Automated Process Control System and Methods for Laboratory Medicine

Abstract

The present invention relates to the field of automation technology of test procedures, specifically to an automated process control system and method for a laboratory department, wherein the system includes a pedigree depth analysis module, an acoustic spectrum precise positioning module, a viral load calculation module, a spectral cell identification module, a cell electrical performance analysis module, a sample comprehensive decision module, and an intelligent sample processing module. The present invention, by combining a number of advanced technologies such as image spectrum analysis, acoustic spectrum analysis, time-frequency domain analysis, spectral technology, and cell electrical impedance analysis, achieves high-precision analysis and deep learning processing of samples, improves the accuracy and efficiency of sample analysis, and especially shows significant advantages in aspects such as viral load calculation, cell type identification, and sample risk assessment. The application of the intelligent sample processing module optimizes the sample processing process, improves work efficiency and processing speed, and greatly improves the automation and intelligence level of the laboratory department's work.

Description

Technical Field

[0001] The present invention relates to the technical field of automated testing procedures, and in particular to automated testing procedure control systems and methods. Background Art

[0002] The field of laboratory process automation technology focuses on automating the workflow of the laboratory through advanced technical means. This field generally involves the application of machine learning, artificial intelligence, robotics, and data processing technology to medical, chemical, biological and other laboratory tests, with the goal of improving the efficiency and accuracy of tests while reducing reliance on human resources. In this field, the focus includes automatic sample processing, automatic collection and analysis of experimental data, and real-time monitoring of the experimental process. Automation technology not only speeds up the inspection process, but also improves the reliability of the results and reduces the possibility of human error.

[0003] Among them, the laboratory automation process control system is an integrated system designed specifically for laboratory testing processes. Its main purpose is to improve the efficiency and accuracy of laboratory work through automation technology. The system usually includes automated equipment, software, and algorithms to manage and execute the entire testing process from sample collection to result analysis. The goal is to reduce manual operations, speed up testing, and ensure the consistency and reliability of test results. By reducing manual intervention, it helps reduce the risk of cross contamination and improve overall laboratory safety.

[0004] Traditional laboratory systems have limitations in sample processing and analysis, relying on manual operations and traditional sample analysis methods, resulting in low sample processing efficiency and limited analysis accuracy. In terms of pathogen detection, cell classification and disease diagnosis, this leads to misdiagnosis or delayed diagnosis. Traditional systems are prone to bottlenecks when processing a large number of samples, making it difficult to cope with urgent or high-intensity testing needs, limiting the development of medical testing technology and the improvement of patient services. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an automated process control system and method for a laboratory.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions: an automated process control system for the laboratory, the system comprising a pedigree depth analysis module, an acoustic spectrum precise positioning module, a viral load calculation module, a spectral cell recognition module, a cell electrical performance analysis module, a sample comprehensive decision module, and an intelligent sample processing module;

[0007] The spectrum deep analysis module uses an image spectrum analysis algorithm to decompose the spectrum features of the sample based on the sample image, uses a graph convolutional network to perform deep learning analysis on the decomposed spectrum features, combines an image feature reconstruction algorithm to classify the sample, and generates a deep classification index;

[0008] The spectrum accurate positioning module is based on the deep classification index, uses the spectrum analysis method to analyze the spectrum characteristics of the sample sound wave data, uses the hidden Markov model to perform pattern recognition on the spectrum data, locates the sample components through the sound wave feature matching algorithm, and generates the component positioning result;

[0009] The viral load calculation module analyzes the viral RNA and DNA amplification signals based on the component positioning results using the time-frequency domain analysis method, extracts the time-frequency characteristics of the signals using wavelet transform, reconstructs the spectrum using fast Fourier transform technology, calculates the load of the signals, and generates viral load calculation data;

[0010] The spectral cell identification module uses spectral analysis technology to analyze the spectral characteristics of cell data collected by the flow cytometer based on the viral load calculation data, and then uses a deep belief network to perform deep learning processing on the spectral data, and identifies the cell type through a cell characteristic reconstruction algorithm to generate cell identification data;

[0011] The cell electrical performance analysis module is based on cell identification data, combines microfluidics technology to control blood cell flow, applies cell electrical impedance analysis technology to measure the electrical impedance of cells flowing through, combines electrical performance analysis algorithms to deeply evaluate cell electrical characteristics, and generates electrical performance analysis indicators;

[0012] The sample comprehensive decision module is based on the deep classification index, component location results, viral load calculation data, cell identification data, and electrical performance analysis index, and uses a comprehensive data analysis method for integration processing, comprehensively evaluates the sample through a decision support algorithm, and conducts a comprehensive decision analysis on the sample in combination with a risk assessment model to generate a comprehensive decision analysis result;

[0013] The intelligent sample processing module optimizes the sample processing process based on the comprehensive decision analysis results using intelligent algorithms, applies process automation control technology to intelligently regulate the sample processing process, adjusts the sample processing process in combination with the processing efficiency optimization model, and generates intelligent regulation results.

[0014] The improvements of the present invention are that the deep classification indicators are specifically the spectral characteristics, image characteristics and deep correlations of the samples, the component positioning results include the acoustic characteristics and component identification of the samples, the viral load calculation data specifically refer to the time-frequency characteristics and load estimation of the viral amplification signal, the cell identification data include the spectral characteristics and type identification of the cells, the electrical performance analysis indicators specifically refer to the electrical impedance characteristics and state analysis of blood cells, the comprehensive decision-making analysis results include a comprehensive evaluation and risk analysis of the samples, and the intelligent regulation results specifically include the optimization and automated adjustment of the sample processing flow.

[0015] The present invention is improved in that the pedigree depth analysis module includes a spectrum analysis submodule, a graph convolutional network application submodule, and a classification index optimization submodule;

[0016] The spectrum analysis submodule uses the Fourier transform algorithm to decompose the spectrum characteristics based on the sample image, converts the time domain signal of the image into the frequency domain signal, and reveals the frequency characteristics of the sample image, including analyzing the periodic fluctuation of the brightness value of the image pixel points, capturing the frequency information of the sample, and performing frequency-based image analysis to generate basic spectrum information;

[0017] The graph convolution network application submodule uses a graph convolution network for deep learning processing based on basic spectrum information, and uses the nonlinear characteristics of image data to capture the relationship between samples through graph convolution operations. The graph convolution network performs weighted aggregation on the features of each node, captures the key features of the image, and generates key feature information;

[0018] The classification index optimization submodule optimizes the classification index based on the key feature information and applies the feature reconstruction algorithm, performs mathematical processing on the key features, including dimensionality reduction, normalization and pattern recognition, distinguishes multiple types of samples, accurately classifies the samples, and generates deep classification indicators.

[0019] The present invention is improved in that the sound spectrum precise positioning module includes a sound wave analysis submodule, a sound spectrum model application submodule, and a component positioning optimization submodule;

[0020] The acoustic wave analysis submodule uses the Fourier transform algorithm to analyze the spectrum characteristics of the sample acoustic wave data based on the deep classification index, analyzes and identifies the multiple frequency components of the sample acoustic wave by converting the acoustic wave signal from the time domain to the frequency domain, calculates the amplitude and phase of each frequency component, reveals the structure of the sample acoustic wave data, identifies key information, and draws the spectrum diagram of the sample acoustic wave to generate the acoustic wave spectrum characteristics;

[0021] The spectrum model application submodule uses a hidden Markov model to perform pattern recognition on spectrum data based on the characteristics of the sound wave spectrum, identifies hidden patterns in the spectrum by calculating the transition probabilities between multiple states in the spectrum data, processes the uncertainty and randomness in the sound wave data, analyzes the changing trend of the sound wave characteristics, identifies specific components in the sample, and generates spectrum pattern recognition results;

[0022] The component localization optimization submodule locates the sample components based on the results of the spectrum pattern recognition and applies the acoustic wave feature matching algorithm. It matches the specific components in the sample by comparing the features of the spectrum data with the pre-established component feature database. It comprehensively analyzes the spectrum data in combination with the sound wave frequency, amplitude and timing features, identifies and locates the target components in the sample, and generates component localization results.

[0023] The present invention is improved in that the viral load calculation module includes a time-frequency domain analysis submodule, a signal calculation submodule, and a load evaluation submodule;

[0024] The time-frequency domain analysis submodule uses short-time Fourier transform to analyze the time-frequency characteristics of viral RNA and DNA amplification signals based on the component positioning results, performs local Fourier transform on the signal within a limited time window, moves the window to cover the entire signal, obtains the spectrum information of the signal at multiple time points, and generates time-frequency domain analysis data;

[0025] The signal calculation submodule applies wavelet transform to extract signal characteristics based on time-frequency domain analysis data, decomposes the signal into a series of wavelet basis functions, each basis function represents a local time-frequency feature of the signal, and captures specific features in the signal, including mutation points and non-stationary characteristics of the virus amplification signal, by adjusting the scale and translation parameters of the wavelet function, to generate signal characteristic data;

[0026] The load assessment submodule reconstructs the signal spectrum and calculates the load based on the signal characteristic data in combination with the fast Fourier transform technology. By analyzing the spectral amplitude and phase information of the signal, the amplification degree of viral RNA and DNA is evaluated, and the viral load is calculated to generate viral load calculation data.

[0027] The present invention is improved in that the spectral cell recognition module includes a spectral analysis optimization submodule, a deep learning processing submodule, and a cell recognition enhancement submodule;

[0028] The spectral analysis optimization submodule uses Raman spectroscopy technology to perform spectral characteristic analysis on the cell data of the flow cytometer based on the viral load calculation data, uses a laser beam to irradiate the cells and captures the scattered light caused by the vibration of the cell molecules, analyzes its spectral characteristics, reveals the chemical and structural information of the cells, obtains the cell spectral characteristics, and generates spectral characteristic data;

[0029] The deep learning processing submodule uses a convolutional neural network to perform deep learning processing on the spectral characteristic data based on the spectral characteristic data, extracts multi-layer features of the spectral characteristic data, and uses multiple convolutional layers and pooling layers of the CNN model to identify and extract key patterns and features in the cell data to generate deep learning features;

[0030] The cell identification enhancement submodule is based on deep learning features and applies a cell characteristic reconstruction algorithm to identify cell types. Combined with the deep learning features, it performs a complex analysis of the multidimensional features of cells to identify cell biological properties, including protein expression and nucleic acid content on the cell surface, distinguish cell types, and generate cell identification data.

[0031] The present invention is improved in that the cell electrical performance analysis module includes a microfluidic regulation submodule, an electrical impedance analysis submodule, and a performance evaluation submodule;

[0032] The microfluidic regulation submodule uses microfluidic technology to control the flow of blood cells based on the cell identification data, separates and locates the cells by adjusting the liquid flow parameters in the microchannel, including flow rate and flow direction, and controls the position and flow state of the cells by controlling the design of the microfluidic channel and the physical properties of the fluid, thereby generating microfluidic positioning data;

[0033] The electrical impedance analysis submodule uses the electrical impedance spectrum analysis technology to measure the electrical impedance of cells based on the microfluidic positioning data, obtains the electrical characteristics of the cells by applying current to the cells and measuring the impedance response of the cells to the current, including the evaluation of the membrane integrity of the cells and the activity characteristics of the ion channels inside the cells, and generates electrical impedance measurement data;

[0034] The performance evaluation submodule uses statistical methods and machine learning algorithms based on the electrical impedance measurement data to deeply evaluate the electrical characteristics of cells, perform multi-dimensional analysis of the electrical impedance data, identify the various electrophysiological states and biological characteristics of cells, analyze the health status and physiological activities of cells, and generate electrical performance analysis indicators.

[0035] The present invention is improved in that the sample comprehensive decision module includes an algorithm model fusion submodule, a decision analysis submodule, and a result optimization submodule;

[0036] The algorithm model fusion submodule uses random forest and gradient boosting tree to perform data fusion based on deep classification indicators, component location results, viral load calculation data, cell identification data, and electrical performance analysis indicators, conducts comprehensive analysis of the data, conducts in-depth mining of multiple aspects of the data, extracts key patterns and trends, and generates a comprehensive data model;

[0037] The decision analysis submodule uses a Bayesian network to conduct a comprehensive evaluation of samples based on a comprehensive data model, and builds a probabilistic graph model to quantitatively analyze the causal relationship between data, gain insights into sample characteristics and potential risks, and generate decision support analysis results.

[0038] The result optimization submodule conducts comprehensive decision analysis of samples based on the decision support analysis results in combination with linear and logistic regression models, conducts quantitative analysis of the decision support analysis results, identifies key risk factors and potential trends of samples, evaluates the risk level of samples and the pros and cons of decision plans, optimizes the decision-making process, and generates comprehensive decision analysis results.

[0039] The present invention is improved in that the intelligent sample processing module includes a process optimization submodule, an intelligent algorithm application submodule, and an intelligent regulation submodule;

[0040] The process optimization submodule optimizes the sample processing process based on the comprehensive decision analysis results by using the linear programming method, and optimizes the resource allocation, time arrangement, and operation sequence by establishing a mathematical model to simulate the sample processing link, including setting the objective function, defining the constraints, and using the solver to match the optimal solution to generate the process optimization results;

[0041] The intelligent algorithm application submodule uses a neural network algorithm to intelligently control the sample processing process based on the process optimization results, automatically adjusts the operating parameters and optimizes the processing steps by learning historical data and real-time feedback, processes input data by simulating the working mode of human brain neurons, learns to identify complex patterns and relationships, dynamically adjusts and optimizes process parameters, and generates intelligent processing results;

[0042] The intelligent control submodule uses fuzzy logic control technology based on the intelligent processing results to adjust the sample processing process, and optimizes control decisions by processing uncertainty and approximate reasoning, including defining fuzzy variables, establishing fuzzy rule sets, applying fuzzy reasoning engines to make decisions, matching uncertain factors in the processing process, and generating intelligent control results.

[0043] The method for controlling the automated process of the laboratory department is implemented based on the automated process control system of the laboratory department, and comprises the following steps:

[0044] S1: Based on the sample image, the Fourier transform algorithm is used to decompose the spectrum features, the graph convolutional network is used for deep learning processing, nonlinear features and key feature information are extracted, and the feature reconstruction algorithm is applied to optimize the classification index to generate deep classification index;

[0045] S2: Based on the deep classification index, the Fourier transform algorithm is used to analyze the spectrum characteristics of the sample sound wave data, the hidden Markov model is used to perform pattern recognition on the sound spectrum data, and the sound wave feature matching algorithm is used to locate the sample components to generate component location results;

[0046] S3: Based on the component positioning results, short-time Fourier transform is used to analyze the time-frequency characteristics of viral RNA and DNA amplification signals, wavelet transform is used to extract signal characteristics, and fast Fourier transform technology is combined to perform spectrum reconstruction and load calculation to generate viral load calculation data;

[0047] S4: Based on the viral load calculation data, Raman spectroscopy is used to analyze the spectral characteristics of the cell data, a convolutional neural network is used to perform deep learning processing on the spectral data, deep learning features are extracted, and a cell characteristic reconstruction algorithm is applied to identify the cell type to generate cell identification data;

[0048] S5: Based on the cell identification data, use microfluidics technology to control the flow of cells, use electrical impedance spectroscopy analysis technology to measure the electrical impedance of cells, apply statistical methods and machine learning algorithms to deeply evaluate the electrical characteristics of cells, and generate electrical performance analysis indicators;

[0049] S6: Based on the deep classification indicators, component location results, viral load calculation data, cell identification data and electrical performance analysis indicators, random forests and gradient boosting trees are used for data fusion, Bayesian networks are applied for comprehensive evaluation, and linear and logistic regression models are combined for comprehensive decision analysis to optimize the decision-making process. Linear programming methods, neural network algorithms, and fuzzy logic control techniques are used to optimize and intelligently control the sample processing process to generate intelligent control results.

[0050] Compared with the prior art, the advantages and positive effects of the present invention are:

[0051] In the present invention, a number of advanced technologies such as image spectrum analysis, acoustic spectrum analysis, time-frequency domain analysis, spectral technology and cell electrical impedance analysis are combined to achieve high-precision analysis and deep learning processing of samples, improve the accuracy and efficiency of sample analysis, and show significant advantages in viral load calculation, cell type identification and sample risk assessment. The application of intelligent sample processing modules optimizes the sample processing process, improves work efficiency and processing speed, and greatly improves the degree of automation and intelligence of the work of the laboratory. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 The present invention proposes a module diagram of the laboratory automation process control system;

[0053] Figure 2The framework diagram of the automated process control system of the laboratory is proposed for the present invention;

[0054] Figure 3 The present invention proposes a schematic diagram of a pedigree depth analysis module in an automated process control system for a laboratory;

[0055] Figure 4 The present invention proposes a schematic diagram of a module for accurate positioning of sound spectra in an automated process control system for a laboratory;

[0056] Figure 5 The present invention provides a schematic diagram of a viral load calculation module in an automated process control system for a laboratory;

[0057] Figure 6 The present invention proposes a schematic diagram of a spectral cell recognition module in an automated process control system for a laboratory;

[0058] Figure 7 The present invention proposes a schematic diagram of a cell electrical performance analysis module in an automated process control system for a laboratory;

[0059] Figure 8 The present invention proposes a schematic diagram of a sample comprehensive decision-making module in an automated process control system of a laboratory;

[0060] Fig. 9 The present invention proposes a schematic diagram of an intelligent sample processing module in an automated process control system of a laboratory;

[0061] Fig.10 The present invention proposes a flow chart of an automated process control method for a laboratory. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0063] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0064] Example

[0065] See also Figure 1-Figure 9, the present invention provides a technical solution: an automated process control system for the laboratory, the system includes a pedigree depth analysis module, an acoustic spectrum precise positioning module, a viral load calculation module, a spectral cell recognition module, a cell electrical performance analysis module, a sample comprehensive decision module, and an intelligent sample processing module;

[0066] The spectrum deep analysis module uses the image spectrum analysis algorithm to decompose the spectrum features of the sample based on the sample image, uses the graph convolutional network to perform deep learning analysis on the decomposed spectrum features, combines the image feature reconstruction algorithm to classify the sample, and generates deep classification indicators;

[0067] The spectrum accurate positioning module is based on the deep classification index, uses the spectrum analysis method to analyze the spectrum characteristics of the sample sound wave data, uses the hidden Markov model to perform pattern recognition on the spectrum data, locates the sample components through the sound wave feature matching algorithm, and generates the component positioning results;

[0068] The viral load calculation module uses the time-frequency domain analysis method to analyze the viral RNA and DNA amplification signals based on the component positioning results, uses wavelet transform to extract the time-frequency characteristics of the signals, combines fast Fourier transform technology to reconstruct the spectrum, calculates the signal load, and generates viral load calculation data;

[0069] The spectral cell recognition module uses spectral analysis technology to analyze the spectral characteristics of cell data collected by the flow cytometer based on the viral load calculation data, and then uses a deep belief network to perform deep learning processing on the spectral data, and uses a cell characteristic reconstruction algorithm to identify cell types and generate cell recognition data;

[0070] The cell electrical performance analysis module is based on cell identification data and combines microfluidics technology to control the flow of blood cells. It uses cell electrical impedance analysis technology to measure the electrical impedance of cells flowing through it, and combines electrical performance analysis algorithms to conduct in-depth evaluation of cell electrical characteristics and generate electrical performance analysis indicators.

[0071] The sample comprehensive decision module is based on deep classification indicators, component location results, viral load calculation data, cell identification data, and electrical performance analysis indicators. It uses comprehensive data analysis methods to integrate and process, conducts a comprehensive evaluation of samples through decision support algorithms, and conducts comprehensive decision analysis on samples in combination with risk assessment models to generate comprehensive decision analysis results.

[0072] The intelligent sample processing module is based on the comprehensive decision-making analysis results, uses intelligent algorithms to optimize the sample processing process, applies process automation control technology to intelligently regulate the sample processing process, and adjusts the sample processing process in combination with the processing efficiency optimization model to generate intelligent regulation results.

[0073] The deep classification indicators specifically include the sample's spectral characteristics, image characteristics and their deep correlations. The component positioning results include the sample's acoustic characteristics and component identification. The viral load calculation data specifically refers to the time-frequency characteristics and load estimation of the viral amplification signal. The cell identification data includes the cell's spectral characteristics and type identification. The electrical performance analysis indicators specifically refer to the electrical impedance characteristics and state analysis of blood cells. The comprehensive decision-making analysis results include a comprehensive evaluation and risk analysis of the samples. The intelligent control results specifically include the optimization and automated adjustment of the sample processing process.

[0074] In the spectral deep analysis module, the image spectrum analysis algorithm is used to process the sample image data. This process first involves converting the image data into spectrum information through Fourier transform. Then, the graph convolutional network is applied to perform deep learning analysis on these spectrum features. In this process, the graph convolutional network obtains a more accurate image feature representation by learning the local connection pattern in the image data. Finally, the samples are classified in combination with the image feature reconstruction algorithm. This step divides different categories more accurately by reconstructing the spectrum features. The generated deep classification index is a detailed description of the sample's spectrum features, image features and their deep correlation. This process effectively improves the accuracy and efficiency of sample classification.

[0075] In the spectrum accurate positioning module, based on the deep classification index, the spectrum analysis method is used to process the sample sound wave data. First, the key features of the sound wave are extracted by analyzing the spectrum characteristics. Then, the hidden Markov model is used to perform pattern recognition on these spectrum data, and different spectrum patterns are identified by analyzing the statistical characteristics in the sound wave data. Subsequently, the sound wave feature matching algorithm is used to accurately locate the sample components. The algorithm determines the sample components by comparing and matching the sound wave features of different samples. This process generates the component positioning results, namely the acoustic features and component identification information of the sample, which provides an accurate positioning basis for subsequent analysis.

[0076] In the viral load calculation module, based on the component positioning results, the time-frequency domain analysis method is used to analyze the amplification signals of viral RNA and DNA. First, the time-frequency characteristics of the signal are extracted using wavelet transform. This step can effectively extract key time-frequency features from complex signals. Then, combined with the fast Fourier transform technology, the spectrum is reconstructed to more accurately analyze the frequency composition of the signal. Finally, the load is calculated using this information to generate viral load calculation data, which includes the time-frequency characteristics and load estimation of the viral amplification signal, providing key data for virus detection.

[0077] In the spectral cell recognition module, spectral analysis technology is used to process cell data based on viral load calculation data. First, the spectral characteristics of the cell data collected by the flow cytometer are analyzed to extract the key spectral features of the cells. Subsequently, a deep belief network is applied to perform deep learning processing on these spectral data. This step improves the accuracy of cell type recognition by learning the spectral characteristics of cells. Finally, cell type recognition is performed through a cell characteristic reconstruction algorithm to generate cell recognition data, which includes the spectral characteristics and type recognition of cells, providing efficient technical support for cell classification.

[0078] In the cell electrical performance analysis module, blood cell flow control is performed based on cell identification data combined with microfluidics technology. During this process, cells are precisely manipulated and positioned. Next, the cell electrical impedance analysis technology is used to measure the electrical impedance of the cells. This technology evaluates the state of the cells by measuring the changes in the electrical impedance of the cells. Then, the electrical characteristics of the cells are deeply evaluated in combination with the electrical performance analysis algorithm. This process analyzes the electrical impedance data of the cells to obtain information on the health status and type of the cells. The generated electrical performance analysis indicators are the electrical impedance characteristics and state analysis of blood cells, which are crucial for understanding the electrical performance and functional status of cells.

[0079] In the sample comprehensive decision-making module, data such as deep classification indicators, component location results, viral load calculation data, cell identification data, and electrical performance analysis indicators are integrated and processed using comprehensive data analysis methods. First, the samples are comprehensively evaluated through decision support algorithms. This process involves the fusion and analysis of multiple data dimensions to provide a comprehensive sample situation. Subsequently, combined with the risk assessment model, a comprehensive decision analysis is performed on the samples, which includes an assessment of the potential risks of the samples and management recommendations. The generated comprehensive decision analysis results are a comprehensive assessment and risk analysis of the samples, which plays an important role in guiding subsequent testing and treatment decisions.

[0080] In the intelligent sample processing module, intelligent algorithms are used to optimize the sample processing process based on the results of comprehensive decision-making analysis. First, process automation control technology is applied to intelligently regulate the sample processing process. This step involves the automated arrangement and execution of workflows to improve the efficiency and accuracy of sample processing. Subsequently, the sample processing process is adjusted in combination with the processing efficiency optimization model, which includes optimizing and accelerating the key steps in the process to ensure the efficiency and accuracy of the entire sample processing process. The generated intelligent regulation result is the optimization and automated adjustment of the sample processing process, which not only improves the efficiency of the inspection process, but also ensures the accuracy and reliability of the processing results.

[0081] The pedigree deep analysis module includes a spectrum analysis submodule, a graph convolutional network application submodule, and a classification index optimization submodule;

[0082] The spectrum analysis submodule uses the Fourier transform algorithm to decompose the spectrum characteristics based on the sample image, converts the time domain signal of the image into the frequency domain signal, and reveals the frequency characteristics of the sample image, including analyzing the periodic fluctuation of the brightness value of the image pixel points, capturing the frequency information of the sample, and performing frequency-based image analysis to generate basic spectrum information;

[0083] The graph convolutional network application submodule uses graph convolutional networks for deep learning processing based on basic spectrum information. It uses the nonlinear characteristics of image data and captures the relationship between samples through graph convolution operations. The graph convolutional network performs weighted aggregation on the features of each node, captures the key features of the image, and generates key feature information.

[0084] The classification index optimization submodule optimizes the classification index based on the key feature information and applies the feature reconstruction algorithm. It performs mathematical processing on the key features, including dimensionality reduction, normalization and pattern recognition, distinguishes multiple types of samples, accurately classifies the samples, and generates deep classification indicators.

[0085] In the spectrum analysis submodule, the sample image data is processed by the Fourier transform algorithm. The specific operation is to convert the time domain signal of the image into a frequency domain signal to reveal the frequency characteristics of the sample image. This process involves periodic fluctuation analysis of the brightness values ​​of the image pixels to capture the frequency information of the sample, thereby performing frequency-based image analysis. During the operation, the image data is input in matrix form, and the Fourier transform algorithm transforms the brightness value of each pixel to extract its frequency component. This conversion reveals the subtle spectral features in the image, and the generated basic spectral information reflects the unique frequency properties of the sample image, which is crucial for subsequent in-depth analysis.

[0086] In the graph convolution network application submodule, the graph convolution network is used for deep learning processing based on the basic spectrum information. In this process, the nonlinear characteristics of the image data are used, and the graph convolution operation captures the relationship between samples, specifically including weighted aggregation of the features of each node. In the operation, the basic spectrum information is used as input, and the graph convolution network extracts key information by learning the structural features of this data. Each layer in the network integrates the information of adjacent nodes in a weighted manner to capture the local and overall features of the image. The generated key feature information provides an important foundation for deep learning, enabling the model to more accurately identify and classify samples.

[0087] In the classification index optimization submodule, based on the key feature information, the feature reconstruction algorithm is applied to optimize the classification index. In the operation, the key features are first mathematically processed, including dimension reduction, normalization and pattern recognition. These steps are aimed at simplifying and standardizing the feature data and improving its interpretability. In the dimension reduction process, algorithms such as principal component analysis are used to reduce the number of features while retaining the most important information; normalization ensures that the comparison and combination of different features are more reasonable; pattern recognition uses algorithms such as support vector machines to identify and distinguish different types of samples. The final generated deep classification index provides key information for accurate classification, effectively improving the accuracy and efficiency of classification.

[0088] Suppose a sample image dataset contains images of different types of cells. First, in the spectrum analysis submodule, the image data is Fourier transformed, such as converting a 256x256 pixel image into its corresponding frequency domain representation. The conversion result reveals the main frequency components in the image, such as high-frequency details and low-frequency background information. Then, in the graph convolutional network application submodule, these frequency domain data are used as input, and the graph convolutional network identifies and extracts the key morphological features of cells in the image. Finally, in the classification index optimization submodule, the extracted features are processed by the feature reconstruction algorithm, such as reducing the dimension to 30 main components, normalizing, and classifying the cells into normal cells and abnormal cells through the pattern recognition algorithm. The final generated deep classification index provides a classification label for each sample, such as "normal" or "abnormal", and the corresponding confidence score. This information provides an important basis for subsequent diagnosis and treatment.

[0089] The sound spectrum precise positioning module includes the sound wave analysis submodule, the sound spectrum model application submodule, and the component positioning optimization submodule;

[0090] The sound wave analysis submodule uses the Fourier transform algorithm to analyze the spectrum characteristics of the sample sound wave data based on the deep classification index. By converting the sound wave signal from the time domain to the frequency domain, it analyzes and identifies the multiple frequency components of the sample sound wave, calculates the amplitude and phase of each frequency component, reveals the structure of the sample sound wave data, identifies key information, and draws the spectrum diagram of the sample sound wave to generate the sound wave spectrum characteristics;

[0091] The spectrum model application submodule uses the hidden Markov model to perform pattern recognition on spectrum data based on the characteristics of sound wave spectrum. It calculates the transition probability between multiple states in spectrum data, identifies hidden patterns in the spectrum, processes uncertainty and randomness in sound wave data, analyzes the changing trend of sound wave characteristics, identifies specific components in samples, and generates spectrum pattern recognition results.

[0092] The component localization optimization submodule locates the sample components based on the results of spectral pattern recognition and applies the acoustic wave feature matching algorithm. It matches the specific components in the sample by comparing the features of the spectral data with the pre-established component feature database. It conducts a comprehensive analysis of the spectral data in combination with the sound wave frequency, amplitude and timing features, identifies and locates the target components in the sample, and generates component localization results.

[0093] In the sound wave analysis submodule, the Fourier transform algorithm is used to perform a detailed spectral characteristic analysis of the sample sound wave data. The process first converts the sound wave signal from the time domain to the frequency domain. This step is achieved by decomposing the sound wave signal into a sine wave sequence and calculating the amplitude and phase of each frequency component. The specific operations include: entering the sound wave sample data, which is usually in the form of a digital array of a time series; performing a fast Fourier transform (FFT) to analyze and identify the multiple frequency components of the sample sound wave; and calculating the amplitude and phase of each frequency component. This process reveals the structure of the sample sound wave data, identifies key information, and finally generates a spectrum diagram of the sound wave to form a specific sound wave spectrum characteristic file. This file shows the frequency distribution of the sound wave sample in detail, providing an accurate data basis for subsequent modules.

[0094] In the spectrum model application submodule, the spectrum data is pattern recognized by using the hidden Markov model. The data format involved in this process is mainly the sound wave spectrum characteristic data, which is generated by the Fourier transform in the previous step. The core of this submodule is to calculate the transition probability between multiple states in the spectrum data, and then identify the hidden patterns in the spectrum. The specific operation process includes: importing the spectrum characteristic data; using the hidden Markov model to calculate the transition probability between each state through statistical analysis; identifying the pattern in the spectrum based on these probabilities, and processing the uncertainty and randomness in the sound wave data. This process not only analyzes the changing trend of the sound wave characteristics, but also identifies the specific components in the sample, and finally generates the result file of the spectrum pattern recognition. This file records the patterns and characteristics in the spectrum data in detail, providing key information for the subsequent component positioning.

[0095] In the component location optimization submodule, the sample components are located by applying the acoustic wave feature matching algorithm. The data format of this submodule includes the acoustic spectrum pattern recognition results generated in the previous step and the pre-established component feature database. The operation process mainly compares the characteristics of the acoustic spectrum data with the component feature database to match the specific components in the sample. The specific operations include: importing the acoustic spectrum pattern recognition results and the component feature database; executing the acoustic wave feature matching algorithm to analyze the frequency, amplitude and timing characteristics of the acoustic wave; and identifying and locating the target components in the sample through algorithm comprehensive analysis. The component location result file generated by this process records the location information of the target component in detail, providing accurate data support for the practical application of acoustic spectrum analysis.

[0096] In the automated process control system of the laboratory, there is a sound sample containing noise components of different frequencies. The sample is input into the acoustic wave analysis submodule in the form of a digital array, where the sample data is such as [0.5, 0.8, 0.3, ...]. Through Fourier transform, a spectrum characteristic file is obtained to show the distribution of different frequencies. Then, in the acoustic spectrum model application submodule, the hidden Markov model is used to process these spectrum data to identify specific patterns, such as the large number of noise components with a frequency near 1000Hz. Finally, in the component positioning optimization submodule, the acoustic wave feature matching algorithm is used to determine the exact location and characteristics of these noise components. The final generated positioning result file records the specific information of these noise components in detail, such as frequency, amplitude, etc., providing accurate data basis for subsequent noise control and processing.

[0097] The viral load calculation module includes a time-frequency domain analysis submodule, a signal calculation submodule, and a load assessment submodule;

[0098] The time-frequency domain analysis submodule uses short-time Fourier transform to analyze the time-frequency characteristics of viral RNA and DNA amplification signals based on the component positioning results. It performs local Fourier transform on the signal within a limited time window, moves the window to cover the entire signal, obtains the spectrum information of the signal at multiple time points, and generates time-frequency domain analysis data.

[0099] The signal calculation submodule analyzes data in the time-frequency domain and applies wavelet transform to extract signal characteristics. The signal is decomposed into a series of wavelet basis functions, each of which represents a local time-frequency feature of the signal. By adjusting the scale and translation parameters of the wavelet function, specific features in the signal are captured, including mutation points and non-stationary characteristics of the virus amplification signal, to generate signal characteristic data.

[0100] The load assessment submodule reconstructs the signal spectrum and calculates the load based on the signal characteristic data and combines the fast Fourier transform technology. By analyzing the signal's spectral amplitude and phase information, it evaluates the degree of amplification of viral RNA and DNA, calculates the viral load, and generates viral load calculation data.

[0101] In the time-frequency domain analysis submodule, the time-frequency characteristics of viral RNA and DNA amplification signals are analyzed by short-time Fourier transform (STFT). The data format of this process is time series data based on the component localization results. STFT performs a local Fourier transform on the signal within a limited time window, and gradually moves the window to cover the entire signal to obtain the spectrum information of the signal at multiple time points. In the operation, a time window, such as a Hamming window or a Gaussian window, is first set, and then this window is applied at different time points of the signal to perform a local Fourier transform to obtain the frequency components of the signal at each time point. This process is repeated until the entire signal is covered. The generated time-frequency domain analysis data shows the spectral changes of the signal at each time point, which is crucial for understanding the time-frequency characteristics of the signal.

[0102] In the signal calculation submodule, the characteristics of the time-frequency domain analysis data are extracted by applying wavelet transform. Wavelet transform can decompose the signal into a series of wavelet basis functions, each of which represents a local time-frequency feature of the signal. In this process, an appropriate wavelet mother function, such as Daubechies wavelet, is first selected, and then the scale and translation parameters of the wavelet function are adjusted to capture specific features in the signal, such as the mutation points and non-stationary characteristics of the virus amplification signal. Through this fine-grained decomposition, important features in the signal can be more clearly identified. The generated signal characteristic data describes the local time-frequency characteristics of the signal in detail, which is crucial for understanding the complexity of the virus amplification signal.

[0103] In the load assessment submodule, based on the signal characteristic data and combined with the fast Fourier transform technology, the signal spectrum is reconstructed and the load is calculated. In the operation, the fast Fourier transform is first applied to the signal characteristic data to analyze the spectral amplitude and phase information of the signal, which helps to understand the characteristics of the viral RNA and DNA amplification signals. Then, through the comprehensive analysis of these spectral information, the degree of viral amplification is evaluated and the viral load is calculated. The generated viral load calculation data records the quantitative information of the viral load in detail, which is crucial for virus detection and treatment.

[0104] In the automated process control system of the laboratory, there is a viral RNA sample whose amplification signal data format is a time series, such as [0.2, 0.4, 0.6, ...]. In the time-frequency domain analysis submodule, STFT is used to analyze the signal, obtain spectrum information at different time points, and display the time-frequency changes of the signal. Subsequently, in the signal calculation submodule, wavelet transform is applied to extract signal characteristics, such as the mutation point of the amplification signal. Finally, in the load evaluation submodule, the signal spectrum is reconstructed by FFT, the viral load is calculated and evaluated, and viral load data such as [2.5, 3.0, 4.0, ...] are generated, indicating the changes in the viral load at different time points, providing important information for the monitoring and treatment of the virus.

[0105] The spectral cell recognition module includes a spectral analysis optimization submodule, a deep learning processing submodule, and a cell recognition enhancement submodule;

[0106] The spectral analysis optimization submodule uses Raman spectroscopy to analyze the spectral characteristics of the cell data of the flow cytometer based on the viral load calculation data. It uses a laser beam to illuminate the cells and capture the scattered light caused by the vibration of cell molecules, analyzes its spectral characteristics, reveals the chemical and structural information of the cells, obtains the cell spectral characteristics, and generates spectral characteristic data;

[0107] The deep learning processing submodule uses a convolutional neural network to perform deep learning processing on the spectral characteristic data based on the spectral characteristic data, extracts multi-layer features of the spectral characteristic data, and uses multiple convolutional layers and pooling layers of the CNN model to identify and extract key patterns and features in the cell data to generate deep learning features;

[0108] The cell recognition enhancement submodule is based on deep learning features and uses a cell characteristic reconstruction algorithm to identify cell types. It combines deep learning features to perform a complex analysis of the multi-dimensional characteristics of cells, identify cell biological properties, including protein expression and nucleic acid content on the cell surface, distinguish cell types, and generate cell recognition data.

[0109] In the spectral analysis optimization submodule, the spectral characteristics of the cell data of the flow cytometer are analyzed by Raman spectroscopy. During the operation, the cell sample is first irradiated with a laser beam to capture the scattered light caused by the vibration of the cell molecules. The data format is Raman spectral data, which reflects the chemical and structural information of the cell. The specific operations include: setting laser parameters to optimize the irradiation effect, receiving and recording scattered light signals, and processing these signals through Raman spectroscopy analysis software. The processing process involves converting the received spectral signal into a digital signal, and performing noise filtering and baseline correction to reveal the spectral characteristics of the cell. The generated spectral characteristic data describes the chemical composition and structural characteristics of the cell in detail, providing an accurate data basis for subsequent deep learning processing and cell recognition.

[0110] In the deep learning processing submodule, the spectral characteristic data is deep-learned by applying a convolutional neural network (CNN). The data format is the spectral characteristic data generated by the spectral analysis optimization submodule. During the operation, these spectral data are first imported into the CNN model, and multiple convolutional layers and pooling layers are used to identify and extract key patterns and features in the cell data. The convolutional layer is responsible for extracting local features, while the pooling layer is used to reduce the feature dimension, thereby reducing the amount of calculation. In this process, CNN adjusts network parameters (such as weights and biases) and trains through the back-propagation algorithm to finally generate deep learning feature data. These data reveal the key patterns and features in the cell data, providing important information for further identification of cells.

[0111] In the cell identification enhancement submodule, the cell characteristics reconstruction algorithm is applied to identify the cell type based on the deep learning features. In the operation, the deep learning feature data is first imported, and then these features are combined to perform a composite analysis of the multidimensional features of the cell. This includes biological properties such as protein expression and nucleic acid content on the cell surface. Using the cell characteristics reconstruction algorithm, these multidimensional features are comprehensively considered to distinguish cell types. The generated cell identification data describes the characteristics and type of each cell in detail, making the identification of cells more accurate and reliable. These data are extremely valuable for pathological analysis and biological research.

[0112] In the automated process control system of the laboratory, spectral cell identification is performed on a batch of cell samples. First, the spectral analysis optimization submodule receives the cell data from the flow cytometer. For example, the Raman spectral data format is [520,540,560,...], which represents the intensity values ​​of different wavelengths. The chemical and structural information of the cells is obtained through Raman spectroscopy analysis. Subsequently, in the deep learning processing submodule, these data are input into the convolutional neural network, and the network extracts the key patterns and features of the cells through training. Finally, the cell identification enhancement submodule uses the cell characteristic reconstruction algorithm based on these features to identify the cell type and generate cell identification data, such as marking different types of cells, such as normal cells, cancer cells, etc., which provides important data support for clinical diagnosis and research.

[0113] The cell electrical performance analysis module includes a microfluidic regulation submodule, an electrical impedance analysis submodule, and a performance evaluation submodule;

[0114] The microfluidic regulation submodule uses microfluidic technology to control the flow of blood cells based on cell identification data. It separates and locates cells by adjusting the liquid flow parameters in the microchannel, including flow rate and flow direction. It controls the position and flow state of cells by controlling the design of the microfluidic channel and the physical properties of the fluid, and generates microfluidic positioning data.

[0115] The impedance analysis submodule uses the impedance spectrum analysis technology to measure the impedance of cells based on the microfluidic positioning data. It obtains the electrical characteristics of cells by applying current to cells and measuring the impedance response of cells to the current, including the evaluation of cell membrane integrity and the activity characteristics of ion channels inside cells, and generates impedance measurement data.

[0116] The performance evaluation submodule uses statistical methods and machine learning algorithms based on the electrical impedance measurement data to conduct an in-depth evaluation of the electrical characteristics of cells, perform multi-dimensional analysis of the electrical impedance data, identify the various electrophysiological states and biological characteristics of cells, analyze the health status and physiological activities of cells, and generate electrical performance analysis indicators.

[0117] In the microfluidic regulation submodule, the flow of blood cells is controlled by microfluidic technology and operated based on cell identification data. In the specific process, the geometric parameters of the microchannel and the physical properties of the fluid, such as flow rate and flow direction, are first set to control the movement of cells in the microchannel. The operation process includes using a microfluidic chip to adjust the design of the microchannel according to the cell identification data, such as the width and length of the channel, as well as the viscosity and flow rate of the fluid. By precisely controlling these parameters, the single separation and positioning of cells are achieved. This control strategy relies on the principles of fluid dynamics to ensure that cells move along a predetermined path in the microchannel. The generated microfluidic positioning data records the specific position and flow state of the cells in the microchannel, providing accurate input data for subsequent electrical impedance analysis.

[0118] In the electrical impedance analysis submodule, the electrical impedance of cells is measured by applying the electrical impedance spectroscopy technique. Based on the microfluidic positioning data, the operation first applies current to the cells and then measures the impedance response of the cells to the current. This process involves applying an alternating current of a known frequency to the cell sample using electrodes while recording the impedance changes of the cells. Analyzing this data can reveal the electrical properties of the cells, such as membrane integrity and the activity of ion channels inside the cells. The generated electrical impedance measurement data contains the response characteristics of the cells to currents of different frequencies, providing important information about the physiological state of the cells.

[0119] In the performance evaluation submodule, the electrical properties of cells are deeply evaluated based on the electrical impedance measurement data using statistical methods and machine learning algorithms. Operations include multi-dimensional analysis of the electrical impedance data, such as using machine learning algorithms such as support vector machines or random forests to identify the electrophysiological state and biological characteristics of cells. These algorithms are able to distinguish different types of cell states, such as healthy cells and damaged cells, by learning patterns in the electrical impedance data. The generated electrical performance analysis indicators describe the health status and physiological activities of cells in detail, which is of great significance for biomedical research and clinical diagnosis.

[0120] In the automated process control system of the laboratory, a batch of blood samples were subjected to cell electrical performance analysis. First, the microfluidic regulation submodule received cell identification data, such as cell type and quantity. The parameters of the microfluidic chip were adjusted based on these data to achieve precise positioning of blood cells and generate microfluidic positioning data such as the specific position of cells in the microchannel. Next, in the electrical impedance analysis submodule, current is applied to these located cells, their impedance response is measured, and electrical impedance measurement data is generated, such as the response characteristics of different cells to currents of specific frequencies. Finally, the performance evaluation submodule uses machine learning algorithms to analyze these data and generate electrical performance analysis indicators, such as the health status and electrophysiological activity of cells, providing detailed analysis results for diagnosis and research.

[0121] The sample comprehensive decision module includes the algorithm model fusion submodule, the decision analysis submodule, and the result optimization submodule;

[0122] The algorithm model fusion submodule uses random forest and gradient boosting trees to fuse data based on deep classification indicators, component location results, viral load calculation data, cell identification data, and electrical performance analysis indicators, conducts comprehensive analysis of the data, conducts in-depth mining of multiple aspects of the data, extracts key patterns and trends, and generates a comprehensive data model;

[0123] The decision analysis submodule uses the Bayesian network to conduct a comprehensive evaluation of samples based on a comprehensive data model. By constructing a probabilistic graphical model, it conducts quantitative analysis of the causal relationship between data, provides insights into sample characteristics and potential risks, and generates decision support analysis results.

[0124] The result optimization submodule conducts comprehensive decision analysis of samples based on the decision support analysis results, combining linear and logistic regression models, quantitatively analyzing the decision support analysis results, identifying the key risk factors and potential trends of samples, evaluating the risk level of samples and the pros and cons of decision plans, optimizing the decision-making process, and generating comprehensive decision analysis results.

[0125] In the algorithm model fusion submodule, data fusion is performed through random forest and gradient boosting tree, and operations are performed based on deep classification indicators, component location results, viral load calculation data, cell identification data, and electrical performance analysis indicators. These data usually exist in the form of structured numerical tables and contain multiple feature dimensions. The operation process includes first importing these multidimensional data, and then applying random forest and gradient boosting tree algorithms. Random forest improves the accuracy and robustness of predictions by constructing multiple decision trees and synthesizing their results. Gradient boosting tree gradually improves the predictive ability of the model by iteratively optimizing errors. Through these two algorithms, multiple aspects of the data are comprehensively analyzed to extract key patterns and trends. The generated comprehensive data model describes the complex relationships and patterns between the data in detail, providing a solid foundation for decision analysis.

[0126] In the decision analysis submodule, a comprehensive evaluation of the samples is performed by applying Bayesian networks. Based on the comprehensive data model, the operation process first builds a probabilistic graph model to quantify the causal relationship between the data. Through the Bayesian network, insights into sample characteristics and potential risks can be obtained. In actual operation, it includes selecting appropriate nodes and edges to represent data features and their associations, and then calculating conditional probabilities between nodes based on statistical data. In this way, the Bayesian network reveals the interdependence between different data features. The generated decision support analysis results provide a quantitative method for identifying potential risk factors and sample characteristics, enhancing the scientificity and accuracy of the decision-making process.

[0127] In the result optimization submodule, based on the decision support analysis results, the linear and logistic regression models are combined to conduct a comprehensive decision analysis of the samples. The operation includes quantitative analysis of the decision support analysis results, such as using the linear regression model to analyze the linear relationship of the data and the logistic regression model to identify the relationship between categorical variables. Through these models, the key risk factors and potential trends of the samples can be identified, and the risk level of the samples and the pros and cons of the decision plans can be evaluated. The generated comprehensive decision analysis results provide a quantitative basis for optimizing the decision-making process, ensuring that the decision is more comprehensive and effective.

[0128] In the automated process control system of the laboratory, a series of pathological samples were subjected to comprehensive decision analysis. First, the algorithm model fusion submodule received multidimensional data, such as viral load data [2.5, 3.0, 4.0, ...], cell identification data, etc. Through the random forest and gradient boosting tree algorithms, a comprehensive data model was generated to show the complex relationship between sample features. Then, in the decision analysis submodule, a Bayesian network was constructed to analyze the causal relationship between sample features and generate decision support analysis results. Finally, in the result optimization submodule, linear and logistic regression models were applied to further quantitatively analyze these results, generating comprehensive decision analysis results, such as risk assessment and decision recommendations for samples, which provide a scientific basis for clinical decision-making and research.

[0129] The intelligent sample processing module includes a process optimization submodule, an intelligent algorithm application submodule, and an intelligent regulation submodule;

[0130] The process optimization submodule uses linear programming to optimize the sample processing process based on the comprehensive decision analysis results. It simulates the sample processing link by establishing a mathematical model to optimize resource allocation, time scheduling, and operation sequence, including setting the objective function, defining constraints, and using the solver to match the optimal solution to generate process optimization results.

[0131] Based on the process optimization results, the intelligent algorithm application submodule uses the neural network algorithm to intelligently control the sample processing process. By learning historical data and real-time feedback, it automatically adjusts the operating parameters and optimizes the processing steps. By simulating the working mode of human brain neurons, it processes input data, learns to identify complex patterns and relationships, and dynamically adjusts and optimizes process parameters to generate intelligent processing results.

[0132] Based on the intelligent processing results, the intelligent control submodule applies fuzzy logic control technology to adjust the sample processing process, and optimizes control decisions by processing uncertainty and approximate reasoning, including defining fuzzy variables, establishing fuzzy rule sets, applying fuzzy reasoning engines to make decisions, matching uncertain factors in the processing process, and generating intelligent control results.

[0133] In the process optimization submodule, the sample processing process is optimized by linear programming, and operations are performed based on the results of comprehensive decision analysis. First, a mathematical model is established to simulate the sample processing link. The data format is usually a structured table, containing information such as resource allocation, time scheduling, and operation sequence. The operation process includes setting objective functions, such as minimizing processing time or cost, and defining constraints, including resource limitations and operation sequence requirements. Then, mathematical solvers, such as the simplex method or the interior point method, are used to find the optimal solution and match resources and time to maximize efficiency. The generated process optimization results describe the optimized sample processing process in detail, including the timing and resource allocation of each step, which helps to improve the efficiency and quality of sample processing.

[0134] In the intelligent algorithm application submodule, the sample processing process is intelligently regulated by applying the neural network algorithm based on the process optimization results. The operation includes importing the process optimization results as input data, using the neural network for deep learning, learning historical data and real-time feedback. The neural network processes input data and learns to recognize complex patterns and relationships by simulating the working mode of human brain neurons. By adjusting network parameters such as weights and biases, the neural network dynamically adjusts and optimizes the process parameters. The generated intelligent processing results describe the optimized processing steps and operating parameters in detail, improve the automation and intelligence level of the processing process, and ensure more efficient and accurate sample processing.

[0135] In the intelligent control submodule, fuzzy logic control technology is applied to adjust the sample processing process based on the intelligent processing results. The operation includes defining fuzzy variables, such as processing time and quality level, and establishing fuzzy rule sets to describe the relationship between these variables. Decisions are made using fuzzy reasoning engines, such as Mamdani or Sugeno methods. This method can effectively handle uncertainty and approximate reasoning and optimize control decisions. The generated intelligent control results match the uncertainties in the processing process, providing a more flexible and adaptable process control solution, and improving the quality and efficiency of sample processing.

[0136] In the automated process control system of the laboratory, a batch of blood samples is processed intelligently. First, the process optimization submodule receives the results of comprehensive decision analysis, such as the urgency and resource requirements of each sample, optimizes the processing process through linear programming, and generates an optimized schedule and resource allocation plan. Then, the intelligent algorithm application submodule uses neural networks to analyze these data and automatically adjust the processing steps, such as adjusting the order of certain detection steps to improve efficiency. Finally, the intelligent regulation submodule uses fuzzy logic control technology to adjust the process according to real-time feedback, such as flexibly adjusting the processing order of other samples when the equipment fails. The generated intelligent regulation results ensure the efficiency and accuracy of sample processing, providing strong support for clinical diagnosis.

[0137] See also Fig.10 , the automated process control method of the laboratory department comprises the following steps:

[0138] S1: Based on the sample image, the Fourier transform algorithm is used to decompose the spectrum features, the graph convolutional network is used for deep learning processing, nonlinear features and key feature information are extracted, and the feature reconstruction algorithm is applied to optimize the classification index to generate deep classification index;

[0139] S2: Based on the deep classification index, the Fourier transform algorithm is used to analyze the spectrum characteristics of the sample sound wave data, the hidden Markov model is used to perform pattern recognition on the sound spectrum data, and the sound wave feature matching algorithm is used to locate the sample components to generate component location results;

[0140] S3: Based on the component positioning results, short-time Fourier transform is used to analyze the time-frequency characteristics of viral RNA and DNA amplification signals, wavelet transform is used to extract signal characteristics, and fast Fourier transform technology is combined to perform spectrum reconstruction and load calculation to generate viral load calculation data;

[0141] S4: Based on the viral load calculation data, Raman spectroscopy technology is used to analyze the spectral characteristics of the cell data, and a convolutional neural network is used to perform deep learning processing on the spectral data, extract deep learning features, and apply a cell characteristic reconstruction algorithm to identify the cell type and generate cell identification data;

[0142] S5: Based on the cell identification data, microfluidics technology is used to control the flow of cells, electrical impedance spectroscopy analysis technology is used to measure the electrical impedance of cells, statistical methods and machine learning algorithms are used to deeply evaluate the electrical characteristics of cells, and electrical performance analysis indicators are generated;

[0143] S6: Based on deep classification indicators, component location results, viral load calculation data, cell identification data and electrical performance analysis indicators, random forest and gradient boosting trees are used for data fusion, Bayesian networks are applied for comprehensive evaluation, and linear and logistic regression models are combined for comprehensive decision analysis to optimize the decision-making process. Linear programming, neural network algorithms, and fuzzy logic control technology are used to optimize and intelligently control the sample processing process to generate intelligent control results.

[0144] In step S1, the generation of deep classification indicators through Fourier transform algorithm and graph convolutional network not only improves the accuracy of sample image analysis, but also enhances the meticulousness and reliability of sample classification through nonlinear feature extraction and reconstruction of key feature information. This provides an accurate preliminary data foundation for subsequent sample processing, ensuring that the starting point of the entire process is highly accurate and effective.

[0145] In step S2, the Fourier transform algorithm combined with the hidden Markov model and the acoustic wave feature matching algorithm makes the analysis of acoustic wave data more comprehensive and improves the accuracy of component location. This step can effectively identify the complex components in the sample and is crucial for accurately judging the nature and state of the sample.

[0146] In step S3, by combining short-time Fourier transform, wavelet transform and fast Fourier transform techniques, the amplification signals of viral RNA and DNA are accurately analyzed, which not only improves the accuracy of viral load calculations, but also increases the in-depth understanding of viral behavior, which is of great significance for the accurate identification and quantification of pathogens.

[0147] In step S4, the combination of Raman spectroscopy and convolutional neural networks provides an efficient method for spectral characteristic analysis and deep learning feature extraction of cell data. The execution of this step improves the ability to identify cell types and provides important basic data for subsequent biomedical research and clinical applications.

[0148] In step S5, the combination of microfluidics and electrical impedance spectroscopy, as well as statistical methods and machine learning algorithms, not only optimizes the flow control of cells, but also improves the accuracy of cell electrical property analysis. These refined evaluation methods help to deeply understand the biological characteristics of cells and provide more accurate indicators for diagnosis and treatment.

[0149] In step S6, random forest, gradient boosting tree, Bayesian network, linear and logistic regression models are comprehensively applied to conduct a comprehensive decision analysis of the samples, and intelligent regulation is performed in combination with linear programming, neural network algorithm, and fuzzy logic control technology. This step not only optimizes the automated process control method of the entire laboratory department and improves the efficiency and accuracy of the processing flow, but also improves the ability to handle uncertainty through intelligent regulation, ensuring the flexibility and adaptability of the process. This highly integrated and optimized process control method provides the laboratory department with a comprehensive, accurate, efficient and intelligent operating platform, greatly improving the quality and efficiency of medical testing, and is of great significance for improving diagnostic accuracy, early disease detection and treatment effect evaluation.

[0150] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. The automated process control system of the laboratory department is characterized by: The system includes a pedigree depth analysis module, an acoustic spectrum precision positioning module, a viral load calculation module, a spectral cell recognition module, a cell electrical performance analysis module, a sample comprehensive decision module, and an intelligent sample processing module; The spectrum deep analysis module uses an image spectrum analysis algorithm to decompose the spectrum features of the sample based on the sample image, uses a graph convolutional network to perform deep learning analysis on the decomposed spectrum features, combines an image feature reconstruction algorithm to classify the sample, and generates a deep classification index; The spectrum accurate positioning module is based on the deep classification index, uses the spectrum analysis method to analyze the spectrum characteristics of the sample sound wave data, uses the hidden Markov model to perform pattern recognition on the spectrum data, locates the sample components through the sound wave feature matching algorithm, and generates the component positioning result; The viral load calculation module analyzes the viral RNA and DNA amplification signals based on the component positioning results using the time-frequency domain analysis method, extracts the time-frequency characteristics of the signals using wavelet transform, reconstructs the spectrum using fast Fourier transform technology, calculates the load of the signals, and generates viral load calculation data; The spectral cell identification module uses spectral analysis technology to analyze the spectral characteristics of cell data collected by the flow cytometer based on the viral load calculation data, and then uses a deep belief network to perform deep learning processing on the spectral data, and identifies the cell type through a cell characteristic reconstruction algorithm to generate cell identification data; The cell electrical performance analysis module is based on cell identification data, combines microfluidics technology to control blood cell flow, applies cell electrical impedance analysis technology to measure the electrical impedance of cells flowing through, combines electrical performance analysis algorithms to deeply evaluate cell electrical characteristics, and generates electrical performance analysis indicators; The sample comprehensive decision module is based on the deep classification index, component location results, viral load calculation data, cell identification data, and electrical performance analysis index, and uses a comprehensive data analysis method for integration processing, comprehensively evaluates the sample through a decision support algorithm, and conducts a comprehensive decision analysis on the sample in combination with a risk assessment model to generate a comprehensive decision analysis result; The intelligent sample processing module optimizes the sample processing process based on the comprehensive decision analysis results using intelligent algorithms, applies process automation control technology to intelligently regulate the sample processing process, adjusts the sample processing process in combination with the processing efficiency optimization model, and generates intelligent regulation results.

2. The automated process control system for the laboratory department according to claim 1 is characterized in that: The deep classification indicators specifically include the spectral characteristics, image characteristics and deep correlations of the samples. The component positioning results include the acoustic characteristics and component identification of the samples. The viral load calculation data specifically refer to the time-frequency characteristics and load estimation of the viral amplification signal. The cell identification data include the spectral characteristics and type identification of the cells. The electrical performance analysis indicators specifically refer to the electrical impedance characteristics and state analysis of blood cells. The comprehensive decision-making analysis results include a comprehensive evaluation and risk analysis of the samples. The intelligent regulation results specifically include the optimization and automated adjustment of the sample processing flow.

3. The automated process control system for the laboratory department according to claim 1 is characterized in that: The pedigree depth analysis module includes a spectrum analysis submodule, a graph convolutional network application submodule, and a classification index optimization submodule; The spectrum analysis submodule uses the Fourier transform algorithm to decompose the spectrum characteristics based on the sample image, converts the time domain signal of the image into the frequency domain signal, and reveals the frequency characteristics of the sample image, including analyzing the periodic fluctuation of the brightness value of the image pixel points, capturing the frequency information of the sample, and performing frequency-based image analysis to generate basic spectrum information; The graph convolution network application submodule uses a graph convolution network for deep learning processing based on basic spectrum information, and uses the nonlinear characteristics of image data to capture the relationship between samples through graph convolution operations. The graph convolution network performs weighted aggregation on the features of each node, captures the key features of the image, and generates key feature information; The classification index optimization submodule optimizes the classification index based on the key feature information and applies the feature reconstruction algorithm, performs mathematical processing on the key features, including dimensionality reduction, normalization and pattern recognition, distinguishes multiple types of samples, accurately classifies the samples, and generates deep classification indicators.

4. The automated process control system for the laboratory department according to claim 1, characterized in that: The sound spectrum precise positioning module includes a sound wave analysis submodule, a sound spectrum model application submodule, and a component positioning optimization submodule; The acoustic wave analysis submodule uses the Fourier transform algorithm to analyze the spectrum characteristics of the sample acoustic wave data based on the deep classification index, analyzes and identifies the multiple frequency components of the sample acoustic wave by converting the acoustic wave signal from the time domain to the frequency domain, calculates the amplitude and phase of each frequency component, reveals the structure of the sample acoustic wave data, identifies key information, and draws the spectrum diagram of the sample acoustic wave to generate the acoustic wave spectrum characteristics; The spectrum model application submodule uses a hidden Markov model to perform pattern recognition on spectrum data based on the characteristics of the sound wave spectrum, identifies hidden patterns in the spectrum by calculating the transition probabilities between multiple states in the spectrum data, processes the uncertainty and randomness in the sound wave data, analyzes the changing trend of the sound wave characteristics, identifies specific components in the sample, and generates spectrum pattern recognition results; The component localization optimization submodule locates the sample components based on the results of the spectrum pattern recognition and applies the acoustic wave feature matching algorithm. It matches the specific components in the sample by comparing the features of the spectrum data with the pre-established component feature database. It comprehensively analyzes the spectrum data in combination with the sound wave frequency, amplitude and timing features, identifies and locates the target components in the sample, and generates component localization results.

5. The automated process control system for the laboratory department according to claim 1 is characterized in that: The viral load calculation module includes a time-frequency domain analysis submodule, a signal calculation submodule, and a load assessment submodule; The time-frequency domain analysis submodule uses short-time Fourier transform to analyze the time-frequency characteristics of viral RNA and DNA amplification signals based on the component positioning results, performs local Fourier transform on the signal within a limited time window, moves the window to cover the entire signal, obtains the spectrum information of the signal at multiple time points, and generates time-frequency domain analysis data; The signal calculation submodule applies wavelet transform to extract signal characteristics based on time-frequency domain analysis data, decomposes the signal into a series of wavelet basis functions, each basis function represents a local time-frequency feature of the signal, and captures specific features in the signal, including mutation points and non-stationary characteristics of the virus amplification signal, by adjusting the scale and translation parameters of the wavelet function, to generate signal characteristic data; The load assessment submodule reconstructs the signal spectrum and calculates the load based on the signal characteristic data in combination with the fast Fourier transform technology. By analyzing the spectral amplitude and phase information of the signal, the amplification degree of viral RNA and DNA is evaluated, and the viral load is calculated to generate viral load calculation data.

6. The automated process control system for the laboratory department according to claim 1, characterized in that: The spectral cell recognition module includes a spectral analysis optimization submodule, a deep learning processing submodule, and a cell recognition enhancement submodule; The spectral analysis optimization submodule uses Raman spectroscopy technology to perform spectral characteristic analysis on the cell data of the flow cytometer based on the viral load calculation data, uses a laser beam to irradiate the cells and captures the scattered light caused by the vibration of the cell molecules, analyzes its spectral characteristics, reveals the chemical and structural information of the cells, obtains the cell spectral characteristics, and generates spectral characteristic data; The deep learning processing submodule uses a convolutional neural network to perform deep learning processing on the spectral characteristic data based on the spectral characteristic data, extracts multi-layer features of the spectral characteristic data, and uses multiple convolutional layers and pooling layers of the CNN model to identify and extract key patterns and features in the cell data to generate deep learning features; The cell identification enhancement submodule is based on deep learning features and applies a cell characteristic reconstruction algorithm to identify cell types. Combined with the deep learning features, it performs a complex analysis of the multidimensional features of cells to identify cell biological properties, including protein expression and nucleic acid content on the cell surface, distinguish cell types, and generate cell identification data.

7. The automated process control system for the laboratory department according to claim 1 is characterized by: The cell electrical performance analysis module includes a microfluidic regulation submodule, an electrical impedance analysis submodule, and a performance evaluation submodule; The microfluidic regulation submodule uses microfluidic technology to control the flow of blood cells based on the cell identification data, separates and locates the cells by adjusting the liquid flow parameters in the microchannel, including flow rate and flow direction, and controls the position and flow state of the cells by controlling the design of the microfluidic channel and the physical properties of the fluid, thereby generating microfluidic positioning data; The electrical impedance analysis submodule uses the electrical impedance spectrum analysis technology to measure the electrical impedance of cells based on the microfluidic positioning data, obtains the electrical characteristics of the cells by applying current to the cells and measuring the impedance response of the cells to the current, including the evaluation of the membrane integrity of the cells and the activity characteristics of the ion channels inside the cells, and generates electrical impedance measurement data; The performance evaluation submodule uses statistical methods and machine learning algorithms based on the electrical impedance measurement data to deeply evaluate the electrical characteristics of cells, perform multi-dimensional analysis of the electrical impedance data, identify the various electrophysiological states and biological characteristics of cells, analyze the health status and physiological activities of cells, and generate electrical performance analysis indicators.

8. The automated process control system for the laboratory department according to claim 1 is characterized by: The sample comprehensive decision module includes an algorithm model fusion submodule, a decision analysis submodule, and a result optimization submodule; The algorithm model fusion submodule uses random forest and gradient boosting tree to perform data fusion based on deep classification indicators, component location results, viral load calculation data, cell identification data, and electrical performance analysis indicators, conducts comprehensive analysis of the data, conducts in-depth mining of multiple aspects of the data, extracts key patterns and trends, and generates a comprehensive data model; The decision analysis submodule uses a Bayesian network to conduct a comprehensive evaluation of samples based on a comprehensive data model, and builds a probabilistic graph model to quantitatively analyze the causal relationship between data, gain insights into sample characteristics and potential risks, and generate decision support analysis results. The result optimization submodule conducts comprehensive decision analysis of samples based on the decision support analysis results in combination with linear and logistic regression models, conducts quantitative analysis of the decision support analysis results, identifies key risk factors and potential trends of samples, evaluates the risk level of samples and the pros and cons of decision plans, optimizes the decision-making process, and generates comprehensive decision analysis results.

9. The automated process control system for the laboratory department according to claim 1, characterized in that: The intelligent sample processing module includes a process optimization submodule, an intelligent algorithm application submodule, and an intelligent regulation submodule; The process optimization submodule optimizes the sample processing process based on the comprehensive decision analysis results by using the linear programming method, and optimizes the resource allocation, time arrangement, and operation sequence by establishing a mathematical model to simulate the sample processing link, including setting the objective function, defining the constraints, and using the solver to match the optimal solution to generate the process optimization results; The intelligent algorithm application submodule uses a neural network algorithm to intelligently control the sample processing process based on the process optimization results, automatically adjusts the operating parameters and optimizes the processing steps by learning historical data and real-time feedback, processes input data by simulating the working mode of human brain neurons, learns to identify complex patterns and relationships, dynamically adjusts and optimizes process parameters, and generates intelligent processing results; The intelligent control submodule uses fuzzy logic control technology based on the intelligent processing results to adjust the sample processing process, and optimizes control decisions by processing uncertainty and approximate reasoning, including defining fuzzy variables, establishing fuzzy rule sets, applying fuzzy reasoning engines to make decisions, matching uncertain factors in the processing process, and generating intelligent control results.

10. A method for controlling an automated process in a laboratory, characterized in that: According to any one of claims 1 to 9, the laboratory automation process control system is implemented. The following steps are included: Based on the sample image, the Fourier transform algorithm is used to decompose the spectrum features, the graph convolutional network is used for deep learning processing, nonlinear features and key feature information are extracted, and the feature reconstruction algorithm is applied to optimize the classification index to generate deep classification indicators; Based on the deep classification index, the Fourier transform algorithm is used to analyze the spectrum characteristics of the sample sound wave data, the hidden Markov model is used to perform pattern recognition on the sound spectrum data, and the sound wave feature matching algorithm is used to locate the sample components to generate component positioning results; Based on the component positioning results, short-time Fourier transform is used to analyze the time-frequency characteristics of viral RNA and DNA amplification signals, wavelet transform is used to extract signal characteristics, and fast Fourier transform technology is combined to perform spectrum reconstruction and load calculation to generate viral load calculation data; Based on the viral load calculation data, Raman spectroscopy is used to analyze the spectral characteristics of the cell data, a convolutional neural network is used to perform deep learning processing on the spectral data, deep learning features are extracted, and a cell characteristic reconstruction algorithm is applied to identify the cell type to generate cell identification data; Based on the cell identification data, microfluidics technology is used to control the flow of cells, electrical impedance spectroscopy analysis technology is used to measure the electrical impedance of cells, statistical methods and machine learning algorithms are used to deeply evaluate the electrical characteristics of cells, and electrical performance analysis indicators are generated; Based on the deep classification indicators, component positioning results, viral load calculation data, cell identification data and electrical performance analysis indicators, random forests and gradient boosting trees are used for data fusion, Bayesian networks are applied for comprehensive evaluation, and linear and logistic regression models are combined for comprehensive decision analysis to optimize the decision-making process. Linear programming methods, neural network algorithms, and fuzzy logic control techniques are used to optimize and intelligently control the sample processing process to generate intelligent control results.

Images