IgA nephropathy prediction and classification method and device combined with Fourier transform infrared spectroscopy, electronic equipment and storage medium

Abstract

The invention relates to the technical field of auxiliary prediction and classification, in particular to an IgA nephropathy prediction and classification method and device combined with Fourier transform infrared spectroscopy, electronic equipment and a storage medium, and the method comprises the following steps: acquiring serum infrared spectroscopy data of a patient to be classified; inputting serum infrared spectrum data of a to-be-classified patient to a prediction classification model to obtain an IgA nephropathy prediction classification result of the to-be-classified patient, the prediction classification model being obtained by training a preset KAN-based multi-scale residual network model by using a plurality of samples, each sample comprises serum infrared spectrum data and a corresponding IgA nephropathy prediction classification result label. According to the method, multi-scale global information and fine-grained local details in the serum infrared spectrum data can be obtained by using the multi-scale residual network model based on the KAN, information loss is reduced, the accuracy of IgA nephropathy prediction and classification is improved, and the method does not need professional operation, is wide in application range and is easy to popularize.

Description

Technical Field

[0001] The present invention relates to the technical field of auxiliary prediction and classification, and is a method, device, electronic equipment and storage medium for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy. Background Art

[0002] IgA nephropathy is the most common primary glomerular disease worldwide. Its clinical and pathological manifestations vary, but its primary characteristic is the deposition of immune complexes composed of IgA, accompanied by proliferation of mesangial cells and matrix, and deposition of membrane attack complexes (MACs), such as complement C3, in the glomerular mesangium. IgA nephropathy is also a common cause of end-stage renal disease, with approximately 20%-40% of patients with IgA nephropathy progressing to renal failure within 20 years. Therefore, early diagnosis and effective prevention and treatment of IgA nephropathy are key to reducing the morbidity and mortality of end-stage renal disease.

[0003] Currently, early-stage diagnostic tests for IgA nephropathy are mostly invasive, based on renal biopsy. The gold standard for IgAN diagnosis is renal biopsy. However, in clinical practice, renal biopsy carries risks such as persistent hematuria, subcapsular hematoma, and puncture site infection. In addition to renal biopsy, immunofluorescence or immunohistochemistry results are essential. Serum IgA levels are significantly elevated in patients with IgA nephropathy, making it a valuable diagnostic tool. Urine protein quantification and typing, as well as renal function tests, also play a role in disease assessment and progression prediction. However, these initial tests are not very sensitive, with elevated serum IgA levels detected in only 50% of patients. Furthermore, renal biopsy is invasive, carries the risk of infection, and is expensive, requiring high-level medical expertise, making it difficult to access in economically disadvantaged areas. Therefore, a non-invasive, cost-effective, sensitive, and rapid method is needed for early-stage diagnostic testing of IgA nephropathy. Summary of the Invention

[0004] The present invention provides a method, device, electronic device and storage medium for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy, which can effectively solve the problems that renal tissue biopsy is invasive, prone to infection, and has high testing costs and high technical requirements for doctors, making it difficult to popularize in economically backward areas.

[0005] One of the technical solutions of the present invention discloses a method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy, comprising:

[0006] Acquire serum infrared spectrum data of patients to be classified;

[0007] The serum infrared spectral data of the patient to be classified is input into the predictive classification model to obtain the IgA nephropathy prediction classification result of the patient to be classified. The predictive classification model is obtained by training a preset KAN-based multi-scale residual network model using multiple samples. Each sample includes serum infrared spectral data and the corresponding IgA nephropathy prediction classification result label.

[0008] The optimized process of building a predictive classification model includes:

[0009] Obtain a sample data set and divide it into a training sample set and a test sample set in proportion, wherein the sample data set includes multiple samples, each sample includes serum infrared spectral data and a corresponding IgA classification result label;

[0010] The preset KAN-based multi-scale residual network model is trained using the training sample set to obtain a trained KAN-based multi-scale residual network model, wherein the KAN-based multi-scale residual network model includes a multi-scale residual network and a KAN network. The multi-scale residual network extracts and fuses infrared spectral features at different scales, and the KAN network classifies the fused infrared spectral features at different scales.

[0011] The trained KAN-based multi-scale residual network model is tested using the test set to optimize the model parameters, and the KAN-based multi-scale residual network model that meets the evaluation indicators is output as a prediction classification model.

[0012] The optimized,evaluation metrics include accuracy, sensitivity, precision and F1-score.

[0013] The optimized process of obtaining sample datasets includes:

[0014] Obtain infrared spectral data of N IgA nephropathy patients and infrared spectral data of M healthy controls, wherein the infrared spectral data of each IgA nephropathy patient / healthy control is the average value of the infrared spectrum of the corresponding serum sample at different positions;

[0015] Remove 2290-2385 cm from all IR spectra -1 The characteristics of the band are baseline corrected using iterative adaptive weighted penalized least squares method;

[0016] All infrared spectral data after baseline correction were preprocessed using the Savitzky-Golay algorithm to form a sample data set, where the preprocessing included smoothing, denoising, and normalization.

[0017] The second technical solution of the present invention discloses an IgA nephropathy prediction and classification device combined with Fourier transform infrared spectroscopy, comprising:

[0018] A test data acquisition unit, which acquires serum infrared spectrum data of the patient to be classified;

[0019] The prediction and classification unit inputs the serum infrared spectral data of the patient to be classified into the prediction and classification model to obtain the IgA nephropathy prediction and classification results of the patient to be classified, wherein the prediction and classification model is obtained by training a preset KAN-based multi-scale residual network model using multiple samples, and each sample includes serum infrared spectral data and the corresponding IgA nephropathy prediction and classification result label.

[0020] Optimized, model building units, including:

[0021] A sample acquisition module obtains a sample data set and divides it into a training sample set and a test sample set in proportion, wherein the sample data set includes multiple samples, each of which includes serum infrared spectroscopy data and corresponding IgA classification result labels;

[0022] The model training module uses the training sample set to train the preset KAN-based multi-scale residual network model to obtain the trained KAN-based multi-scale residual network model, wherein the KAN-based multi-scale residual network model includes a multi-scale residual network and a KAN network. The multi-scale residual network extracts and fuses infrared spectral features at different scales, and the KAN network classifies the fused infrared spectral features at different scales.

[0023] The model testing module uses the test set to test the trained KAN-based multi-scale residual network model, optimizes the model parameters, and outputs the KAN-based multi-scale residual network model that meets the evaluation indicators as a prediction classification model.

[0024] Optimized sample acquisition module, including:

[0025] The initial sample acquisition module obtains infrared spectral data of N IgA nephropathy patients and M healthy controls, where the infrared spectral data of each IgA nephropathy patient / healthy control is the average value of the infrared spectrum of the corresponding serum sample at different positions;

[0026] Correction module to remove 2290-2385 cm in all infrared spectral data -1 The characteristics of the band are baseline corrected using iterative adaptive weighted penalized least squares method;

[0027] The preprocessing module uses the Savitzky-Golay algorithm to preprocess all infrared spectral data after baseline correction to form a sample data set, where the preprocessing includes smoothing, denoising and normalization.

[0028] The third technical solution of the present invention discloses an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the computer program is loaded and executed by the processor to implement the steps in the IgA nephropathy prediction and classification method combined with Fourier transform infrared spectroscopy.

[0029] A fourth technical solution of the present invention discloses a storage medium storing a computer program that can be read by a computer, wherein the computer program is configured to execute the steps of a method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy when running.

[0030] The present invention uses multiple samples to train a preset KAN-based multi-scale residual network model to obtain a predictive classification model, and uses the predictive classification model to process the serum infrared spectral data of the patient to be classified to obtain the IgA nephropathy prediction and classification results. In this way, multi-scale global information and fine-grained local details in the serum infrared spectral data are obtained, which reduces information loss, makes up for the shortcomings of traditional deep convolutional networks, improves the accuracy of IgA nephropathy prediction and classification, and provides accurate auxiliary diagnostic information for IgA nephropathy. Moreover, this method does not require professional operation, has a wide range of adaptability, and is easy to popularize. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Attachment Figure 1 A schematic diagram of an implementation environment provided for one embodiment of the present invention.

[0032] Attachment Figure 2 A schematic flow chart of a method for predicting and classifying IgA nephropathy provided in one embodiment of the present invention.

[0033] FIG3( a ) is an average FTIR serum infrared spectrum provided by one embodiment of the present invention.

[0034] FIG3( b ) is a diagram showing the difference in average FTIR serum infrared spectra provided by one embodiment of the present invention.

[0035] Attachment Figure 4 A flowchart of a method for constructing a prediction classification model provided by one embodiment of the present invention.

[0036] Attachment Figure 5 A schematic diagram of the KAN-based multi-scale residual network model structure provided for one embodiment of the present invention.

[0037] Attachment Figure 6 A schematic flow chart of a method for obtaining a sample data set according to an embodiment of the present invention.

[0038] Attachment Figure 7 A loss curve diagram of a KAN-based multi-scale residual network model provided in one embodiment of the present invention.

[0039] Attachment Figure 8 A schematic diagram of the structure of an IgA nephropathy prediction and classification device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention is not limited to the following embodiments, and specific implementation methods can be determined based on the technical solutions of the present invention and actual conditions.

[0041] Those skilled in the art will understand that, unless otherwise specified, a "module" or "unit" in the embodiments of the present application refers to a computer program or portion of a computer program that has a predetermined function and works together with other related components to achieve a predetermined goal. Such a program may be implemented in whole or in part using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memories) may be used to implement one or more modules or units. Furthermore, each module or unit may be part of an overall module or unit that includes the functionality of that module or unit.

[0042] In addition, the “plurality” in the embodiments of the present application refers to two or more, and “first” and “second” are used to distinguish descriptions and should not be understood as implying relative importance.

[0043] An embodiment of the present application provides a method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy, obtaining serum infrared spectral data of a patient to be classified; inputting the serum infrared spectral data of the patient to be classified into a predictive classification model to obtain a predictive classification result of IgA nephropathy for the patient to be classified, wherein the predictive classification model is obtained by training a preset KAN-based multi-scale residual network model using multiple samples, and each sample includes serum infrared spectral data and a corresponding IgA nephropathy predictive classification result label.

[0044] Among them, the method provided in the embodiment of the present application may involve artificial intelligence (AI) technology and can be implemented based on artificial intelligence technology, for example, using deep learning to obtain a corresponding model through sample training.

[0045] Machine Learning (ML) is a multidisciplinary field that encompasses probability theory, statistics, approximation theory, convex analysis, and algorithmic complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is at the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications span all areas of AI.

[0046] Deep learning (DL) specifically refers to machine learning based on deep neural network models and methods. It is developed based on statistical machine learning, artificial neural networks, and other algorithmic models, combined with the development of modern big data and massive computing power. The most important technical feature of deep learning is its ability to automatically extract features.

[0047] As attached Figure 1 FIG. 1 shows a schematic diagram of an implementation environment provided by an embodiment of the present application. The implementation environment may include: training equipment and user equipment.

[0048] Both the training device and the usage device are computer devices; optionally, the computer device is a terminal device, such as a mobile phone, tablet computer, PC (Personal Computer) and other electronic devices; or, the computer device is a server, which can be a single server, a server cluster composed of multiple servers, or a cloud computing service center, which is not limited to this embodiment of the present application.

[0049] A training device refers to a computer device capable of training and learning a KAN-based multi-scale residual network model. Optionally, the training device has the ability to obtain a KAN-based multi-scale residual network model and train and learn it according to application requirements. For example, the training device obtains a KAN-based multi-scale residual network model from other devices through the network, and then trains it through training samples according to application requirements, so that the KAN-based multi-scale residual network model has the ability to obtain IgA nephropathy prediction and classification results; optionally, the training device has the ability to construct a KAN-based multi-scale residual network model, which can construct a KAN-based multi-scale residual network model according to application requirements, and then train and learn it. For example, in order to obtain IgA nephropathy prediction and classification results based on serum infrared spectroscopy data, the training device constructs a KAN-based multi-scale residual network model and then trains and learns it through samples according to application requirements.

[0050] The device used refers to a computer device that has the need to use the KAN-based multi-scale residual network model. Optionally, the device uses the KAN-based multi-scale residual network model from other devices through the network according to application requirements. For example, if the device uses the need to predict and classify IgA nephropathy, it can obtain the KAN-based multi-scale residual network model that has completed training and learning to predict the IgA nephropathy prediction and classification results from other devices through the network, and use the KAN-based multi-scale residual network model to perform IgA nephropathy prediction and classification.

[0051] Based on this, the technical solution of this application will be introduced and explained with reference to several examples below.

[0052] Example 1: As shown in the attached Figure 1 As shown, this embodiment discloses a method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy, comprising:

[0053] Step S110: Acquire serum infrared spectrum data of the patient to be classified.

[0054] The above process of obtaining serum infrared spectrum data of the patient to be classified includes:

[0055] (1) Obtain infrared spectral data of the patient to be classified, as follows:

[0056] Obtain serum samples from patients to be classified. Collect 3 mL of fresh blood from patients to be classified. Centrifuge the blood sample at 4000 rpm for 10 min at 4°C without using any anticoagulant. Take the supernatant to extract the serum sample. Place the serum sample in a centrifuge tube and store it in a -80°C refrigerator for subsequent testing.

[0057] The serum samples of the patients to be classified were extracted using a Fourier transform infrared spectrometer FTIR (FTIR-850, Jiangdong Science, China), with a spectral range of 700 to 4000 cm -1 4 cm -1 The resolution and scan number were 32, and the measurements were performed at constant temperature and humidity. Before measurement, the samples were thawed at room temperature (22 degrees Celsius) for 30 minutes, and then a 5 μL drop of each serum sample was placed on zinc selenide crystals and dried for 8 minutes to form a uniform solid film.

[0058] Corresponding infrared spectrum data are recorded at multiple different positions of the serum sample of the patient to be classified, and the average value thereof is obtained as the serum infrared spectrum data of the patient to be classified.

[0059] (2) Delete the 2290-2385 cm infrared spectrum data of the patient to be classified -1 The characteristics of the bands are baseline corrected using iterative adaptive weighted penalized least squares method.

[0060] (3) The baseline-corrected infrared spectral data are preprocessed using the Savitzky-Golay algorithm to form a sample data set, where the preprocessing includes smoothing, denoising, and normalization.

[0061] Step S120, input the serum infrared spectral data of the patient to be classified into the prediction classification model to obtain the IgA nephropathy prediction classification result of the patient to be classified, wherein the prediction classification model is obtained by training a preset KAN-based multi-scale residual network model using multiple samples, and each sample includes serum infrared spectral data and a corresponding IgA nephropathy prediction classification result label.

[0062] The serum infrared spectrum data extracted by FTIR used in this example can provide information about lipids, proteins, nucleic acids and carbohydrates in serum to assist in identifying the type of disease. Table 1 shows the positions and temporal distribution of characteristic peaks in the serum infrared spectrum data extracted by FTIR. Figure 3 shows the peaks of the control group and the peaks of the 600 cm -1 –4000 cm -1 The average FTIR serum infrared spectra and the difference between the average FTIR serum infrared spectra of IgA nephropathy patients and healthy controls within the range are shown. Their curve waveforms are similar, with the main difference being the amplitude of the curve fluctuations. For the detection of biologically relevant substances, the FTIR region can be divided into "biological fingerprint" bands, in which strong absorption bands can be directly associated with specific components. 3000-2800 cm -1 It is dominated by fatty acids and cholesterol, 1750-1500 cm -1 The amide region is composed of the carboxyl groups of proteins with amide groups, 1500-1300 cm -1 The region is the carboxyl group of protein, protein with amide II band, free amino acid and polysaccharide in the "mixed zone". It can be seen that the peaks at 1080, 1396, 1541, 1637, 2931 cm -1 The characteristic peak of the healthy control is higher than that of the IgA nephropathy patients. There are significant differences in the FTIR serum infrared spectrum data between the IgA nephropathy patients and the healthy control group. Therefore, the change in the intensity of the infrared characteristic peak reflects the change in the blood components of the IgA nephropathy patients. These intensity differences are the theoretical basis for subsequent model classification. Therefore, the present invention constructs a predictive classification model based on serum infrared spectrum data, and then uses the predictive classification model to obtain the serum infrared spectrum data of the patient to be classified to predict and classify IgA nephropathy.

[0063] Table 1 Characteristic peak distribution table

[0064]

[0065] The above-mentioned method uses a preset KAN-based multi-scale residual network model for training to construct a prediction and classification model. The KAN-based multi-scale residual network model includes a multi-scale residual network and a KAN network. It uses convolution kernels of different sizes to adaptively detect infrared spectral features of different scales, so that features of different scales can interact with each other to obtain the most effective infrared spectral information, so that the model can better understand and process the complex information in infrared spectral data. The classification accuracy of the KAN network is better than that of the traditional multi-layer perceptron.

[0066] In summary, this embodiment discloses a method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy. A preset KAN-based multi-scale residual network model is trained using multiple samples to obtain a predictive classification model. The predictive classification model is used to process the serum infrared spectral data of the patient to be classified to obtain the IgA nephropathy prediction and classification results. In this way, multi-scale global information and fine-grained local details in the serum infrared spectral data are obtained, information loss is reduced, the shortcomings of traditional deep convolutional networks are compensated, the accuracy of IgA nephropathy prediction and classification is improved, and accurate auxiliary diagnostic information is provided for IgA nephropathy. Moreover, this method does not require professional operation, has a wide range of adaptability, and is easy to popularize.

[0067] Example 2: As shown in the attached Figure 4 As shown, this embodiment is a further optimization of the above embodiment, wherein the process of constructing the prediction classification model includes:

[0068] Step S210, obtaining a sample data set and dividing it into a training sample set and a test sample set in proportion, wherein the sample data set includes a plurality of samples, each of which includes serum infrared spectral data and a corresponding IgA classification result label;

[0069] Step S220: training a preset KAN-based multi-scale residual network model using the training sample set to obtain a trained KAN-based multi-scale residual network model, wherein the KAN-based multi-scale residual network model includes a multi-scale residual network and a KAN network, the multi-scale residual network extracts and fuses infrared spectral features at different scales, and the KAN network classifies the fused infrared spectral features at different scales;

[0070] Step S230 , using the test set to test the trained KAN-based multi-scale residual network model, optimize the model parameters, and output the KAN-based multi-scale residual network model that meets the evaluation indicators as a prediction classification model.

[0071] The above preset KAN-based multi-scale residual network model structure is as shown in the attached Figure 5 As shown, it includes multi-scale residual network and KAN network.

[0072] The residual blocks in the multi-scale residual network use multi-scale feature extraction, improving the network's ability to model multi-scale features by increasing the scale dimension. Convolution modules with kernel sizes of 3, 5, and 7, and the best channel numbers of 64, 128, and 256 are used. The output of each residual block is used as a feature at different levels for global feature fusion to ensure the effective integration of all features.

[0073] The KAN network is a shallow KAN network with a hidden number of 128, which is used to classify the infrared spectral features at different scales after fusion.

[0074] The above training sample set is used to train the preset KAN-based multi-scale residual network model. The AdamW optimizer is used for model parameter optimization. The learning rate can be set to 0.001, the learning rate adjustment method adopts exponential decay, and the gamma value can be set to 0.8. The cross-entropy loss function is used for classification loss calculation. The epoch value can be set to 200, and the batch size can be set to 5.

[0075] The above evaluation indicators include accuracy, sensitivity, precision and F1-score, and the corresponding formulas are as follows:

[0076]

[0077]

[0078]

[0079]

[0080] Example 3: As shown in the attached Figure 6 As shown, this embodiment is a further optimization of the above embodiment, wherein the process of obtaining the sample data set includes:

[0081] Step S310 , obtaining infrared spectral data of N IgA nephropathy patients and infrared spectral data of M healthy controls, wherein the infrared spectral data of each IgA nephropathy patient / healthy control is the average value of the infrared spectrum of the corresponding serum sample at different positions.

[0082] Serum samples of IgA nephropathy patients / healthy controls were obtained as described above. 3 mL of fresh blood was collected from IgA nephropathy patients / healthy controls. Without using any anticoagulant, the blood samples were high-speed centrifuged at 4000 rpm for 10 min at 4°C. The supernatant layer was taken to extract serum samples. The serum samples were then placed in centrifuge tubes and refrigerated in a -80°C refrigerator for subsequent experiments.

[0083] Serum samples from IgA nephropathy patients and healthy controls were extracted using a Fourier transform infrared spectrometer (FTIR-850, Jiangdong Science, China) in the spectral range of 700 to 4000 cm -1 4cm -1 The resolution and scan number were 32, and the measurements were performed at constant temperature and humidity. Before measurement, all samples were thawed at room temperature (22 degrees Celsius) for 30 minutes, and then a 5 μL drop of each serum sample was placed on zinc selenide crystals and dried for 8 minutes to form a uniform solid film.

[0084] Step S320: Delete 2290-2385 cm from all infrared spectrum data. -1 The characteristics of the bands are baseline corrected using iterative adaptive weighted penalized least squares method.

[0085] Although infrared spectra are collected in a relatively closed environment, they are still affected by the air content. Existing research shows that at 2350 cm -1 The band near the wavelength represents CO2-related substances. In order to reduce the interference of air background on the spectrum, the 2290-2385 cm band was deleted from all infrared spectrum data. -1 Characteristics of the band.

[0086] In step S330, all baseline-corrected infrared spectral data are preprocessed using the Savitzky-Golay algorithm to form a sample dataset. This preprocessing includes smoothing, denoising, and normalization. The Savitzky-Golay algorithm is used for smoothing and denoising the spectra. The window size can be 7, the polynomial order can be 2, and the smoothed spectra are normalized to the range [0, 1]. This preprocessing step can be performed in MATLAB R2021a.

[0087] Example 4: The same test sample set was tested using MLP, Kan, ResNet, ResNet-Kan, MSR, and the KAN-based multi-scale residual network model used in the present invention. Accuracy, sensitivity, precision, and F1-score were introduced for evaluation. The evaluation results are shown in Table 2.

[0088] Table 2 Experimental results of each model

[0089]

[0090] The KAN-based multi-scale residual network model, MSRKan, achieved the highest accuracy of 86.21%. Its precision, recall, and F1-score were 84.62%, 91.67%, and 88%, respectively. MSRKan also achieved the best AUC value of 0.85. MSRKan's metrics improved by 6.21%, 3.37%, 5%, and 4.13%, respectively, compared to the shallow Kan network. The shallow Kan network achieved 4.14% higher classification accuracy for IgAN than the MLP. Replacing the classification layer of a ResNet with a Kan network also significantly improved the model's performance, with accuracy increasing by 3.45%. Similarly, replacing the classification layer with a Kan in a multi-scale residual network improved the accuracy and precision by 6.9% and 8.43%, respectively.

[0091] The loss of the multi-scale residual network model MSRKan based on KAN is as follows Figure 7 The multi-scale residual network (MSR) also shows significant improvements over ResNet, with an accuracy increase of 3.45%. This demonstrates the effectiveness of multi-scale network models in processing spectral data.

[0092] Example 5: As shown in the attached Figure 8 As shown, this embodiment discloses a device for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy, comprising:

[0093] A test data acquisition unit, which acquires serum infrared spectrum data of the patient to be classified;

[0094] The prediction and classification unit inputs the serum infrared spectral data of the patient to be classified into the prediction and classification model to obtain the IgA nephropathy prediction and classification results of the patient to be classified, wherein the prediction and classification model is obtained by training a preset KAN-based multi-scale residual network model using multiple samples, and each sample includes serum infrared spectral data and the corresponding IgA nephropathy prediction and classification result label.

[0095] The model building unit includes:

[0096] A sample acquisition module obtains a sample data set and divides it into a training sample set and a test sample set in proportion, wherein the sample data set includes multiple samples, each of which includes serum infrared spectroscopy data and corresponding IgA classification result labels;

[0097] The model training module uses the training sample set to train the preset KAN-based multi-scale residual network model to obtain the trained KAN-based multi-scale residual network model, wherein the KAN-based multi-scale residual network model includes a multi-scale residual network and a KAN network. The multi-scale residual network extracts and fuses infrared spectral features at different scales, and the KAN network classifies the fused infrared spectral features at different scales.

[0098] The model testing module uses the test set to test the trained KAN-based multi-scale residual network model, optimizes the model parameters, and outputs the KAN-based multi-scale residual network model that meets the evaluation indicators as a prediction classification model.

[0099] The sample acquisition module includes:

[0100] The initial sample acquisition module obtains infrared spectral data of N IgA nephropathy patients and M healthy controls, where the infrared spectral data of each IgA nephropathy patient / healthy control is the average value of the infrared spectrum of the corresponding serum sample at different positions;

[0101] Correction module to remove 2290-2385 cm in all infrared spectral data -1 The characteristics of the band are baseline corrected using iterative adaptive weighted penalized least squares method;

[0102] The preprocessing module uses the Savitzky-Golay algorithm to preprocess all infrared spectral data after baseline correction to form a sample data set, where the preprocessing includes smoothing, denoising and normalization.

[0103] Example 6: This example discloses a storage medium having a computer program readable by a computer stored thereon, wherein the computer program is configured to execute a method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy when running.

[0104] The above storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory, a mobile hard disk, a magnetic disk, or an optical disk.

[0105] Example 7: This example discloses an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the computer program is loaded and executed by the processor to implement a method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy.

[0106] The processor may be a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an ASIC, an FPGA, or other programmable logic device, a transistor logic device, a hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. It may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and so on. Memory may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memories, removable hard drives, magnetic disks, or optical disks.

[0107] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application may be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0108] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0109] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0110] The above content is only a specific implementation method of the present application, which has strong adaptability and implementation effect, but the protection scope of the present application is not limited to this. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be included in the protection scope of the present application. Therefore, equivalent changes made according to the claims of this application are still within the scope covered by this application.

Claims

1. A method for predicting and classifying IgA nephropathy using Fourier transform infrared spectroscopy, characterized in that: include: Acquire serum infrared spectrum data of patients to be classified; The serum infrared spectral data of the patient to be classified is input into the predictive classification model to obtain the IgA nephropathy prediction classification result of the patient to be classified. The predictive classification model is obtained by training a preset KAN-based multi-scale residual network model using multiple samples. Each sample includes serum infrared spectral data and the corresponding IgA nephropathy prediction classification result label.

2. The method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy according to claim 1, characterized in that: The process of constructing the prediction classification model includes: Obtain a sample data set and divide it into a training sample set and a test sample set in proportion, wherein the sample data set includes multiple samples, each sample includes serum infrared spectral data and a corresponding IgA classification result label; The preset KAN-based multi-scale residual network model is trained using the training sample set to obtain a trained KAN-based multi-scale residual network model, wherein the KAN-based multi-scale residual network model includes a multi-scale residual network and a KAN network. The multi-scale residual network extracts and fuses infrared spectral features at different scales, and the KAN network classifies the fused infrared spectral features at different scales. The trained KAN-based multi-scale residual network model is tested using the test set to optimize the model parameters, and the KAN-based multi-scale residual network model that meets the evaluation indicators is output as a prediction classification model.

3. The method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy according to claim 2, characterized in that: The evaluation indicators include accuracy, sensitivity, precision and F1-score.

4. The method for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy according to claim 1, 2 or 3, wherein: The process of obtaining the sample data set includes: Obtain infrared spectral data of N IgA nephropathy patients and infrared spectral data of M healthy controls, wherein the infrared spectral data of each IgA nephropathy patient / healthy control is the average value of the infrared spectrum of the corresponding serum sample at different positions; Remove 2290-2385 cm from all IR spectra -1 The characteristics of the band are baseline corrected using iterative adaptive weighted penalized least squares method; All infrared spectral data after baseline correction were preprocessed using the Savitzky-Golay algorithm to form a sample data set, where the preprocessing included smoothing, denoising, and normalization.

5. A device for predicting and classifying IgA nephropathy using the method according to any one of claims 1 to 4 in combination with Fourier transform infrared spectroscopy, characterized in that: include: A test data acquisition unit, which acquires serum infrared spectrum data of the patient to be classified; The prediction and classification unit inputs the serum infrared spectral data of the patient to be classified into the prediction and classification model to obtain the IgA nephropathy prediction and classification results of the patient to be classified, wherein the prediction and classification model is obtained by training a preset KAN-based multi-scale residual network model using multiple samples, and each sample includes serum infrared spectral data and the corresponding IgA nephropathy prediction and classification result label.

6. The device for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy according to claim 5, characterized in that: The model building unit includes: A sample acquisition module obtains a sample data set and divides it into a training sample set and a test sample set in proportion, wherein the sample data set includes multiple samples, each of which includes serum infrared spectroscopy data and corresponding IgA classification result labels; The model training module uses the training sample set to train the preset KAN-based multi-scale residual network model to obtain the trained KAN-based multi-scale residual network model, wherein the KAN-based multi-scale residual network model includes a multi-scale residual network and a KAN network. The multi-scale residual network extracts and fuses infrared spectral features at different scales, and the KAN network classifies the fused infrared spectral features at different scales. The model testing module uses the test set to test the trained KAN-based multi-scale residual network model, optimizes the model parameters, and outputs the KAN-based multi-scale residual network model that meets the evaluation indicators as a prediction classification model.

7. The device for predicting and classifying IgA nephropathy in combination with Fourier transform infrared spectroscopy according to claim 5 or 6, characterized in that: The sample acquisition module includes: The initial sample acquisition module obtains infrared spectral data of N IgA nephropathy patients and M healthy controls, where the infrared spectral data of each IgA nephropathy patient / healthy control is the average value of the infrared spectrum of the corresponding serum sample at different positions; Correction module to remove 2290-2385 cm in all infrared spectral data -1 The characteristics of the band are baseline corrected using iterative adaptive weighted penalized least squares method; The preprocessing module uses the Savitzky-Golay algorithm to preprocess all infrared spectral data after baseline correction to form a sample data set, where the preprocessing includes smoothing, denoising and normalization.

8. An electronic device, characterized in that: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and the computer program is loaded and executed by the processor to implement the steps in the method according to any one of claims 1 to 4.

9. A storage medium, characterized in that: The storage medium stores a computer program that can be read by a computer, and the computer program is configured to execute the steps of the method according to any one of claims 1 to 4 when run.

Images