Brain glioma auxiliary decision-making method based on terahertz signals

Abstract

The invention relates to a brain glioma auxiliary decision-making method based on terahertz signals. The brain glioma auxiliary decision-making method comprises the following steps: carrying out terahertz spectrum detection on brain glioma of different pathological levels to obtain terahertz spectrum data; converting the terahertz wave spectrum data of each wave band into pixel points, and arranging the pixel points according to a preset sequence to obtain a terahertz wave spectrum image; extracting a characteristic value of the terahertz wave spectrum image; inputting the characteristic value of the terahertz wave spectrum image as a training sample into a deep learning model for training to obtain a brain glioma pathology level prediction model; and predicting the pathological level of the target brain glioma by using the brain glioma pathological level prediction model to obtain a prediction result. According to the method, the terahertz spectrum data is converted into the image, the features are extracted, and the deep learning model is combined, so that key features for distinguishing different pathological levels can be automatically learned from a complex spectrum mode, multi-scale and multi-angle comprehensive analysis is realized, and the accuracy of predicting the pathological level of the brain glioma is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical technology, and in particular to a terahertz signal-based brain glioma decision-making auxiliary method. Background Art

[0002] Glioma is a common and malignant brain tumor originating from glial cells in the brain. It is characterized by strong invasiveness, high recurrence rate and poor prognosis. Clinically, early and accurate diagnosis and effective treatment of glioma are directly related to the patient's survival rate and quality of life. Existing glioma diagnosis mainly relies on imaging techniques such as magnetic resonance imaging (MRI), computed tomography (CT) and histopathological examination. Although these methods have been widely used, they still have problems such as limited imaging resolution, insufficient information on the molecular level of the tumor, complex operation and high invasiveness, which limit the precise analysis of the microstructure and biological characteristics of glioma.

[0003] With the development of terahertz technology, terahertz waves have shown broad application prospects in the biomedical field due to their non-ionizing radiation, good penetrability, and high sensitivity to biological tissues. Terahertz signals can provide multi-dimensional information such as tissue moisture content, molecular vibrations, and dynamic changes, and are expected to achieve non-destructive, rapid, and highly sensitive detection of glioma tissue. In addition, imaging and diagnostic technologies based on terahertz signals, combined with modern signal processing and machine learning algorithms, can enhance the differential diagnosis of gliomas and provide more objective and accurate auxiliary decision-making basis for clinicians. However, terahertz signal detection and auxiliary diagnosis methods for gliomas are still in the research stage, and there is still a lack of systematic, practical solutions suitable for clinical promotion. Summary of the Invention

[0004] To solve the above problems, an embodiment of the present invention aims to provide a terahertz signal-based glioma decision-making auxiliary method.

[0005] A terahertz signal-based brain glioma decision-making aid method, comprising:

[0006] Step 1: Perform terahertz spectroscopy on gliomas of different pathological grades to obtain terahertz spectrum data;

[0007] Step 2: Convert the terahertz spectrum data of each band into pixel points, and arrange the pixel points in a preset order to obtain a terahertz spectrum image;

[0008] Step 3: Extract the characteristic values ​​of the terahertz spectrum image;

[0009] Step 4: Input the characteristic values ​​of the terahertz spectrum image as training samples into the deep learning model for training to obtain a glioma pathology grade prediction model;

[0010] Step 5: Use the glioma pathology grade prediction model to predict the pathology grade of the target glioma and obtain the prediction result.

[0011] Preferably, the step 2: converting the terahertz spectrum data of each band into pixel points, and arranging the pixel points in a preset order to obtain a terahertz spectrum image, includes:

[0012] Use the pixel conversion function to convert the terahertz spectrum data of each band into pixel points, and put the converted pixel points into the white frame in order from left to right to form a terahertz spectrum image;

[0013] Wherein, the pixel conversion function is:

[0014]

[0015] In the formula, P(m,n) represents the pixel after conversion, round represents an integer, and x i represents the value of the i-th sampling point on the terahertz spectrum data, x min Indicates the minimum value of the sampling point on the terahertz spectrum data, x max Indicates the maximum value of the sampling point on the terahertz spectrum data.

[0016] Preferably, the step 3: extracting characteristic values ​​of the terahertz spectrum image includes:

[0017] The terahertz spectrum image is processed using convolution kernels of different sizes to obtain the eigenvalues ​​of the terahertz spectrum image. The eigenvalue extraction process of the terahertz spectrum image is as follows:

[0018]

[0019] Among them, F represents the terahertz spectrum image, F1 represents the first feature extraction value, F2 represents the second feature extraction value, F3 represents the third feature extraction value, F4 represents the fourth feature extraction value, F″ represents the eigenvalue of the terahertz spectrum image, Conv1, Conv3, Conv4 and Conv6 represent 1×1×1 convolution, Conv2 and Conv5 represent 3×3×3 convolution, Cat represents splicing, and Maxpool represents maximum pooling.

[0020] Preferably, the step 4: inputting the characteristic values ​​of the terahertz spectrum image as training samples into the deep learning model for training to obtain a glioma pathology grade prediction model includes:

[0021] Step 4.1: Input the characteristic values ​​of the terahertz spectrum image as training samples into the BP neural network;

[0022] Step 4.2: Replace the loss function in the training process with a genetic loss function, and use the genetic loss function to optimize the BP neural network to obtain a glioma pathology grade prediction model; wherein the genetic loss function is:

[0023]

[0024] Among them, n represents the number of samples, F represents the fitness of chromosome k, is the network parameter, and f ik represents the pathological grade of glioma of the i-th sample under the k-th chromosome output by the BP neural network, f(x ik ) represents the true pathological grade of glioma.

[0025] Preferably, in step 4.2, the BP neural network is optimized using the genetic loss function to obtain a glioma pathology grade prediction model, including:

[0026] Step 4.2.1: Determine the probability of chromosome selection for the next generation based on the fitness of the chromosome;

[0027] Step 4.2.2: Based on the probability of next generation chromosome selection, cross the parent chromosome to obtain a new chromosome, and return to step 4.2.1 until the fitness of a chromosome is within the set range;

[0028] Step 4.2.3: Extract the network parameters corresponding to the chromosomes with fitness within the set range;

[0029] Step 4.2.4: Substitute the network parameters into the BP neural network to obtain a glioma pathological grade prediction model.

[0030] Preferably, the probability of chromosome selection in the next generation is:

[0031]

[0032] Among them, p i represents the probability of chromosome selection in the next generation, F i represents the fitness of chromosome i, and N represents the number of chromosomes;

[0033] Preferably, in step 4.2.2, the formula is used:

[0034]

[0035] Cross the parent chromosome to get a new chromosome; where a′ ij represents the parent chromosome after crossover, a kj represents the first parent chromosome, a lj represents the second parent chromosome, b represents a random number between [0,1], represents a new chromosome, a min represents the minimum threshold, a max represents the maximum threshold, g represents the number of iterations, G max Indicates the maximum number of evolutions, G min Indicates the minimum number of evolutions.

[0036] The present invention also provides a terahertz signal-based brain glioma decision-making auxiliary system, comprising:

[0037] The terahertz spectrum detection module is used to perform terahertz spectrum detection on brain gliomas of different pathological levels to obtain terahertz spectrum data;

[0038] A pixel conversion module is used to convert the terahertz spectrum data of each band into pixels and arrange the pixels in a preset order to obtain a terahertz spectrum image;

[0039] A feature extraction module is used to extract the characteristic values ​​of the terahertz spectrum image;

[0040] A training module is used to input the characteristic values ​​of the terahertz spectrum image as training samples into the deep learning model for training to obtain a brain glioma pathology grade prediction model;

[0041] The prediction module is used to predict the pathological grade of the target glioma using the glioma pathological grade prediction model to obtain a prediction result.

[0042] The present invention also provides an electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, and wherein the computer program, when executed by the processor, implements the steps of the above-mentioned method for assisting decision-making in glioma based on terahertz signals.

[0043] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps in the above-mentioned terahertz signal-based brain glioma decision-making auxiliary method are implemented.

[0044] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0045] The present invention relates to a terahertz signal-based decision-making method for gliomas. Compared with the existing technology, the present invention converts terahertz spectral data into images and extracts features. Combined with a deep learning model, it can automatically learn key features for distinguishing different pathological levels from complex spectral patterns, achieve multi-scale and multi-angle comprehensive analysis, and improve the accuracy of predicting the pathological level of gliomas.

[0046] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 A flow chart of a glioma decision-making assistance method based on terahertz signals provided by the present invention;

[0049] Figure 2 Schematic diagram of feature extraction provided by the present invention;

[0050] Figure 3 This is a schematic diagram of a terahertz signal-based brain glioma decision-making support system provided by the present invention. DETAILED DESCRIPTION

[0051] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0052] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature identified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0053] In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediary; or internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0054] See also Figure 1 , a terahertz signal-based brain glioma decision-making assistance method, comprising:

[0055] Step 1: Perform terahertz spectroscopy on gliomas of different pathological grades to obtain terahertz spectrum data;

[0056] Step 2: Convert the terahertz spectrum data of each band into pixel points, and arrange the pixel points in a preset order to obtain a terahertz spectrum image;

[0057] In step 2, the terahertz spectrum data of each band is converted into pixels using a pixel conversion function, and the converted pixels are placed into white frames in order from left to right to form a terahertz spectrum image; wherein the pixel conversion function is:

[0058]

[0059] In the formula, P(m,n) represents the pixel after conversion, round represents an integer, and x i represents the value of the i-th sampling point on the terahertz spectrum data, x min Indicates the minimum value of the sampling point on the terahertz spectrum data, x max Indicates the maximum value of the sampling point on the terahertz spectrum data.

[0060] The processing method of the present invention, which uses a pixel conversion function to convert terahertz spectrum data into image samples, can convert time-frequency data into a pixel matrix with image properties. This can enhance the invisible information in the original signal, thereby helping the deep learning network to better learn features and reduce the risk of overfitting.

[0061] Step 3: Extract the characteristic values ​​of the terahertz spectrum image;

[0062] like Figure 2 As shown in FIG, the terahertz spectrum image is processed using convolution kernels of different sizes to obtain the eigenvalues ​​of the terahertz spectrum image; wherein, the eigenvalue extraction process of the terahertz spectrum image is:

[0063]

[0064] Among them, F represents the terahertz spectrum image, F1 represents the first feature extraction value, F2 represents the second feature extraction value, F3 represents the third feature extraction value, F4 represents the fourth feature extraction value, F″ represents the eigenvalue of the terahertz spectrum image, Conv1, Conv3, Conv4 and Conv6 represent 1×1×1 convolution, Conv2 and Conv5 represent 3×3×3 convolution, Cat represents splicing, and Maxpool represents maximum pooling.

[0065] Convolution kernels of different sizes can capture information at different scales in terahertz spectral images. Small convolution kernels help extract fine-grained local features (such as tiny fluctuations and texture details), while large convolution kernels can capture a wider range of global structures and contextual information. By combining features extracted by convolution kernels of multiple sizes, this paper enriches the model's expressive power, making it more sensitive to spectral differences between gliomas of different pathological grades and improving classification and recognition accuracy.

[0066] Step 4: Input the characteristic values ​​of the terahertz spectrum image as training samples into the deep learning model for training to obtain a glioma pathology grade prediction model;

[0067] Furthermore, step 4 includes:

[0068] Step 4.1: Input the characteristic values ​​of the terahertz spectrum image as training samples into the BP neural network;

[0069] Step 4.2: Replace the loss function in the training process with a genetic loss function, and use the genetic loss function to optimize the BP neural network to obtain a glioma pathology grade prediction model; wherein the genetic loss function is:

[0070]

[0071] Among them, n represents the number of samples, F represents the fitness of chromosome k, is the network parameter, and f ik represents the pathological grade of glioma of the i-th sample under the k-th chromosome output by the BP neural network, f(x ik ) represents the true pathological grade of glioma.

[0072] In the above step 4.2, the BP neural network is optimized using the genetic loss function to obtain a glioma pathology grade prediction model, including:

[0073] Step 4.2.1: Determine the probability of chromosome selection for the next generation based on the fitness of the chromosome;

[0074] Among them, the probability of chromosome selection in the next generation is:

[0075]

[0076] Among them, p i represents the probability of chromosome selection in the next generation, F i represents the fitness of chromosome i, and N represents the number of chromosomes;

[0077] Step 4.2.2: Based on the probability of next generation chromosome selection, cross the parent chromosome to obtain a new chromosome, and return to step 4.2.1 until the fitness of a chromosome is within the set range;

[0078] In step 4.2.2, use the formula:

[0079]

[0080] Cross the parent chromosome to get a new chromosome; where a′ ij represents the parent chromosome after crossover, a kj represents the first parent chromosome, a lj represents the second parent chromosome, b represents a random number between [0,1], represents a new chromosome, a min represents the minimum threshold, a max represents the maximum threshold, g represents the number of iterations, G max Indicates the maximum number of evolutions, G min Indicates the minimum number of evolutions.

[0081] Step 4.2.3: Extract the network parameters corresponding to the chromosomes with fitness within the set range;

[0082] Step 4.2.4: Substitute the network parameters into the BP neural network to obtain a glioma pathological grade prediction model.

[0083] Because BP neural networks are prone to falling into local optimal solutions during training, resulting in limited model performance, and traditional gradient descent methods are relatively sensitive to initial weights, which easily affects the convergence speed and accuracy of the network. The present invention introduces a genetic algorithm (GA) and utilizes its powerful global search capabilities to optimize the initial weights and thresholds of the BP neural network through genetic operations such as selection, crossover, and mutation. The genetic algorithm searches globally, making it easier to find parameter combinations close to the global optimal solution, thereby helping the network escape the local optimum.

[0084] Specifically, the genetic algorithm first randomly generates an initial population with a set of weights and thresholds. Using a fitness function, it evaluates the impact of each parameter set on training error. Then, through continuous iterations of genetic operations, it selects outstanding individuals, gradually approaching the optimal solution. The optimized parameters are then passed to the BP neural network as initial weights and thresholds, significantly improving network training results. This method not only improves the generalization ability of the BP neural network and enhances the accuracy of data prediction, but also effectively accelerates network convergence and reduces training time.

[0085] Step 5: Use the glioma pathology grade prediction model to predict the pathology grade of the target glioma and obtain the prediction result.

[0086] By converting terahertz spectral data into images and extracting features, combined with a deep learning model, the present invention can automatically learn the key features for distinguishing different pathological levels from complex spectral patterns, achieve multi-scale and multi-angle comprehensive analysis, and improve the accuracy of predicting the pathological level of brain gliomas.

[0087] See also Figure 3 The present invention also provides a glioma decision-making support system based on terahertz signals, comprising:

[0088] The terahertz spectrum detection module is used to perform terahertz spectrum detection on brain gliomas of different pathological levels to obtain terahertz spectrum data;

[0089] A pixel conversion module is used to convert the terahertz spectrum data of each band into pixels and arrange the pixels in a preset order to obtain a terahertz spectrum image;

[0090] A feature extraction module is used to extract the characteristic values ​​of the terahertz spectrum image;

[0091] A training module is used to input the characteristic values ​​of the terahertz spectrum image as training samples into the deep learning model for training to obtain a brain glioma pathology grade prediction model;

[0092] The prediction module is used to predict the pathological grade of the target glioma using the glioma pathological grade prediction model to obtain a prediction result.

[0093] Compared with the prior art, the beneficial effects of the terahertz signal-based glioma decision-making support system provided by the present invention are the same as the beneficial effects of the terahertz signal-based glioma decision-making support method described in the above technical solution, and will not be repeated here.

[0094] The present invention also provides an electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, and is characterized in that when the computer program is executed by the processor, the steps in the above-mentioned method for assisting decision-making for gliomas based on terahertz signals are implemented. Compared with the prior art, the beneficial effects of the electronic device provided by the present invention are the same as the beneficial effects of the method for assisting decision-making for gliomas based on terahertz signals described in the above-mentioned technical solution, and are not elaborated here.

[0095] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the above-mentioned method for assisting decision-making for gliomas based on terahertz signals are implemented. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present invention are the same as the beneficial effects of the method for assisting decision-making for gliomas based on terahertz signals described in the above-mentioned technical solution, and will not be elaborated here.

[0096] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technical solution that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A terahertz signal-based glioma decision-making aid method, characterized in that: include: Step 1: Perform terahertz spectroscopy on gliomas of different pathological grades to obtain terahertz spectrum data; Step 2: Convert the terahertz spectrum data of each band into pixel points, and arrange the pixel points in a preset order to obtain a terahertz spectrum image; Step 3: Extract the characteristic values ​​of the terahertz spectrum image; Step 4: Input the characteristic values ​​of the terahertz spectrum image as training samples into the deep learning model for training to obtain a glioma pathology grade prediction model; Step 5: Use the glioma pathology grade prediction model to predict the pathology grade of the target glioma and obtain the prediction result.

2. The terahertz signal-based brain glioma decision-making auxiliary method according to claim 1, characterized in that: The step 2: converting the terahertz spectrum data of each band into pixel points, and arranging the pixel points in a preset order to obtain a terahertz spectrum image, includes: Use the pixel conversion function to convert the terahertz spectrum data of each band into pixel points, and put the converted pixel points into the white frame in order from left to right to form a terahertz spectrum image; Wherein, the pixel conversion function is: In the formula, P(m,n) represents the pixel after conversion, round represents an integer, and x i represents the value of the i-th sampling point on the terahertz spectrum data, x min Indicates the minimum value of the sampling point on the terahertz spectrum data, x max Indicates the maximum value of the sampling point on the terahertz spectrum data.

3. The terahertz signal-based brain glioma decision-making auxiliary method according to claim 2, characterized in that: The step 3: extracting characteristic values ​​of the terahertz spectrum image includes: The terahertz spectrum image is processed using convolution kernels of different sizes to obtain the eigenvalues ​​of the terahertz spectrum image. The eigenvalue extraction process of the terahertz spectrum image is as follows: Among them, F represents the terahertz spectrum image, F1 represents the first feature extraction value, F2 represents the second feature extraction value, F3 represents the third feature extraction value, F4 represents the fourth feature extraction value, F″ represents the eigenvalue of the terahertz spectrum image, Conv1, Conv3, Conv4 and Conv6 represent 1×1×1 convolution, Conv2 and Conv5 represent 3×3×3 convolution, Cat represents splicing, and Maxpool represents maximum pooling.

4. The terahertz signal-based glioma decision-making assistance method according to claim 3, characterized in that: Step 4: Inputting the characteristic values ​​of the terahertz spectrum image as training samples into the deep learning model for training to obtain a glioma pathology grade prediction model, including: Step 4.1: Input the characteristic values ​​of the terahertz spectrum image as training samples into the BP neural network; Step 4.2: Replace the loss function in the training process with a genetic loss function, and use the genetic loss function to optimize the BP neural network to obtain a glioma pathology grade prediction model; wherein the genetic loss function is: Among them, n represents the number of samples, F represents the fitness of chromosome k, is the network parameter, and f ik represents the pathological grade of glioma of the i-th sample under the k-th chromosome output by the BP neural network, f(x ik ) represents the true pathological grade of glioma.

5. The terahertz signal-based glioma decision-making assistance method according to claim 4, characterized in that: In step 4.2, the BP neural network is optimized using the genetic loss function to obtain a glioma pathology grade prediction model, including: Step 4.2.1: Determine the probability of chromosome selection for the next generation based on the fitness of the chromosome; Step 4.2.2: Based on the probability of next generation chromosome selection, cross the parent chromosome to obtain a new chromosome, and return to step 4.2.1 until the fitness of a chromosome is within the set range; Step 4.2.3: Extract the network parameters corresponding to the chromosomes with fitness within the set range; Step 4.2.4: Substitute the network parameters into the BP neural network to obtain a glioma pathological grade prediction model.

6. The terahertz signal-based glioma decision-making assistance method according to claim 5, characterized in that: in, The probability of chromosome selection in the next generation is: Among them, p i represents the probability of chromosome selection in the next generation, F i represents the fitness of chromosome i, and N represents the number of chromosomes.

7. The terahertz signal-based glioma decision-making assistance method according to claim 6, characterized in that: In step 4.2.2, use the formula: Cross the parent chromosome to get a new chromosome; where a′ ij represents the parent chromosome after crossover, a kj represents the first parent chromosome, a lj represents the second parent chromosome, b represents a random number between [0,1], represents a new chromosome, a min represents the minimum threshold, a max represents the maximum threshold, g represents the number of iterations, G max Indicates the maximum number of evolutions, G min Indicates the minimum number of evolutions.

8. A terahertz signal-based brain glioma decision-making support system, characterized in that: include: The terahertz spectrum detection module is used to perform terahertz spectrum detection on brain gliomas of different pathological levels to obtain terahertz spectrum data; A pixel conversion module is used to convert the terahertz spectrum data of each band into pixels and arrange the pixels in a preset order to obtain a terahertz spectrum image; A feature extraction module is used to extract the characteristic values ​​of the terahertz spectrum image; A training module is used to input the characteristic values ​​of the terahertz spectrum image as training samples into the deep learning model for training to obtain a brain glioma pathology grade prediction model; The prediction module is used to predict the pathological grade of the target glioma using the glioma pathological grade prediction model to obtain a prediction result.

9. An electronic device comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, wherein: When the computer program is executed by the processor, the steps of the terahertz signal-based glioma decision-making assistance method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the terahertz signal-based glioma decision-making assistance method according to any one of claims 1 to 7 are implemented.

Images