Fatty liver related model construction method and device based on nuclear magnetic resonance imaging data, equipment and medium

Abstract

The invention discloses a fatty liver related model construction method and device based on nuclear magnetic resonance imaging data, equipment and a medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining initial image data; the initial image data comprises an initial proton density fat fraction image and an initial transverse relaxation rate image; extracting a first initial feature of the initial proton density fat fraction image and a second initial feature of the initial transverse relaxation rate image; determining a numerical label of the initial image data; determining a target feature based on the first initial feature and the second initial feature; respectively constructing a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver efficacy evaluation model based on a machine learning algorithm, the numerical labels and the target features; and outputting an evaluation result of the to-be-detected object by using the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver efficacy evaluation model. According to the invention, accurate and effective evaluation of the fatty liver can be realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and medium for constructing a fatty liver-related model based on magnetic resonance imaging data. Background Art

[0002] In recent decades, the global prevalence of metabolically associated fatty liver disease (MASLD) has rapidly increased, closely associated with obesity and type 2 diabetes. Although liver biopsy is considered the gold standard for assessing histological features, it is limited by invasiveness and potential observer error. Non-invasive methods, such as magnetic resonance elastography and magnetic resonance-derived proton density fat fraction (PDFF), have become valuable tools for assessing liver characteristics. PDFF quantifies fat deposition and content by analyzing proton density ratios without the need for additional equipment and is widely used in clinical research to diagnose and evaluate the efficacy of MASLD treatment. In addition to fat deposition, impaired iron metabolism is also implicated in the development and progression of fatty liver disease, with iron deficiency and iron overload associated with disease severity and progression. It is considered a reliable indicator for evaluating liver iron content and can be The current assessment of the severity of fatty liver disease is mostly based on single-modality imaging data. However, using one assessment method in subjects with different causes and mechanisms of the disease cannot distinguish between fatty liver subtypes, lacks a comprehensive assessment of the progression and severity of fatty liver disease, and may lead to a one-sided evaluation of the efficacy of drug intervention. Measuring the fat and iron content of the liver at the same time can more comprehensively reflect the severity of fatty liver disease and facilitate the multi-dimensional assessment of severity and drug efficacy, but the simple sum of measurements cannot represent and predict the true severity of fatty liver disease.

[0003] In summary, how to achieve accurate and effective assessment of fatty liver is a technical problem that needs to be solved urgently. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for constructing a fatty liver-related model based on MRI data, which can achieve accurate and effective assessment of fatty liver. The specific scheme is as follows:

[0005] In a first aspect, the present application provides a method for constructing a fatty liver-related model based on magnetic resonance imaging data, which is applied to a computer device and comprises:

[0006] Acquiring initial image data of the target object based on a preset nuclear magnetic resonance imaging technique; the initial image data includes an initial proton density fat fraction image and an initial transverse relaxation rate image;

[0007] extracting a first initial feature of the initial proton density fat fraction image and a second initial feature of the initial transverse relaxation rate image;

[0008] Determining target numerical labels corresponding to the initial image data; the target numerical labels include target fatty liver subclass numerical labels, target fatty liver severity numerical labels, and target fatty liver efficacy evaluation numerical labels;

[0009] Determining a target feature based on a preset feature engineering method, the first initial feature, and the second initial feature;

[0010] Based on the preset machine learning algorithm, the target numerical label and the target feature, a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver pharmacodynamic evaluation model are respectively constructed, so that the target fatty liver classification model is used to output the classification result of the fatty liver subclass of the subject to be detected, the target fatty liver severity scoring model is used to output the fatty liver severity scoring result of the subject to be detected, and the target fatty liver pharmacodynamic evaluation model is used to output the fatty liver pharmacodynamic evaluation result of the subject to be detected.

[0011] Optionally, acquiring initial image data of the target object based on a preset magnetic resonance imaging technology includes:

[0012] Scanning the target object based on a preset three-dimensional gradient echo sequence, a preset region growing algorithm, a preset scan trigger mechanism, and preset scan parameters to obtain a target amplitude map and a target phase map; wherein the preset scan parameters include a preset field of view angle, a preset acquisition matrix, a preset acquisition bandwidth, a preset repetition time, a preset flip angle, a preset voxel, a preset echo time, and a preset number of scan slices;

[0013] The target amplitude map and the target phase map are fitted according to a preset echo condition and a preset fitting method to obtain the initial proton density fat fraction image and the initial transverse relaxation rate image of the target object.

[0014] Optionally, extracting the first initial feature of the initial proton density fat fraction image and the second initial feature of the initial transverse relaxivity image includes:

[0015] determining a target liver layer in the target object according to a first preset number condition, and determining a target region from the target liver layer according to a second preset number condition and a preset layer region area condition;

[0016] determining the initial proton density fat fraction image corresponding to the target area as a target proton density fat fraction image, and determining the initial transverse relaxation rate image corresponding to the target area as a target transverse relaxation rate image;

[0017] The number of targets corresponding to the target area is determined, and based on the number of targets, the target proton density fat fraction images are averaged to obtain the first initial feature, and based on the number of targets, the target transverse relaxation rate images are averaged to obtain the second initial feature.

[0018] Optionally, determining a target numerical label corresponding to the initial image data includes:

[0019] determining, based on the initial image data, a target fatty liver subclass of the target subject, a first classification rating corresponding to a target fatty liver severity, and a second classification rating corresponding to a target fatty liver drug efficacy evaluation;

[0020] The target fatty liver subclass is converted into a target fatty liver subclass numerical label, and the first classification rating corresponding to the target fatty liver severity is converted into a target fatty liver severity numerical label, and the second classification rating corresponding to the target fatty liver efficacy evaluation is converted into a target fatty liver efficacy evaluation numerical label.

[0021] Optionally, determining the target feature based on a preset feature engineering method, the first initial feature, and the second initial feature includes:

[0022] performing data preprocessing on the first initial features and the second initial features based on a preset outlier removal method, a preset missing value removal method, and a preset standardization processing method, respectively, and generating a third initial feature based on the preset feature engineering method, the first initial features after data preprocessing, and the second initial features after data preprocessing;

[0023] The target feature is determined from the third initial feature according to a preset minimum absolute value shrinkage and selection operator method, or a preset recursive feature elimination method.

[0024] Optionally, the target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver efficacy evaluation model are constructed based on a preset machine learning algorithm, the target numerical label, and the target feature, respectively, including:

[0025] constructing an initial fatty liver classification model based on a preset classification algorithm, the target fatty liver subclass numerical label, and the target feature;

[0026] constructing an initial fatty liver severity scoring model based on a preset regression algorithm, the target fatty liver severity numerical label, and the target feature;

[0027] Based on a preset logistic regression algorithm or a preset neural network model, and using the target fatty liver pharmacodynamic evaluation numerical label and the target feature, an initial fatty liver pharmacodynamic evaluation model is constructed;

[0028] The initial fatty liver classification model, the initial fatty liver severity scoring model and the initial fatty liver drug efficacy evaluation model are optimized to obtain the corresponding target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver drug efficacy evaluation model.

[0029] Optionally, optimizing the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver pharmacodynamic evaluation model to obtain the corresponding target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver pharmacodynamic evaluation model includes:

[0030] Dividing the target features into a target training set, a target validation set, and a target test set based on preset division conditions, and respectively evaluating the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver efficacy evaluation model based on a preset cross-validation method and the target training set to obtain respective evaluation results;

[0031] According to the preset hyperparameter tuning method, the evaluation results, the preset cross-validation method, the target training set and the target validation set, the hyperparameters of the initial fatty liver classification model, the initial fatty liver severity scoring model and the initial fatty liver pharmacodynamic evaluation model are adjusted respectively to obtain the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver pharmacodynamic evaluation model, and the target test set is used to test the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver pharmacodynamic evaluation model respectively.

[0032] In a second aspect, the present application provides a device for constructing a fatty liver-related model based on magnetic resonance imaging data, which is applied to a computer device and includes:

[0033] An image data acquisition module is used to acquire initial image data of the target object based on a preset nuclear magnetic resonance imaging technology; the initial image data includes an initial proton density fat fraction image and an initial transverse relaxation rate image;

[0034] a feature extraction module, configured to extract a first initial feature of the initial proton density fat fraction image and a second initial feature of the initial transverse relaxivity image;

[0035] a target numerical label determination module, configured to determine a target numerical label corresponding to the initial image data; the target numerical label comprising a target fatty liver subclass numerical label, a target fatty liver severity numerical label, and a target fatty liver efficacy evaluation numerical label;

[0036] a target feature determination module, configured to determine a target feature based on a preset feature engineering method, the first initial feature, and the second initial feature;

[0037] A model construction module is used to construct a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver pharmacodynamic evaluation model based on a preset machine learning algorithm, the target numerical label and the target feature, so as to output the classification result of the fatty liver subclass of the subject to be detected using the target fatty liver classification model, output the fatty liver severity scoring result of the subject to be detected using the target fatty liver severity scoring model, and output the fatty liver pharmacodynamic evaluation result of the subject to be detected using the target fatty liver pharmacodynamic evaluation model.

[0038] In a third aspect, the present application provides an electronic device, comprising:

[0039] Memory, used to store computer programs;

[0040] A processor is used to execute the computer program to implement the aforementioned method for constructing a fatty liver-related model based on magnetic resonance imaging data.

[0041] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the aforementioned method for constructing a fatty liver-related model based on magnetic resonance imaging data is implemented.

[0042] In the present application, first, the initial image data of the target object is acquired based on the preset nuclear magnetic resonance imaging technology; the initial image data includes an initial proton density fat fraction image and an initial transverse relaxivity image; then, the first initial feature of the initial proton density fat fraction image and the second initial feature of the initial transverse relaxivity image are extracted; then, the target numerical label corresponding to the initial image data is determined; the target numerical label includes a target fatty liver subclass numerical label, a target fatty liver severity numerical label and a target fatty liver pharmacodynamic evaluation numerical label; then, the target feature is determined based on the preset feature engineering method, the first initial feature and the second initial feature; finally, based on the preset machine learning algorithm, the target numerical label and the target feature, a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver pharmacodynamic evaluation model are respectively constructed, so as to use the target fatty liver classification model to output the classification result of the fatty liver subclass of the object to be detected, use the target fatty liver severity scoring model to output the fatty liver severity scoring result of the object to be detected, and use the target fatty liver pharmacodynamic evaluation model to output the fatty liver pharmacodynamic evaluation result of the object to be detected. As can be seen from the above, in the present application, the PDFF image and Image, and extract PDFF image and The features of the image, then based on the feature engineering method, the features of the PDFF image and The features of the image are used to determine the target features, and the machine learning method, the target features and the target numerical labels are used to respectively determine the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver efficacy evaluation model, so as to evaluate fatty liver using the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver efficacy evaluation model. In this way, the present application combines PDFF and Two quantitative parameters make full use of complementary information to improve the accuracy and robustness of the model; the target fatty liver classification model is used to realize the automatic classification of fatty liver subclasses, avoiding the subjectivity of traditional methods that rely on manual experience; the target fatty liver severity scoring model is used to quantitatively score the severity of fatty liver, providing a more accurate diagnostic basis for the clinic; the target fatty liver efficacy evaluation model is used to evaluate the effect of drug intervention in advance, providing a scientific basis for personalized treatment. In this way, this application can achieve accurate and effective evaluation of fatty liver, and can provide intelligent tools for early diagnosis, disease monitoring and efficacy evaluation of fatty liver, which has important clinical application value and social and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] 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 merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0044] Figure 1 A flowchart of a method for constructing a fatty liver-related model based on magnetic resonance imaging data provided in this application;

[0045] Figure 2 A schematic diagram of a specific random forest algorithm formula provided for this application;

[0046] Figure 3 A schematic diagram of a specific support vector machine algorithm formula provided in this application;

[0047] Figure 4 A schematic diagram of a specific XGBoost algorithm formula provided for this application;

[0048] Figure 5 A schematic diagram of a specific linear regression algorithm formula provided in this application;

[0049] Figure 6 A schematic diagram of a specific support vector regression algorithm formula provided in this application;

[0050] Figure 7 A schematic diagram of a specific gradient boosting regression tree algorithm formula provided in this application;

[0051] Figure 8 A schematic diagram of a specific logistic regression formula provided for this application;

[0052] Figure 9 A schematic diagram of a specific neural network formula provided for this application;

[0053] Figure 10 A schematic diagram of a specific fatty liver mouse model, MRI acquisition device, and original image provided in this application;

[0054] Figure 11 A specific PDFF image provided for this application and image;

[0055] Figure 12 This is a schematic diagram of the structure of a device for constructing a fatty liver-related model based on magnetic resonance imaging data provided in this application;

[0056] Figure 13 This is a structural diagram of an electronic device provided in this application. DETAILED DESCRIPTION

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] In recent decades, the global prevalence of metabolically associated fatty liver disease (MASLD) has rapidly increased, closely associated with obesity and type 2 diabetes. Although liver biopsy is considered the gold standard for assessing histological features, it is limited by invasiveness and potential observer error. Non-invasive methods, such as magnetic resonance elastography and magnetic resonance-derived proton density fat fraction (PDFF), have become valuable tools for assessing liver characteristics. PDFF quantifies fat deposition and content by analyzing proton density ratios without the need for additional equipment and is widely used in clinical research to diagnose and evaluate the efficacy of MASLD treatment. In addition to fat deposition, impaired iron metabolism is also implicated in the development and progression of fatty liver disease, with iron deficiency and iron overload associated with disease severity and progression. It is considered a reliable indicator for evaluating liver iron content and can be to reflect the changes in liver iron content. Currently, the assessment of the severity of fatty liver is mostly based on single-modality imaging data, but using one assessment method in subjects with different causes and mechanisms of the disease cannot distinguish between fatty liver subtypes, lacks a comprehensive assessment of the progression and severity of fatty liver, and may be a one-sided evaluation of the efficacy of drug intervention. Measuring the fat content and iron content of the liver at the same time can more comprehensively reflect the severity of fatty liver, which is conducive to multi-dimensional assessment of severity and drug efficacy, but a simple sum of measurements cannot represent and predict the true severity of fatty liver. To this end, the present application provides a scheme for constructing a fatty liver-related model based on magnetic resonance imaging data, which can achieve accurate and effective assessment of fatty liver.

[0059] See also Figure 1 As shown, an embodiment of the present invention discloses a method for constructing a fatty liver-related model based on magnetic resonance imaging data, which is applied to a computer device and may include:

[0060] Step S11 : acquiring initial image data of the target object based on a preset nuclear magnetic resonance imaging technique; the initial image data includes an initial proton density fat fraction image and an initial transverse relaxation rate image.

[0061] In this embodiment, the target object can be scanned based on a three-dimensional gradient echo sequence, such as a FACT sequence (Fat Analysis Calculation Technique), a region-growing algorithm, a preset scan trigger mechanism, and preset scan parameters to obtain a target amplitude map and a target phase map; wherein the preset scan parameters include a preset field of view, a preset acquisition matrix, a preset acquisition bandwidth, a preset repetition time, a preset flip angle, a preset voxel, a preset echo time, and a preset number of scan layers. In a specific embodiment, the preset scan trigger mechanism can be set to be triggered based on the respiration of the target object, and the preset scan parameters can be set to FOV (Field of View) = 30×30 ; Acquisition matrix = 144 × 220; Acquisition bandwidth = 1390 Hz / pixel; Repetition time = 45 ms; Flip angle = ; voxel = 0.21 × 0.21 × 1 TE (Echo Time) = 1.01, 1.96, 2.91, 3.86, 4.81, 5.76ms; scan 8 slices. Then, the target amplitude map and the target phase map can be fitted according to the preset echo conditions and the preset fitting method to obtain the initial proton density fat fraction image and the initial transverse relaxation rate image of the target object. Specifically, the initial PDFF and initial transverse relaxation rate image of the target object can be obtained by fitting the amplitude map and phase map under each echo. Quantitative graph.

[0062] Step S12: extracting a first initial feature of the initial proton density fat fraction image and a second initial feature of the initial transverse relaxivity image.

[0063] In this embodiment, first, the target liver layer in the target object can be determined according to the first preset number condition, and the target area can be determined from the target liver layer according to the second preset number condition and the preset layer area condition; then, the initial proton density fat fraction image corresponding to the target area is determined as the target proton density fat fraction image, and the initial transverse relaxation rate image corresponding to the target area is determined as the target transverse relaxation rate image; finally, the target number corresponding to the target area is determined, and the target proton density fat fraction image is averaged based on the target number to obtain the first initial feature, and the target transverse relaxation rate image is averaged based on the target number to obtain the second initial feature. In a specific embodiment, 3 layers can be selected from the liver of the target object, and then 3 0.3 layers can be drawn from the liver cross section of each layer to avoid liver blood vessels, stomach and other organs. The target area, and finally take the average value of the 9 areas as the initial PDFF and initial The quantitative value of the initial PDFF is obtained by The second initial feature of .

[0064] Step S13: determining a target numerical label corresponding to the initial image data; the target numerical label includes a target fatty liver subclass numerical label, a target fatty liver severity numerical label, and a target fatty liver efficacy evaluation numerical label.

[0065] In this embodiment, the target fatty liver subclass, the first classification rating corresponding to the target fatty liver severity, and the second classification rating corresponding to the target fatty liver efficacy evaluation of the target subject can first be determined based on the initial image data; then, the target fatty liver subclass is converted into a target fatty liver subclass numerical label, and the first classification rating corresponding to the target fatty liver severity is converted into a target fatty liver severity numerical label, and the second classification rating corresponding to the target fatty liver efficacy evaluation is converted into a target fatty liver efficacy evaluation numerical label. Specifically, fatty liver subclasses include simple fatty liver, non-alcoholic steatohepatitis, etc., the classification rating corresponding to the fatty liver severity can include mild fatty liver, moderate fatty liver, and severe fatty liver, and the classification rating corresponding to the fatty liver efficacy evaluation can include markedly effective, effective, and ineffective.

[0066] Step S14: determining target features based on a preset feature engineering method, the first initial features, and the second initial features.

[0067] In this embodiment, the first initial feature and the initial PDFF can be respectively processed based on the preset outlier removal method, the preset missing value removal method and the preset standardization processing method. The first initial feature of the initial PDFF is preprocessed, and a third initial feature is generated based on the preset feature engineering method, the first initial feature after data preprocessing, and the second initial feature after data preprocessing; then the target feature is determined from the third initial feature according to the preset minimum absolute value shrinkage and selection operator method, or the preset recursive feature elimination method. Specifically, the first initial feature and The initial second initial feature is preprocessed to remove outliers and missing values and perform normalization. Then the third initial feature can be generated by combining feature engineering, for example ratio, and then LASSO regression (Least Absolute Shrinkage and Selection Operator) or recursive feature elimination can be used to select the target feature from the third initial feature.

[0068] Step S15: Based on the preset machine learning algorithm, the target numerical label and the target feature, a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver pharmacodynamic evaluation model are respectively constructed, so as to output the classification result of the fatty liver subclass of the subject to be detected by using the target fatty liver classification model, output the fatty liver severity scoring result of the subject to be detected by using the target fatty liver severity scoring model, and output the fatty liver pharmacodynamic evaluation result of the subject to be detected by using the target fatty liver pharmacodynamic evaluation model.

[0069] It should be noted that this embodiment can construct an initial fatty liver classification model based on a preset classification algorithm, target fatty liver subclass numerical labels and target features. Figure 2 、 Figure 3 and Figure 4 As shown, specifically, a classification algorithm such as random forest, support vector machine, or XGBoost (eXtreme Gradient Boosting, i.e., extreme gradient boosting algorithm) can be used to construct an initial fatty liver classification model. An initial fatty liver severity scoring model can be constructed based on a preset regression algorithm, a target fatty liver severity numerical label, and target features. Figure 5 、 Figure 6 and Figure 7As shown, specifically, a regression algorithm such as linear regression, support vector regression (SVR), or gradient boosting regression tree (GBRT) can be used to construct an initial fatty liver severity scoring model. The initial fatty liver efficacy evaluation model can be constructed based on a preset logistic regression algorithm or a preset neural network model, and using the target fatty liver efficacy evaluation numerical label and target features. See Figure 8 and Figure 9 Specifically, a logistic regression or neural network model can be used to construct an initial fatty liver disease efficacy evaluation model. Finally, the initial fatty liver disease classification model, the initial fatty liver disease severity scoring model, and the initial fatty liver disease efficacy evaluation model can be optimized to obtain a corresponding target fatty liver disease classification model, target fatty liver disease severity scoring model, and target fatty liver disease efficacy evaluation model.

[0070] In this embodiment, the above-mentioned optimization of the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver pharmacodynamic evaluation model to obtain the corresponding target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver pharmacodynamic evaluation model may include: first, dividing the target feature into a target training set, a target validation set, and a target test set based on a preset division condition, and evaluating the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver pharmacodynamic evaluation model based on a preset cross-validation method and the target training set to obtain respective evaluation results; then, adjusting the hyperparameters of the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver pharmacodynamic evaluation model according to a preset hyperparameter tuning method, the respective evaluation results, the preset cross-validation method, the target training set, and the target validation set to obtain the target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver pharmacodynamic evaluation model, and using the target test set to test the target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver pharmacodynamic evaluation model. Specifically, the training set usually accounts for 60%-80% of the total data set and is used to train the model, allowing the model to learn the features and patterns in the data to adjust the model's hyperparameters, such as weights. The validation set usually accounts for 10%-20% of the total data set and is used to evaluate the performance of the model during model training, adjust and select the model's hyperparameters, and prevent the model from overfitting. The test set usually accounts for 10%-20% of the total data set and is used to finally evaluate the generalization ability of the model after model training and tuning are completed. In a specific embodiment, the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver efficacy evaluation model can be evaluated based on the k-fold cross-validation method and the training set. Specifically, the training set is first further divided into k subsets of similar size, and each subset keeps the data distribution characteristics as similar as possible to the original training set. Then, k rounds of training and validation are performed: in the i-th round, the i-th subset is used as the validation set, and the remaining k-1 subsets are combined as new training sets for training the model. The trained model is then used to perform predictions on the validation set, and corresponding performance metrics, such as precision, recall, and F1 value, are calculated. This continues until k rounds of training and validation are completed, resulting in k performance metrics. Statistics, such as the mean and standard deviation, are calculated for these k performance metrics to generate the corresponding evaluation results for the initial fatty liver classification model, initial fatty liver severity scoring model, and initial fatty liver efficacy assessment model.The hyperparameters of the model can then be adjusted based on the evaluation results corresponding to each model through grid search or Bayesian optimization. During the adjustment process, the performance of the model is evaluated based on the k-fold cross-validation method, the training set, and the validation set. The hyperparameter combination with the optimal objective function is then used as the final hyperparameter of the model to obtain the target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver efficacy evaluation model. Finally, the target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver efficacy evaluation model can be tested separately using the test set. Afterwards, the model can be applied to new data, and the target fatty liver classification model can be used to output the classification results of the fatty liver subclasses of the subject to be tested, the target fatty liver severity scoring model can be used to output the fatty liver severity scoring results of the subject to be tested, and the target fatty liver efficacy evaluation model can be used to output the fatty liver efficacy evaluation results of the subject to be tested.

[0071] As can be seen from the above, in this embodiment, the initial image data of the target object is first acquired based on the preset nuclear magnetic resonance imaging technology; the initial image data includes an initial proton density fat fraction image and an initial transverse relaxivity image; then the first initial feature of the initial proton density fat fraction image and the second initial feature of the initial transverse relaxivity image are extracted; then the target numerical label corresponding to the initial image data is determined; the target numerical label includes a target fatty liver subclass numerical label, a target fatty liver severity numerical label and a target fatty liver pharmacodynamic evaluation numerical label; then the target feature is determined based on the preset feature engineering method, the first initial feature and the second initial feature; finally, based on the preset machine learning algorithm, the target numerical label and the target feature, a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver pharmacodynamic evaluation model are respectively constructed, so as to use the target fatty liver classification model to output the classification result of the fatty liver subclass of the object to be detected, use the target fatty liver severity scoring model to output the fatty liver severity scoring result of the object to be detected, and use the target fatty liver pharmacodynamic evaluation model to output the fatty liver pharmacodynamic evaluation result of the object to be detected. As can be seen from the above, in this embodiment, the PDFF image and Image, and extract PDFF image and The features of the image, then based on the feature engineering method, the features of the PDFF image and The features of the image are used to determine the target features, and the machine learning method, the target features and the target numerical labels are used to respectively determine the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver efficacy evaluation model, so as to evaluate fatty liver using the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver efficacy evaluation model. In this way, in this embodiment, the PDFF and The two quantitative parameters fully utilize complementary information to improve the accuracy and robustness of the model. The target fatty liver classification model is used to automatically classify fatty liver subtypes, avoiding the subjectivity of traditional methods that rely on manual experience. The target fatty liver severity scoring model is used to quantitatively score the severity of fatty liver disease, providing a more accurate diagnostic basis for clinicians. The target fatty liver efficacy evaluation model is used to pre-evaluate the effects of drug interventions, providing a scientific basis for personalized treatment. In this way, this embodiment can achieve accurate and effective assessment of fatty liver disease and provide intelligent tools for early diagnosis, disease monitoring, and efficacy evaluation of fatty liver disease, with important clinical application value and socioeconomic benefits.

[0072] In one embodiment, a multi-parameter imaging method based on ultra-high field magnetic resonance imaging can be used to simultaneously obtain PDFF and The technical solution in this application is described in conjunction with different mouse model fatty liver image databases. Specifically, the following steps are included:

[0073] (1) Selection of fatty liver mouse model: 8-week-old mice with C57BL / 6J genetic background were selected and fed with 60% high-fat diet for 16 weeks, so that the liver of the mice showed fatty degeneration. Alternatively, 12-week-old ob mice and db mice with leptin gene deficiency were selected.

[0074] (2) Use drugs to treat fatty liver for 4 weeks: The fatty liver mice were divided into the fatty liver group + PBS group, the fatty liver group + drug 1 treatment group, and the fatty liver group + drug 2 treatment group. At the same time, age- and sex-matched normal diet PBS-treated mice were set up and subcutaneously injected with equal doses of drugs (15 nmol / kg concentration) and PBS (Phosphate-Buffered Saline) once every 3 days for 4 weeks.

[0075] (3) MRI data were collected from mice after 4 weeks of treatment, see Figure 10 As shown, a 42mm inner diameter transceiver coil was used for data acquisition on the mouse abdomen. To reduce respiratory motion artifacts and improve the signal-to-noise ratio, a respiratory-triggered 3D gradient echo sequence, namely the FACT sequence, was used. The FACT sequence integrates a region growing algorithm based on region growing, synchronizes MRI acquisition with the respiratory cycle, and triggers data acquisition during stable respiratory phases. After the positioning phase acquisition and 3D field shimming, a 3D gradient echo sequence scan was performed: FOV = 30×30 ; Acquisition matrix = 144 × 220; Acquisition bandwidth = 1390 Hz / pixel; Repetition time = 45 ms; Flip angle = ; voxel = 0.21 × 0.21 × 1 ;TE=1.01, 1.96, 2.91, 3.86, 4.81, 5.76ms; scan 8 layers. Figure 11 As shown in Figure 2, by fitting the amplitude diagram and phase diagram under each echo, the PDFF and Quantitative graph. After scanning, three slices were selected for each mouse liver, and three 0.3 The average value of the 9 regions was taken as the first feature and The second characteristic.

[0076] (4) After the scan, the mice were euthanized and liver tissues were collected. Some of the tissues were used to detect the content of triglycerides and iron in the liver, and some were used for tissue sectioning for hematoxylin-eosin staining, oil red O staining, enhanced Prussian blue staining, and Masson staining. The hematoxylin-eosin-stained liver tissue sections were scored for fatty degeneration, the oil red O-stained lipid droplet area was statistically analyzed, and the score results and lipid droplet area were recorded. The Masson staining was used to score fibrosis, and the severity of fatty liver was scored in multiple dimensions based on the histological results.

[0077] (5) For the slices and clinical information collected from different models and patients, the fatty liver subclasses, such as simple fatty liver, non-alcoholic steatohepatitis, etc., the classification ratings corresponding to the severity of fatty liver, and the classification ratings corresponding to the fatty liver efficacy evaluation are converted into numerical labels, and the numerical labels of fatty liver subclasses, fatty liver severity, and fatty liver efficacy evaluation are obtained.

[0078] (6) The first characteristic of PDFF and The second feature of is preprocessed to remove outliers and missing values, and is standardized, and new features are generated by combining feature engineering, such as ratio and use LASSO regression or recursive feature elimination to select important features.

[0079] (7) Construct a model for fatty liver subclass classification: Use classification algorithms such as random forest, support vector machine or XGBoost to construct a fatty liver classification model to predict fatty liver subclasses, such as simple fatty liver, non-alcoholic steatohepatitis, etc.; Predict the severity of fatty liver: Use regression algorithms such as linear regression, support vector regression or gradient boosting regression tree to construct a fatty liver severity scoring model to predict the severity of fatty liver; Drug efficacy evaluation: Use logistic regression or neural network model to construct a fatty liver efficacy evaluation model to predict the efficacy after drug intervention.

[0080] (8) Model training and optimization: The multi-parameter imaging dataset obtained under intervention conditions is divided into a training set, a validation set, and a test set. Cross-validation, such as k-fold cross-validation, is used to evaluate the performance of each model. The hyperparameters of the model are adjusted through grid search or Bayesian optimization to improve the prediction accuracy, and the corresponding target fatty liver classification model, target fatty liver severity scoring model, and target fatty liver efficacy evaluation model are obtained.

[0081] (9) Model application: The trained target fatty liver classification model, target fatty liver severity scoring model and target fatty liver efficacy evaluation model are applied to new data to output the classification results of fatty liver subclasses of the subjects to be tested, the fatty liver severity scoring results of the subjects to be tested and the fatty liver efficacy evaluation results of the subjects to be tested.

[0082] Accordingly, see Figure 12 As shown, the embodiment of the present application further provides a device for constructing a fatty liver-related model based on magnetic resonance imaging data, which is applied to a computer device and may include:

[0083] An image data acquisition module 11 is configured to acquire initial image data of a target object based on a preset nuclear magnetic resonance imaging technique; the initial image data includes an initial proton density fat fraction image and an initial transverse relaxivity image;

[0084] a feature extraction module 12, configured to extract a first initial feature of the initial proton density fat fraction image and a second initial feature of the initial transverse relaxivity image;

[0085] A target numerical label determination module 13 is configured to determine a target numerical label corresponding to the initial image data; the target numerical label includes a target fatty liver subclass numerical label, a target fatty liver severity numerical label, and a target fatty liver efficacy evaluation numerical label;

[0086] a target feature determination module 14, configured to determine a target feature based on a preset feature engineering method, the first initial feature, and the second initial feature;

[0087] The model construction module 15 is used to construct a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver pharmacodynamic evaluation model based on a preset machine learning algorithm, the target numerical label and the target feature, so as to output the classification result of the fatty liver subclass of the subject to be detected using the target fatty liver classification model, output the fatty liver severity scoring result of the subject to be detected using the target fatty liver severity scoring model, and output the fatty liver pharmacodynamic evaluation result of the subject to be detected using the target fatty liver pharmacodynamic evaluation model.

[0088] As can be seen from the above, in this application, the initial image data of the target object is first obtained based on the preset nuclear magnetic resonance imaging technology; the initial image data includes an initial proton density fat fraction image and an initial transverse relaxivity image; then the first initial feature of the initial proton density fat fraction image and the second initial feature of the initial transverse relaxivity image are extracted; then the target numerical label corresponding to the initial image data is determined; the target numerical label includes a target fatty liver subclass numerical label, a target fatty liver severity numerical label and a target fatty liver pharmacodynamic evaluation numerical label; then the target feature is determined based on the preset feature engineering method, the first initial feature and the second initial feature; finally, based on the preset machine learning algorithm, the target numerical label and the target feature, a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver pharmacodynamic evaluation model are respectively constructed, so as to use the target fatty liver classification model to output the classification result of the fatty liver subclass of the object to be detected, use the target fatty liver severity scoring model to output the fatty liver severity scoring result of the object to be detected, and use the target fatty liver pharmacodynamic evaluation model to output the fatty liver pharmacodynamic evaluation result of the object to be detected. As can be seen from the above, in this application, the PDFF image and Image, and extract PDFF image and The features of the image, then based on the feature engineering method, the features of the PDFF image and The features of the image are used to determine the target features, and the machine learning method, the target features and the target numerical labels are used to respectively determine the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver efficacy evaluation model, so as to evaluate fatty liver using the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver efficacy evaluation model. In this way, the present application combines PDFF and Two quantitative parameters make full use of complementary information to improve the accuracy and robustness of the model; the target fatty liver classification model is used to realize the automatic classification of fatty liver subclasses, avoiding the subjectivity of traditional methods that rely on manual experience; the target fatty liver severity scoring model is used to quantitatively score the severity of fatty liver, providing a more accurate diagnostic basis for the clinic; the target fatty liver efficacy evaluation model is used to evaluate the effect of drug intervention in advance, providing a scientific basis for personalized treatment. In this way, this application can achieve accurate and effective evaluation of fatty liver, and can provide intelligent tools for early diagnosis, disease monitoring and efficacy evaluation of fatty liver, which has important clinical application value and social and economic benefits.

[0089] In some specific implementations, the image data acquisition module 11 may include:

[0090] a target amplitude map acquisition unit, configured to scan the target object based on a preset three-dimensional gradient echo sequence, a preset region growing algorithm, a preset scan trigger mechanism, and preset scan parameters to obtain a target amplitude map and a target phase map; wherein the preset scan parameters include a preset field of view angle, a preset acquisition matrix, a preset acquisition bandwidth, a preset repetition time, a preset flip angle, a preset voxel, a preset echo time, and a preset number of scan slices;

[0091] An initial image data acquisition unit is used to fit the target amplitude map and the target phase map according to a preset echo condition and a preset fitting method to obtain the initial proton density fat fraction image and the initial transverse relaxation rate image of the target object.

[0092] In some specific implementations, the feature extraction module 12 may include:

[0093] a target region determining unit, configured to determine a target liver layer in the target object according to a first preset number condition, and determine a target region from the target liver layer according to a second preset number condition and a preset layer region area condition;

[0094] a target image data determining unit, configured to determine the initial proton density fat fraction image corresponding to the target area as a target proton density fat fraction image, and to determine the initial transverse relaxation rate image corresponding to the target area as a target transverse relaxation rate image;

[0095] a feature extraction unit, configured to determine the number of targets corresponding to the target area, and perform averaging processing on the target proton density fat fraction image based on the number of targets to obtain the first initial feature, and perform averaging processing on the target transverse relaxation rate image based on the number of targets to obtain the second initial feature.

[0096] In some specific implementations, the target value label determination module 13 may include:

[0097] a target fatty liver subclass determination unit, configured to determine a target fatty liver subclass of the target subject, a first classification rating corresponding to a target fatty liver severity, and a second classification rating corresponding to a target fatty liver drug efficacy evaluation based on the initial image data;

[0098] A target numerical label determination unit is used to convert the target fatty liver subclass into a target fatty liver subclass numerical label, convert the first classification rating corresponding to the target fatty liver severity into a target fatty liver severity numerical label, and convert the second classification rating corresponding to the target fatty liver efficacy evaluation into a target fatty liver efficacy evaluation numerical label.

[0099] In some specific implementations, the target feature determination module 14 may include:

[0100] a third initial feature generating unit, configured to perform data preprocessing on the first initial feature and the second initial feature based on a preset outlier removal method, a preset missing value removal method, and a preset standardization processing method, respectively, and generate a third initial feature based on the preset feature engineering method, the first initial feature after data preprocessing, and the second initial feature after data preprocessing;

[0101] A target feature determination unit is configured to determine the target feature from the third initial feature according to a preset minimum absolute value shrinkage and selection operator method, or a preset recursive feature elimination method.

[0102] In some specific implementations, the model building module 15 may include:

[0103] An initial fatty liver classification model construction submodule is used to construct an initial fatty liver classification model based on a preset classification algorithm, the target fatty liver subclass numerical label and the target feature;

[0104] An initial fatty liver severity scoring model construction submodule is used to construct an initial fatty liver severity scoring model based on a preset regression algorithm, the target fatty liver severity numerical label and the target feature;

[0105] An initial fatty liver pharmacodynamic evaluation model construction submodule is used to construct an initial fatty liver pharmacodynamic evaluation model based on a preset logistic regression algorithm or a preset neural network model and using the target fatty liver pharmacodynamic evaluation numerical label and the target feature;

[0106] The model optimization submodule is used to optimize the initial fatty liver classification model, the initial fatty liver severity scoring model and the initial fatty liver pharmacodynamic evaluation model to obtain the corresponding target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver pharmacodynamic evaluation model.

[0107] In some specific implementations, the model optimization submodule may include:

[0108] a model evaluation unit, configured to divide the target features into a target training set, a target validation set, and a target test set based on a preset division condition, and respectively evaluate the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver efficacy evaluation model based on a preset cross-validation method and the target training set to obtain respective evaluation results;

[0109] A target model determination unit is used to adjust the hyperparameters of the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver pharmacodynamic evaluation model according to a preset hyperparameter tuning method, the evaluation results, the preset cross-validation method, the target training set, and the target validation set, to obtain the target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver pharmacodynamic evaluation model, and use the target test set to test the target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver pharmacodynamic evaluation model.

[0110] Furthermore, the embodiment of the present application also discloses an electronic device, Figure 13 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram should not be considered as any limitation on the scope of use of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the method for constructing a fatty liver-related model based on magnetic resonance imaging data disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0111] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device. The communication protocol it follows is any communication protocol that can be applied to the technical solution of this application and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world. Its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0112] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or CD, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0113] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, and can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of implementing the method for constructing a fatty liver-related model based on magnetic resonance imaging data, which is disclosed in any of the aforementioned embodiments and executed by the electronic device 20, the computer program 222 may further include a computer program capable of performing other specific tasks.

[0114] Furthermore, this application discloses a computer-readable storage medium for storing a computer program; when executed by a processor, the computer program implements the aforementioned method for constructing a fatty liver-related model based on magnetic resonance imaging data. The specific steps of this method can be found in the corresponding content disclosed in the aforementioned embodiments and will not be further described here.

[0115] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Reference can be made to the descriptions of the identical or similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions of the methods.

[0116] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0117] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0118] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0119] The above is a detailed introduction to the technical solution provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for those skilled in the art, according to the ideas of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for constructing a fatty liver-related model based on magnetic resonance imaging data, characterized in that: Applicable to computer devices, including: Acquiring initial image data of the target object based on a preset nuclear magnetic resonance imaging technique; the initial image data includes an initial proton density fat fraction image and an initial transverse relaxation rate image; extracting a first initial feature of the initial proton density fat fraction image and a second initial feature of the initial transverse relaxation rate image; Determining target numerical labels corresponding to the initial image data; the target numerical labels include target fatty liver subclass numerical labels, target fatty liver severity numerical labels, and target fatty liver efficacy evaluation numerical labels; Determining a target feature based on a preset feature engineering method, the first initial feature, and the second initial feature; Based on the preset machine learning algorithm, the target numerical label and the target feature, a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver pharmacodynamic evaluation model are respectively constructed, so that the target fatty liver classification model is used to output the classification result of the fatty liver subclass of the subject to be detected, the target fatty liver severity scoring model is used to output the fatty liver severity scoring result of the subject to be detected, and the target fatty liver pharmacodynamic evaluation model is used to output the fatty liver pharmacodynamic evaluation result of the subject to be detected.

2. The method for constructing a fatty liver-related model based on magnetic resonance imaging data according to claim 1, characterized in that: The method of obtaining initial image data of the target object based on a preset magnetic resonance imaging technique includes: Scanning the target object based on a preset three-dimensional gradient echo sequence, a preset region growing algorithm, a preset scan trigger mechanism, and preset scan parameters to obtain a target amplitude map and a target phase map; wherein the preset scan parameters include a preset field of view angle, a preset acquisition matrix, a preset acquisition bandwidth, a preset repetition time, a preset flip angle, a preset voxel, a preset echo time, and a preset number of scan slices; The target amplitude map and the target phase map are fitted according to a preset echo condition and a preset fitting method to obtain the initial proton density fat fraction image and the initial transverse relaxation rate image of the target object.

3. The method for constructing a fatty liver-related model based on magnetic resonance imaging data according to claim 1, characterized in that: The extracting the first initial feature of the initial proton density fat fraction image and the second initial feature of the initial transverse relaxation rate image includes: determining a target liver layer in the target object according to a first preset number condition, and determining a target region from the target liver layer according to a second preset number condition and a preset layer region area condition; determining the initial proton density fat fraction image corresponding to the target area as a target proton density fat fraction image, and determining the initial transverse relaxation rate image corresponding to the target area as a target transverse relaxation rate image; The number of targets corresponding to the target area is determined, and based on the number of targets, the target proton density fat fraction images are averaged to obtain the first initial feature, and based on the number of targets, the target transverse relaxation rate images are averaged to obtain the second initial feature.

4. The method for constructing a fatty liver-related model based on magnetic resonance imaging data according to claim 1, characterized in that: The determining of the target numerical label corresponding to the initial image data includes: determining, based on the initial image data, a target fatty liver subclass of the target subject, a first classification rating corresponding to a target fatty liver severity, and a second classification rating corresponding to a target fatty liver drug efficacy evaluation; The target fatty liver subclass is converted into a target fatty liver subclass numerical label, and the first classification rating corresponding to the target fatty liver severity is converted into a target fatty liver severity numerical label, and the second classification rating corresponding to the target fatty liver efficacy evaluation is converted into a target fatty liver efficacy evaluation numerical label.

5. The method for constructing a fatty liver-related model based on magnetic resonance imaging data according to claim 1, characterized in that: The determining of the target feature based on the preset feature engineering method, the first initial feature, and the second initial feature includes: performing data preprocessing on the first initial features and the second initial features based on a preset outlier removal method, a preset missing value removal method, and a preset standardization processing method, respectively, and generating a third initial feature based on the preset feature engineering method, the first initial features after data preprocessing, and the second initial features after data preprocessing; The target feature is determined from the third initial feature according to a preset minimum absolute value shrinkage and selection operator method, or a preset recursive feature elimination method.

6. The method for constructing a fatty liver-related model based on magnetic resonance imaging data according to any one of claims 1 to 5, characterized in that: The target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver efficacy evaluation model are constructed based on the preset machine learning algorithm, the target numerical label, and the target feature, respectively, including: constructing an initial fatty liver classification model based on a preset classification algorithm, the target fatty liver subclass numerical label, and the target feature; constructing an initial fatty liver severity scoring model based on a preset regression algorithm, the target fatty liver severity numerical label, and the target feature; Based on a preset logistic regression algorithm or a preset neural network model, and using the target fatty liver pharmacodynamic evaluation numerical label and the target feature, an initial fatty liver pharmacodynamic evaluation model is constructed; The initial fatty liver classification model, the initial fatty liver severity scoring model and the initial fatty liver drug efficacy evaluation model are optimized to obtain the corresponding target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver drug efficacy evaluation model.

7. The method for constructing a fatty liver-related model based on magnetic resonance imaging data according to claim 6, characterized in that: The optimizing the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver pharmacodynamic evaluation model to obtain the corresponding target fatty liver classification model, the target fatty liver severity scoring model, and the target fatty liver pharmacodynamic evaluation model includes: Dividing the target features into a target training set, a target validation set, and a target test set based on preset division conditions, and respectively evaluating the initial fatty liver classification model, the initial fatty liver severity scoring model, and the initial fatty liver efficacy evaluation model based on a preset cross-validation method and the target training set to obtain respective evaluation results; According to the preset hyperparameter tuning method, the evaluation results, the preset cross-validation method, the target training set and the target validation set, the hyperparameters of the initial fatty liver classification model, the initial fatty liver severity scoring model and the initial fatty liver pharmacodynamic evaluation model are adjusted respectively to obtain the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver pharmacodynamic evaluation model, and the target test set is used to test the target fatty liver classification model, the target fatty liver severity scoring model and the target fatty liver pharmacodynamic evaluation model respectively.

8. A device for constructing a fatty liver-related model based on magnetic resonance imaging data, characterized in that: Applicable to computer devices, including: An image data acquisition module is used to acquire initial image data of the target object based on a preset nuclear magnetic resonance imaging technology; the initial image data includes an initial proton density fat fraction image and an initial transverse relaxation rate image; a feature extraction module, configured to extract a first initial feature of the initial proton density fat fraction image and a second initial feature of the initial transverse relaxivity image; a target numerical label determination module, configured to determine a target numerical label corresponding to the initial image data; the target numerical label comprising a target fatty liver subclass numerical label, a target fatty liver severity numerical label, and a target fatty liver efficacy evaluation numerical label; a target feature determination module, configured to determine a target feature based on a preset feature engineering method, the first initial feature, and the second initial feature; A model construction module is used to construct a target fatty liver classification model, a target fatty liver severity scoring model and a target fatty liver pharmacodynamic evaluation model based on a preset machine learning algorithm, the target numerical label and the target feature, so as to output the classification result of the fatty liver subclass of the subject to be detected using the target fatty liver classification model, output the fatty liver severity scoring result of the subject to be detected using the target fatty liver severity scoring model, and output the fatty liver pharmacodynamic evaluation result of the subject to be detected using the target fatty liver pharmacodynamic evaluation model.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory; wherein the memory is used to store a computer program, and the computer program is loaded and executed by the processor to implement the method for constructing a fatty liver-related model based on magnetic resonance imaging data as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that Used to store a computer program, which, when executed by a processor, implements the method for constructing a fatty liver-related model based on magnetic resonance imaging data as described in any one of claims 1 to 7.

Images