A mass spectrometer lens parameter fitting method based on radial basis function network

Through the lens parameter fitting method based on the radial basis function network, the problem of long time to find the lens parameters of the mass spectrometer is solved, and the effect of quickly finding the global optimal solution is achieved, and the fitting accuracy and flexibility are improved.

CN113780428BInactive Publication Date: 2024-11-22TIANJIN GUOKE MEDICAL ENG & TECH DEV CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111074354.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-09-14
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing mass spectrometer lens parameter optimization method has the problem of long search time, low efficiency and difficulty in converging to the global optimal solution.

Method used

The lens parameter fitting method based on the radial basis function network is adopted, and the radial basis function network is combined with the prior information set to obtain the fitting surface of lens parameters and peak intensity, and the optimal weight vector is calculated through the weight vector trainer and iterative algorithm. The lens parameters at the highest peak intensity are obtained.

Benefits of technology

It achieves rapid obtaining the best combination of lens parameters, avoids edge effects in traditional methods, improves fitting accuracy and flexibility, and can quickly find the fitting surface of the lens parameters-peak intensity of the mass spectrometer with very little prior information. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113780428B_ABST
    Figure CN113780428B_ABST
Patent Text Reader

Abstract

The present invention provides a mass spectrometer lens parameter fitting method based on a radial basis function network, which uses a radial basis function network combined with a priori information set to obtain a fitting surface of lens parameters and peak intensity, and obtains lens parameters corresponding to the highest peak intensity according to the fitting surface. The present invention adopts a mass spectrometer lens parameter-peak intensity surface for fitting, without actually traversing all mass spectrometer lens parameter-peak intensity pairs; after the surface fitting is completed, the maximum value of the peak intensity can be obtained according to the input-output explicit expression, and the optimal lens parameters are obtained based on this, which has the advantage of being able to quickly obtain the optimal lens parameter combination.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of mass spectrometers, and in particular relates to a mass spectrometer lens parameter fitting method based on a radial basis function network. Background Art

[0002] The triple quadrupole mass spectrometer is an important instrument currently used by analysts to conduct qualitative and quantitative analysis of target compounds. It consists of four parallel metal rods. The DC voltage and radio frequency voltage applied to the rods generate a magnetic field, causing the ions to move in a spiral orbit along the axis between the metal rods. Depending on the voltage applied to the electrodes, ions with a specific m / z value will pass through the quadrupole, and other ions with larger or smaller m / z values ​​will fly outward and cannot pass through the quadrupole.

[0003] Before using the mass spectrometer, we need to obtain the best quality spectrometer lens parameters for the compound to be tested. Currently, there are two ways to obtain the optimal lens parameters for mass spectrometry. One is manual acquisition, which requires manual tuning by debuggers or experimentalists. The disadvantage is that it requires a certain amount of manpower, which involves the energy of experimentalists and reduces experimental efficiency. At the same time, the optimization of lens parameters has a certain threshold for debuggers, and it is not easy to find the global optimal solution. The other is automatic optimization. Automatic optimization generally uses the method of changing a single parameter to fix the remaining lens parameters for parameter optimization. This is a manifestation of the control variable method. Automatic optimization has two disadvantages. First, the optimization time is usually long because multiple iterations are required; second, the optimal lens parameters determined by the automatic optimization algorithm are often not the global optimal. In summary, the current lens parameter optimization method has the disadvantages of long optimization time, low efficiency, and difficulty in converging to the global optimal solution. Summary of the invention

[0004] In view of this, in order to overcome the above defects, the present invention aims to propose a mass spectrometer lens parameter fitting method based on radial basis function network.

[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] A mass spectrometer lens parameter fitting method based on a radial basis function network is disclosed. The radial basis function network is combined with a priori information set to obtain a fitting surface of lens parameters and peak intensity, and the lens parameters corresponding to the highest peak intensity are obtained according to the fitting surface.

[0007] Further, the specific method is as follows:

[0008] S1. Calculate the weight vector of the radial basis function network using a weight vector trainer combined with a priori set;

[0009] S2, select the activation function of the radial basis function network;

[0010] S3. Based on the weight vector and activation function, the input-output explicit expression of the radial basis function network is obtained.

[0011] Furthermore, when calculating the weight vector in step S1, it is necessary to combine the loss function, and the formula of the loss function is as follows:

[0012]

[0013] Where w is the weight vector of the radial basis function network, d(i) is the test set element, which represents the actual spectrum intensity in the test, parameter i = 1,…,N, and λ is the learning rate parameter to prevent overfitting. middle, μ K The activation function centered on i is the input parameter of the i-th element d(i) in the test set; let the loss function ε av The weight vector that minimizes (w) is the final desired weight vector.

[0014] Furthermore, an iterative algorithm is used to obtain the weight vector that minimizes the above loss function, as follows:

[0015] Step 1: Initialize the weight vector and intermediate matrix as shown below:

[0016] w(0)=0 and P(0)=λ -1 I

[0017] Initially, the weight vector w(0) of the radial basis function network is a zero vector, and the intermediate matrix P(0) is a diagonal matrix.

[0018] Step 2: Calculate the intermediate matrix P(n) of the nth iteration;

[0019]

[0020] Among them, P(n) is the intermediate matrix of the nth iteration in the calculation process, and P(n-1) is the intermediate matrix of the n-1th iteration in the calculation process. is the radial basis function vector, is the activation function centered on μ1, x n Represents the input lens parameters during the nth iteration.

[0021] Step 3: Calculate the gradient vector g(n) and the gradient coefficient α(n). The calculation formula is:

[0022] g(n)=P(n)φ(n)

[0023] α(n)=d(n)-w T (n-1)φ(n)

[0024] Where P(n) is the intermediate matrix calculated in the previous step, and φ(n) is the radial basis function vector, as introduced in the previous step. T (n-1) is the weight vector calculated in the previous iteration, and d(n) is the signal strength in the test set.

[0025] Step 4: Calculate the radial basis function network weight vector w(n) for the nth iteration, the iteration is as follows:

[0026] w(n)=w(n-1)+g(n)α(n)

[0027] Among them, g(n) represents the gradient vector, α(n) represents the gradient coefficient;

[0028] When n<N, update the number of iterations n=n+1, and then return to step 2; when n≥N, the algorithm ends, and the output w(N) of the Nth iteration is the final weight vector.

[0029] Furthermore, in step S2, the activation function adopts a Gaussian function or an Inverse multiquadrics function.

[0030] Furthermore, the input-output explicit expression of the radial basis function network is:

[0031]

[0032] Where y is the spectrum signal intensity when the lens parameter x is input, w n is the nth element of the weight vector w, is the activation function.

[0033] Compared with the prior art, the mass spectrometer lens parameter fitting method based on radial basis function network described in the present invention has the following advantages:

[0034] (1) The mass spectrometer lens parameter fitting method based on the radial basis function network of the present invention fits the mass spectrometer lens parameter-peak intensity surface without actually traversing all the mass spectrometer lens parameter-peak intensity pairs; after the surface fitting is completed, the maximum value of the peak intensity can be obtained according to the explicit expression of input-output, and the optimal lens parameters can be obtained based on this. The advantage of this is that the optimal lens parameter combination can be quickly obtained.

[0035] (2) The radial basis function network method is used to fit the mass spectrometer lens parameter-peak intensity surface. The use of the radial basis function network can avoid the edge effect caused by the traditional fitting method. In addition, different activation functions can be flexibly selected according to the characteristics of the mass spectrometer during use. Therefore, the fitting method based on the radial basis function network is more flexible and has higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0037] Figure 1 A block diagram for implementing the mass spectrometer lens parameter-peak intensity surface fitting method;

[0038] Figure 2 It is the basic structure diagram of radial basis function network;

[0039] Figure 3 This is a three-dimensional diagram of the input-output mapping relationship described in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0041] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0042] This patent proposes a mass spectrometer lens parameter fitting method based on a radial basis function network. This is a fast method for mass spectrometer lens parameter fitting. It can quickly obtain a fitting surface of lens parameter-peak intensity with very little prior information. After obtaining this fitting surface, the lens parameters corresponding to the highest peak intensity can be easily obtained.

[0043] The following characteristics of the mass spectrometer lens parameters are prerequisites for the derivation and use of the fitting algorithm, and further illustrate the rationality of the fitting algorithm used in this application.

[0044] 1) When adjusting the lens parameters, the effect on the resolution and mass axis of the mass spectrometer is minimal;

[0045] 2) Slight changes in lens parameters will not cause sudden changes in mass spectrometer peak intensity;

[0046] 3) When the lens parameters are fixed, the spectrum is scanned multiple times, and this process is a stable random process;

[0047] This application is based on the radial basis function network. Figure 1As shown in the figure, the radial basis function (RBF) network has three elements: activation function, weight vector and input vector. The RBF network weight vector trainer trains the RBF weight vector from the provided prior set data. The activation function uses but is not limited to the commonly used Gaussian function and Inverse multiquadrics function. The activation function can be flexibly selected according to the characteristics of the system, which is also one of the advantages of the RBF network. After the activation function and weight vector are determined, the RBF network is also determined, and the corresponding output is generated after the input vector enters the network.

[0048] Figure 2 is the basic structure of the radial basis function network, where μ1,μ2,…,μ K is the center value of the activation function.

[0049] The input-output explicit expression of the radial basis function network is as follows:

[0050]

[0051] Where y is the spectrum signal intensity when the lens parameter x is input, w n is the nth element of the weight vector w, is the activation function;

[0052] Defining vectors When calculating the weight vector, the loss function for evaluating the radial basis function network is defined as:

[0053]

[0054] Where w is the weight vector of the radial basis function network, d(i) is the test set element, which represents the actual spectrum intensity in the test, parameter i = 1,…,N, and λ is the learning rate parameter to prevent overfitting. middle, μ K The activation function centered on i is the input parameter of the i-th element d(i) in the test set; let the loss function ε av The weight vector that minimizes (w) is the final desired weight vector.

[0055] Step 1: Initialize the weight vector and intermediate matrix as shown below:

[0056] w(0)=0 and P(0)=λ -1 I (3)

[0057] Initially, the weight vector w(0) of the radial basis function network is a zero vector, and the intermediate matrix P(0) is a diagonal matrix.

[0058] Step 2: Calculate the intermediate matrix P(n) for the nth iteration

[0059]

[0060] P(n) is the intermediate matrix of the nth iteration in the calculation process, and P(n-1) is the intermediate matrix of the n-1th iteration in the calculation process. is the radial basis function vector, is the activation function centered on μ1, x n Represents the input lens parameters during the nth iteration.

[0061] Step 3: Calculate the gradient vector g(n) and the gradient coefficient α(n). The calculation formula is:

[0062] g(n)=P(n)φ(n) (5)

[0063] α(n)=d(n)-w T (n-1)φ(n) (6)

[0064] P(n) is the intermediate matrix calculated in the previous step, and φ(n) is the radial basis function vector, as introduced in the previous step. T (n-1) is the weight vector calculated in the previous iteration, and d(n) is the signal strength in the test set.

[0065] Step 4: Calculate the radial basis function network weight vector w(n) for the nth iteration, the iteration is as follows:

[0066] w(n)=w(n-1)+g(n)α(n) (7)

[0067] Step 4 is the updating process of the weight vector, where g(n) represents the gradient vector and α(n) represents the gradient coefficient.

[0068] When n<N, update the number of iterations n=n+1, and then return to step 2; when n≥N, the algorithm ends, and the output w(N) of the Nth iteration is the final weight vector.

[0069] After the weight vector is calculated using the above iterative algorithm, the explicit relationship of formula (1) is determined. In this way, the corresponding fitting relationship is actually determined.

[0070] The following is an example of the Q1 POS mode of the HTQ-2020 triple quadrupole mass spectrometer. The network input vector x = [a DP ,a EP ] T (represents the two lens parameters in Q1POS mode), the output is the peak intensity I(x=[a DP ,aEP ] T ), so the mapping that needs to be fitted should actually be First, we use a mass spectrometer to obtain 12 sets of training samples in a very short time, input them into the weight vector iteration algorithm, and calculate the weight vector of the network. After obtaining the weight vector, all possible inputs are input into the radial basis function network, the peak intensity is calculated, and the relevant mapping is obtained. At this time, we can get the mapping we need like Figure 3 As shown, the surface I(x=[a DP ,a EP ] T ), this maximum value I max ([a DP-OP ,a EP-OP ] T ) can be expressed as follows: in the lens parameter combination [a DP-OP ,a EP-OP ] T The peak intensity of this compound in Q1 POS mode is the highest.

[0071] Those of ordinary skill in the art will appreciate that the units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0072] In the several embodiments provided in the present application, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the division of the units described above is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The above-mentioned units may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A mass spectrometer lens parameter fitting method based on radial basis function network, Features: The radial basis function network is combined with a priori information set to obtain a fitting surface of lens parameters and peak intensity, and the lens parameters corresponding to the highest peak intensity are obtained based on the fitting surface; The specific method is as follows: S1. Calculate the weight vector of the radial basis function network using a weight vector trainer combined with a priori set; S2, select the activation function of the radial basis function network; S3. Based on the weight vector and activation function, the input-output explicit expression of the radial basis function network is obtained; When calculating the weight vector in step S1, it is necessary to combine the loss function. The formula of the loss function is as follows: Where w is the weight vector of the radial basis function network, d(i) is the test set element, which represents the actual spectrum intensity in the test, parameter i = 1,…,N, and λ is the learning rate parameter to prevent overfitting. middle, μ K The activation function centered on i is the input parameter of the i-th element d(i) in the test set; let the loss function ε av (w) The minimized weight vector is the final desired weight vector; The iterative algorithm is used to obtain the weight vector that minimizes the above loss function, as follows: Step 1: Initialize the weight vector and intermediate matrix as shown below: w(0)=0and P(0)=λ -1 I Initially, the weight vector w(0) of the radial basis function network is a zero vector, and the intermediate matrix P(0) is a diagonal matrix; Step 2: Calculate the intermediate matrix P(n) of the nth iteration; Among them, P(n) is the intermediate matrix of the nth iteration in the calculation process, and P(n-1) is the intermediate matrix of the n-1th iteration in the calculation process. is the radial basis function vector, μ 1 The activation function centered on n represents the input lens parameters during the nth iteration; Step 3: Calculate the gradient vector g(n) and the gradient coefficient α(n). The calculation formula is: g(n)=P(n)φ(n) α(n)=d(n)-w T (n-1)φ(n) Where P(n) is the intermediate matrix calculated in the previous step, φ(n) is the radial basis function vector, as introduced in the previous step, w T (n-1) is the weight vector calculated in the previous iteration, and d(n) is the signal strength in the test set; Step 4: Calculate the radial basis function network weight vector w(n) for the nth iteration, the iteration is as follows: w(n)=w(n-1)+g(n)α(n) Among them, g(n) represents the gradient vector, α(n) represents the gradient coefficient; When n<N, update the number of iterations n=n+1, and then return to step 2; when n≥N, the algorithm ends, and the output w(N) of the Nth iteration is the final weight vector.

2. The mass spectrometer lens parameter fitting method based on radial basis function network according to claim 1, Features: In step S2, the activation function adopts a Gaussian function or an Inverse multiquadrics function.

3. The mass spectrometer lens parameter fitting method based on radial basis function network according to claim 1, It is characterized in that The input-output explicit expression of the radial basis function network is: Where y is the spectrum signal intensity when the lens parameter x is input, w n is the nth element of the weight vector w, is the activation function.