Artificial intelligence-based organic sample processing method, device, equipment and storage medium

Abstract

The application provides an organic sample processing method and device based on artificial intelligence, electronic equipment and computer readable storage medium; relates to artificial intelligence technology; the method comprises the following steps: obtaining the fingerprint spectrum and protein marker information of the organic sample; performing feature extraction processing on the fingerprint spectrum of the organic sample to obtain the metabolic characteristics of the organic sample; performing prediction processing based on the metabolic characteristics of the organic sample to obtain the metabolic abnormal information of the organic sample; fusing the metabolic abnormal information of the organic sample and the protein marker information of the organic sample to obtain the fusion information of the organic sample; and performing prediction processing based on the fusion information of the organic sample to obtain the classification information of the organic sample. Through the application, the fingerprint spectrum and the protein marker of the organic sample can be combined for classification, and the accurate classification of the automatic organic sample can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to artificial intelligence technology, and in particular to an organic sample processing method and device based on artificial intelligence, an electronic device, and a computer readable storage medium. BACKGROUND

[0002] Artificial intelligence (AI) is a comprehensive technology of computer science, which enables machines to have perception, reasoning and decision-making functions by studying the design principles and implementation methods of various intelligent machines. Artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, such as natural language processing technology and machine learning / deep learning. With the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0003] In related technologies, there is a lack of effective solutions for classifying organic samples based on artificial intelligence. SUMMARY

[0004] The embodiments of the present application provide an organic sample processing method and device based on artificial intelligence, an electronic device, and a computer readable storage medium, which can classify organic samples in combination with fingerprint maps and protein markers to improve the accuracy of automatic classification of organic samples.

[0005] The technical solutions of the embodiments of the present application are as follows:

[0006] The embodiments of the present application provide an organic sample processing method based on artificial intelligence, comprising:

[0007] Obtaining fingerprint maps and protein marker information of the organic sample;

[0008] Performing feature extraction processing on the fingerprint maps of the organic sample to obtain metabolic characteristics of the organic sample;

[0009] Performing prediction processing based on the metabolic characteristics of the organic sample to obtain metabolic abnormal information of the organic sample;

[0010] Performing fusion processing on the metabolic abnormal information of the organic sample and the protein marker information of the organic sample to obtain fusion information of the organic sample;

[0011] Performing prediction processing based on the fusion information of the organic sample to obtain classification information of the organic sample.

[0012] In the above technical solution, the fusion processing is realized by a fusion network, and the fusion network includes a linear mapping layer and a fully connected layer.

[0013] The fusion information of the organic matter sample is obtained by fusing the metabolic abnormal information of the organic matter sample and the protein marker information of the organic matter sample.

[0014] The metabolic abnormal information of the organic matter sample and the protein marker information of the organic matter sample are weighted by a linear mapping layer in the fusion network to obtain weighted metabolic abnormal information and weighted protein marker information.

[0015] The weighted metabolic abnormal information and the weighted protein marker information are spliced by a full connection layer in the fusion network to obtain the fusion information of the organic matter sample.

[0016] After the fusion information of the organic matter sample is obtained by fusing the metabolic abnormal information of the organic matter sample and the protein marker information of the organic matter sample, the method further comprises:

[0017] Based on the classification information of the organic matter sample, the organic matter sample is clustered to obtain the category to which the organic matter sample belongs.

[0018] According to the category to which the organic matter sample belongs, the fingerprint spectrum of the organic matter sample and the related data of the organic matter sample are stored in the corresponding document in the database.

[0019] In the above technical solution, the method further comprises:

[0020] The metabolic characteristics of the pathological training sample are extracted by a prediction model to obtain the metabolic characteristics of the pathological training sample.

[0021] The metabolic abnormal information of the pathological training sample is obtained by predicting based on the metabolic characteristics of the pathological training sample.

[0022] The fusion information of the pathological training sample is obtained by fusing the metabolic abnormal information of the pathological training sample and the protein marker information of the pathological training sample.

[0023] The prediction classification information of the pathological training sample is obtained by predicting based on the fusion information of the pathological training sample.

[0024] Based on the prediction abnormal information and the abnormal information label of the pathological training sample, a loss function of the prediction model is constructed.

[0025] The parameters of the prediction model are updated until the loss function converges, and the updated parameters of the prediction model when the loss function converges are used as the parameters of the trained prediction model.

[0026] The embodiment of the present application provides a kind of organic sample processing device, comprising:

[0027] The acquisition module is used to acquire the fingerprint spectrum and protein marker information of the organic sample;

[0028] The extraction module is used to extract the feature of the fingerprint spectrum of the organic sample, to obtain the metabolic characteristics of the organic sample;

[0029] The first prediction module is used to predict based on the metabolic characteristics of the organic sample, to obtain the metabolic abnormal information of the organic sample;

[0030] The fusion module is used to fuse the metabolic abnormal information of the organic sample with the protein marker information of the organic sample, to obtain the fusion information of the organic sample;

[0031] The second prediction module is used to predict based on the fusion information of the organic sample, to obtain the classification information of the organic sample.

[0032] In the above technical solution, the acquisition module is also used to extract the spectrum of the organic sample, to obtain the original spectrum of the organic sample;

[0033] The original spectrum of the organic sample is preprocessed to obtain the fingerprint spectrum of the organic sample;

[0034] The protein marker detection processing is carried out on the organic sample to obtain the protein marker information of the organic sample.

[0035] In the above technical solution, the acquisition module is also used to execute the following processing through the automatic sample detection system:

[0036] The organic sample is moved to the mass spectrometry target plate, and the matrix is covered on the organic sample;

[0037] The mass spectrometer is controlled to detect the organic sample including the matrix, to obtain the original spectrum of the organic sample.

[0038] In the above technical solution, the acquisition module is also used to resample the original spectrum of the organic sample,

[0039] The obtained sampling spectrum is denoised to obtain the smoothed spectrum;

[0040] The baseline correction processing is carried out on the smoothed spectrum to obtain the corrected spectrum;

[0041] The peak extraction processing is carried out based on the corrected spectrum to obtain the fingerprint spectrum of the organic sample.

[0042] In the technical solution, the prediction processing is implemented by a first prediction network, and the first prediction network includes a full connection layer and an activation layer; the first prediction module is further configured to map the metabolic characteristics of the organic sample to a latent vector space through the full connection layer in the first prediction network to obtain a latent vector of the organic sample;

[0043] The latent vector of the organic sample is processed by the activation layer in the first prediction network to obtain a metabolic abnormality probability of the organic sample as metabolic abnormality information of the organic sample.

[0044] In the technical solution, the feature extraction processing is implemented by a feature extraction network, and the feature extraction network includes a plurality of cascaded encoding layers; the extraction module is further configured to perform feature encoding processing on the fingerprint spectrum of the organic sample through a first encoding layer in the plurality of cascaded encoding layers;

[0045] The encoding result of the first encoding layer is output to a subsequent cascaded encoding layer, so that feature encoding and encoding result output are continuously performed in the subsequent cascaded encoding layer until the last encoding layer, and

[0046] The encoding result output by the last encoding layer is taken as the metabolic characteristics of the organic sample.

[0047] In the technical solution, the encoding layer includes a full connection layer, a hidden layer and an activation layer; the extraction module is further configured to perform the following processing through an i th encoding layer in the plurality of cascaded encoding layers:

[0048] The encoding result of an i-1 th encoding layer is mapped through the full connection layer in the i th encoder to obtain an i th mapping vector;

[0049] The i th mapping vector is regularized through the hidden layer in the i th encoder to obtain an i th regularized vector;

[0050] The i th regularized vector is processed by the activation layer in the i th encoder to obtain the encoding result of the i th encoding layer, and the encoding result of the i th encoding layer is output to an i+1 th encoding layer;

[0051] Wherein, i is a natural number starting from 2 and increasing, and the value satisfies 2≤i≤N-1, and N is the number of the plurality of cascaded encoding layers.

[0052] In the technical solution, the extraction module is further configured to perform the following processing through the hidden layer in the i th encoding layer:

[0053] performing nonlinear mapping processing on the i th mapping vector to obtain an i th nonlinear mapping vector;

[0054] performing weighting processing on the i th nonlinear mapping vector, and taking a weighting result as the i th regularization vector.

[0055] In the technical solution, the extraction module is further configured to perform the following processing through a fully connected layer in the i th encoder:

[0056] performing connection processing on the encoding result of the i-1 th encoding layer and the encoding result of the i-2 th encoding layer to obtain an i th connection vector;

[0057] performing linear mapping processing on the i th connection vector to obtain the i th mapping vector.

[0058] In the technical solution, the fusion processing is implemented through a fusion network, and the fusion network comprises a linear mapping layer and a fully connected layer.

[0059] The fusion processing of the metabolic abnormality information of the organic sample and the protein marker information of the organic sample to obtain the fusion information of the organic sample comprises:

[0060] performing weighting on the metabolic abnormality information of the organic sample and the protein marker information of the organic sample through a linear mapping layer in the fusion network to obtain weighted metabolic abnormality information and weighted protein marker information;

[0061] performing splicing processing on the weighted metabolic abnormality information and the weighted protein marker information through a fully connected layer in the fusion network to obtain the fusion information of the organic sample.

[0062] In the technical solution, the prediction processing is implemented through a second prediction network, and the second prediction network comprises a first nonlinear network and a second nonlinear network; the second prediction module is further configured to perform down-sampling encoding processing on the fusion information of the organic sample through the first nonlinear network in the second prediction network to obtain latent space information of the organic sample.

[0063] performing up-sampling decoding processing on the latent space information through the second nonlinear network in the second prediction network to obtain classification information of the organic sample.

[0064] In the technical solution, the first nonlinear network comprises a fully connected layer, a hidden layer and an activation layer; the second prediction module is further configured to perform the following processing through the first nonlinear network in the second prediction network:

[0065] mapping processing on the fusion information of the organic sample through a full connection layer in the first nonlinear network to obtain a mapping vector of the fusion information;

[0066] regularization processing on the mapping vector of the fusion information through a hidden layer in the first nonlinear network to obtain a regularization vector of the fusion information;

[0067] nonlinear mapping processing on the regularization vector of the fusion information through an activation layer in the first nonlinear network to obtain hidden space information of the organic sample.

[0068] In the technical solution, the second nonlinear network includes a full connection layer and an activation layer; the second prediction module is further configured to perform mapping processing on the hidden space information through a full connection layer in the second nonlinear network to obtain a mapping vector of the organic sample;

[0069] perform nonlinear mapping processing on the mapping vector of the organic sample through an activation layer in the second nonlinear network, and take the obtained comprehensive abnormal probability of the organic sample as classification information of the organic sample.

[0070] An electronic device for organic sample processing is provided in an embodiment of the present application, and the electronic device comprises:

[0071] a memory configured to store executable instructions;

[0072] a processor configured to execute the executable instructions stored in the memory to implement the method for organic sample processing based on artificial intelligence provided in the embodiments of the present application.

[0073] A computer readable storage medium is provided in an embodiment of the present application, and the computer readable storage medium stores executable instructions, which are used to cause a processor to execute the method for organic sample processing based on artificial intelligence provided in the embodiments of the present application.

[0074] The embodiments of the present application have the following beneficial effects:

[0075] The method for organic sample processing based on artificial intelligence provided in the embodiments of the present application can predict the metabolic abnormal information of the organic sample based on the fingerprint spectrum of the organic sample, and determine the classification information of the organic sample by combining the metabolic abnormal information of the organic sample and the protein marker information of the organic sample, so that the precise classification process of the organic sample is realized, and the rapid grouping is realized. BRIEF DESCRIPTION OF DRAWINGS

[0076] Figure 1 is an application scenario diagram of a medical system provided in the embodiments of the present application;

[0077] Figure 2FIG. 1 is a structural schematic diagram of an electronic device for organic sample processing provided by an embodiment of the present application;

[0078] Figures 3-5 FIG. 2 is a flow schematic diagram of an organic sample processing method based on artificial intelligence provided by an embodiment of the present application;

[0079] Figure 6 FIG. 3 is a structural schematic diagram of a feature extraction network provided by an embodiment of the present application;

[0080] Figure 7 FIG. 4 is a structural schematic diagram of an encoding layer provided by an embodiment of the present application;

[0081] Figure 8 FIG. 5 is a structural schematic diagram of a first prediction network provided by an embodiment of the present application;

[0082] Figure 9 FIG. 6 is a structural schematic diagram of a first nonlinear network provided by an embodiment of the present application;

[0083] Figure 10 FIG. 7 is a flow schematic diagram of model classification provided by the related art;

[0084] Figure 11 FIG. 8 is a structural schematic diagram of a neural network provided by an embodiment of the present application. DETAILED DESCRIPTION

[0085] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be described in further detail below with reference to the drawings, and the described embodiments should not be regarded as limiting the present application, and all other embodiments obtained by those skilled in the art without making creative efforts fall within the scope of protection of the present application.

[0086] In the following description, the terms "first\second" are merely to distinguish similar objects, and do not represent a specific order of the objects. Understandably, the "first\second" can be interchanged in a specific order or sequence as allowed, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0087] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used herein are merely for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0088] The related data collection and processing in the embodiments of the present application should strictly comply with the requirements of relevant laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of authorization of laws and regulations and the personal information subject.

[0089] Before the embodiments of the present application are further described in detail, the terms and names involved in the embodiments of the present application are explained, and the terms and names involved in the embodiments of the present application are applicable to the following explanations.

[0090] 1) Deep Learning (DL): a new research direction in the field of Machine Learning (ML), which is introduced into machine learning to make it closer to the original goal-artificial intelligence (AI). Deep learning is to learn the internal rules and representation levels of sample data, and the information obtained in the learning process is very helpful for the interpretation of data such as text, images and sound. The ultimate goal of deep learning is to enable machines to have analysis and learning ability like people, and to be able to recognize text, images and sound data.

[0091] 2) Convolutional Neural Networks (CNN): a class of feedforward neural networks (FNN) containing convolutional computation and having deep structure, which is one of the representative algorithms of deep learning. Convolutional neural networks have representation learning ability and can perform shift-invariant classification on input images according to their hierarchical structure.

[0092] 3) Organic sample: a sample including various organic compounds collected from a living body, for example, urine, blood and the like collected from a human body.

[0093] 4) Protein marker: a protein index indicating changes or possible changes in a system, organ, tissue, cell and subcellular structure or function. Protein markers can be used to assist doctors and researchers in diagnosing diseases, rechecking and researching treatment methods.

[0094] 5) Fingerprint: refers to a chromatogram or spectrum that can indicate the chemical characteristics of certain complex substances, such as traditional Chinese medicine, DNA of a certain organism or a certain tissue or cell, after being properly treated and using certain analysis means. Among them, the metabolic fingerprint refers to a chromatogram or spectrum generated based on metabolites, for example, a chromatogram or spectrum generated based on blood.

[0095] The embodiment of the present application provides an organic sample processing method and device based on artificial intelligence, electronic equipment and computer readable storage medium, which can classify organic samples based on fingerprint spectrum and protein markers, and improve the accuracy of automatic classification of organic samples.

[0096] The organic sample processing method based on artificial intelligence provided by the embodiment of the present application can be realized by a terminal / server alone, or by a terminal and a server cooperatively, for example, the terminal alone bears the organic sample processing method based on artificial intelligence described below, or the terminal sends a classification request for an organic sample (including fingerprint spectrum and protein marker information of the organic sample) to the server, the server executes the organic sample processing method based on artificial intelligence according to the received classification request for the organic sample, and performs subsequent organic research, analysis and the like based on the classification information of the organic sample.

[0097] The electronic equipment for organic sample processing provided by the embodiment of the present application can be various types of terminal equipment or servers, wherein the server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services; the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal and the server can be directly or indirectly connected through wired or wireless communication, which is not limited in the present application.

[0098] Taking the server as an example, for example, it can be a server cluster deployed in the cloud, which opens an artificial intelligence cloud service (AI as a Service, AIaaS) to users, the AIaaS platform will split several common AI services, and provide independent or packaged services in the cloud. This service mode is similar to an AI theme mall, all users can access and use one or more artificial intelligence services provided by the AIaaS platform through the application programming interface.

[0099] For example, one of the artificial intelligence cloud services can be an organic sample processing service, i.e., a server in the cloud encapsulates a program for processing an organic sample provided by the embodiments of the present application. A user invokes the organic sample processing service in the cloud service through a terminal (running a client, such as a medical client, etc.) to enable the server deployed in the cloud to invoke the encapsulated program for processing an organic sample, perform prediction processing based on the fingerprint of the organic sample, obtain metabolic abnormality information (score for the fingerprint) of the organic sample, and combine the metabolic abnormality information of the organic sample and the protein marker information of the organic sample to determine the classification information of the organic sample. Subsequently, the classification request for the organic sample is responded based on the classification information of the organic sample, for example, for medical applications, the fingerprint and protein marker information of blood collected from an animal body are obtained, prediction processing is performed based on the fingerprint of the blood, metabolic abnormality information (such as abnormal urea index) of the blood is obtained, and the classification information (such as normal, abnormal, high blood lipids, low blood lipids, etc.) of the blood is determined in combination with the metabolic abnormality information of the blood and the protein marker information of the blood. The organic sample processing method based on artificial intelligence provided by the embodiments of the present application is not for objects with life or animal bodies, and is not directly aimed at obtaining disease diagnosis results or health conditions. The classification information of the organic sample cannot be directly used to obtain the diagnosis results of diseases or health conditions, and is only used to assist doctors and researchers in disease diagnosis, re-examination, and research on treatment methods.

[0100] Referring to Figure 1 , Figure 1 is a schematic diagram of an application scenario of a medical system 10 provided by the embodiments of the present application. The terminal 200 is connected to the server 100 through the network 300. The network 300 can be a wide area network or a local area network, or a combination of the two.

[0101] The terminal 200 (running a client, such as a medical client, etc.) can be used to obtain the fingerprint and protein marker information of the organic sample, for example, by using the automatic sample detection system of the terminal 200 to detect the organic sample to obtain the fingerprint and protein marker information of the organic sample, and automatically generate a classification request for the organic sample.

[0102] In some embodiments, the client running in the terminal can be implanted with an organic sample processing plug-in to implement an artificial intelligence-based organic sample processing method locally in the client. For example, after the terminal 200 obtains a classification request for an organic sample (including the fingerprint spectrum of the organic sample and the protein marker information of the organic sample), the organic sample processing plug-in is called to implement an artificial intelligence-based organic sample processing method, to perform prediction processing based on the fingerprint spectrum of the organic sample, to obtain metabolic abnormality information of the organic sample, and to determine classification information of the organic sample in combination with the metabolic abnormality information of the organic sample and the protein marker information of the organic sample, and to subsequently respond to the classification request for the organic sample based on the classification information of the organic sample.

[0103] In some embodiments, after the terminal 200 obtains a classification request for an organic sample, the organic sample processing interface of the server 100 (which can be provided in the form of a cloud service, i.e., an organic sample processing service) is called, the server 100 performs prediction processing based on the fingerprint spectrum of the organic sample, to obtain metabolic abnormality information of the organic sample, and to determine classification information of the organic sample in combination with the metabolic abnormality information of the organic sample and the protein marker information of the organic sample, and to subsequently respond to the classification request for the organic sample based on the classification information of the organic sample. For example, for a medical application, the medical application automatically obtains a classification request for an organic sample (including the fingerprint spectrum of the organic sample and the protein marker information of the organic sample), and calls the organic sample processing interface of the server 100, performs prediction processing based on the fingerprint spectrum of the organic sample, to obtain metabolic abnormality information of the organic sample (e.g., abnormal urea index), and to determine classification information of the organic sample (e.g., normal, abnormal, high blood lipids, low blood lipids, etc.) in combination with the metabolic abnormality information of the organic sample and the protein marker information of the organic sample. The organic sample processing method provided in the embodiments of the present application is not targeted at living or animal bodies, and is not directly aimed at obtaining disease diagnosis results or health conditions. The classification information of the organic sample cannot be directly used to obtain disease diagnosis results or health conditions, i.e., the classification information of the organic sample is not directly used for disease diagnosis, but only serves as intermediate data to assist patients in predicting diseases, and to assist doctors and researchers in diagnosing diseases, rechecking, and researching treatment methods.

[0104] The structure of the electronic device for organic sample processing provided in the embodiments of the present application is described below. Referring to Figure 2 , Figure 2 is a structural schematic diagram of an electronic device 500 for organic sample processing provided in the embodiments of the present application, which is taken as an example of a server, Figure 2The electronic device 500 shown for organic sample processing includes at least one processor 510, a memory 550, at least one network interface 520, and a user interface 530. The various components in the electronic device 500 are coupled together by a bus system 540. It is understood that the bus system 540 is used for communicating data between the components. The bus system 540 includes a data bus, a power bus, a control bus, and a state signal bus, although for the sake of clarity, the various buses are shown as the bus system 540. Figure 2

[0105] The processor 510 can be an integrated circuit chip that has processing capability, such as a general purpose processor, a digital signal processor (DSP), or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc.

[0106] The memory 550 includes volatile memory or non-volatile memory, and can include both volatile and non-volatile memory. The non-volatile memory can be read only memory (ROM), and the volatile memory can be random access memory (RAM). The memory 550 described in the embodiments of the present application is intended to include any suitable type of memory. The memory 550 can optionally include one or more storage devices that are physically separate from the processor 510.

[0107] In some embodiments, the memory 550 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, which are described below.

[0108] The operating system 551 includes system programs for processing various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks.

[0109] The network communication module 552 is used to communicate with other computing devices via one or more (wired or wireless) network interfaces 520, examples of which include Bluetooth, wireless fidelity (WiFi), universal serial bus (USB), etc.

[0110] ​In some embodiments, the organic sample processing apparatus provided by the embodiments of the present application can be implemented in a software manner, for example, can be an organic sample processing plug-in in the terminal described above, or can be an organic sample processing service in the server described above. Of course, the organic sample processing apparatus provided by the embodiments of the present application can be provided in various software embodiments, including various forms such as application programs, software, software modules, scripts or codes.

[0111] Figure 2 An organic sample processing apparatus 555 stored in the memory 550 is shown, which can be software in the form of programs and plug-ins, for example, an organic sample processing plug-in, and includes a series of modules, including an acquisition module 5551, an extraction module 5552, a first prediction module 5553, a fusion module 5554 and a second prediction module 5555; wherein the acquisition module 5551, the extraction module 5552, the first prediction module 5553, the fusion module 5554 and the second prediction module 5555 are used to implement the organic sample processing function provided by the embodiments of the present application.

[0112] As described above, the organic sample processing method based on artificial intelligence provided by the embodiments of the present application can be implemented by various types of electronic devices. Referring to Figure 3 , Figure 3 is a flowchart of the organic sample processing method based on artificial intelligence provided by the embodiments of the present application, which is described in combination with Figure 3 the steps shown.

[0113] In the following steps, the fingerprint spectrum and the protein marker information of the organic sample can be detected by the automatic sample feeding detection system of the terminal, the terminal automatically generates a classification request for the organic sample (including the fingerprint spectrum and the protein marker information of the organic sample), and sends the classification request for the organic sample to the server, and the server parses the classification request for the organic sample to obtain the fingerprint spectrum and the protein marker information of the organic sample. The fingerprint spectrum and the protein marker information of the organic sample can also be detected by the automatic sample feeding detection system of the server.

[0114] In step 101, the fingerprint spectrum and the protein marker information of the organic sample are acquired.

[0115] For example, the user sends the organic sample to the automatic sample feeding detection system of the server, and the automatic sample feeding detection system will automatically perform a detection task to obtain the fingerprint spectrum and the protein marker information of the organic sample, so as to subsequently perform a prediction task based on the fingerprint spectrum and the protein marker information of the organic sample.

[0116] In some embodiments, the original spectrum of the organic sample is obtained through spectrum extraction processing of the organic sample by the automatic sample feeding detection system; the fingerprint spectrum of the organic sample is obtained through pretreatment of the original spectrum of the organic sample; and the protein marker information of the organic sample is obtained through protein marker detection processing of the organic sample.

[0117] In the automatic sample feeding detection system, the organic sample is detected in a fully automated manner, so as to completely eliminate manual operation and greatly reduce labor costs.

[0118] In some embodiments, the original spectrum of the organic sample is obtained through spectrum extraction processing of the organic sample, including: moving the organic sample to a mass spectrometry target plate and covering a matrix on the organic sample through the automatic sample feeding detection system; and detecting the organic sample including the matrix through a mass spectrometer to obtain the original spectrum of the organic sample.

[0119] For example, the organic sample is transmitted to the mass spectrometry target plate through a mechanical transmission device in the automatic sample feeding detection system, and is left for a period of time to be air-dried. Then, the matrix is covered on the organic sample through a mechanical arm in the automatic sample feeding detection system, and is left for a period of time to be air-dried. Then, a completion signal is sent to the matrix-assisted laser desorption ionization mass spectrometer to trigger the matrix-assisted laser desorption ionization mass spectrometer to start detection, and the mass spectrum of the metabolic small molecule is collected to obtain the original spectrum of the organic sample.

[0120] In some embodiments, the fingerprint spectrum of the organic sample is obtained through pretreatment of the original spectrum of the organic sample, including: performing resampling processing on the original spectrum of the organic sample, performing denoising processing on the obtained sampled spectrum to obtain a smoothed spectrum, performing baseline correction processing on the smoothed spectrum to obtain a corrected spectrum, and performing spectrum peak extraction processing based on the corrected spectrum to obtain the fingerprint spectrum of the organic sample.

[0121] For example, equidistant binning is used to reduce the data dimension to realize resampling, and the sampled spectrum is obtained. Specifically, the m / z section in the original spectrum is divided into N equal parts, and N is a positive integer greater than 2. After obtaining the sampled spectrum, Gaussian filtering is used to smooth the noise to obtain the smoothed spectrum, for example, the sigma parameter is set to 1 and the window radius is set to 20. After obtaining the smoothed spectrum, the top-hat method in morphology is used for baseline correction. The implementation process of the top-hat method is as follows: morphological erosion and dilation operations are performed on the smoothed spectrum to obtain an open operated spectrum, and the smoothed spectrum is subtracted from the open operated spectrum to obtain the fingerprint spectrum.

[0122] In step 102, feature extraction processing is performed on the fingerprint of the organic matter sample to obtain the metabolic characteristics of the organic matter sample.

[0123] After the server obtains the fingerprint of the organic matter sample, feature extraction processing is performed on the fingerprint of the organic matter sample to obtain the metabolic characteristics of the organic matter sample, so that subsequent prediction processing can be performed through the metabolic characteristics of the organic matter sample.

[0124] In some embodiments, the feature extraction processing is implemented through a feature extraction network, and the feature extraction network includes a plurality of cascaded encoding layers. The feature extraction processing on the fingerprint of the organic matter sample to obtain the metabolic characteristics of the organic matter sample includes: performing feature encoding processing on the fingerprint of the organic matter sample through a first encoding layer in the plurality of cascaded encoding layers; outputting the encoding result of the first encoding layer to a subsequent encoding layer for continuing feature encoding and encoding result output in the subsequent encoding layer until the last encoding layer, and outputting the encoding result of the last encoding layer as the metabolic characteristics of the organic matter sample.

[0125] As shown in Figure 6 , the feature extraction network includes a plurality of cascaded encoding layers, the first encoding layer performs feature encoding on the fingerprint of the organic matter sample and outputs the encoding result to the second encoding layer, the second encoding layer continues feature encoding and encoding result output until the Nth encoding layer, where N is the total number of the plurality of cascaded encoding layers. Through the cascaded encoding layers, features can be extracted hierarchically to obtain accurate metabolic characteristics.

[0126] In some embodiments, the encoding layer includes a fully connected layer, a hidden layer, and an activation layer. The feature encoding and encoding result output are continued in the subsequent cascaded encoding layer, including: performing the following processing through the i th encoding layer of the plurality of cascaded encoding layers: mapping the encoding result of the i-1 th encoding layer through the fully connected layer in the i th encoder to obtain an i th mapping vector; performing regularization processing on the i th mapping vector through the hidden layer in the i th encoder to obtain an i th regularization vector; performing nonlinear mapping processing on the i th regularization vector through the activation layer in the i th encoder to obtain the encoding result of the i th encoding layer, and outputting the encoding result of the i th encoding layer to the i+1 th encoding layer; where i is a natural number starting from 2 and taking values satisfying 2≤i≤N-1, and N is the number of the plurality of cascaded encoding layers.

[0127] As shown in the above example, Figure 7 , the hidden layer is a Dropout layer, and the Dropout layer is used for regularization processing to alleviate the overfitting problem; the activation layer uses a ReLU activation function for nonlinear mapping processing to complete the nonlinear grouping task.

[0128] For example, the i-th mapping vector is regularized by the hidden layer in the i-th encoder to obtain the i-th regularized vector, including: performing the following processing by the hidden layer in the i-th encoding layer: performing nonlinear mapping processing on the i-th mapping vector to obtain the i-th nonlinear mapping vector; performing weighting processing on the i-th nonlinear mapping vector, and taking the weighting result as the i-th regularized vector.

[0129] For example, the encoding result of the i-1-th encoding layer is linearly mapped by the fully connected layer in the i-th encoder to obtain the i-th mapping vector, including: performing the following processing by the fully connected layer in the i-th encoder: connecting the encoding result of the i-1-th encoding layer and the encoding result of the i-2-th encoding layer to obtain the i-th connection vector; performing linear mapping processing on the i-th connection vector to obtain the i-th mapping vector. The skip connection mechanism is adopted between the fully connected layers to alleviate the gradient disappearance and strengthen the information transmission.

[0130] In step 103, the prediction processing is performed based on the metabolic characteristics of the organic matter sample to obtain the metabolic abnormal information of the organic matter sample.

[0131] After the server obtains the metabolic characteristics of the organic matter sample, the prediction model is combined with the metabolic characteristics of the organic matter sample to perform prediction processing to obtain the metabolic abnormal information of the organic matter sample, so as to subsequently perform classification processing based on the metabolic abnormal information of the organic matter sample, for example, the metabolic abnormal information is that the urea is too high and the snow routine is too high.

[0132] Referring to Figure 4 , Figure 4 is an optional flowchart of the organic matter sample processing method based on artificial intelligence provided by the embodiment of the present application, Figure 4 illustrates Figure 3 Step 103 in the embodiment can be implemented by step 1031-step 1032: the prediction processing is implemented by a first prediction network, and the first prediction network includes a fully connected layer and an activation layer; in step 1031, the metabolic characteristics of the organic matter sample are mapped to a hidden vector space by the fully connected layer in the first prediction network to obtain the hidden vector of the organic matter sample; in step 1032, the hidden vector of the organic matter sample is nonlinearly mapped by the activation layer in the first prediction network, and the obtained metabolic abnormal probability of the organic matter sample is taken as the metabolic abnormal information of the organic matter sample.

[0133] As Figure 8As shown, the first prediction network includes a two-layer structure, i.e., a fully connected layer and an activation layer, wherein the activation layer adopts a Sigmoid activation function, and the Sigmoid activation function forces the first prediction network to complete a nonlinear grouping task to obtain metabolic abnormality information, which is a metabolic abnormality probability of the organic sample, for example, a metabolic abnormality probability of the organic sample is 0.6, and a metabolic normality probability of the organic sample is 0.4.

[0134] In step 104, the metabolic abnormality information of the organic sample is fused with the protein marker information of the organic sample to obtain fusion information of the organic sample.

[0135] For example, after the server obtains the metabolic abnormality information of the organic sample and the protein marker information of the organic sample, the metabolic abnormality information of the organic sample and the protein marker information of the organic sample can be spliced to obtain the fusion information of the organic sample, so as to perform prediction processing based on the fusion information of the organic sample subsequently, and realize accurate classification of the organic sample.

[0136] For example, the fusion processing is realized through a fusion network, and the fusion network includes a linear mapping layer and a fully connected layer; the metabolic abnormality information of the organic sample and the protein marker information of the organic sample are weighted through the linear mapping layer in the fusion network to obtain weighted metabolic abnormality information and weighted protein marker information; and the weighted metabolic abnormality information and the weighted protein marker information are spliced through the fully connected layer in the fusion network to obtain the fusion information of the organic sample.

[0137] In step 105, prediction processing is performed based on the fusion information of the organic sample to obtain classification information of the organic sample.

[0138] After the server obtains the fusion information of the organic sample, prediction is performed based on the fusion information of the organic sample through a second prediction network to obtain classification information of the organic sample, for example, the organic sample is normal or abnormal.

[0139] Referring to Figure 5 , Figure 5 is an optional flowchart of an organic sample processing method based on artificial intelligence provided by the embodiment of the present application, Figure 5 is shown Figure 3The step 105 in the method 100 can be implemented by steps 1051-1052: the prediction processing is implemented by a second prediction network, the second prediction network comprising a first nonlinear network and a second nonlinear network; in step 1051, the fusion information of the organic matter sample is subjected to down-sampling encoding processing by the first nonlinear network in the second prediction network, to obtain the latent space information of the organic matter sample; in step 1052, the latent space information is subjected to up-sampling decoding processing by the second nonlinear network in the second prediction network, to obtain the classification information of the organic matter sample.

[0140] In some embodiments, the first nonlinear network comprises a fully connected layer, a hidden layer, and an activation layer; the down-sampling encoding processing of the fusion information of the organic matter sample by the first nonlinear network in the second prediction network comprises: performing the following processing by the first nonlinear network in the second prediction network: mapping processing of the fusion information of the organic matter sample by the fully connected layer in the first nonlinear network, to obtain a mapping vector of the fusion information; regularization processing of the mapping vector of the fusion information by the hidden layer in the first nonlinear network, to obtain a regularization vector of the fusion information; nonlinear mapping processing of the regularization vector of the fusion information by the activation layer in the first nonlinear network, to obtain the latent space information of the organic matter sample.

[0141] As shown in the above examples, the hidden layer is a Dropout layer, and the Dropout layer is used for regularization processing to alleviate the overfitting problem; the activation layer uses a ReLU activation function for nonlinear mapping processing to complete the nonlinear grouping task. Figure 9 As shown in the above examples, the hidden layer is a Dropout layer, and the Dropout layer is used for regularization processing to alleviate the overfitting problem; the activation layer uses a ReLU activation function for nonlinear mapping processing to complete the nonlinear grouping task.

[0142] For example, the hidden layer in the first nonlinear network performs the following processing: nonlinear mapping processing of the mapping vector of the fusion information to obtain a nonlinear mapping vector, and weighting processing of the nonlinear mapping vector, with the weighted result being taken as the regularization vector of the fusion information.

[0143] As shown in the above examples, the hidden layer is a Dropout layer, and the Dropout layer is used for regularization processing to alleviate the overfitting problem; the activation layer uses a ReLU activation function for nonlinear mapping processing to complete the nonlinear grouping task.

[0144] In some embodiments, after the prediction processing based on the fusion information of the organic matter sample is performed to obtain the classification information of the organic matter sample, the method further includes: performing clustering processing on the organic matter sample based on the classification information of the organic matter sample to obtain a category to which the organic matter sample belongs; and storing the fingerprint spectrum of the organic matter sample and related data of the organic matter sample into a corresponding document in the database according to the category to which the organic matter sample belongs.

[0145] After the server obtains the classification information of the organic matter sample, the fingerprint spectrum of the organic matter sample, the protein marker information of the organic matter sample, and related data of the organic matter sample (such as the vital signs and daily habits of the organism corresponding to the organic matter sample) are stored into a corresponding document in the database according to the category to which the organic matter sample belongs (such as a positive sample or a negative sample), so that the required organic matter sample material can be quickly located according to the category to which the image belongs for subsequent related research.

[0146] For example, when the number of organic matter samples of the same category in the database exceeds a threshold value, the database can be periodically cleaned to avoid a situation of insufficient storage space in the database. For example, the organic matter samples that have been stored for a long time (exceeding a time threshold value) can be removed, and the organic matter samples that have been referenced a low number of times (less than a threshold value) can also be removed.

[0147] In some embodiments, in order to classify the organic matter sample through the prediction model, the prediction model needs to be trained, where the prediction model includes a feature extraction network, a first prediction network, a fusion network, and a second prediction network. The training process includes: performing metabolic feature extraction processing on the fingerprint spectrum of a pathological training sample through the feature extraction network in the prediction model to obtain metabolic features of the pathological training sample; performing prediction processing on the metabolic features of the pathological training sample based on the first prediction network to obtain metabolic abnormal information of the pathological training sample; performing fusion processing on the metabolic abnormal information of the pathological training sample and the protein marker information of the pathological training sample through the fusion network to obtain fusion information of the pathological training sample; performing prediction processing on the fusion information of the pathological training sample based on the second prediction network to obtain prediction classification information of the pathological training sample; constructing a loss function of the prediction model based on the prediction classification information and a classification information label of the pathological training sample; updating parameters of the prediction model until the loss function converges; and taking the updated parameters of the prediction model when the loss function converges as the parameters of the trained prediction model.

[0148] For example, the organic matter training sample is classified by the prediction model to obtain prediction classification information of the organic matter training sample. After the value of the loss function (for example, a cross-entropy loss function) of the prediction model is determined according to the prediction classification information and the classification information label of the pathological training sample, it can be judged whether the value of the loss function of the prediction model exceeds a preset threshold. When the value of the loss function of the prediction model exceeds the preset threshold, the error signal of the prediction model is determined based on the loss function of the prediction model, the error information is back propagated in the prediction model, and the model parameters of each layer are updated in the process of propagation.

[0149] Here, the back propagation is described. The training sample data is input to the input layer of the neural network model, passes through the hidden layer, finally reaches the output layer and outputs the result, which is the forward propagation process of the neural network model. Since there is an error between the output result of the neural network model and the actual result, the error between the output result and the actual value is calculated, and the error is back propagated from the output layer to the hidden layer until it propagates to the input layer. In the process of back propagation, the value of the model parameter is adjusted according to the error; the above process is iterated continuously until convergence. The prediction model belongs to the neural network model.

[0150] In the following, an exemplary application of the embodiments of the present application in a practical medical application scenario will be described.

[0151] The embodiments of the present application can be applied in a medical application scenario. For example, a terminal (running a medical application) automatically acquires a classification request (including a fingerprint spectrum of a sample and protein marker information) for an organic matter sample (a sample such as urine, blood, etc.), and calls an organic matter sample processing interface of a server to determine classification information (for example, normal, abnormal, etc.) of the organic matter sample based on the fingerprint spectrum of the sample and the protein marker of the sample. The organic matter sample processing method provided by the embodiments of the present application is not targeted at a living or animal body, and is not directly aimed at obtaining a disease diagnosis result or a health condition. The disease diagnosis result or the health condition cannot be directly obtained according to the classification information of the organic matter sample, that is, the classification information of the organic matter sample is not directly used for disease diagnosis, but only serves as intermediate data to assist patients in predicting diseases, and assist doctors and researchers in diagnosing diseases, rechecking, and researching treatment methods.

[0152] In the related art, as shown in Figure 10 As shown, the urine metabolic fingerprint spectrum is extracted by using the polymer silver nanoparticle, the urine metabolic fingerprint spectrum in the control data set and the reference data set is extracted, and the model is trained by using the urine metabolic fingerprint spectrum in the control data set and the reference data set, so as to classify by using the trained model to obtain the classification result.

[0153] However, the related art only considers the decoding information of the metabolic fingerprint, and the grouping performance still has room for improvement. In addition, the related art fuses all omics data at one time, ignores the inconsistency and difference between different platform data, and the comprehensive analysis result leads to a large classification error.

[0154] To solve the above problems, in the biomedical field, the molecular phenotypes presented by different disease subtypes and different cell types are complex, and the fusion of metabolic fingerprint and protein marker decoding information can provide new ideas for researchers to better utilize multi-omics data to study the molecular typing of different molecular phenotypes. In-depth study of the expression characteristics of metabolic fingerprint and protein marker, and development of a new method suitable for the fusion of the two types of data. The embodiment of the application proposes a method of fusing protein markers and metabolic fingerprints based on neural network technology, relying on matrix-assisted laser desorption ionization mass spectrometry technology to detect sample metabolic fingerprints and protein markers, and realizing rapid grouping, so as to classify the metabolic fingerprint and protein marker information based on artificial intelligence, and construct a sample grouping model to realize artificial intelligence assisted mass spectrometry data mining.

[0155] The method of fusing protein markers and metabolic fingerprints based on neural network technology according to the embodiment of the application will be described in detail below, and the processing process is shown in the following steps:

[0156] Step 1: The sample is sucked onto the mass spectrometry target plate and dried as a thin layer for subsequent mass spectrometry analysis. Specifically, a pipette is used to take 1 μL of sample and point it onto the mass spectrometry target plate, and then air dry; 1 μL of matrix is taken and covered on the sample, and then air dry.

[0157] Step 2: A matrix-assisted laser desorption ionization mass spectrometer is used to collect metabolic small molecule mass spectra of each sample to obtain an original spectrum, for example, the collection mode is positive ion mode, and the mass-to-charge ratio range is 100-1000.

[0158] Step 3: The signal peaks of the original spectrum are extracted for pretreatment as metabolic fingerprints. The specific operations include resampling, spectral peak smoothing, baseline correction, and spectral peak extraction. Among them, the resampling part uses equidistant binning to reduce the data dimension, for example, the m / z segment in the original spectrum is 100-1000, which is divided into 20000 equal parts, and the length of each equal part is 0.045; the spectral peak smoothing uses Gaussian filter to smooth the noise, for example, the specific parameter setting: sigma is 1, and the window radius is 20; the baseline correction uses the top-hat method in morphology for baseline correction, and the top-hat includes subtracting the spectrum after opening operation from the spectrum after equidistant binning and Gaussian smoothing, and the specific formula is shown in formulas (1)-(2):

[0159] x_o=dilation(erosion(x_input))(1)

[0160] y = x_input - x_o(2)

[0161] Where x_input represents the spectrum after equal-interval binning and Gaussian smoothing, erosion represents the erosion operation, dilation represents the dilation operation, x_o represents the spectrum after the opening operation, and the final baseline-corrected spectrum y is the difference between the spectrum after equal-interval binning and Gaussian smoothing and the spectrum after the opening operation.

[0162] In the peak extraction process, the signal peaks are first roughly extracted using the maximum value, and then the signal peaks are refined using the signal-to-noise ratio (SNR). For example, the threshold for the SNR-refined peaks is 3.

[0163] Step 4: Use enzyme-linked immunosorbent assay (ELISA) or electrochemical sensing methods to detect and collect protein biomarker concentration information and other information from the sample.

[0164] Step 5: Input the data from Steps 3 and 4 into the neural network for sample grouping processing.

[0165] The neural network structure used in step 5 to process the data extracted from the samples is as follows:

[0166] like Figure 11 As shown, the multilayer neural network includes: network input, network body, and network output; the network body includes a metabolic scoring part (including a feature extraction network (for extracting metabolic features) and a first prediction network) and a metabolism-protein nonlinear fusion part (including a fusion network and a second prediction network). The metabolic fingerprint map serves as the input to the metabolic scoring part, undergoing metabolic feature extraction processing. The output of the metabolic feature extraction part (metabolic features) and protein biomarker information serve as the input to the metabolism-protein nonlinear fusion part, ultimately outputting the grouping probabilities after nonlinear fusion.

[0167] The underlying formulas from the network input to the nonlinear fusion output are as follows: (3)-(6)

[0168] x_met_feature=met_feature_extract(x_met)(3)

[0169] x_met_pro=concatenate(x_met_feature,x_pro)(4)

[0170] x_f=feature_extract(x_met_pro)(5)

[0171] y_pred = sigmoid(x_f) (6)

[0172] Wherein, x_met represents metabolic fingerprint information (metabolic fingerprint map), x_met_feature represents the score of the neural network for metabolic fingerprint information, x_pro represents protein marker information, x_met_pro represents the score of metabolic fingerprint information and protein marker information, x_f represents the features extracted by the neural network for the score of metabolic fingerprint information and protein marker information, y_pred represents the nonlinear fusion comprehensive score of metabolic fingerprint information and protein (the final output probability of grouping, i.e. classification information).

[0173] The input of the neural network is metabolic fingerprint information (such as 1-10000 dimensions), 2-dimensional gender, age information and protein marker information (such as 1-20 dimensions). The neural network is mainly divided into two parts: metabolic scoring part, metabolic-protein nonlinear fusion part.

[0174] The metabolic scoring part is stacked by fully connected layers (such as 1-20 layers), which is divided into two levels, the first level: such as the first 10 layers of fully connected layer followed by Dropout layer and ReLU activation function, and the number of neurons is 1 to 10000; the second level: only fully connected layer followed by Sigmoid layer, and the number of neurons is 2, outputting the prediction probability of positive and negative sample grouping (i.e. disease, normal), and the probability sum is 1. The skip connection mechanism is adopted between the fully connected layers to relieve the gradient vanishing and strengthen the information transmission.

[0175] The metabolic-protein nonlinear fusion part includes two layers of nonlinear network, i.e. two layers of fully connected layer, the first layer of fully connected layer followed by ReLU activation function and the number of neurons is 32, and the second layer only has fully connected layer followed by Sigmoid layer and the number of neurons is 2; finally, two layers of nonlinear network are established according to the score of metabolic fingerprint information, protein marker information and sample basic information (such as gender, age information, etc.) to decode the sample state, and get the nonlinear fusion comprehensive score (the final output probability of grouping, i.e. classification information). The overall neural network design adopts Dropout layer to relieve the overfitting problem, and adopts ReLU activation function to force the neural network to complete the nonlinear grouping task.

[0176] In summary, the embodiment of the present application provides a more reasonable method for classifying and typing multi-omics data, which not only considers the influence of different types of data on typing, but also reasonably integrates the different influences of multi-omics data. The laser desorption ionization mass spectrometry is simple and convenient to operate. Compared with decoding only the metabolic fingerprint, the data preprocessing and analysis of the embodiment of the present application have certain improvement in grouping performance after fusing protein markers and other information, and the result reliability is relatively improved.

[0177] The organic sample processing method based on artificial intelligence provided by the embodiment of the present application has been described in combination with the exemplary application and implementation of the server provided by the embodiment of the present application. The embodiment of the present application also provides an organic sample processing device. In actual application, each functional module in the organic sample processing device can be cooperatively implemented by hardware resources of an electronic device (such as a terminal device, a server or a server cluster), such as a processor and other computing resources, communication resources (such as those used to support various communication modes such as optical cable and cellular network), and a memory. Figure 2 The organic sample processing device 555 stored in the memory 550 is shown, which can be software in the form of programs and plug-ins, for example, software modules designed using software C / C++, Java and other programming languages, special software modules in application software or large software systems designed using C / C++, Java and other programming languages, application program interfaces, plug-ins, cloud services and other implementation manners. Different implementation manners are exemplarily described below.

[0178] Example 1: The organic sample processing device is a mobile terminal application program and module

[0179] The organic sample processing device 555 in the embodiment of the present application can be provided as a software module designed using software C / C++, Java and other programming languages, and embedded into various mobile terminal applications based on Android or iOS systems (stored in the storage medium of the mobile terminal in the form of executable instructions and executed by the processor of the mobile terminal), so as to directly use the computing resources of the mobile terminal itself to complete the related information recommendation task, and periodically or irregularly transmit the processing result to a remote server through various network communication modes, or save it locally on the mobile terminal.

[0180] Example 2: The organic sample processing device is a server application program and platform

[0181] The organic sample processing device 555 in the embodiment of the present application can be provided as a special software module in an application software or a large software system designed using C / C++, Java and other programming languages, and run on a server side (stored in the storage medium of the server side in the form of executable instructions and run by the processor of the server side), so as to use the computing resources of the server itself to complete the related information recommendation task.

[0182] The embodiment of the present application can also be provided as a distributed, parallel computing platform composed of multiple servers, with customized, easy-to-interact web interfaces or other user interfaces (UIs) mounted thereon, forming an information recommendation platform (for recommendation lists) for individuals, groups or units.

[0183] Example Three, the organic sample processing device is a server-side application program interface (API) and plug-in

[0184] The organic sample processing device 555 in the embodiment of the present application can be provided as a server-side API or plug-in, called by a user to execute the artificial intelligence-based organic sample processing method of the embodiment of the present application, and embedded into various application programs.

[0185] Example Four, the organic sample processing device is a mobile device client-side API and plug-in

[0186] The organic sample processing device 555 in the embodiment of the present application can be provided as a mobile device-side API or plug-in, called by a user to execute the artificial intelligence-based organic sample processing method of the embodiment of the present application.

[0187] Example Five, the organic sample processing device is a cloud-side open service

[0188] The organic sample processing device 555 in the embodiment of the present application can be provided as an information recommendation cloud service developed for users, for individuals, groups or units to obtain recommendation lists.

[0189] The organic sample processing device 555 includes a series of modules, including an acquisition module 5551, an extraction module 5552, a first prediction module 5553, a fusion module 5554 and a second prediction module 5555. The following continues to describe the cooperation of each module in the organic sample processing device 555 provided by the embodiment of the present application to implement the organic sample processing scheme.

[0190] The acquisition module 5551 is configured to acquire fingerprint spectrum and protein marker information of an organic sample; the extraction module 5552 is configured to perform feature extraction processing on the fingerprint spectrum of the organic sample to obtain metabolic characteristics of the organic sample; the first prediction module 5553 is configured to perform prediction processing based on the metabolic characteristics of the organic sample to obtain metabolic abnormal information of the organic sample; the fusion module 5554 is configured to perform fusion processing on the metabolic abnormal information of the organic sample and the protein marker information of the organic sample to obtain fusion information of the organic sample; and the second prediction module 5555 is configured to perform prediction processing based on the fusion information of the organic sample to obtain classification information of the organic sample.

[0191] In some embodiments, the acquisition module 5551 is further configured to perform spectrum extraction processing on the organic matter sample to obtain an original spectrum of the organic matter sample; perform preprocessing on the original spectrum of the organic matter sample to obtain a fingerprint spectrum of the organic matter sample; and perform protein marker detection processing on the organic matter sample to obtain protein marker information of the organic matter sample.

[0192] In some embodiments, the acquisition module 5551 is further configured to perform the following processing by using an automatic sample delivery detection system: moving the organic matter sample to a mass spectrometry target plate and covering a matrix on the organic matter sample; and controlling a mass spectrometer to detect the organic matter sample including the matrix to obtain an original spectrum of the organic matter sample.

[0193] In some embodiments, the acquisition module 5551 is further configured to perform resampling processing on the original spectrum of the organic matter sample, perform denoising processing on the obtained sampled spectrum to obtain a smoothed spectrum, perform baseline correction processing on the smoothed spectrum to obtain a corrected spectrum, and perform spectrum peak extraction processing based on the corrected spectrum to obtain a fingerprint spectrum of the organic matter sample.

[0194] In some embodiments, the prediction processing is implemented by using a first prediction network, and the first prediction network includes a fully connected layer and an activation layer; the first prediction module 5553 is further configured to map the metabolic characteristics of the organic matter sample to a hidden vector space by using the fully connected layer in the first prediction network to obtain a hidden vector of the organic matter sample, and perform nonlinear mapping processing on the hidden vector of the organic matter sample by using the activation layer in the first prediction network, and use the obtained metabolic abnormality probability of the organic matter sample as the metabolic abnormality information of the organic matter sample.

[0195] In some embodiments, the feature extraction processing is implemented by using a feature extraction network, and the feature extraction network includes a plurality of cascaded encoding layers; the extraction module 5552 is further configured to perform feature encoding processing on the fingerprint spectrum of the organic matter sample by using a first encoding layer in the plurality of cascaded encoding layers; output the encoding result of the first encoding layer to a subsequent cascaded encoding layer, so as to continue the feature encoding and the encoding result output in the subsequent cascaded encoding layer until the last encoding layer, and use the encoding result output by the last encoding layer as the metabolic characteristics of the organic matter sample.

[0196] In some embodiments, the encoding layers include a full connection layer, a hidden layer, and an activation layer; the extraction module 5552 is further configured to perform the following processing through the i-th encoding layer of the plurality of cascaded encoding layers: mapping the encoding result of the i-1-th encoding layer through the full connection layer in the i-th encoder to obtain an i-th mapping vector; performing regularization processing on the i-th mapping vector through the hidden layer in the i-th encoder to obtain an i-th regularization vector; performing nonlinear mapping processing on the i-th regularization vector through the activation layer in the i-th encoder to obtain the encoding result of the i-th encoding layer, and outputting the encoding result of the i-th encoding layer to the i+1-th encoding layer; wherein i is a natural number starting from 2 and taking values satisfying 2≤i≤N-1, and N is the number of the plurality of cascaded encoding layers.

[0197] In some embodiments, the extraction module 5552 is further configured to perform the following processing through the hidden layer in the i-th encoding layer: performing nonlinear mapping processing on the i-th mapping vector to obtain an i-th nonlinear mapping vector; and performing weighting processing on the i-th nonlinear mapping vector to obtain the i-th regularization vector.

[0198] In some embodiments, the extraction module 5552 is further configured to perform the following processing through the full connection layer in the i-th encoder: performing connection processing on the encoding result of the i-1-th encoding layer and the encoding result of the i-2-th encoding layer to obtain an i-th connection vector; and performing linear mapping processing on the i-th connection vector to obtain the i-th mapping vector.

[0199] In some embodiments, the fusion processing is implemented through a fusion network, and the fusion network includes a linear mapping layer and a full connection layer; the fusion module 5554 is further configured to perform weighting on the metabolic abnormality information of the organic sample and the protein marker information of the organic sample through the linear mapping layer in the fusion network to obtain weighted metabolic abnormality information and weighted protein marker information; and perform splicing processing on the weighted metabolic abnormality information and the weighted protein marker information through the full connection layer in the fusion network to obtain the fusion information of the organic sample.

[0200] In some embodiments, the prediction processing is implemented through a second prediction network, and the second prediction network includes a first nonlinear network and a second nonlinear network; the second prediction module 5555 is further configured to perform down-sampling encoding processing on the fusion information of the organic sample through the first nonlinear network in the second prediction network to obtain hidden space information of the organic sample; and perform up-sampling decoding processing on the hidden space information through the second nonlinear network in the second prediction network to obtain classification information of the organic sample.

[0201] In some embodiments, the first nonlinear network includes a fully connected layer, a hidden layer, and an activation layer; the second prediction module 5555 is further configured to perform the following processing by the first nonlinear network in the second prediction network: mapping the fusion information of the organic sample by the fully connected layer in the first nonlinear network to obtain a mapping vector of the fusion information; performing regularization processing on the mapping vector of the fusion information by the hidden layer in the first nonlinear network to obtain a regularization vector of the fusion information; and performing nonlinear mapping processing on the regularization vector of the fusion information by the activation layer in the first nonlinear network to obtain hidden space information of the organic sample.

[0202] In some embodiments, the second nonlinear network includes a fully connected layer and an activation layer; the second prediction module 5555 is further configured to perform the following processing by the fully connected layer in the second nonlinear network: mapping the hidden space information to obtain a mapping vector of the organic sample; and performing nonlinear mapping processing on the mapping vector of the organic sample by the activation layer in the second nonlinear network, and taking the obtained comprehensive anomaly probability of the organic sample as classification information of the organic sample.

[0203] The embodiment of the present application provides a computer program product or a computer program, which comprises computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the artificial intelligence-based organic sample processing method provided in the embodiment of the present application.

[0204] The embodiment of the present application provides a computer readable storage medium storing executable instructions, wherein the executable instructions are stored in the computer readable storage medium. When the executable instructions are executed by a processor, the processor will execute the artificial intelligence-based question and answer processing method provided in the embodiment of the present application, for example, as shown in the artificial intelligence-based organic sample processing method. Figures 3-5 The embodiment of the present application provides a computer readable storage medium storing executable instructions, wherein the executable instructions are stored in the computer readable storage medium. When the executable instructions are executed by a processor, the processor will execute the artificial intelligence-based question and answer processing method provided in the embodiment of the present application, for example, as shown in the artificial intelligence-based organic sample processing method.

[0205] In some embodiments, the computer readable storage medium can be FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disc, or CD-ROM memory, etc. It can also be various devices including one or any combination of the above storage devices.

[0206] In some embodiments, executable instructions can be in the form of programs, software, software modules, scripts, or code that is written in any form of programming language, including compiled or interpreted languages, or declarative or procedural languages; and that is deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0207] By way of example, executable instructions can, but need not, correspond directly with files in a file system, can be stored in a portion of a file that holds other programs or data, e.g., one or more scripts stored in a markup language document, in a single file dedicated to the program in question, or in multiple coordinated files, e.g., files that store one or more modules, sub programs, or portions of code.

[0208] By way of example, executable instructions can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communication network.

[0209] The above description is only a summary of the application. It is not intended to limit the scope of the application. Any modification, equivalent replacement and improvement made within the spirit and principle of the application shall fall within the scope of the application.

Claims

1. A method for processing organic matter samples based on artificial intelligence, characterized in that, The method includes: Obtain fingerprint patterns and protein biomarker information from organic samples; The fingerprint spectrum of the organic sample is subjected to feature extraction processing to obtain the metabolic characteristics of the organic sample; Based on the metabolic characteristics of the organic sample, predictive processing is performed to obtain metabolic abnormality information of the organic sample. The metabolic abnormality information and protein biomarker information of the organic sample are weighted by a linear mapping layer in the fusion network to obtain weighted metabolic abnormality information and weighted protein biomarker information; the weighted metabolic abnormality information and weighted protein biomarker information are then spliced ​​together by a fully connected layer in the fusion network to obtain the fusion information of the organic sample. Based on the fusion information of the organic samples, prediction processing is performed to obtain the classification information of the organic samples.

2. The method according to claim 1, characterized in that, The acquisition of fingerprint patterns and protein biomarker information from organic samples includes: The organic sample is subjected to spectral extraction processing to obtain the original spectral data of the organic sample; the original spectral data of the organic sample is preprocessed to obtain the fingerprint spectral data of the organic sample; the organic sample is subjected to protein biomarker detection processing to obtain the protein biomarker information of the organic sample.

3. The method according to claim 2, characterized in that, The process of extracting spectral data from organic samples to obtain the original spectral data of the organic samples includes: The following processes are performed using the automated sample delivery and testing system: The organic sample is moved onto a mass spectrometry target plate, and a matrix is ​​applied to the organic sample. The mass spectrometer is used to detect organic samples including the matrix to obtain the raw spectrum of the organic samples.

4. The method according to claim 2, characterized in that, The preprocessing of the original spectrum of the organic sample to obtain the fingerprint spectrum of the organic sample includes: The original spectra of the organic sample were resampled. The sampled spectrum is then denoised to obtain a smoothed spectrum. The smoothed spectrum is subjected to baseline correction to obtain the corrected spectrum; Based on the corrected spectrum, peak extraction processing is performed to obtain the fingerprint spectrum of the organic sample.

5. The method according to claim 1, characterized in that, The prediction process is implemented through a first prediction network, which includes a fully connected layer and an activation layer. The predictive processing based on the metabolic characteristics of the organic sample to obtain metabolic abnormality information of the organic sample includes: The metabolic features of the organic samples are mapped to the latent vector space through the fully connected layer in the first prediction network to obtain the latent vector of the organic samples. The latent vector of the organic sample is nonlinearly mapped through the activation layer in the first prediction network, and the resulting metabolic abnormality probability of the organic sample is used as the metabolic abnormality information of the organic sample.

6. The method according to claim 1, characterized in that, The feature extraction process is implemented through a feature extraction network, which includes multiple cascaded coding layers. The step of extracting features from the fingerprint spectrum of the organic sample to obtain its metabolic characteristics includes: The fingerprint spectrum of the organic sample is processed by feature encoding through the first encoding layer of the multiple cascaded encoding layers; The encoding result of the first encoding layer is output to subsequent concatenated encoding layers, so that feature encoding and encoding result output continue in the subsequent concatenated encoding layers, until the last encoding layer is reached. The encoding result output from the last encoding layer is used as the metabolic feature of the organic sample.

7. The method according to claim 6, characterized in that, The coding layer includes a fully connected layer, a hidden layer, and an activation layer; The step of continuing feature encoding and outputting the encoding result in the subsequent concatenated encoding layers includes: The following processing is performed through the i-th coding layer of the multiple cascaded coding layers: The encoding result of the (i-1)th encoding layer is mapped through the fully connected layer in the i-th encoder to obtain the i-th mapping vector; The i-th mapping vector is regularized by the hidden layer in the i-th encoder to obtain the i-th regularized vector; The activation layer in the i-th encoder performs nonlinear mapping on the i-th regularization vector to obtain the encoding result of the i-th encoding layer, and outputs the encoding result of the i-th encoding layer to the (i+1)-th encoding layer. Where i is a natural number that increments from 2 and satisfies 2≤i≤N-1, and N is the number of the multiple cascaded coding layers.

8. The method according to claim 7, characterized in that, The step of performing regularization processing on the i-th mapping vector through the hidden layer in the i-th encoder to obtain the i-th regularized vector includes: The following processing is performed through the hidden layer in the i-th coding layer: The i-th mapping vector is subjected to nonlinear mapping processing to obtain the i-th nonlinear mapping vector; The i-th nonlinear mapping vector is weighted, and the weighted result is used as the i-th regularization vector.

9. The method according to claim 7, characterized in that, The process of mapping the encoding result of the (i-1)th encoding layer through the fully connected layer in the i-th encoder to obtain the i-th mapping vector includes: The following processing is performed through the fully connected layer in the i-th encoder: The encoding results of the (i-1)th encoding layer and the encoding results of the (i-2)th encoding layer are concatenated to obtain the i-th concatenation vector; The i-th connection vector is linearly mapped to obtain the i-th mapping vector.

10. The method according to claim 1, characterized in that, The prediction process is implemented through a second prediction network, which includes a first nonlinear network and a second nonlinear network. The prediction processing based on the fusion information of the organic samples to obtain the classification information of the organic samples includes: The fusion information of the organic sample is downsampled and encoded by the first nonlinear network in the second prediction network to obtain the latent space information of the organic sample. The latent space information is upsampled and decoded using the second nonlinear network in the second prediction network to obtain the classification information of the organic sample.

11. The method according to claim 10, characterized in that, The first nonlinear network includes a fully connected layer, a hidden layer, and an activation layer; The step of downsampling and encoding the fusion information of the organic sample through the first nonlinear network in the second prediction network to obtain the latent space information of the organic sample includes: The following processing is performed through the first nonlinear network in the second prediction network: The fusion information of the organic samples is mapped through the fully connected layer in the first nonlinear network to obtain the mapping vector of the fusion information; The mapping vector of the fused information is regularized by the hidden layer in the first nonlinear network to obtain the regularized vector of the fused information. The latent space information of the organic sample is obtained by performing nonlinear mapping on the regularized vector of the fused information through the activation layer in the first nonlinear network.

12. The method according to claim 10, characterized in that, The second nonlinear network includes a fully connected layer and an activation layer; The step of upsampling and decoding the latent space information through the second nonlinear network in the second prediction network to obtain the classification information of the organic sample includes: The latent space information is mapped through the fully connected layer in the second nonlinear network to obtain the mapping vector of the organic sample; The mapping vector of the organic sample is nonlinearly mapped by the activation layer in the second nonlinear network, and the resulting comprehensive anomaly probability of the organic sample is used as the classification information of the organic sample.

13. An organic sample processing device, characterized in that, The device includes: The acquisition module is used to acquire fingerprint patterns and protein biomarker information of organic samples; The extraction module is used to perform feature extraction processing on the fingerprint spectrum of the organic sample to obtain the metabolic characteristics of the organic sample; The first prediction module is used to perform prediction processing based on the metabolic characteristics of the organic sample to obtain metabolic abnormality information of the organic sample. The fusion module is used to weight the metabolic abnormality information and protein biomarker information of the organic sample through a linear mapping layer in the fusion network to obtain weighted metabolic abnormality information and weighted protein biomarker information; and to splice the weighted metabolic abnormality information and weighted protein biomarker information through a fully connected layer in the fusion network to obtain the fusion information of the organic sample. The second prediction module is used to perform prediction processing based on the fusion information of the organic samples to obtain the classification information of the organic samples.

14. The apparatus according to claim 13, characterized in that, The device further includes: The acquisition module is also used to perform the following processing through an automatic sample delivery and detection system: moving the organic sample onto a mass spectrometry target plate and covering the organic sample with a matrix; controlling the mass spectrometer to detect the organic sample including the matrix, and obtaining the original spectrum of the organic sample.

15. The apparatus according to claim 13, characterized in that, The device further includes: The acquisition module is further configured to resample the original spectrum of the organic sample, denoise the sampled spectrum to obtain a smoothed spectrum, perform baseline correction on the smoothed spectrum to obtain a corrected spectrum, and perform peak extraction based on the corrected spectrum to obtain the fingerprint spectrum of the organic sample.

16. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the artificial intelligence-based organic sample processing method according to any one of claims 1 to 12.

17. A computer-readable storage medium, characterized in that, It stores executable instructions for use by a processor to implement the artificial intelligence-based organic sample processing method according to any one of claims 1 to 12.

18. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the artificial intelligence-based organic sample processing method according to any one of claims 1 to 12.

Images