Deep learning-based vocs mass spectrum ion fragment peak region identification method and device

Abstract

The application provides a Voc mass spectrum ion fragment peak region identification method and device based on deep learning, which comprises the following steps: obtaining a target Voc mass spectrum, inputting the target Voc mass spectrum into a trained target DeepGCMSPeak model, wherein the target DeepGCMSPeak model comprises a first CNN network and a second CNN network; identifying a target ion fragment peak region from the target Voc mass spectrum through the first CNN network; inputting the target ion fragment peak region into the second CNN network, identifying a target region area of the target ion fragment peak region through the second CNN network; and determining the target ion fragment peak region and the target region area as a target identification result of the target Voc mass spectrum. According to the technical scheme of the embodiment of the application, the target ion fragment peak region and the target region area can be respectively identified through the two deep learning networks of the target DeepGCMSPeak model, the identification efficiency of the ion fragment peak region is effectively improved, and a data basis for improving the analysis efficiency of the Voc mass spectrum is provided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of neural networks, and in particular to a Vocs mass spectrum ion fragment peak region identification method and device based on deep learning. BACKGROUND

[0002] A mass spectrum is the main basis for analysis of volatile organic compounds (Vocs), and a Vocs mass spectrum usually records ion fragment peaks of Vocs. How to identify an ion fragment peak region from a Vocs mass spectrum has an important influence on the analysis efficiency of the Vocs mass spectrum.

[0003] In related technologies, researchers mainly analyze data distribution of a Vocs mass spectrum based on their professional knowledge to find an ion fragment peak region, and calculate an area of the ion fragment peak region by integration, which consumes a large amount of manpower and time, and the analysis efficiency of the Vocs mass spectrum is low. SUMMARY

[0004] The present application aims to at least solve one of the technical problems in the prior art. To this end, the present application provides a Vocs mass spectrum ion fragment peak region identification method and device based on deep learning, which can improve the identification efficiency of the ion fragment peak region and improve the analysis efficiency of the Vocs mass spectrum.

[0005] In a first aspect, an embodiment of the present application provides a Vocs mass spectrum ion fragment peak region identification method based on deep learning, comprising:

[0006] obtaining a target Vocs mass spectrum, inputting the target Vocs mass spectrum into a trained target DeepGCMSPeak model, the target DeepGCMSPeak model comprising a first CNN network and a second CNN network;

[0007] identifying a target ion fragment peak region from the target Vocs mass spectrum by the first CNN network;

[0008] inputting the target ion fragment peak region into the second CNN network, and identifying a target region area of the target ion fragment peak region by the second CNN network;

[0009] determining the target ion fragment peak region and the target region area as a target identification result of the target Vocs mass spectrum.

[0010] According to some embodiments of the present application, the training step of the target DeepGCMSPeak model comprises:

[0011] constructing an initial DeepGCMSPeak model, configuring initial first network hyperparameters of the first CNN network and initial second network hyperparameters of the second CNN network;

[0012] obtaining a mass spectrum data training set, the mass spectrum data training set comprising a plurality of sample Voc mass spectra, the sample Voc mass spectra being pre-labeled with first region information, second region information and sample peak region area, the first region information being used to indicate a sample ion fragment peak region, the second region information being used to indicate a region excluding the sample ion fragment peak region, and the sample peak region area being an area of the sample ion fragment peak region;

[0013] training the first CNN network based on the first region information and the second region information to adjust the initial first network hyperparameters to target first network hyperparameters;

[0014] training the second CNN network based on the first region information and the sample peak region area to adjust the initial second network hyperparameters to target second network hyperparameters;

[0015] configuring the target first network hyperparameters and the target second network hyperparameters to the initial DeepGCMSPeak model to obtain an intermediate DeepGCMSPeak model;

[0016] performing data verification on the intermediate DeepGCMSPeak model, and determining the intermediate DeepGCMSPeak model as the target DeepGCMSPeak model after verification.

[0017] According to some embodiments of the present application, the loss function of the first CNN network is a cross-entropy loss function, and the loss function of the second CNN network is a mean square error loss function.

[0018] According to some embodiments of the present application, the data verification on the intermediate DeepGCMSPeak model comprises:

[0019] obtaining a preset mass spectrum test set, the mass spectrum test set comprising a plurality of test Voc mass spectra and reference data, the reference data comprising reference ion fragment peak regions and reference peak region areas of the test Voc mass spectra;

[0020] inputting the mass spectrum test set into the intermediate DeepGCMSPeak model to obtain test ion fragment peak regions and test peak region areas output by the intermediate DeepGCMSPeak model;

[0021] When the test ion fragment peak region matches the reference ion fragment peak region, and the test peak region area matches the reference peak region area, it is determined that the intermediate DeepGCMSPeak model passes the verification.

[0022] According to some embodiments of the present application, when the data verification of the intermediate DeepGCMSPeak model fails, the method further comprises continuing to train the intermediate DeepGCMSPeak model, and before the continuing to train the intermediate DeepGCMSPeak model, the method further comprises:

[0023] adjusting the target first network hyperparameters and the target second network hyperparameters according to preset adjustment values to obtain optimized first network hyperparameters and optimized target second network hyperparameters and configure them to the intermediate DeepGCMSPeak model;

[0024] Alternatively, the network structure of the first CNN network or the second CNN network is adjusted to obtain an optimized first CNN network or an optimized second CNN network.

[0025] Alternatively, the number of samples of the mass spectrum data training set is increased.

[0026] According to some embodiments of the present application, after the mass spectrum data training set is obtained, the method further comprises:

[0027] The sample Vocsmass spectrum is preprocessed, and the preprocessing includes denoising processing and background baseline removal processing.

[0028] The initial DeepGCMSPeak model is trained based on the sample Vocsmass spectrum obtained by preprocessing.

[0029] According to some embodiments of the present application, the optimization algorithm of the initial DeepGCMSPeak model is an ADAM optimization algorithm.

[0030] In a second aspect, embodiments of the present application provide a deep learning-based Vocsmass spectrum ion fragment peak region identification device, which includes at least one control processor and a memory in communication connection with the at least one control processor; the memory stores instructions executable by the at least one control processor, and the instructions are executed by the at least one control processor to enable the at least one control processor to execute the deep learning-based Vocsmass spectrum ion fragment peak region identification method as described in the first aspect above.

[0031] In a third aspect, embodiments of the present application provide an electronic device comprising the deep learning-based Vocsmass spectrum ion fragment peak region identification device as described in the second aspect above.

[0032] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium storing computer executable instructions for performing the deep learning based Vocs mass spectrum ion fragment peak area identification method according to the first aspect.

[0033] The deep learning based Vocs mass spectrum ion fragment peak area identification method according to the embodiment of the present application has at least the following beneficial effects: a target Vocs mass spectrum is obtained, the target Vocs mass spectrum is input into a trained target DeepGCMSPeak model, the target DeepGCMSPeak model includes a first CNN network and a second CNN network; the target ion fragment peak area is identified from the target Vocs mass spectrum through the first CNN network; the target ion fragment peak area is input into the second CNN network, and the target area area of the target ion fragment peak area is identified through the second CNN network; and the target ion fragment peak area and the target area area are determined as the target identification result of the target Vocs mass spectrum. According to the technical solution of the embodiment of the present application, the target ion fragment peak area and the target area area can be respectively identified through the two deep learning networks of the target DeepGCMSPeak model, the identification efficiency of the ion fragment peak area is effectively improved, and a data basis for improving the analysis efficiency of the Vocs mass spectrum is provided. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 is a flowchart of the deep learning based Vocs mass spectrum ion fragment peak area identification method provided by an embodiment of the present application;

[0035] Figure 2 is a structural schematic diagram of a DeepGCMSPeak model provided by another embodiment of the present application;

[0036] Figure 3 is a training flowchart of a DeepGCMSPeak model provided by another embodiment of the present application;

[0037] Figure 4 is a test flowchart of a DeepGCMSPeak model provided by another embodiment of the present application;

[0038] Figure 5 is an optimization flowchart of a DeepGCMSPeak model provided by another embodiment of the present application;

[0039] Figure 6 is a flowchart of sample preprocessing provided by another embodiment of the present application;

[0040] Figure 7is a structural diagram of a deep learning-based Voc mass spectrum ion fragment peak region identification device provided by another embodiment of the present application. DETAILED DESCRIPTION

[0041] Embodiments of the present application are described in detail below with reference to examples thereof illustrated in the attached drawings, in which like or similar designations denote like or similar elements or components throughout the drawings. The embodiments described below with reference to the drawings are merely exemplary for the purpose of explaining the present application and should not be construed as limiting the present application.

[0042] In the description of the present application, it should be understood that the orientation description, such as the orientation or position relationship indicated by up, down, front, back, left, right, etc., is based on the orientation or position relationship shown in the drawings, and is only for the purpose of facilitating the description of the present application and simplifying the description, and therefore cannot be understood as limiting the present application, which indicates or implies that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0043] In the description of the present application, the meaning of several is one or more, the meaning of multiple is two or more, greater than, less than, more than, etc. are understood as not including the number, above, below, etc. are understood as including the number. If it is described as first, second, it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or the order of indicated technical features.

[0044] In the description of the present application, unless otherwise explicitly limited, the words such as setting, installing, connecting, etc. should be broadly understood, and those skilled in the art can reasonably determine the specific meaning of the above words in the present application in combination with the specific content of the technical solution.

[0045] Before further detailing the embodiments of the present application, the terms and terms involved in the embodiments of the present application are explained, and the terms and terms involved in the embodiments of the present application are applicable to the following explanations:

[0046] Deep learning: deep learning is a kind of machine learning, and machine learning is a necessary path to realize artificial intelligence. The concept of deep learning is derived from the research of artificial neural network, and the multilayer perceptron containing multiple hidden layers is a kind of deep learning structure. Deep learning forms more abstract high-level representation attribute classes or features by combining low-level features to discover distributed feature representation of data. The motivation of studying deep learning is to establish a neural network that simulates the brain to analyze and learn, which simulates the mechanism of the human brain to explain data such as images, sounds and texts, etc.

[0047] Convolutional Neural Networks (CNN): CNN is a kind of feedforward neural networks containing convolution calculation 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 information according to its hierarchical structure, so it is also called "shift-invariant artificial neural networks (SIANN)".

[0048] Vocs: volatile organic compounds, sometimes also represented by TVOC. According to the definition of the World Health Organization, if the boiling point of the compound is 50-250℃ at a pressure of 101.32kPa, it is a volatile organic compound. They exist in gaseous form at room temperature. According to their chemical structure, they can be further divided into eight categories: alkanes, aromatic hydrocarbons, alkenes, halogenated hydrocarbons, esters, aldehydes, ketones and others.

[0049] Mass spectrum: mass spectrum is an analytical technique that ionizes chemical substances and sorts them according to their mass-to-charge ratio. In simple terms, mass spectrometry measures the mass within a sample. Mass spectrometry is used in many different fields and is used for both pure samples and complex mixtures. Mass spectrum is a plot of ion signal as a function of mass-to-charge ratio. These spectra are used to determine the elements or isotopes, particles and molecules of a sample, and to elucidate the chemical structure of molecules such as peptides and other compounds.

[0050] Mass spectrum: after ions of different mass-to-charge ratios are separated by mass analyzer, they are detected by detector and recorded, and after computer processing, they are represented in the form of mass spectrum. In the mass spectrum, the horizontal coordinate represents the mass-to-charge ratio (m / z) value of the ion, and the mass-to-charge ratio value increases from left to right. For ions with single charge, the horizontal coordinate value represents the mass of the ion; the vertical coordinate represents the intensity of the ion current, which is usually represented by relative intensity, that is, the strongest ion current intensity is defined as 100%, and the intensity of other ion currents is represented by its percentage. Sometimes, the total ion current intensity of all recorded ions is taken as 100%, and various ions are represented by their percentage.

[0051] Ion fragment peak: when the energy of electron bombardment exceeds the energy required for molecular ionization (about 50-70eV), the chemical bond of the molecular ion can be further broken, producing a fragment with lower mass number, called fragment ion. The corresponding peak appears on the mass spectrum, called fragment ion peak. The fragment ion peak is located on the left side of the molecular ion peak on the mass spectrum.

[0052] DeepGCMSPeak model: a deep network model for monitoring and identifying GC-MS Vocsmolecular mass spectrum ion fragment peaks.

[0053] The embodiment of the application provides a kind of based on deep learning's Vocs mass spectrum ion fragment peak area identification method and device, wherein based on deep learning's Vocs mass spectrum ion fragment peak area identification method includes: obtaining target Vocs mass spectrum, the target Vocs mass spectrum is input to the target DeepGCMSPeak model trained, the target DeepGCMSPeak model includes first CNN network and second CNN network;Target ion fragment peak area is identified from the target Vocs mass spectrum by the first CNN network;The target ion fragment peak area is input to the second CNN network, and the target area area of the target ion fragment peak area is identified by the second CNN network;The target ion fragment peak area and the target area area are determined as the target identification result of the target Vocs mass spectrum.According to the technical scheme of the embodiment of the application, target ion fragment peak area and target area area can be identified by two deep learning networks of target DeepGCMSPeak model respectively, effectively improve the identification efficiency of ion fragment peak area, provide data basis for improving the analysis efficiency of Vocs mass spectrum.

[0054] The control method of the embodiment of the application will be further described below with reference to the drawings.

[0055] Reference Figure 1 , Figure 1 A flow chart of a Vocs mass spectrum ion fragment peak area identification method based on deep learning provided by the embodiment of the application is provided, and the method includes but is not limited to the following steps:

[0056] In step S110, the target Vocs mass spectrum is obtained, and the target Vocs mass spectrum is input to the target DeepGCMSPeak model trained, and the target DeepGCMSPeak model includes first CNN network and second CNN network.

[0057] In step S120, the target ion fragment peak area is identified from the target Vocs mass spectrum by the first CNN network.

[0058] Step S130, input the target ion fragment peak region to the second CNN network, and identify the target region area of the target ion fragment peak region through the second CNN network;

[0059] Step S140, determine the target ion fragment peak region and the target region area as the target recognition result of the target Voc spectrum.

[0060] It should be noted that the network structure of the target DeepGCMSPeak model is as shown in Figure 2 The first CNN network and the second CNN network are not limited in structure in this embodiment, and can realize the corresponding recognition function after training,

[0061] For example, referring to Figure 2 The first CNN network and the second CNN network each include three convolutional layers. Since the fragment ion peak is located on the left side of the molecular ion peak on the spectrum, the data of the fragment ion peak and the molecular ion peak can have different characteristics after convolution. The first CNN network extracts the characteristics of each region (for example, a columnar graph formed by all ion masses corresponding to each mass-to-charge ratio) of the target Voc spectrum from the target Voc spectrum through convolution calculation of the convolutional layers, and classifies each characteristic through a configured classification network, so as to determine the type of the region where each characteristic is located. In order to facilitate identification, the type of the region can be simply set as an ion fragment peak region and a non-ion fragment peak region. Of course, more types can be set according to actual needs, which is not limited here. The recognition result obtained by the first CNN network is the target ion fragment peak region. The target ion fragment peak region is simultaneously an output of the target DeepGCMSPeak model and an input of the second CNN network. The second CNN network identifies the region for which the area needs to be calculated from the target Voc spectrum based on the target ion fragment peak region and calculates the target region area. Finally, the target DeepGCMSPeak model simultaneously outputs the target ion fragment peak region and the target region area, and completes the identification of the ion fragment peak.

[0062] It should be noted that two CNN networks are provided in the target DeepGCMSPeak model in this embodiment, which can simplify the training of the CNN network. Only the first CNN network needs to be trained to identify the ion fragment peak region, and the second CNN network needs to be trained to calculate the area according to the input region, thereby effectively reducing the complexity of the target DeepGCMSPeak model and improving the data processing efficiency.

[0063] In addition, in an embodiment, referring to Figure 3 The training steps of the target DeepGCMSPeak model include but are not limited to the following steps:

[0064] Step S310, constructing an initial DeepGCMSPeak model, configuring initial first network hyperparameters of the first CNN network and initial second network hyperparameters of the second CNN network;

[0065] Step S320, obtaining a mass spectrum data training set, the mass spectrum data training set including a plurality of sample Voc mass spectrum, the sample Voc mass spectrum being pre-labeled with first region information, second region information and sample peak region area, the first region information being used to indicate a sample ion fragment peak region, the second region information being used to indicate a region not including the sample ion fragment peak region, and the sample peak region area being an area of the sample ion fragment peak region;

[0066] Step S330, training the first CNN network based on the first region information and the second region information to adjust the initial first network hyperparameters to target first network hyperparameters;

[0067] Step S340, training the second CNN network based on the first region information and the sample peak region area to adjust the initial second network hyperparameters to target second network hyperparameters;

[0068] Step S350, configuring the target first network hyperparameters and the target second network hyperparameters to the initial DeepGCMSPeak model to obtain an intermediate DeepGCMSPeak model;

[0069] Step S360, performing data verification on the intermediate DeepGCMSPeak model, and determining the intermediate DeepGCMSPeak model as a target DeepGCMSPeak model after verification passes.

[0070] It should be noted that the construction of the initial DeepGCMSPeak model can be achieved by obtaining pre-set first CNN network and second CNN network, and configuring the input and output thereof, for example, setting the input of the initial DeepGCMSPeak model as the input of the first CNN network, setting the output of the first CNN network as the input of the second CNN network, and configuring the first network hyperparameters of the first CNN network and the second network hyperparameters of the second CNN network, the first network hyperparameters and the second network hyperparameters can be learning rate, iteration number and other parameters of network training, which are not limited herein.

[0071] It should be noted that in the training process of the CNN network, the network hyperparameters need to be adjusted to realize the convergence of the data and the performance standard, and obtain the trained network, so when the CNN network is adjusted by adjusting the network hyperparameters to converge and the performance indicators meet the preset conditions, it can be determined that the initial DeepGCMSPeak model is initially trained, and the intermediate DeepGCMSPeak model is obtained. Under the premise of having mass spectrum data training set and configuring initial DeepGCMSPeak model, the training method of CNN network is a technology known to those skilled in the art, and will not be described here.

[0072] It should be noted that the sample labeling method is a technology known to those skilled in the art, and will not be described here. The labeling of the present embodiment includes the sample ion fragment peak area and the sample peak area, so that the first CNN network can learn the characteristics of the ion fragment peak area, and the second CNN network can learn the calculation method of the area, effectively improving the accuracy of the target DeepGCMSPeak model.

[0073] It should be noted that after obtaining the intermediate DeepGCMSPeak model, the indicators and performance of the intermediate DeepGCMSPeak model need to be verified by test data, and after verification, it is determined as the target DeepGCMSPeak model, to ensure the accuracy of the target DeepGCMSPeak model.

[0074] In addition, in an embodiment, the loss function of the first CNN network is a cross-entropy loss function, and the loss function of the second CNN network is a mean square error loss function.

[0075] It should be noted that since the first CNN network is used for classification, the cross-entropy loss function can effectively improve the accuracy of classification.

[0076] It should be noted that since the second CNN network is used for calculating the area, the mean square error loss function can effectively improve the calculation efficiency and accuracy.

[0077] In addition, in an embodiment, the reference data of the test Vocs mass spectrum includes the reference ion fragment peak area and the reference peak area of the test Vocs mass spectrum. Figure 4 The data verification of the intermediate DeepGCMSPeak model includes:

[0078] Step S410, obtaining a preset mass spectrum test set, the mass spectrum test set including a plurality of test Vocs mass spectrum and reference data, the reference data including a reference ion fragment peak area and a reference peak area of the test Vocs mass spectrum;

[0079] Step S420, input the mass spectrum test set to the intermediate DeepGCMSPeak model to obtain the test ion fragment peak region and the test peak region area output by the intermediate DeepGCMSPeak model;

[0080] Step S430, when the test ion fragment peak region matches the reference ion fragment peak region and the test peak region area matches the reference peak region area, it is determined that the intermediate DeepGCMSPeak model passes the verification.

[0081] It should be noted that the intermediate DeepGCMSPeak model has completed preliminary training and thus has certain recognition and calculation capabilities. The intermediate DeepGCMSPeak model is verified by the mass spectrum test set in this embodiment, which can ensure that the performance and accuracy of the intermediate DeepGCMSPeak model meet the standards, and the intermediate DeepGCMSPeak model that meets the performance and accuracy standards is determined as the target DeepGCMSPeak model.

[0082] It should be noted that after the mass spectrum test set is input to the intermediate DeepGCMSPeak model, the test ion fragment peak region is recognized by the first CNN network, the test peak region area is calculated by the second CNN network, and the reference ion fragment peak region and the reference peak region area in the reference data are compared. The matching condition of this embodiment can be equal, or the error can be within a preset error value. For example, the difference between the reference peak region area and the test peak region area is within a preset area difference, and it can be determined that the two are matched.

[0083] In addition, in an embodiment, the reference Figure 5 When the data verification of the intermediate DeepGCMSPeak model does not pass, the method further includes continuing to train the intermediate DeepGCMSPeak model. Before continuing to train the intermediate DeepGCMSPeak model, the method further includes but is not limited to the following steps:

[0084] Step S510, adjusting the target first network hyperparameters and the target second network hyperparameters according to a preset adjustment value to obtain optimized first network hyperparameters and optimized target second network hyperparameters and configuring them to the intermediate DeepGCMSPeak model;

[0085] Step S520, adjusting the network structure of the first CNN network or the second CNN network to obtain an optimized first CNN network or an optimized second CNN network;

[0086] Step S530, increasing the number of samples of the mass spectrum data training set.

[0087] It should be noted that when the verification of the intermediate DeepGCMSPeak model fails, it may be that the threshold of the loss function is not reached or the maximum number of iterations is reached, or the accuracy and efficiency of the model do not reach the preset indicators. Based on this, the embodiment adjusts the network hyperparameters before retraining the intermediate DeepGCMSPeak model, for example, increasing the learning rate and the number of iterations; or adjusting the structure of the CNN network, for example, increasing or decreasing the convolutional layer; or increasing the number of samples in the mass spectrum data training set, so that the CNN network can learn features from more samples. Those skilled in the art have the motivation to select optimization strategies and specific optimization values according to actual needs, and the embodiment does not make further limitations here.

[0088] In addition, in an embodiment, referring to Figure 6 After obtaining the mass spectrum data training set, the following steps are included but not limited to:

[0089] In step S610, the sample Vocs mass spectrum is preprocessed, and the preprocessing includes denoising processing and background baseline removal processing.

[0090] In step S620, the initial DeepGCMSPeak model is trained based on the preprocessed sample Vocs mass spectrum.

[0091] It should be noted that since the noise of the initial image is large, the embodiment pre-processes the image through common denoising processing methods and background baseline removal processing methods after obtaining the sample Vocs mass spectrum, effectively improving the training efficiency and accuracy of the model.

[0092] In addition, in an embodiment, the optimization algorithm of the initial DeepGCMSPeak model is the ADAM optimization algorithm.

[0093] It should be noted that the ADAM optimization algorithm is beneficial to speed up the optimization process, for example, reducing the number of function evaluations required to reach the optimal state, or improving the function of the optimization algorithm. Therefore, the embodiment configures the ADAM optimization algorithm in the initial DeepGCMSPeak model, which can improve the training efficiency and accuracy of the model.

[0094] As Figure 7 shown, Figure 7 is a structural diagram of a Vocs mass spectrum ion fragment peak region recognition device based on deep learning provided by an embodiment of the application. The application also provides a Vocs mass spectrum ion fragment peak region recognition device based on deep learning, which comprises:

[0095] The processor 701 can be implemented by a general-purpose central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), or one or more integrated circuits, and is configured to execute related programs to implement the technical solutions provided by the embodiments of the present application.

[0096] The memory 702 can be implemented by a read only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), and the like. The memory 702 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are stored in the memory 702 and are called and executed by the processor 701 to implement the deep learning based Voc mass spectrum ion fragment peak region identification method of the embodiments of the present application.

[0097] The input / output interface 703 is configured to realize information input and output.

[0098] The communication interface 704 is configured to realize the communication interaction between the device and other devices. The communication can be realized by a wired manner (for example, a USB, a network cable, and the like) or a wireless manner (for example, a mobile network, WIFI, Bluetooth, and the like).

[0099] The bus 705 is configured to transmit information between the components (for example, the processor 701, the memory 702, the input / output interface 703, and the communication interface 704) of the device.

[0100] The processor 701, the memory 702, the input / output interface 703, and the communication interface 704 are connected to each other through the bus 705 to realize the communication connection between the components in the device.

[0101] The embodiments of the present application also provide an electronic device including the deep learning based Voc mass spectrum ion fragment peak region identification apparatus as described above.

[0102] The embodiments of the present application also provide a storage medium, which is a computer readable storage medium. The storage medium stores a computer program. The computer program is executed by a processor to implement the deep learning based Voc mass spectrum ion fragment peak region identification method.

[0103] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory disposed remotely with respect to the processor, which can be connected to the processor through a network. Examples of the above network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The above-described device embodiments are only illustrative, and units described as separate components can or can not be physically separated, implemented in one place, or distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment.

[0104] Those of ordinary skill in the art can understand that all or some steps in the above disclosed method and system can be implemented as software, firmware, hardware, and appropriate combinations thereof. Some or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transitory media). As known to those of ordinary skill in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage, or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as known to those of ordinary skill in the art, communication media typically includes computer readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transport mechanisms, and can include any information delivery medium.

[0105] The above is a specific description of the preferred embodiment of the present application, but the present application is not limited to the above-described embodiments, and those skilled in the art can make various equivalent modifications or replacements without departing from the spirit of the present application. These equivalent modifications or replacements are all included in the scope defined by the claims of the present application.

Claims

1. A method for identifying ion fragmentation peak regions of a Voc mass spectrum based on deep learning, characterized in that, The method comprises the following steps: acquiring a target Voc spectrum, and inputting the target Voc spectrum into a trained target DeepGCMSPeak model, wherein the target DeepGCMSPeak model comprises a first CNN network and a second CNN network; identifying a target ion fragment peak region from the target Voc spectrum through the first CNN network; inputting the target ion fragment peak region into the second CNN network, and identifying a target region area of the target ion fragment peak region through the second CNN network; determining the target ion fragment peak region and the target region area as a target identification result of the target Voc spectrum; the training steps of the target DeepGCMSPeak model comprise the following steps: constructing an initial DeepGCMSPeak model, and configuring initial first network hyperparameters of the first CNN network and initial second network hyperparameters of the second CNN network; acquiring a spectrum data training set, wherein the spectrum data training set comprises a plurality of sample Voc spectrums, the sample Voc spectrums are pre-labeled with first region information, second region information and a sample peak region area, the first region information is used to indicate a sample ion fragment peak region, the second region information is used to indicate a region excluding the sample ion fragment peak region, and the sample peak region area is an area of the sample ion fragment peak region; training the first CNN network based on the first region information and the second region information, so as to adjust the initial first network hyperparameters into target first network hyperparameters; training the second CNN network based on the first region information and the sample peak region area, so as to adjust the initial second network hyperparameters into target second network hyperparameters; configuring the target first network hyperparameters and the target second network hyperparameters to the initial DeepGCMSPeak model, so as to obtain an intermediate DeepGCMSPeak model; performing data verification on the intermediate DeepGCMSPeak model, and determining the intermediate DeepGCMSPeak model as the target DeepGCMSPeak model after verification; a loss function of the first CNN network is a cross-entropy loss function, and a loss function of the second CNN network is a mean square error loss function.

2. The deep learning-based Voc mass spectrum ion fragment peak region identification method according to claim 1, characterized in that, the data verification on the intermediate DeepGCMSPeak model comprises the following steps: acquiring a preset spectrum test set, wherein the spectrum test set comprises a plurality of test Voc spectrums and reference data, and the reference data comprises reference ion fragment peak regions and reference peak region areas of the test Voc spectrums; inputting the spectrum test set into the intermediate DeepGCMSPeak model, and acquiring test ion fragment peak regions and test peak region areas output by the intermediate DeepGCMSPeak model; When the test ion fragment peak region matches the reference ion fragment peak region, and the test peak region area matches the reference peak region area, it is determined that the intermediate DeepGCMSPeak model passes the verification. 3.The deep learning-based Voc mass spectrum ion fragment peak region identification method according to claim 2, characterized in that, When the data verification of the intermediate DeepGCMSPeak model fails, the method further comprises continuing to train the intermediate DeepGCMSPeak model, and before the continuing to train the intermediate DeepGCMSPeak model, the method further comprises: adjusting the target first network hyperparameters and the target second network hyperparameters according to a preset adjustment value to obtain optimized first network hyperparameters and optimized target second network hyperparameters and configure them to the intermediate DeepGCMSPeak model; or, adjusting the network structure of the first CNN network or the second CNN network to obtain optimized first CNN network or optimized second CNN network; or, increasing the number of samples of the mass spectrum data training set. 4.The deep learning-based Voc mass spectrum ion fragment peak region identification method according to claim 1, characterized in that, After the mass spectrum data training set is obtained, the method further comprises: preprocessing the sample Voc mass spectrum, the preprocessing comprising denoising processing and background baseline removal processing; training the initial DeepGCMSPeak model based on the sample Voc mass spectrum obtained by preprocessing. 5.The Voc mass spectrum ion fragment peak region identification method based on deep learning according to claim 1, characterized in that, The optimization algorithm of the initial DeepGCMSPeak model is ADAM optimization algorithm.

6. A deep learning-based Voc mass spectrum ion fragment peak region identification device, characterized in that, The computer readable storage medium stores computer executable instructions for causing a computer to execute the deep learning based Voc mass spectrum ion fragment peak region identification method according to any one of claims 1 to 5.

7. An electronic device, comprising: The deep learning based Voc mass spectrum ion fragment peak region identification device according to claim 6.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer executable instructions for causing a computer to execute the deep learning based Voc mass spectrum ion fragment peak region identification method according to any one of claims 1 to 5.

Images