A burn grading and typing prediction method, system, device and storage medium

Abstract

The application provides a burn grading and typing prediction method, system, device and storage medium, a rat burn model database containing different burn types and grades is constructed, a rich and structured training basis is provided for machine learning, a triple feature data screening method is used for preprocessing of machine learning data set, the data quality is improved, and subsequent model training is more efficient and reliable; a burn grading and typing prediction model is obtained through training of a random forest model, so that deep representation of the burn state is realized; the model can be used for assisting medical staff to perform burn grading and typing operation after final deployment, and provides objective and quantitative prediction reference for the medical staff, effectively alleviates subjective deviation and misjudgment risk caused by traditional naked eye observation and experience judgment, and helps doctors to perform burn grading and typing earlier and more accurately through the high-credibility prediction result, and reduces the problems of delayed treatment or excessive treatment caused by misjudgment.

Description

Technical Field

[0001] This invention belongs to the technical field of burn grading and classification judgment, specifically relating to a burn grading and classification prediction method, system, device and storage medium. Background Technology

[0002] In clinical medicine, burns are a common type of trauma. Their occurrence is often directly related to exogenous factors such as heat, chemicals, electric current, or radiation. These factors cause damage to the skin and even deep tissues, which not only triggers acute pathophysiological reactions such as severe pain, fluid exudation, and increased risk of infection, but may also pose a long-term threat to the patient's quality of life and safety due to problems such as scar formation, functional limitation, and even organ failure. Therefore, early and accurate assessment and classification of burns are always key links in clinical treatment, directly affecting the selection and judgment of treatment plans.

[0003] In existing technologies, the "three-degree four-classification method" or "four-degree five-classification method" is widely used to classify and grade burns. These methods rely heavily on the clinical experience and visual observation of doctors, and have inherent limitations such as strong subjectivity, vague quantitative standards, and difficulty in dynamic assessment. For example, in the early stage of burns, deep second-degree burns and third-degree burns have similar clinical manifestations, both of which may present as pale or waxy appearances. It is difficult to accurately distinguish them by visual observation alone. Therefore, different doctors may have different definitions of the boundary between "deep second-degree" and "third-degree". In clinical application, this uncertainty in judgment often leads to misdiagnosis of burn patients in the early stage of burns, affecting early treatment decisions and causing technical problems such as delaying critical treatment opportunities or increasing the risk of overtreatment. Summary of the Invention

[0004] To address the inherent limitations of the widely used "three-degree four-classification" or "four-degree five-classification" methods for grading and classifying burns in the background art, such as strong subjectivity, vague quantitative standards, and difficulties in dynamic assessment, which often lead to misdiagnosis of burn patients in the early stages of burns, affecting early treatment decisions, and causing delays in critical treatment or increasing the risk of overtreatment, this invention provides a burn grading and classification prediction method, system, device, and storage medium.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides a method for predicting the grading and classification of burns, comprising:

[0007] S1: Data Collection: Multiple rat burn models were constructed according to different burn types and burn grades. Burn data were collected based on the multiple rat burn models to construct a burn model database. The burn model database includes burn images and clinical burn diagnostic factors corresponding to different burn types and burn grades.

[0008] S2: Data processing: Process the burn data in the burn model database to obtain a machine learning dataset;

[0009] S3: Pre-training: A triple feature data filtering method is used to filter the machine learning dataset to obtain the model training dataset;

[0010] S4: Training the model: Use the model training dataset as input parameters to train the random forest model and obtain the burn grading and classification prediction model.

[0011] S5: Deploy the model and perform burn grading and classification prediction: Deploy the trained burn grading and classification prediction model, use the burn grading and classification prediction model to perform burn grading and classification, and obtain the corresponding burn grading and classification prediction results.

[0012] Optionally, in step S1, the multiple rat burn models constructed are shallow second-degree burn, deep second-degree burn, third-degree burn, flame burn, and electric shock burn models constructed using multiple rats with the same growth conditions.

[0013] Optionally, step S2 specifically includes:

[0014] S2.1: Input the burn images corresponding to different burn types and burn grades into the VGG-11 model to extract features and obtain the image feature set;

[0015] S2.2: For clinical burn diagnostic factors corresponding to different burn types and burn grades, the target protein was quantified using the ELISA method to obtain a clinical burn diagnostic factor dataset;

[0016] S2.3: The image feature set obtained in step S2.1 and the clinical burn diagnostic factor dataset obtained in step S2.2 are directly merged into a machine learning dataset according to the sample naming.

[0017] Optionally, step S2.1 specifically involves inputting burn images corresponding to different burn types and burn grades into the VGG-11 model, and extracting the output content of the second-to-last non-task layer as the image feature set.

[0018] Optionally, step S3 includes:

[0019] S3.1: The machine learning dataset is processed using three feature selection methods: Boruta, RFE, and Pearson, respectively, to obtain the first set, the second set, and the third set.

[0020] S3.2: Take the intersection of the first set, the second set, and the third set as the model training dataset.

[0021] Optionally, in step S4, the model training dataset is divided into a training set and a validation set, and hyperparameter optimization is performed using a grid search method to obtain a burn grading and classification prediction model.

[0022] Optionally, step S5 specifically includes:

[0023] S5.1: Deploy a burn grading and classification prediction model;

[0024] S5.2: Collect images of burns to be analyzed and / or clinical burn diagnostic factors, and obtain the dataset of burns to be analyzed by referring to the methods in steps S2.1 to S2.3;

[0025] S5.3: Input the dataset of burns to be analyzed into the burn grading and classification prediction model to obtain the corresponding burn grading and classification prediction results.

[0026] Secondly, the present invention provides a burn grading and classification prediction system, the burn grading and classification prediction system comprising:

[0027] The data acquisition module is used to construct multiple rat burn models according to different burn types and burn grades, collect burn data based on multiple rat burn models, and construct a burn model database. The burn model database includes burn images and clinical burn diagnostic factors corresponding to different burn types and burn grades.

[0028] The data processing module is used to process burn data in the burn model database to obtain machine learning datasets;

[0029] The pre-training module is used to filter the machine learning dataset using a triple feature data filtering method to obtain the model training dataset;

[0030] The training module is used to take the model training dataset as input parameters and input it into the random forest model for training to obtain a burn grading and classification prediction model.

[0031] The prediction module is used to deploy the trained burn grading and classification prediction model, perform burn grading and classification operations using the burn grading and classification prediction model, and obtain the corresponding burn grading and classification prediction results.

[0032] Thirdly, the present invention provides a burn grading and classification prediction device, which includes a data input unit, a data processing unit and an output unit.

[0033] The data input unit is used to acquire the burn analysis dataset to be detected and output it to the data processing unit.

[0034] The data processing unit is equipped with a burn grading and classification prediction model based on the burn grading and classification prediction method described above, and obtains the corresponding burn grading and classification prediction results based on the burn analysis dataset to be examined.

[0035] The output unit is used to output the corresponding burn grading and classification prediction results.

[0036] Fourthly, the present invention provides a storage device, characterized in that the storage device stores instructions, which are generated based on any of the above-described burn grading and classification prediction methods.

[0037] The beneficial effects of this invention are:

[0038] This invention provides a method for predicting burn grading and classification. By constructing a rat burn model database containing different burn types and grades, and collecting multimodal data of burn images and clinical burn diagnostic factors, a rich and structured training foundation for machine learning is provided. A triple feature data filtering method is used to preprocess the machine learning dataset, effectively removing redundant and noisy features, improving data quality, and making subsequent model training more efficient and reliable. The burn grading and classification prediction model trained using a random forest model can comprehensively learn the latent feature patterns in images and clinical burn diagnostic factors, thereby achieving a deep representation of the burn state. After deployment, this model can assist medical personnel in burn grading and classification, providing them with objective and quantitative predictive references, effectively mitigating the subjective bias and misjudgment risks caused by traditional visual observation and experience-based judgment. This invention helps doctors to grade and classify burns earlier and more accurately by providing highly reliable prediction results, thereby supporting the development of more reasonable early treatment plans and reducing the problems of delayed treatment or overtreatment caused by misjudgment. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the burn grading and classification prediction method in this invention;

[0040] Figure 2 This is a schematic diagram illustrating the construction of burn models of different levels and types in this invention;

[0041] Figure 3 This is a schematic diagram of the result of processing the machine learning dataset using the Boruta feature selection method in step S3 of the present invention.

[0042] Figure 4 This is a schematic diagram of the result of processing the machine learning dataset using the RFE feature selection method in step S3 of the present invention;

[0043] Figure 5This is a schematic diagram showing the result of processing the machine learning dataset using the Pearson feature selection method in step S3 of the present invention. Detailed Implementation

[0044] To provide a detailed description of the burn grading and classification prediction method, system, device, and storage medium of this invention, the following description is provided in conjunction with embodiments.

[0045] In clinical medicine, burns are a common type of trauma, and their occurrence is often directly related to exogenous factors such as heat, chemicals, electric current, or radiation. Current techniques widely employ the "three-degree four-classification" or "four-degree five-classification" method for grading and classifying burns. These methods heavily rely on the clinical physician's experience and visual observation, and have inherent limitations such as high subjectivity, vague quantitative standards, and difficulty in dynamic assessment.

[0046] Existing studies on auxiliary burn grading and classification often focus on single genomic or morphological analyses. Genomics cannot directly correlate with the dynamic fluctuations of metabolites, metabolomics lacks spatiotemporal correlation with histopathological phenotypes, and morphological studies are insufficient in elucidating molecular mechanisms. All of these approaches fail to fully elucidate the complex mechanisms of burns, and their predictive accuracy remains significantly insufficient in practical applications.

[0047] Example 1

[0048] See Figure 1 The diagram illustrates the burn grading and classification prediction method of the present invention, including:

[0049] S1: Data Collection: Multiple rat burn models were constructed according to different burn types and burn grades. Burn data were collected based on the multiple rat burn models to construct a burn model database. The burn model database includes burn images and clinical burn diagnostic factors corresponding to different burn types and burn grades.

[0050] S2: Data processing: Process the burn data in the burn model database to obtain a machine learning dataset;

[0051] S3: Pre-training: A triple feature data filtering method is used to filter the machine learning dataset to obtain the model training dataset;

[0052] S4: Training the model: Use the model training dataset as input parameters to train the random forest model and obtain the burn grading and classification prediction model.

[0053] S5: Deploy the model and perform burn grading and classification prediction: Deploy the trained burn grading and classification prediction model, use the burn grading and classification prediction model to perform burn grading and classification, and obtain the corresponding burn grading and classification prediction results.

[0054] In this embodiment, a rat burn model database containing different burn types and grades is constructed, and multimodal data of burn images and clinical burn diagnostic factors are collected, providing a rich and structured training foundation for machine learning. A triple feature data screening method is used to preprocess the machine learning dataset, effectively removing redundant and noisy features, improving data quality, and making subsequent model training more efficient and reliable. The burn grading and classification prediction model trained using a random forest model can comprehensively learn the latent feature patterns in images and clinical burn diagnostic factors, thereby achieving a deep representation of the burn state. After final deployment, this model can assist medical personnel in burn grading and classification, providing them with objective and quantitative predictive references, effectively mitigating the subjective bias and misjudgment risks caused by traditional visual observation and experience-based judgment. This invention helps doctors to grade and classify burns earlier and more accurately by providing highly reliable prediction results, thereby supporting the development of more reasonable early treatment plans and reducing the problems of delayed treatment or overtreatment caused by misjudgment.

[0055] Optionally, in step S1 of this invention, the multiple rat burn models constructed are shallow second-degree burn, deep second-degree burn, third-degree burn, flame burn, and electric shock burn models constructed using multiple rats with the same growth conditions. (Refer to...) Figure 2 The images show finished models of normal skin, superficial second-degree burns, deep second-degree burns, third-degree burns, flame burns, and electrical burns, respectively. Models a to f are normal skin, superficial second-degree burns, deep second-degree burns, third-degree burns, flame burns, and electrical burns, respectively.

[0056] In this embodiment, multiple burn models, including superficial second-degree burns, deep second-degree burns, third-degree burns, flame burns, and electrical burns, were constructed using multiple rats grown under identical conditions. This significantly improved data consistency and comparability, and reduced noise interference caused by individual differences. The highly standardized modeling approach ensured that the subsequently collected burn images and clinical burn diagnostic factors were highly representative and generalizable, providing a stable and diverse input source for the machine learning model. The burn model database built upon this foundation can more comprehensively cover common burn types and severity in clinical practice, thereby enabling the trained predictive model to possess stronger adaptability and accuracy.

[0057] Furthermore, in this embodiment, male SD rats aged 6-8 weeks were used to construct the rat burn model. All rats were purchased from the same experimental biological company and were raised in the same environment.

[0058] To construct superficial second-degree burns, deep second-degree burns, and third-degree burns, a 100℃ electric cautery iron was used. After hair removal on the back of the rats, pressureless contact was performed at 1 to 3 points (total area ≤ 10% of body surface area) for different durations to establish burn models of different depths: superficial second-degree burns (2 to 4 seconds), deep second-degree burns (5 to 6 seconds), and third-degree burns (8 to 10 seconds).

[0059] When constructing flame burn and electric shock burn models, flame burn models are constructed by burning the modeling point with the outer flame for 2-3 seconds; or by applying conductive paste evenly to the back of a rat and then applying positive and negative electrodes for 2-3 seconds to establish an electric shock burn model.

[0060] Optionally, step S2 in this invention specifically includes:

[0061] S2.1: Input the burn images corresponding to different burn types and burn grades into the VGG-11 model to extract features and obtain the image feature set;

[0062] S2.2: For clinical burn diagnostic factors corresponding to different burn types and burn grades, the target protein was quantified using the ELISA method to obtain a clinical burn diagnostic factor dataset;

[0063] S2.3: The image feature set obtained in step S2.1 and the clinical burn diagnostic factor dataset obtained in step S2.2 are directly merged into a machine learning dataset according to the sample naming.

[0064] In this embodiment, the image feature set extracted by the VGG-11 model is directly merged with the clinical burn diagnostic factor dataset obtained by ELISA, naming the samples directly. This fuses image morphological information with the clinical burn diagnostic factors, a biomarker, to construct a multimodal machine learning dataset. Utilizing the complementarity between visual features of the burn area and the in vivo biochemical response, this dataset more comprehensively and accurately reflects the actual severity and type of burns. Image feature extraction relies on deep convolutional networks to capture the texture, color, and structural changes of the burn surface, while the clinical burn diagnostic factor dataset reveals the body's stress and inflammation state at the molecular level. The combination of these two significantly enhances the model's representational dimension of burn conditions. The constructed machine learning dataset provides rich and valuable information input for subsequent feature selection and model training, enabling the final trained prediction model to exhibit higher reliability and interpretability when assisting medical personnel in burn assessment.

[0065] Furthermore, in this embodiment, the data are directly merged into a machine learning dataset according to the sample names. This involves naming the image feature set and the clinical burn diagnostic factor dataset based on the samples in the experiment, and then merging them according to the sample names to obtain the machine learning dataset.

[0066] Furthermore, the clinical burn diagnostic factor dataset in this embodiment includes a serum protein feature set.

[0067] Furthermore, in step S2.2 of this embodiment, when obtaining the protein feature set, the target protein is quantified using ELISA. Specifically, the contents of EGF, MDA, IL-1β, IL-6, TNF-α, VEGF-α, TGF-β1, and SOD in rat clinical burn diagnostic factors are detected using ELISA. The specific procedure is as follows: the clinical burn diagnostic factor sample is diluted and added to a pre-coated plate, incubated at 37°C, washed, and then biotinylated antibody, enzyme conjugate, and chromogenic substrate are added sequentially. The absorbance at 450 nm is measured using an ELISA reader, and the concentration of each indicator is calculated using a standard curve.

[0068] Optionally, step S2.1 in this invention specifically involves inputting burn images corresponding to different burn types and burn grades into the VGG-11 model, and extracting the output content of the second-to-last non-task layer as the image feature set.

[0069] Optionally, step S3 in this invention includes:

[0070] S3.1: Reference Figure 3 , Figure 4 and Figure 5 The machine learning dataset was processed using three feature selection methods: Boruta, RFE, and Pearson, respectively, resulting in the first set, the second set, and the third set.

[0071] S3.2: Take the intersection of the first set, the second set, and the third set as the model training dataset.

[0072] In this embodiment, three feature selection methods—Boruta, RFE, and Pearson—are used to process the machine learning dataset, and their intersection is taken as the model training dataset. This effectively integrates the advantages of various selection strategies, achieving multiple verification and reinforcement selection of high-value features. Specifically, the Boruta method comprehensively evaluates the contribution of each feature by comparing the importance of the original features and the shadow features; RFE optimizes the feature subset by recursively eliminating redundant features; and the Pearson method focuses on eliminating highly linearly correlated features to reduce collinearity interference. The use of the intersection of the three methods significantly improves the discriminative power and generalization of the feature set, reduces the risk of overfitting, and results in a higher quality model training dataset with less noise. This allows the subsequent training of the random forest model to focus more on key features, improving the stability and interpretability of the model in burn grading and classification prediction, and providing medical staff with clearer and more reliable decision-making references.

[0073] Specifically, by Figure 3It can be seen that, through Boruta feature selection, a total of 28 features were selected, with no features being eliminated or left as candidates.

[0074] Optionally, in step S4 of the present invention, the model training dataset is divided into a training set and a validation set in a 7:3 ratio, and hyperparameter optimization is performed using a grid search method to obtain a burn grading and classification prediction model.

[0075] In this embodiment, by dividing the model training dataset into training and validation sets proportionally and optimizing hyperparameters, the standardization and scientific nature of the model training process are significantly enhanced, overfitting or underfitting is avoided, the model training process is optimized, and its performance in actual burn grading and classification prediction is improved. This provides medical personnel with a high-performance and consistent auxiliary prediction tool, effectively supporting them in making rapid and accurate assessments of burn severity.

[0076] Optionally, step S5 in this invention specifically includes:

[0077] S5.1: Deploy a burn grading and classification prediction model;

[0078] S5.2: Collect images of burns to be analyzed and / or clinical burn diagnostic factors, and obtain the dataset of burns to be analyzed by referring to the methods in steps S2.1 to S2.3;

[0079] S5.3: Input the dataset of burns to be analyzed into the burn grading and classification prediction model to obtain the corresponding burn grading and classification prediction results.

[0080] In this embodiment, the burn grading and classification prediction model trained in this embodiment is deployed in practical use. Images of the burns to be analyzed and / or clinical burn diagnostic factors are collected from the patient to establish a dataset for burn analysis. This dataset is then input into the burn grading and classification prediction model to obtain the corresponding burn grading and classification prediction results. This helps medical staff quickly obtain the grading and classification prediction results corresponding to the burn images and clinical burn diagnostic factors. The output prediction results are not directly used as a diagnostic conclusion, but rather as an objective and quantitative reference to assist doctors in making comprehensive judgments based on clinical experience. This significantly reduces the subjective bias of relying solely on visual observation, improves the early differentiation ability of burns, supports more accurate early treatment decisions, reduces the risk of misjudgment, optimizes treatment outcomes, and provides an efficient and reliable intelligent auxiliary tool for clinical burn management.

[0081] Specifically, in this embodiment, the trained burn grading and classification prediction model can be deployed on a web page or a standalone device, making it convenient for medical staff to use in real time in various scenarios.

[0082] Example 2

[0083] Secondly, the present invention also provides a burn grading and classification prediction system, comprising:

[0084] The data acquisition module is used to construct multiple rat burn models according to different burn types and burn grades, collect burn data based on multiple rat burn models, and construct a burn model database. The burn model database includes burn images and clinical burn diagnostic factors corresponding to different burn types and burn grades.

[0085] The data processing module is used to process burn data in the burn model database to obtain machine learning datasets;

[0086] The pre-training module is used to filter the machine learning dataset using a triple feature data filtering method to obtain the model training dataset;

[0087] The training module is used to take the model training dataset as input parameters and input it into the random forest model for training to obtain a burn grading and classification prediction model.

[0088] The prediction module is used to deploy the trained burn grading and classification prediction model, perform burn grading and classification operations using the burn grading and classification prediction model, and obtain the corresponding burn grading and classification prediction results.

[0089] Example 3

[0090] Thirdly, the present invention also provides a burn grading and classification prediction device, including a data input unit, a data processing unit and an output unit;

[0091] The data input unit is used to acquire the burn analysis dataset to be detected and output it to the data processing unit.

[0092] The data processing unit is equipped with a burn grading and classification prediction model based on the burn grading and classification prediction method in Embodiment 1, and obtains the corresponding burn grading and classification prediction results based on the burn analysis dataset to be examined.

[0093] The output unit is used to output the corresponding burn grading and classification prediction results.

[0094] This embodiment provides a dedicated burn grading and classification prediction device, integrating a data input unit, a data processing unit, and an output unit. It achieves hardware-based encapsulation and convenient deployment of the burn grading and classification prediction model. The device can directly receive the dataset of burns to be analyzed and quickly analyze it using the built-in burn grading and classification prediction model, outputting the corresponding prediction results. The device's design facilitates integration into existing clinical workflows, providing medical personnel with immediate, on-site assistance in judgment without relying on complex remote computing resources. Its output unit clearly presents the prediction results, assisting doctors in making comprehensive diagnostic decisions rather than replacing doctors in direct diagnosis. The application of this device significantly improves the automation level and accessibility of burn assessment, and is particularly suitable for primary healthcare institutions or emergency scenarios, providing timely and reliable technical support for early intervention of burn patients.

[0095] Example 4

[0096] Fourthly, the present invention provides a storage device storing instructions generated based on the burn grading and classification prediction method described in Embodiment 1.

[0097] Specifically, the storage medium described in this invention may include random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, or CD-ROMs. It should be noted that those skilled in the art can select the form and type of storage medium according to actual production and usage needs; this embodiment does not impose further limitations.

[0098] It should be noted that those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and storage media described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

Claims

1. A method for predicting the grading and classification of burns, characterized in that, include: S1: Data Collection: Multiple rat burn models were constructed according to different burn types and burn grades. Burn data were collected based on the multiple rat burn models to construct a burn model database. The burn model database includes burn images and clinical burn diagnostic factors corresponding to different burn types and burn grades. S2: Data Processing: Process the burn data in the burn model database to obtain the image feature set and the clinical burn diagnostic factor dataset, and directly merge them according to the sample naming to obtain the machine learning dataset; among them, the clinical burn diagnostic factor dataset is obtained by quantifying the target protein of the clinical burn diagnostic factors using the ELISA method. S3: Pre-training: A triple feature data filtering method is used to filter the machine learning dataset to obtain the model training dataset; S4: Training the model: Use the model training dataset as input parameters to train the random forest model and obtain the burn grading and classification prediction model. S5: Deploy the model and perform burn grading and classification prediction: Deploy the trained burn grading and classification prediction model, use the burn grading and classification prediction model to perform burn grading and classification, and obtain the corresponding burn grading and classification prediction results.

2. The burn grading and classification prediction method according to claim 1, characterized in that, In step S1, the multiple rat burn models constructed are shallow second-degree burn, deep second-degree burn, third-degree burn, flame burn, and electric shock burn models constructed using multiple rats with the same growth conditions.

3. The burn grading and classification prediction method according to claim 2, characterized in that, Step S2 specifically includes: S2.1: Input the burn images corresponding to different burn types and burn grades into the VGG-11 model to extract features and obtain the image feature set; S2.2: For clinical burn diagnostic factors corresponding to different burn types and burn grades, the target protein was quantified using the ELISA method to obtain a clinical burn diagnostic factor dataset; S2.3: The image feature set obtained in step S2.1 and the clinical burn diagnostic factor dataset obtained in step S2.2 are directly merged into a machine learning dataset according to the sample naming.

4. The burn grading and classification prediction method according to claim 3, characterized in that, Specifically, step S2.1 involves inputting burn images corresponding to different burn types and burn grades into the VGG-11 model and extracting the output content of the second-to-last non-task layer as the image feature set.

5. The burn grading and classification prediction method according to claim 3, characterized in that, Step S3 includes: S3.1: The machine learning dataset is processed using three feature selection methods: Boruta, RFE, and Pearson, respectively, to obtain the first set, the second set, and the third set. S3.2: Take the intersection of the first set, the second set, and the third set as the model training dataset.

6. The burn grading and classification prediction method according to claim 5, characterized in that, In step S4, the model training dataset is divided into a training set and a validation set, and hyperparameter optimization is performed using a grid search method to obtain a burn grading and classification prediction model.

7. The burn grading and classification prediction method according to claim 6, characterized in that, Step S5 specifically includes: S5.1: Deploy a burn grading and classification prediction model; S5.2: Collect images of burns to be analyzed and clinical burn diagnostic factors, and obtain the dataset of burns to be analyzed by referring to the methods in steps S2.1 to S2.3; S5.3: Input the dataset of burns to be analyzed into the burn grading and classification prediction model to obtain the corresponding burn grading and classification prediction results.

8. A burn grading and classification prediction system, characterized in that, The burn grading and classification prediction system includes: The data acquisition module is used to construct multiple rat burn models according to different burn types and burn grades, collect burn data based on multiple rat burn models, and construct a burn model database. The burn model database includes burn images and clinical burn diagnostic factors corresponding to different burn types and burn grades. The data processing module is used to process burn data in the burn model database, obtain image feature sets and clinical burn diagnostic factor datasets, and directly merge them according to sample naming to obtain machine learning datasets; among them, the clinical burn diagnostic factor dataset is obtained by quantifying target proteins of clinical burn diagnostic factors using the ELISA method. The pre-training module is used to filter the machine learning dataset using a triple feature data filtering method to obtain the model training dataset; The training module is used to take the model training dataset as input parameters and input it into the random forest model for training to obtain a burn grading and classification prediction model. The prediction module is used to deploy the trained burn grading and classification prediction model, perform burn grading and classification operations using the burn grading and classification prediction model, and obtain the corresponding burn grading and classification prediction results.

9. A burn grading and classification prediction device, characterized in that, The burn grading and classification prediction device includes a data input unit, a data processing unit, and an output unit; The data input unit is used to acquire the burn analysis dataset to be detected and output it to the data processing unit. The data processing unit is equipped with a burn grading and classification prediction model obtained from the burn grading and classification prediction method according to any one of claims 1 to 7, and obtains the corresponding burn grading and classification prediction results based on the burn analysis dataset to be examined. The output unit is used to output the corresponding burn grading and classification prediction results.

10. A storage device, characterized in that, The storage device stores instructions generated based on the burn grading and classification prediction method according to any one of claims 1 to 7.

Images