A single-molecule detection method based on AI and electrochemiluminescence

Abstract

The present invention relates to the field of electrochemiluminescence detection applications and discloses a single-molecule detection method based on AI and electrochemiluminescence. The method comprises: collecting electrochemiluminescence signal data from a sample to generate a data set, receiving real-time signals to construct a matrix; processing the signal data matrix, extracting features based on the data set, predicting single-molecule distribution based on signal intensity gradients and detection path information, outputting spatiotemporal distribution features through an AI model and updating the data set; and dynamically adjusting detection parameters, such as sensor sensitivity, light source intensity, and sampling frequency, based on the distribution features. By combining AI with electrochemiluminescence, this method enables the determination of single-molecule detection features and the dynamic adjustment of detection parameters, thereby improving the accuracy and efficiency of single-molecule detection.

Description

Technical Field

[0001] The present invention relates to the field of electrochemiluminescence detection technology, and specifically to a single molecule detection method based on AI and electrochemiluminescence. Background Art

[0002] In the field of biological detection, single-molecule detection technology, due to its ability to detect biomolecules at the single-molecule level with high sensitivity and specificity, has shown great application potential in many fields such as disease diagnosis, drug development, and life science research. However, current single-molecule detection technology still faces many challenges.

[0003] While traditional single-molecule detection methods, such as fluorescence microscopy, can achieve single-molecule detection to a certain extent, they suffer from low sensitivity and significant background noise due to limitations in their detection principles. In actual detection, the presence of background noise can severely impact the accuracy of test results, making it difficult to accurately identify and analyze single-molecule signals.

[0004] As an emerging detection technology, electrochemiluminescence (ECL) detection offers advantages such as high sensitivity and good stability, and holds considerable promise for single-molecule detection. However, single-use ECL detection still has some drawbacks when applied to complex biological samples. For example, fluctuations in signal intensity and uneven molecular distribution within the detection area can adversely affect test results.

[0005] Furthermore, existing single-molecule detection systems often lack effective data processing and analysis tools when dealing with large amounts of detection data. Traditional data processing methods struggle to quickly and accurately extract useful information from massive amounts of data, making it impossible to monitor and dynamically adjust the single-molecule detection process in real time, thus impacting detection efficiency and accuracy.

[0006] With the rapid development of artificial intelligence (AI) technology, combining it with electrochemiluminescence (ECL) detection has provided new insights into solving these problems. However, effectively applying AI to ECL single-molecule detection to achieve real-time signal processing, feature extraction, and single-molecule distribution prediction remains an urgent challenge.

[0007] Existing single-molecule detection research based on AI and electrochemiluminescence (ECL) suffers from issues such as suboptimal signal processing algorithms and inadequate integration between AI models and ECL detection systems. These issues lead to unstable detection system performance, making it difficult to meet the needs of practical applications. Summary of the Invention

[0008] The purpose of the present invention is to provide a single molecule detection method based on AI and electrochemiluminescence to solve the problems raised in the above background technology.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a single molecule detection method based on AI and electrochemiluminescence, the method comprising:

[0010] Multiple electrochemiluminescence sensors are deployed in a detection area and an AI processing unit connected to the electrochemiluminescence sensors, with two adjacent electrochemiluminescence sensors being a set distance apart. The method comprises: collecting electrochemiluminescence signal data of samples in the area through the electrochemiluminescence sensor to generate a signal data set; receiving real-time signal data of multiple electrochemiluminescence sensors through the AI ​​processing unit to construct a signal data matrix; determining the characteristics of single-molecule detection based on the signal data set and the real-time data of each electrochemiluminescence sensor, wherein determining the characteristics of single-molecule detection comprises: processing the signal data matrix, performing feature extraction in combination with the signal data set, and predicting single-molecule distribution based on signal intensity gradient and detection path information, outputting the spatiotemporal distribution characteristics of single-molecule detection through the AI ​​model, and updating the signal data set based on the spatiotemporal distribution characteristics; and dynamically adjusting the detection parameters based on the distribution characteristics.

[0011] Preferably, the characteristics of determining single-molecule detection include: processing the signal data matrix to extract signal intensity distribution, noise interference characteristics and single-molecule trends; performing signal intensity modeling on the signal data matrix according to the signal intensity distribution and noise interference characteristics, dividing the detection area into multiple sub-areas and marking the area identifiers, performing correlation matching with the signal data set based on the signal intensity of the sub-area, and marking the area identifiers in the signal data set; calculating the signal intensity gradient according to the position of the electrochemiluminescence sensor, predicting the distribution of single-molecule detection according to the signal intensity gradient and the single-molecule trend, and calculating the single-molecule prediction information of each sub-area; constructing an AI model, using the single-molecule prediction information as the input parameter of the AI ​​model, performing spatial correlation modeling on the single-molecule prediction information through the AI ​​model, and outputting the spatiotemporal distribution characteristics of single-molecule detection; updating the signal data set according to the spatiotemporal distribution characteristics to obtain the distribution characteristics of single-molecule detection.

[0012] Preferably, the processing of the signal data matrix includes: standardizing the signal data matrix, intercepting the signal hotspot area in the matrix through a sliding window, filtering the background noise of the hotspot area, and calculating the single molecule trend through a feature decomposition algorithm; calculating the spatial correlation characteristics of the signal data matrix, calculating the interference intensity, signal stability coefficient and detection blind area index between regions based on the spatial correlation characteristics, constructing a feature fusion network, and calculating the noise interference characteristics through the feature fusion network; extracting the time domain characteristics and frequency domain characteristics collected by each electrochemiluminescence sensor, calculating the signal characteristic vector of the sensor based on the phase difference between the time domain characteristics and the frequency domain characteristics, performing feature matching on the electrochemiluminescence sensors at different positions based on the signal characteristic vector, and calculating the single molecule trend.

[0013] Preferably, the signal strength modeling of the signal data matrix includes: extracting signal strength sampling points in each frame of data according to the signal strength distribution, performing correlation mapping based on the sampling points and noise interference characteristics, generating a signal strength spectrum, spatially aligning the signal strength spectra collected by multiple sensors, and calculating the signal strength distribution of the area; setting an interference threshold value, locating the interference source according to the signal strength values ​​of the multi-frame signal data matrix, calculating the interference strength difference, if the interference strength difference is greater than or equal to the interference threshold value, indicating that there is a detection blind spot in the area, performing detection model constraint compensation on the current area, iteratively correcting the signal strength distribution of the current area according to the path loss model corresponding to the current area, and calculating the strength compensation value of the blind spot signal according to the correction result; performing signal strength modeling on the signal data matrix according to the signal strength distribution, and performing interference marking on the signal model of the area through the noise interference characteristics.

[0014] Preferably, the signal intensity gradient is calculated according to the position of the electrochemiluminescence sensor, including: extracting signal intensity change points according to multiple sets of signal detection data, and mapping the change points to a unified detection coordinate system according to the deployment position of the sensor, fitting the change points by a spatial interpolation algorithm, and generating a signal field model of the region; sampling at equal intervals along the detection path of the signal field model, calculating the signal attenuation rate, interference fluctuation index and signal change slope of the path according to the sampling results, and calculating the signal change parameter according to the signal attenuation rate, interference fluctuation index and signal change slope; according to the deployment parameters and acquisition accuracy of the electrochemiluminescence sensor, the single molecule detection is placed in each frame. The distribution characteristics of the data are projected onto a signal field model, which is partitioned along the detection direction of the signal field model according to the number of sensors. The change law of single-molecule detection in the partition is analyzed, and the distribution characteristics are calculated based on the change law. The signal intensity gradient is calculated based on the signal change parameter and the distribution characteristics. The calculation process of the signal intensity gradient includes: based on the detection position range from the first electrochemiluminescence sensor to the last electrochemiluminescence sensor, selecting spatial coordinate points in the sensor deployment direction, accumulating the product of the signal field strength characteristic weight value and the single-molecule distribution characteristic weight value within the spatial resolution range, and superimposing the influence value of the sensor acquisition frequency on the signal intensity change rate.

[0015] Preferably, the calculation of the single molecule prediction information for each sub-region includes: using the detection main diameter of the signal field model as a baseline, the peak position of the single molecule detection in each frame of data as a reference point, calculating the distribution offset, and drawing a distribution curve according to the detection coordinates; correcting the growth rate and direction in the single molecule trend according to the signal intensity gradient; starting from the most recent distribution point, continuing to draw the distribution curve according to the correction results of the growth rate and direction, generating the distribution points for the next time period, until the distribution points cover the entire target area, and generating the single molecule prediction information.

[0016] Preferably, the AI ​​model includes:

[0017] The input layer is used to organize the single-molecule prediction information into spatial distribution data and perform normalization;

[0018] The feature fusion layer is used to extract the detected regional correlation features by processing the spatial distribution data and build the dependency relationship between the detection units;

[0019] The parameter optimization layer is used to integrate the correlation relationship of single-molecule detection in spatial units and generate detection parameter adjustment strategies.

[0020] Preferably, the obtaining of the distribution characteristics of single-molecule detection includes: according to the spatiotemporal distribution characteristics of single-molecule detection output by the AI ​​model, the identification of the sub-region is matched with the spatiotemporal distribution characteristics; the regional data in the signal data set is reorganized according to the spatiotemporal characteristics to generate a regional distribution map sorted by the single-molecule detection intensity; based on the reorganized regional distribution map, the optimized single-molecule detection distribution characteristics are output.

[0021] Preferably, the dynamic adjustment of the detection parameters includes: mapping the region identifiers to the regions of the distribution characteristics of the single molecule detection one by one; controlling the execution of the detection parameters according to the spatiotemporal distribution characteristics of the single molecule detection, including sensor sensitivity adjustment, light source intensity control and sampling frequency operation; and dynamically allocating the detection parameters to the corresponding detection areas according to the spatial distribution of the single molecule detection and the preset parameter adjustment strategy.

[0022] Preferably, the execution action of controlling the detection parameters includes: triggering a sensitivity enhancement instruction of a neighboring sensor when the single-molecule detection reaches a preset intensity threshold in the target area; dynamically combining available light source resources according to the light source intensity strategy to generate a sampling frequency vector; and adjusting the optical parameters of the target area detection node based on the sampling frequency vector.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] In terms of signal processing, this method effectively removes background noise interference and extracts more accurate single-molecule signal features through a series of operations, including normalizing the signal data matrix, capturing hotspot areas with a sliding window, filtering background noise, and calculating single-molecule trends using a feature decomposition algorithm. Furthermore, by calculating the spatial correlation characteristics of the signal data matrix, constructing a feature fusion network to calculate noise interference characteristics, extracting time-domain and frequency-domain features, and calculating the sensor's signal eigenvectors for feature matching, the accuracy and reliability of signal processing are further improved, enabling the detection system to more accurately identify and analyze single-molecule signals.

[0025] In terms of determining the characteristics of single-molecule detection, this method models the signal intensity distribution, divides the detection area into multiple sub-areas and marks the area identifiers, combines the signal intensity gradient and single-molecule trend to predict the distribution of single-molecule detection, constructs an AI model for spatial correlation modeling, and outputs spatiotemporal distribution characteristics. This method can accurately determine the characteristics of single-molecule detection and achieve precise prediction of the spatiotemporal distribution of single molecules within the detection area. This method can fully utilize the powerful data processing and analysis capabilities of the AI ​​model to extract valuable information from large amounts of detection data, providing a reliable basis for the subsequent dynamic adjustment of detection parameters.

[0026] In terms of dynamic adjustment of detection parameters, this method maps regional identifiers with distribution characteristics based on the spatiotemporal distribution characteristics output by the AI ​​model, controls the execution of detection parameters such as sensor sensitivity adjustment, light source intensity control, and sampling frequency operation, and dynamically allocates detection parameters to corresponding areas based on spatial distribution and preset strategies. This dynamic adjustment mechanism enables the detection system to optimize detection parameters in real time based on the actual distribution of single molecules, thereby improving the sensitivity and accuracy of detection. For example, when the single-molecule detection reaches a preset intensity threshold in the target area, the sensitivity enhancement instruction of the adjacent sensor is triggered, which can more accurately capture the single-molecule signal; by dynamically combining available light source resources to generate a sampling frequency vector and adjust the optical parameters, the detection process can be optimized and the detection efficiency can be improved.

[0027] In terms of data processing and analysis, this method constructs a signal data matrix and utilizes an AI processing unit to process real-time signal data from multiple electrochemiluminescence sensors, enabling efficient processing and analysis of massive amounts of detection data. The introduction of AI models enables the detection system to quickly and accurately extract useful information from large amounts of data, enabling real-time monitoring and dynamic adjustment of the single-molecule detection process, significantly improving detection efficiency and accuracy.

[0028] Furthermore, through precise calculation of signal intensity gradients, including extraction of signal intensity change points, mapping to a unified coordinate system, fitting and generating a signal field model, and calculating signal change parameters and distribution characteristics, this method can more accurately describe the distribution patterns and trends of single molecules within the detection region, providing a more reliable basis for distribution prediction in single-molecule detection. Furthermore, by identifying and compensating for blind spots, the method can effectively improve the overall performance of the detection system and reduce the impact of blind spots on detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a diagram showing the working principle of the single-molecule detection method based on AI and electrochemiluminescence according to the present invention;

[0030] Figure 2 Flowchart for characterization of single-molecule detection;

[0031] Figure 3 Flowchart for signal data matrix processing;

[0032] Figure 4 Flowchart for the generation of prediction information for single molecules. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0034] See also Figures 1-4 The present invention relates to a single molecule detection method based on AI and electrochemiluminescence, and the specific implementation steps are as follows:

[0035] The electrochemiluminescence signal data of samples in the area are collected through the electrochemiluminescence sensor to generate a signal data set; the real-time signal data of multiple electrochemiluminescence sensors are received through the AI ​​processing unit to construct a signal data matrix; the characteristics of single-molecule detection are determined based on the signal data set and the real-time data of each electrochemiluminescence sensor. This process includes processing the signal data matrix, extracting features based on the signal data set, and predicting the single-molecule distribution based on the signal intensity gradient and detection path information. The spatiotemporal distribution characteristics of single-molecule detection are output through the AI ​​model, and then the signal data set is updated according to the spatiotemporal distribution characteristics; finally, the detection parameters are dynamically adjusted according to the distribution characteristics.

[0036] Example 1:

[0037] This example details a specific implementation method for determining single-molecule detection characteristics in a single-molecule detection method based on AI and electrochemiluminescence. This process is performed in a biological detection system that includes multiple electrochemiluminescence sensors deployed in a detection area and connected to an AI processing unit. Adjacent sensors are spaced a predetermined distance apart.

[0038] The signal data matrix is ​​processed to extract key features. The signal data matrix constructed from the real-time signal data of multiple electrochemiluminescence sensors received by the AI ​​processing unit is standardized, and the signal hotspot areas in the matrix are intercepted using a sliding window. This allows focus on the parts where the signal changes significantly. Background noise is filtered in the hotspot areas to remove interfering signals and make the effective signals more prominent. The single-molecule trend is then calculated using the feature decomposition algorithm to extract the overall direction of single-molecule behavior from the complex signal.

[0039] Calculate the spatial correlation characteristics of the signal data matrix. Based on these characteristics, calculate the interference strength between regions, signal stability coefficient, and detection blind zone index. Then, construct a feature fusion network, and use it to calculate the noise interference characteristics, thereby fully understanding the spatial interaction of signals and noise conditions.

[0040] The time-domain and frequency-domain features collected by each ECL sensor are extracted. The sensor's signal feature vector is calculated based on the phase difference between the time-domain and frequency-domain features. This feature vector is used to match the features of ECL sensors at different locations, and then single-molecule trends are calculated to analyze the dynamic changes of single molecules from different dimensions.

[0041] Signal strength modeling is performed on the signal data matrix based on the extracted signal strength distribution and noise interference characteristics. Based on the signal strength distribution, signal strength sampling points are extracted from each frame of data. These sampling points are then mapped and correlated with the noise interference characteristics to generate a signal strength map. The signal strength maps collected by multiple sensors are spatially aligned, and the regional signal strength distribution is calculated using a unified spatial reference system, ensuring comparability and spatial consistency of data from each sensor.

[0042] An interference threshold is set, and interference sources are located based on the signal strength values ​​of the multi-frame signal data matrix. The interference strength difference is calculated. If the interference strength difference is greater than or equal to the interference threshold, it indicates a detection blind spot in the area. Detection model constraint compensation is then applied to the current area. Based on the path loss model corresponding to the current area, the signal strength distribution in the current area is iteratively corrected. Based on the correction results, a strength compensation value for the blind spot signal is calculated to compensate for the signal loss caused by the detection blind spot.

[0043] The signal strength modeling of the signal data matrix is ​​carried out according to the signal strength distribution, and the signal model of the region is interfered with by the noise interference characteristics to clarify the degree and type of interference of the signal in each region.

[0044] The signal intensity gradient is then calculated based on the position of the electrochemiluminescence sensor. Signal intensity change points are extracted from multiple sets of signal detection data. These change points are mapped to a unified detection coordinate system based on the sensor deployment location. A spatial interpolation algorithm is used to fit the change points to generate a regional signal field model that intuitively reflects the signal distribution and change trends in space.

[0045] Sampling is performed at equal intervals along the detection path of the signal field model. The signal attenuation rate, interference fluctuation index, and signal change slope of the path are calculated based on the sampling results. The signal change parameters are then calculated from these parameters to quantify the signal changes along the detection path.

[0046] Based on the deployment parameters and acquisition accuracy of the electrochemiluminescence sensor, the distribution characteristics of single-molecule detection in each frame of data are projected onto the signal field model. The signal field model is then partitioned along the detection direction according to the number of sensors. The changing patterns of single-molecule detection in each partition are analyzed, and the distribution characteristics are calculated to gain a deeper understanding of the distribution characteristics of single molecules in different regions.

[0047] The signal intensity gradient is calculated based on signal variation parameters and distribution characteristics. The specific process is as follows: Based on the detection position range from the first to the last electrochemiluminescence sensor, spatial coordinate points are selected in the sensor deployment direction. The product of the signal field intensity characteristic weight value and the single molecule distribution characteristic weight value within the spatial resolution range is cumulatively calculated. At the same time, the effect of the sensor acquisition frequency on the signal intensity change rate is added. An accurate signal intensity gradient is obtained by comprehensively considering multiple factors.

[0048] The distribution of single-molecule detections is predicted based on the signal intensity gradient and single-molecule trend, and single-molecule prediction information is calculated for each subregion. Using the main detection path of the signal field model as the baseline and the peak position of the single-molecule detection in each frame of data as the reference point, the distribution offset is calculated and a distribution curve is plotted according to the detection coordinates, visually displaying the current distribution of single molecules.

[0049] According to the signal intensity gradient, the growth rate and direction in the single molecule trend are corrected to make the prediction of the single molecule trend more consistent with the actual signal changes.

[0050] Starting from the most recent distribution point, the distribution curve is continued to be drawn according to the correction results of growth rate and direction, and the distribution points of the next time period are gradually generated until the distribution points cover the entire target area. Finally, single-molecule prediction information is generated to realize the prediction of the future distribution of single molecules.

[0051] Build an AI model and use the single-molecule prediction information as input parameters. The input layer of the AI ​​model is used to organize the single-molecule prediction information into spatially distributed data and perform standardization processing to make the input data format uniform and easier for model processing.

[0052] The feature fusion layer processes spatially distributed data, extracts the detected regional correlation features, builds the dependency relationship between detection units, and explores the intrinsic connections between different regions.

[0053] The parameter optimization layer integrates the correlation between single-molecule detection in spatial units, generates detection parameter adjustment strategies, and provides a basis for the dynamic adjustment of subsequent detection parameters.

[0054] The AI ​​model is used to perform spatial correlation modeling on the single-molecule prediction information, output the spatiotemporal distribution characteristics of the single-molecule detection, and then update the signal data set based on the spatiotemporal distribution characteristics to obtain the distribution characteristics of the single-molecule detection, completing the entire process of determining the single-molecule detection characteristics.

[0055] Example 2:

[0056] This example details a specific implementation method for signal intensity modeling of a signal data matrix in a single-molecule detection method based on AI and electrochemiluminescence. This process is implemented in a biological detection system consisting of multiple electrochemiluminescence sensors deployed in a detection area and connected to an AI processing unit. Adjacent sensors are spaced a predetermined distance apart. The goal is to accurately reflect the signal intensity distribution through modeling, providing a reliable data foundation for subsequent single-molecule detection.

[0057] Based on the signal strength distribution extracted from the signal data matrix, signal strength sampling points are extracted from each frame of data. These sampling points are selected based on the variation in signal strength and can effectively characterize the signal characteristics of that frame of data. Next, the extracted sampling points are correlated with noise interference characteristics. These noise interference characteristics, derived from processing the signal data matrix, include inter-regional interference intensity and signal stability coefficients. This correlation mapping clearly identifies the degree of noise interference at the sampling points, providing a basis for subsequent processing. The signal strength maps collected by multiple sensors are then spatially aligned. Due to the different deployment locations of each sensor, the collected data varies spatially. Spatial alignment integrates the data from each sensor within a unified spatial reference system, making the signal strength maps from different sensors comparable. This allows the regional signal strength distribution to be calculated, visually presenting the distribution of signal strength within the detection area.

[0058] An interference threshold is set, and the setting of this threshold is based on factors such as the actual needs of the detection system and the noise level. The interference source is located according to the signal strength value of the multi-frame signal data matrix, and the possible interference source position is determined by analyzing the change in signal strength in the multi-frame data. The interference intensity difference is calculated, and the interference intensity of the current frame is compared with the reference frame or the adjacent frame to obtain the difference. If the interference intensity difference is greater than or equal to the interference threshold, it indicates that there is a detection blind spot in the area. At this time, the detection model constraint compensation is performed on the current area. According to the path loss model corresponding to the current area, which describes the attenuation law of the signal intensity during transmission with factors such as distance, the signal intensity distribution of the current area is iteratively corrected. Through multiple iterations, the signal strength distribution is gradually adjusted to make it closer to the actual situation. The intensity compensation value of the blind spot signal is calculated based on the correction result to make up for the signal loss caused by the detection blind spot and improve the accuracy of the signal strength distribution.

[0059] After completing the correction and compensation of the signal intensity distribution, the signal data matrix is ​​modeled for signal intensity based on the finalized signal intensity distribution. During the modeling process, the spatial distribution characteristics of the signal within the detection area, the acquisition characteristics of each sensor, and noise interference are fully considered to construct a model that can accurately describe changes in signal intensity. At the same time, the regional signal model is annotated using noise interference characteristics to clarify the type and degree of signal interference in each area. For example, some areas are labeled as having high interference intensity, while others are labeled as having low interference intensity. This allows the model to reflect not only signal intensity but also interference conditions, providing more comprehensive information for subsequent single-molecule detection and analysis.

[0060] Throughout the signal strength modeling process, each step is closely linked and mutually influential. Extracting signal strength sampling points is the foundation of modeling. Only by accurately selecting sampling points can subsequent correlation mapping and spatial alignment provide reliable data. Spatial alignment ensures the consistency and comparability of multi-sensor data, making the calculated regional signal strength distribution more reasonable. Setting interference thresholds and locating interference sources are key to addressing detection blind spots. By properly setting thresholds and accurately locating interference sources, blind spot signals can be promptly detected and compensated. Iterative correction and intensity compensation further improve the accuracy of signal strength distribution, making the model more accurate to actual detection conditions. Interference annotation adds more detailed information to the model, facilitating a more in-depth analysis of signals and noise.

[0061] It is important to ensure temporal consistency when processing multi-frame signal data to accurately analyze changes in signal strength and the location of interference sources. When performing spatial alignment, it is important to select an appropriate reference coordinate system and alignment method to ensure that the signal strength maps of each sensor are accurately mapped into a unified space. When calculating interference intensity differences and performing iterative corrections, appropriate parameters must be set to avoid over-correction or under-correction.

[0062] Example 3:

[0063] This example details a specific implementation for calculating the signal intensity gradient based on the position of the ECL sensor in a single-molecule detection method based on AI and ECL. This process is performed within a biological detection system consisting of multiple ECL sensors deployed in a detection area and connected to an AI processing unit. Adjacent sensors are spaced a predetermined distance apart. The purpose of calculating the signal intensity gradient is to provide a key parameter for single-molecule distribution prediction.

[0064] Based on multiple sets of signal detection data, signal strength change points are extracted. These change points are locations where signal strength changes significantly and reflect the spatial characteristics of signal variation. These change points are then mapped to a unified detection coordinate system based on the sensor deployment locations. Since each sensor is deployed in a different location, a unified coordinate system ensures that all change points are in the same spatial reference system, facilitating subsequent processing. A spatial interpolation algorithm is used to fit the change points to generate a signal field model for the region. Based on discrete change points, the spatial interpolation algorithm infers the signal strength distribution across the entire region, forming a continuous signal field model that intuitively displays the spatial distribution of the signal.

[0065] Sampling is performed at equal intervals along the detection path of the signal field model. The detection path is set according to the detection requirements. Equally spaced sampling ensures that evenly distributed sample points are obtained along the path to accurately reflect signal changes along the path. Based on the sampling results, the signal attenuation rate, interference fluctuation index, and signal change slope of the path are calculated. The signal attenuation rate indicates the degree of signal strength attenuation during transmission, the interference fluctuation index reflects the fluctuation of the signal due to interference, and the signal change slope reflects the rate of change of signal strength with spatial position. Based on these parameters, the signal change parameter is calculated. This parameter integrates the attenuation, interference, and change rate characteristics of the signal along the path to describe the overall signal change along the detection path.

[0066] Based on the deployment parameters and acquisition accuracy of the electrochemiluminescence (ECL) sensors, the distribution characteristics of single-molecule detections in each frame of data are projected onto the signal field model. Deployment parameters include sensor position and spacing, while acquisition accuracy affects data accuracy. Projection allows the distribution of single molecules to be correlated with the signal field model. Based on the number of sensors, the signal field model is partitioned along the detection direction. The number of partitions is determined by the number of sensors, allowing the signal characteristics within each partition to be correlated with the corresponding sensor data for easier analysis. The variation patterns of single-molecule detection within each partition are analyzed, such as the distribution density and movement direction of single molecules within each partition. Based on these variation patterns, distribution characteristics are calculated, which describe the distribution and variation characteristics of single molecules within different partitions.

[0067] The signal intensity gradient is calculated based on the signal change parameters and distribution characteristics. The calculation process of the signal intensity gradient is as follows: Based on the detection position range from the first electrochemiluminescence sensor to the last electrochemiluminescence sensor, a spatial coordinate point is selected in the sensor deployment direction. For each spatial coordinate point, the product of the signal field intensity characteristic weight value and the single molecule distribution characteristic weight value within the spatial resolution range is cumulatively calculated, where the signal field intensity characteristic weight value reflects the degree of influence of the signal field intensity at the point, and the single molecule distribution characteristic weight value reflects the effect of the single molecule distribution on the point. At the same time, the influence of the sensor acquisition frequency on the signal intensity change rate is superimposed. The higher the acquisition frequency, the greater the influence on the signal intensity change rate. It can be expressed as:

[0068]

[0069] in, represents the signal intensity gradient; is the number of selected spatial coordinate points; For the The signal field strength characteristic weight value of a spatial coordinate point is related to factors such as the stability and intensity of the signal field strength at that point, and is used to measure the contribution of the signal field strength characteristic to the gradient; For the The signal field strength value of a spatial coordinate point represents the signal strength at that point; For the The weight value of the single molecule distribution characteristics at a spatial coordinate point is related to the distribution density and uniformity of the single molecule near the point, and is used to reflect the weight of the influence of the single molecule distribution characteristics on the gradient; For the The single molecule distribution characteristic value of a spatial coordinate point describes the distribution of single molecules at that point; The sensor acquisition frequency is in Hertz (Hz), which indicates the number of times the sensor collects data per unit time. The signal strength change rate, measured in volts per second (V / s), reflects how quickly the signal strength changes over time.

[0070] During the entire calculation process, the following points need to be noted: when extracting signal intensity change points, the accuracy of the change points must be ensured to avoid erroneous extraction due to factors such as noise; when mapping change points to a unified coordinate system, the coordinate system must be reasonably selected to accurately reflect the spatial relationship between the sensor deployment position and the detection area; the choice of spatial interpolation algorithm should be determined based on factors such as the shape of the detection area and the complexity of the signal change to generate an accurate signal field model; the interval of equal-interval sampling should be set according to the detection accuracy requirements and the speed of signal change to ensure that the sampling points can fully reflect the signal characteristics; when calculating the signal attenuation rate, interference fluctuation index and signal change slope, appropriate algorithms and parameters should be used to obtain reliable results; when determining the signal field strength characteristic weight value and the single molecule distribution characteristic weight value, the performance of the sensor, the characteristics of the detection area and other factors should be comprehensively considered and assigned through a reasonable method; when superimposing the influence of the sensor acquisition frequency on the signal intensity change rate, the product of the two should be correctly calculated to reflect the influence of the acquisition frequency on the signal intensity gradient.

[0071] Example 4:

[0072] This example describes in detail the specific implementation of calculating single-molecule prediction information for each subregion in a single-molecule detection method based on AI and electrochemiluminescence, as well as the construction and operation of the AI ​​model. This process is performed within a biological detection system consisting of multiple electrochemiluminescence sensors deployed in the detection area and connected to an AI processing unit. Adjacent sensors are spaced a set distance apart. The system aims to accurately predict single-molecule distributions using a signal field model and AI model.

[0073] When calculating single-molecule prediction information, the main detection path of the signal field model is first used as the baseline. The main detection path is the main path for signal transmission in the signal field model. Using it as a benchmark can ensure the uniformity and accuracy of subsequent calculations. The peak position of the single-molecule detection in each frame of data is used as the reference point. The peak position is the position where the single-molecule signal intensity is the highest, which can intuitively reflect the concentrated area of ​​the single molecule in that frame. The distribution offset is calculated, that is, the degree of deviation of the peak position of each frame relative to the main detection path, and the distribution curve is drawn according to the detection coordinates. For example, assuming that the main detection path is the X-axis, in a certain frame of data, the single-molecule peak position appears at the coordinates (10,5), and the coordinates of the main detection path in this area are the X-axis, then the distribution offset is 5 units. Based on this, the distribution points of the frame are drawn in the detection coordinate system, and the distribution points of multiple frames are connected in sequence to form a distribution curve, which intuitively shows the distribution change trend of single molecules in the detection area.

[0074] Based on the calculated signal intensity gradient, the growth rate and direction of the single-molecule trend are corrected. The signal intensity gradient reflects the rate and direction of change of the signal in space. When the signal intensity gradient is large, it means that the signal changes dramatically, and the single molecule may have a strong movement or distribution change in this area. In this case, the growth rate and direction of the single-molecule trend need to be adjusted accordingly. For example, if the original single-molecule trend predicts a growth rate of 2 units per second in the positive direction of the X-axis, and the current signal intensity gradient shows a large change in the Y-axis direction, then the growth rate may be adjusted to 2.5 units per second and the direction adjusted to the composite direction of the X-axis and Y-axis to make the single-molecule trend more consistent with the actual signal changes.

[0075] Starting from the most recent distribution point, the distribution curve is drawn based on the correction results of the growth rate and direction. For example, the latest distribution point is (20, 8), the corrected growth rate is 3 units per second, and the direction is 30 degrees with respect to the X-axis. Then the distribution point for the next time period can be calculated as moving 3 units along the 30-degree angle based on the current point, resulting in a new distribution point (20 + 3cos30°, 8 + 3sin30°). And so on, the distribution points for the next time period are continuously generated until the distribution points cover the entire target area, generating complete single-molecule prediction information, which includes the possible distribution position and density of single molecules in the future.

[0076] When building an AI model, the first layer is the input layer, which is used to organize single-molecule prediction information into spatially distributed data and perform normalization processing. For example, single-molecule prediction information may contain data such as the coordinates and intensity of distribution points in multiple sub-regions. The input layer organizes this data into a matrix or vector form to give it a unified format. Normalization is to adjust the data to a specific range, such as adjusting the coordinate value to the interval [0,1], to prevent data of different dimensions from affecting model training and ensure that the model can accurately process the input data.

[0077] Next comes the feature fusion layer, which processes the spatial distribution data to extract regional correlation features and construct dependencies between detection units. For example, in spatial distribution data, the single-molecule distribution features of adjacent sub-regions may be correlated. By analyzing this data, the feature fusion layer extracts features such as the correlation of single-molecule intensities in adjacent regions and the consistency of distribution trends. This layer then constructs a dependency model between detection units, enabling the model to understand the mutual influence between different regions.

[0078] Finally, the parameter optimization layer integrates the correlations between single-molecule detections across spatial units and generates detection parameter adjustment strategies. For example, the feature fusion layer identifies the dependencies between regions. Based on these relationships and preset optimization goals, such as improving detection sensitivity or reducing noise interference, the parameter optimization layer calculates the required detection parameter adjustments for each spatial unit, such as sensor sensitivity and light source intensity. This generates a detailed parameter adjustment strategy, providing a basis for subsequent dynamic adjustment of detection parameters.

[0079] During the entire implementation process, the following points need to be noted: When determining the main detection path, the distribution characteristics of the signal field model should be fully considered, and a path that can represent the main transmission direction of the signal should be selected; when extracting the peak position, the peak of the signal should be accurately identified to avoid misjudgment caused by noise interference; when calculating the distribution offset and drawing the distribution curve, the consistency and accuracy of the detection coordinate system should be guaranteed; when correcting the single-molecule trend, the information of the signal intensity gradient should be reasonably utilized to avoid over-correction or under-correction; when organizing and standardizing spatial distribution data, the integrity and correctness of the data should be ensured to avoid data loss or errors affecting model performance; when extracting regional correlation features and constructing dependency relationships, various possible correlation factors should be fully considered to ensure that the model can accurately reflect the actual situation; when generating the detection parameter adjustment strategy, the actual needs and performance limitations of the detection system should be combined to make the strategy feasible and effective.

[0080] Example 5:

[0081] When obtaining the single-molecule detection distribution characteristics, the sub-region identifiers are matched to the spatiotemporal distribution characteristics based on the spatiotemporal distribution characteristics of the single-molecule detection output by the AI ​​model. For example, assuming the detection area is divided into 10 sub-regions, each with a unique identifier such as R1, R2, etc., and the AI ​​model outputs a single molecule with a higher intensity in region R3 and a more concentrated distribution in region R5 at a certain moment, the region identifiers of R3 and R5 are associated with the corresponding intensity and distribution characteristics, respectively, to establish a clear correspondence so that the identifier of each sub-region can accurately reflect its spatiotemporal distribution.

[0082] The regional data in the signal dataset is reorganized according to spatiotemporal characteristics. The signal dataset stores signal data from different regions at different times. During the reorganization, the spatiotemporal characteristics are used as indexes to organize data from different regions at the same time, as well as data from the same region at different times. For example, the single-molecule detection intensity data from regions R1 to R10 at time t1 is extracted and arranged in regional order. The data at time t2 is then processed in the same way to form a regional data set arranged in time series, generating a regional distribution map sorted by single-molecule detection intensity. In this distribution map, each region uses a different color or pattern to represent the intensity level, such as darker colors represent higher intensity, thus intuitively displaying the distribution and changing trend of single-molecule detection intensity within the entire detection area.

[0083] Based on the reorganized regional distribution map, the optimized single-molecule detection distribution characteristics are output. For example, by analyzing the distribution map, it is found that the R3 and R5 regions show high-intensity distribution at multiple time points and the distribution pattern is relatively stable, while the R7 region has lower intensity and greater fluctuations. The output distribution characteristics can be described as follows: R3 and R5 regions are the main single-molecule distribution areas, with high and stable intensity, while the R7 region has less single-molecule distribution and is unstable, providing clear distribution information for subsequent detection.

[0084] When dynamically adjusting detection parameters, we first map region identifiers to regions with single-molecule detection distribution characteristics. For example, region R3 corresponds to a single-molecule high-intensity distribution characteristic, and region R5 corresponds to a single-molecule concentrated distribution characteristic. A mapping table between region identifiers and distribution characteristics is established to ensure that parameter adjustments for each region are based on its actual distribution characteristics.

[0085] The execution of detection parameters is controlled based on the spatiotemporal distribution characteristics of single-molecule detection. For example, when the intensity of a single molecule detected in region R3 reaches a preset intensity threshold, a sensitivity increase instruction is triggered for neighboring sensors such as S4 and S5, enabling these sensors to more accurately collect signals from region R3. Based on the light source intensity strategy, available light source resources are dynamically combined. For example, at time t1, light sources L1 and L2 are used, and at time t2, based on the distribution characteristics of region R5, they are switched to L2 and L3. This generates a sampling frequency vector containing the sampling frequencies of different regions at different times, such as 10Hz for region R3 and 8Hz for region R5. Based on the sampling frequency vector, the optical parameters of the detection node in the target region are adjusted, such as adjusting the light source intensity of the detection node in region R3 to 50% and the exposure time to 100ms to accommodate the high signal intensity in that region.

[0086] Based on the spatial distribution of single-molecule detection and the preset parameter adjustment strategy, detection parameters are dynamically assigned to the corresponding detection areas. For example, if the preset strategy dictates that areas with high single-molecule intensity require a 10% increase in sensor sensitivity and a 20% decrease in light source intensity, these parameter adjustments are assigned to detection nodes in regions R3 and R5. This ensures that parameter adjustments in each region align with their distribution characteristics and the preset strategy, improving detection accuracy and efficiency.

[0087] The following points should be noted: when matching sub-region identifiers with spatiotemporal distribution characteristics, the uniqueness of the identifiers and the accuracy of the corresponding relationship should be ensured to avoid confusion; when reorganizing signal data sets, the integrity of the data and the correctness of the time sequence should be ensured to prevent data loss or confusion; when generating regional distribution maps, appropriate representation methods should be selected to make the distribution characteristics intuitive and easy to read; when mapping regional identifiers and distribution characteristics, the mapping table should be updated in a timely manner to reflect the latest distribution situation; when controlling the execution of detection parameters, the corresponding instructions should be triggered accurately to avoid misoperation; when dynamically allocating detection parameters, the preset strategies should be strictly followed to ensure the rationality and effectiveness of parameter adjustments.

[0088] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0089] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A single-molecule detection method based on AI and electrochemiluminescence, applied to a biological detection system, the system comprising a plurality of electrochemiluminescence sensors deployed in a detection area and an AI processing unit connected to the electrochemiluminescence sensors, wherein two adjacent electrochemiluminescence sensors are separated by a set distance, characterized in that: The method includes: collecting electrochemiluminescence signal data of samples in a region through the electrochemiluminescence sensor to generate a signal data set; receiving real-time signal data of multiple electrochemiluminescence sensors through the AI ​​processing unit to construct a signal data matrix; determining the characteristics of single-molecule detection based on the signal data set and the real-time data of each electrochemiluminescence sensor, wherein determining the characteristics of single-molecule detection includes: processing the signal data matrix, performing feature extraction in combination with the signal data set, and predicting the single-molecule distribution based on the signal intensity gradient and detection path information, outputting the spatiotemporal distribution characteristics of the single-molecule detection through the AI ​​model, and updating the signal data set according to the spatiotemporal distribution characteristics; and dynamically adjusting the detection parameters according to the distribution characteristics. The characteristics of determining single-molecule detection include: processing the signal data matrix to extract signal intensity distribution, noise interference characteristics and single-molecule trends; performing signal intensity modeling on the signal data matrix according to the signal intensity distribution and noise interference characteristics, dividing the detection area into multiple sub-areas and marking the area identifiers, performing correlation matching with the signal data set according to the signal intensity of the sub-area, and marking the area identifiers in the signal data set; calculating the signal intensity gradient according to the position of the electrochemiluminescence sensor, predicting the distribution of single-molecule detection according to the signal intensity gradient and the single-molecule trend, and calculating the single-molecule prediction information of each sub-area; constructing an AI model, using the single-molecule prediction information as the input parameter of the AI ​​model, performing spatial correlation modeling on the single-molecule prediction information through the AI ​​model, and outputting the spatiotemporal distribution characteristics of single-molecule detection; updating the signal data set according to the spatiotemporal distribution characteristics to obtain the distribution characteristics of single-molecule detection; The processing of the signal data matrix includes: standardizing the signal data matrix, intercepting the signal hotspot area in the matrix through a sliding window, filtering the background noise of the hotspot area, and calculating the single molecule trend through a feature decomposition algorithm; calculating the spatial correlation characteristics of the signal data matrix, calculating the interference intensity, signal stability coefficient and detection blind area index between regions based on the spatial correlation characteristics, constructing a feature fusion network, and calculating the noise interference characteristics through the feature fusion network; extracting the time domain characteristics and frequency domain characteristics collected by each electrochemiluminescence sensor, calculating the signal feature vector of the sensor based on the phase difference between the time domain characteristics and the frequency domain characteristics, performing feature matching on the electrochemiluminescence sensors at different positions based on the signal feature vector, and calculating the single molecule trend.

2. The single molecule detection method based on AI and electrochemiluminescence according to claim 1, characterized in that: The signal strength modeling of the signal data matrix includes: extracting signal strength sampling points in each frame of data according to the signal strength distribution, performing correlation mapping based on the sampling points and noise interference characteristics to generate a signal strength spectrum, spatially aligning the signal strength spectrum collected by multiple sensors, and calculating the signal strength distribution of the area; setting an interference threshold value, locating the interference source according to the signal strength values ​​of the multi-frame signal data matrix, calculating the interference strength difference, if the interference strength difference is greater than or equal to the interference threshold value, indicating that there is a detection blind spot in the area, performing detection model constraint compensation on the current area, iteratively correcting the signal strength distribution of the current area according to the path loss model corresponding to the current area, and calculating the strength compensation value of the blind spot signal according to the correction result; performing signal strength modeling on the signal data matrix according to the signal strength distribution, and interference marking the signal model of the area through the noise interference characteristics.

3. The single molecule detection method based on AI and electrochemiluminescence according to claim 2, characterized in that: The method for calculating the signal intensity gradient according to the position of the electrochemiluminescence sensor includes: extracting signal intensity change points according to multiple sets of signal detection data, mapping the change points to a unified detection coordinate system according to the deployment position of the sensor, fitting the change points through a spatial interpolation algorithm, and generating a signal field model of the region; sampling at equal intervals along the detection path of the signal field model, calculating the signal attenuation rate, interference fluctuation index and signal change slope of the path according to the sampling results, and calculating the signal change parameter according to the signal attenuation rate, interference fluctuation index and signal change slope; and adding single molecule detection to each frame of data according to the deployment parameters and acquisition accuracy of the electrochemiluminescence sensor. The distribution characteristics are projected onto the signal field model, and the signal field model is partitioned along the detection direction according to the number of sensors. The change law of single-molecule detection in the partition is analyzed, and the distribution characteristics are calculated according to the change law; the signal intensity gradient is calculated according to the signal change parameter and the distribution characteristics. The calculation process of the signal intensity gradient includes: based on the detection position range from the first electrochemiluminescence sensor to the last electrochemiluminescence sensor, spatial coordinate points are selected in the sensor deployment direction, and the product of the signal field strength characteristic weight value and the single-molecule distribution characteristic weight value within the spatial resolution range is cumulatively calculated, and the influence value of the sensor acquisition frequency on the signal intensity change rate is superimposed.

4. The single molecule detection method based on AI and electrochemiluminescence according to claim 3, characterized in that: The calculation of the single-molecule prediction information for each sub-region includes: using the detection main diameter of the signal field model as a baseline, the peak position of the single-molecule detection in each frame of data as a reference point, calculating the distribution offset, and drawing a distribution curve according to the detection coordinates; correcting the growth rate and direction in the single-molecule trend according to the signal intensity gradient; starting from the most recent distribution point, continuing to draw the distribution curve according to the correction results of the growth rate and direction, generating the distribution points for the next time period, until the distribution points cover the entire target area, and generating the single-molecule prediction information.

5. The single molecule detection method based on AI and electrochemiluminescence according to claim 1, characterized in that: The AI ​​model includes: The input layer is used to organize the single-molecule prediction information into spatial distribution data and perform normalization; The feature fusion layer is used to extract the detected regional correlation features by processing the spatial distribution data and build the dependency relationship between the detection units; The parameter optimization layer is used to integrate the correlation relationship of single-molecule detection in spatial units and generate detection parameter adjustment strategies.

6. The single molecule detection method based on AI and electrochemiluminescence according to claim 1, characterized in that: The obtaining of the distribution characteristics of single-molecule detection includes: according to the spatiotemporal distribution characteristics of the single-molecule detection output by the AI ​​model, matching the identifier of the sub-region with the spatiotemporal distribution characteristics; reorganizing the regional data in the signal data set according to the spatiotemporal characteristics to generate a regional distribution map sorted by the single-molecule detection intensity; and outputting the optimized single-molecule detection distribution characteristics based on the reorganized regional distribution map.

7. The single molecule detection method based on AI and electrochemiluminescence according to claim 1, characterized in that: The dynamic adjustment of detection parameters includes: mapping the regional identifiers to the regions of the distribution characteristics of single-molecule detection one by one; controlling the execution of the detection parameters according to the spatiotemporal distribution characteristics of the single-molecule detection, including sensor sensitivity adjustment, light source intensity control and sampling frequency operation; and dynamically allocating the detection parameters to the corresponding detection regions based on the spatial distribution of the single-molecule detection and a preset parameter adjustment strategy.

8. The single molecule detection method based on AI and electrochemiluminescence according to claim 7, characterized in that: The execution actions of controlling the detection parameters include: triggering a sensitivity increase instruction for a neighboring sensor when the single-molecule detection reaches a preset intensity threshold in the target area; dynamically combining available light source resources according to the light source intensity strategy to generate a sampling frequency vector; and adjusting the optical parameters of the target area detection node based on the sampling frequency vector.

Images