the difficulty in effectively integrating multi-source data in traditional methods.

[0016] 2. Based on the gradient threshold set by the predicted residual distribution, it can automatically judge the consistency of the real-time acquired morphological and mechanical data, effectively identify and eliminate abnormal data in the measurement process, and significantly improve data quality and reliability.

[0017] 3. Through systematic data processing steps, from historical data analysis to real-time data verification, a complete quality control closed loop is formed, ensuring the consistency and repeatability of the sample feature extraction process and providing reliable technical support for sample characterization. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the precise feature extraction method based on AFM multi-source data analysis of the present invention. Figure 2 This is a schematic diagram of the feature extraction system based on AFM multi-source data analysis of the present invention. Figure 3 This is a schematic diagram of the gradient threshold determination module of the feature extraction system based on AFM multi-source data analysis of the present invention. Detailed Implementation

[0019] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example: Figure 1 , Figure 2 and Figure 3 As shown, this invention provides a technical solution: a feature extraction system and method based on AFM multi-source data analysis. The methods include: Step S1: Obtain the historical dataset of the physical feature quantitative characterization experiment based on three-probe atomic force nanomanipulation. The historical dataset contains several sets of synchronously acquired historical morphological data and historical mechanical data. The historical morphological data includes historical sample surface height information obtained by probe scanning. Step S2: Traverse the historical morphology data to identify regions with significant historical morphology gradients; based on these regions, simultaneously extract the morphology gradient values ​​and mechanical gradient values ​​to form several sets of morphology and mechanical gradient correlation data; based on this correlation data, establish a baseline mapping relationship between the morphology gradient and the mechanical gradient through regression analysis, and determine the gradient threshold based on the predicted residual distribution of the baseline mapping relationship. Step S3: Simultaneously acquire real-time morphological data and real-time mechanical data of the biological sample to be tested; perform gradient calculation on the real-time morphological data to identify regions with significant real-time morphological gradients; Step S4: Using the gradient threshold, perform consistency judgment on real-time topography data and real-time mechanical data. If the data exceeds the gradient threshold, it is marked as invalid data of real-time topography data and real-time mechanical data. Remove the invalid data marked as invalid data of real-time topography data and real-time mechanical data from the real-time data to generate a compliant dataset.

[0021] Furthermore, step S1 includes: The samples were scanned using a three-probe atomic force nanomanipulation system under standard culture conditions. Ten sets of simultaneous data were acquired, each set containing: Historical topographic data: 256×256 pixel height image; Historical mechanical data: Distribution maps of the corresponding regions, obtained through force versus distance curves.

[0022] Furthermore, step S2 includes: Step S2-1: Define a reference coordinate system with the bottom of the petri dish as the reference plane and the system's default fast scan direction as the reference direction.

[0023] Step S2-2: The historical morphology data is processed into a grid, dividing the sample surface into 32×32 grid regions. The average gradient magnitude within all grid regions is calculated, and regions with average gradient magnitudes greater than a preset gradient magnitude threshold are identified. The grid region is considered as a region with significant historical morphological gradients. Among them, , This represents the median of the gradient magnitude across all regions in the historical topographic data. According to calculations, nm / pixel nm / pixel, identifying gradient magnitude greater than The region of nm / pixel is considered as a region with significant historical morphological gradient; Step S2-3: Select m=500 sampling locations within the historical significant topographic gradient region, and simultaneously extract topographic gradient values ​​and mechanical gradient values. Specific process: The process of calculating the topographic gradient value is as follows: For each sampling position i, the topographic gradient value is calculated. : Obtain the topographic height values ​​of sampling point i and its neighboring points: the current position height is The height of adjacent points in the reference direction is The height of the adjacent point on the other side of the reference direction is... ; The gradient is calculated using the central difference method: ; when Pixel = 0.5nm nm, When nm, nm / pixel; The process of calculating the mechanical gradient value: For the same sampling position i, calculate the mechanical gradient value. : Obtain the mechanical values ​​of sampling point i and its neighboring points: the mechanical value at the current position is... The mechanical values ​​of adjacent points in the reference direction are The mechanical value of the adjacent point on the other side of the reference direction is... ; The gradient is calculated using the central difference method: ; when kPa, When kPa, kPa / pixel; Forming a set of data pairs ; Step S2-4: Based on the data set The least squares method is used to establish the baseline mapping relationship: Calculation process: calculate Average value: nm / pixel; calculate Average value: kPa / pixel; Calculate covariance: ; calculate variance: ; Calculate the regression coefficients: ; Calculate the intercept: ; The regression equation is obtained as follows: ; Step S2-5: For each data pair Calculate the predicted residuals: To form the predicted residual set ; Step S2-6: Sort all elements in the predicted residual set E by numerical value: the sorted set is {0.01, 0.03, 0.05, ..., 0.45, 0.52, 0.61}. Taking the 95th quantile position: 500 × 0.95 = 475, the 475th element in the sorted set, 0.35, is taken as the gradient threshold. .

[0024] Furthermore, step S3 includes: Step S3-1: Synchronously acquire real-time morphological data and real-time mechanical data of the sample; Step S3-2: Calculate the gradient of the real-time topographic data and identify gradient magnitudes greater than 1. The region of nm / pixel is used as the region with significant real-time topographic gradient.

[0025] Furthermore, step S4 includes: Step S4-1: Select q=300 sampling locations within the salient region: Extract real-time topographic gradient values. and real-time mechanical gradient values The gradient value is calculated using the same central difference method as in steps S2-3, forming a real-time data pair set. ; Step S4-2: For each data pair ,calculate: To form a real-time prediction residual set ; Step S4-3: ... Each element in The comparison showed that 35 elements exceeded the threshold, and these 35 data points were identified as invalid data. Step S4-4: Remove 35 invalid data points from the 300 data points, leaving 265 valid data points to form a compliant dataset; The compliant dataset containing 265 data points is input into a pre-trained convolutional neural network enhancement model, which outputs a super-resolution topographic image of the sample and the corresponding mechanical feature distribution map.

[0026] The system includes: a historical data acquisition module, a gradient threshold determination module, a real-time data acquisition and preprocessing module, and a mechanical verification and compliance data generation module; The historical data acquisition module is used to acquire multimodal historical data of biological samples collected simultaneously by three probes, providing a data foundation for subsequent analysis; The gradient threshold determination module is used to establish a quantitative mapping relationship based on the spatial correlation between historical morphology and mechanical data, and to determine the judgment criteria for mechanical anomalies. The real-time data acquisition and preprocessing module is used to simultaneously acquire real-time topographic data and real-time mechanical data, and to perform gradient calculation on the real-time topographic data to identify regions with significant gradients in the real-time topographic data. The mechanical verification and compliance data generation module is used to verify the consistency of morphology and mechanics based on gradient thresholds and generate high-quality compliance datasets.

[0027] The gradient threshold determination module includes: The coordinate system establishment unit is used to establish a unified coordinate reference datum to ensure spatial alignment of multi-source data; Historical gradient significant region identification unit is used to identify regions with significant historical morphological changes and locate key analysis sites; The gradient data acquisition unit is used to simultaneously acquire topographic and mechanical gradient data and establish spatially correlated data pairs. The benchmark mapping establishment unit is used to establish the quantitative mapping relationship between the topographic gradient and the mechanical gradient; The prediction residual calculation unit is used to calculate the deviation between the actual measured value and the predicted value, and to quantify the degree of mechanical response anomaly. The threshold setting unit is used to set gradient thresholds based on the residual distribution to determine the anomaly judgment criteria.

[0028] The real-time data acquisition and preprocessing module includes: The real-time data acquisition unit is used to synchronously acquire real-time multimodal data to ensure data spatiotemporal consistency. The real-time gradient saliency region identification unit is used to identify regions with significant real-time topographic gradients.

[0029] The real-time data acquisition and mechanical verification module includes: The real-time gradient data acquisition unit is used to acquire real-time topographic and mechanical gradient data pairs. The real-time prediction residual calculation unit is used to calculate the prediction residual of real-time data; The invalid mechanical data identification unit is used to identify abnormal mechanical data based on a gradient threshold. The compliance dataset generation unit is used to remove all invalid data and generate high-quality compliance datasets.

[0030] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for precise feature extraction based on AFM multi-source data analysis, characterized by: Includes the following steps: Step S1: Obtain the historical dataset of the physical feature quantitative characterization experiment based on three-probe atomic force nanomanipulation. The historical dataset contains several sets of synchronously acquired historical morphological data and historical mechanical data. The historical morphological data includes historical sample surface height information obtained by probe scanning. Step S2: Traverse the historical morphology data to identify regions with significant historical morphology gradients; based on these regions, simultaneously extract the morphology gradient values ​​and mechanical gradient values ​​to form several sets of morphology and mechanical gradient correlation data; based on this correlation data, establish a baseline mapping relationship between the morphology gradient and the mechanical gradient through regression analysis, and determine the gradient threshold based on the predicted residual distribution of the baseline mapping relationship. Step S3: Simultaneously acquire real-time morphological data and real-time mechanical data of the biological sample to be tested; perform gradient calculation on the real-time morphological data to identify regions with significant real-time morphological gradients; Step S4: Using the gradient threshold, perform consistency judgment on real-time topography data and real-time mechanical data. If the data exceeds the gradient threshold, it is marked as invalid data of real-time topography data and real-time mechanical data. Remove the invalid data marked as invalid data of real-time topography data and real-time mechanical data from the real-time data to generate a compliant dataset.

2. The feature extraction method based on AFM multi-source data analysis according to claim 1, characterized in that: Step S2 includes: Step S2-1: Define the plane of the sample-supporting substrate as the xOy reference plane, define the scanning direction of the physical characteristic quantitative characterization experiment based on three-probe atomic force nanomanipulation as the reference direction as the positive x-axis, the direction perpendicular to the x-axis as the positive y-axis, and the direction perpendicular to the xOy reference plane upward as the positive z-axis to establish the sample reference coordinate system. Step S2-2: Divide the sample surface into several grid regions; calculate the average gradient magnitude in all grid regions, and identify those whose average gradient magnitude is greater than a preset gradient magnitude threshold. The grid region is considered as a region with significant historical morphological gradient; among which, , This represents the median gradient magnitude across all grid regions in the historical topographic data. Scaling factor indicating the determination of significant regions in historical morphological gradients; Step S2-3: Select m sampling locations within the historical morphological gradient significant region, and simultaneously extract the morphological gradient values ​​and mechanical gradient values ​​at the m sampling locations. The morphological gradient value specifically refers to the rate of change of the morphological data in the reference direction, and the mechanical gradient value specifically refers to the rate of change of the mechanical data in the reference direction; form a data pair set composed of morphological gradient values ​​and mechanical gradient values. ,in, , , ..., , ..., These represent data pairs consisting of the topographic gradient value at the 1st, 2nd, ..., ith, ..., mth sampling positions and the mechanical gradient value at the 1st, 2nd, ..., ith, ..., mth sampling positions, respectively. Step S2-4: Based on the data, adjust the set The least squares method is used to establish a baseline mapping relationship between the topographic gradient and the mechanical gradient. This baseline mapping relationship is represented by the following linear equation: ,in, Represents the morphological gradient. This represents the predicted mechanical gradient. The regression intercept coefficient, The regression slope coefficient; Step S2-5: Calculate the data pair set The i-th data pair Predicted residuals under the reference mapping relationship : The predicted residual set is obtained. ,in, , , ..., , ..., Let represent the prediction residuals of the 1st, 2nd, ..., i-th, ..., m-th data pairs, respectively; Step S2-6: Set the predicted residuals All elements are sorted numerically, and the values ​​at preset quantile positions among the sorted elements are set as the gradient threshold. .

3. The feature extraction method based on AFM multi-source data analysis according to claim 2, characterized in that: Step S3 includes: Step S3-1: Simultaneously acquire real-time morphological data and real-time mechanical data of the biological sample to be tested; Step S3-2: Perform gradient calculation on the real-time topography data, and identify gradient magnitudes that are greater than a preset gradient magnitude. The region is considered as the region with significant real-time topographic gradient.

4. The feature extraction method based on AFM multi-source data analysis according to claim 3, characterized in that: Step S4 includes: Step S4-1: Select q sampling locations within the significant region of the real-time topography gradient, and extract the real-time topography gradient value at the q-th sampling location. Simultaneously extract the real-time mechanical gradient value at the q-th sampling position. This forms a set of real-time topographic and mechanical gradient data. ,in, , , ..., These represent real-time data pairs consisting of the real-time topographic gradient value at the 1st, 2nd, ..., qth sampling positions and the real-time mechanical gradient value at the 1st, 2nd, ..., qth sampling positions, respectively. Step S4-2: Calculate the set of real-time topography and mechanical gradient data pairs. The q-th data pair Prediction residuals under the baseline mapping relationship : To obtain the real-time predicted residual set ,in, , , ..., Let represent the real-time prediction residuals of the 1st, 2nd, ..., qth real-time data pairs, respectively; Step S4-3: Set the real-time prediction residuals All predicted residuals and gradient thresholds Compare, if greater than If so, it is marked as invalid data for real-time topography and real-time mechanical data; Step S4-4: Remove invalid data identified as real-time topographic data and real-time mechanical data from the real-time data; use the data obtained after removing invalid data as compliant data to form a compliant dataset.

5. A feature extraction system based on AFM multi-source data analysis, used to execute the feature extraction method based on AFM multi-source data analysis as described in any one of claims 1-4, characterized in that: The system includes: Historical data acquisition module, gradient threshold determination module, real-time data acquisition and preprocessing module, and mechanical verification and compliance data generation module; The historical data acquisition module is used to acquire multimodal historical data of biological samples collected simultaneously by three probes, providing a data foundation for subsequent analysis; The gradient threshold determination module is used to establish a quantitative mapping relationship based on the spatial correlation between historical morphology and mechanical data, and to determine the judgment criteria for mechanical anomalies. The real-time data acquisition and preprocessing module is used to simultaneously acquire real-time topographic data and real-time mechanical data, and to perform gradient calculation on the real-time topographic data to identify regions with significant gradients in the real-time topographic data. The mechanical verification and compliance data generation module is used to verify the consistency of morphology and mechanics based on gradient thresholds and generate high-quality compliance datasets.

6. The feature extraction system based on AFM multi-source data analysis according to claim 5, characterized in that: The gradient threshold determination module includes: The coordinate system establishment unit is used to establish a unified coordinate reference datum to ensure spatial alignment of multi-source data; Historical gradient significant region identification unit is used to identify regions with significant historical morphological changes and locate key analysis sites; The gradient data acquisition unit is used to simultaneously acquire topographic and mechanical gradient data and establish spatially correlated data pairs. The benchmark mapping establishment unit is used to establish the quantitative mapping relationship between the topographic gradient and the mechanical gradient; The prediction residual calculation unit is used to calculate the deviation between the actual measured value and the predicted value, and to quantify the degree of mechanical response anomaly. The threshold setting unit is used to set gradient thresholds based on the residual distribution to determine the anomaly judgment criteria.

7. The feature extraction system based on AFM multi-source data analysis according to claim 6, characterized in that: The real-time data acquisition and preprocessing module includes: The real-time data acquisition unit is used to synchronously acquire real-time multimodal data to ensure data spatiotemporal consistency. The real-time gradient saliency region identification unit is used to identify regions with significant real-time topographic gradients.

8. The feature extraction system based on AFM multi-source data analysis according to claim 7, characterized in that: The real-time data acquisition and mechanical verification module includes: The real-time gradient data acquisition unit is used to acquire real-time topographic and mechanical gradient data pairs. The real-time prediction residual calculation unit is used to calculate the prediction residual of real-time data; The invalid mechanical data identification unit is used to identify abnormal mechanical data based on a gradient threshold. The compliance dataset generation unit is used to remove all invalid data and generate high-quality compliance datasets.

Images