18 Method for predicting the risk of physiological myocardial uptake of F-FDG PET / CT tumor imaging

Abstract

The application discloses a kind of 18 The application discloses a risk prediction method for myocardial physiological uptake of F-FDG PET / CT tumor imaging, candidate clinical data of malignant tumor patients who have undergone examination are collected, independent risk factors for myocardial physiological uptake are determined by using single factor and multi-factor logistic regression analysis, and a logistic regression prediction model and multiple machine learning prediction models are established. The ROC curve analysis is used to select the pending prediction model with the largest AUC value. Furthermore, the application evaluates whether the pending prediction model is the best prediction model through calibration curve analysis and decision curve analysis. The best prediction model can be used to predict whether a new patient has the risk of myocardial physiological uptake before the examination, which has the advantages of fast prediction speed, high accuracy and reliability.

Description

Technical Field

[0001] This invention relates to the field of nuclear medicine imaging technology, specifically a... 18 F-FDG PET / CT tumor imaging: a method for predicting the risk of myocardial physiological uptake. Background Technology

[0002] 18 F-FDG PET / CT, as a functional imaging method, is widely used in the diagnosis, staging, and efficacy evaluation of malignant tumors. However, in routine... 18 In F-FDG PET / CT tumor imaging, unpredictable changes often occur in the myocardium. 18 Physiological uptake of F-FDG. This uptake not only reduces the contrast between the target and non-target areas of the lesion, but also limits the diagnostic sensitivity of 18F-FDG PET imaging for vulnerable coronary plaques and infective endocarditis.

[0003] Current clinical practice primarily suppresses myocardial contrast by controlling blood glucose levels and prolonging fasting time. However, considering that malignant tumor lesions are highly metabolic tissues, they can directly affect the distribution of tracers throughout the body through "competitive inhibition (steal effect)".

[0004] Clinical prediction models offer the possibility of accurate prediction, but few studies have quantified tumor metabolic burden (lesion SUVmax) and combined it with clinical indicators to construct a specific prediction model for cancer patients. Summary of the Invention

[0005] The technical problem to be solved by the present invention is: to provide a 18 F-FDG PET / CT tumor imaging risk prediction method for myocardial physiological uptake, to address existing 18 The issue of unnecessary repeat examinations before F-FDG PET / CT tumor imaging is due to patients' inability to predict the risk of myocardial physiological uptake.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0007] A sort of 18 A method for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging, characterized by comprising:

[0008] Step S1: Based on the preset arrangement standards, from the... 18 Target patients were screened from patients with malignant tumors undergoing F-FDG PET / CT tumor imaging examination;

[0009] Step S2: Obtain the basic variable dataset, which includes candidate clinical data for each target patient, and the target patient's clinical data during the treatment process. 18 The assessment of whether myocardial physiological uptake occurs during F-FDG PET / CT tumor imaging; generally, the basic variable dataset is divided into training and test sets in a 7:3 ratio for subsequent model training, which will not be elaborated here.

[0010] Step S3: Perform univariate logistic regression analysis and multivariate logistic regression analysis (LR) on the aforementioned basic variable dataset to determine from the candidate clinical data: 18 Independent risk factors for myocardial physiological uptake in patients with malignant tumors during F-FDG PET / CT tumor imaging were identified; and a logistic regression prediction model was obtained to predict whether myocardial physiological uptake would occur based on the independent risk factors of the patients.

[0011] Step S4: Based on the aforementioned basic variable dataset, train the models using various machine learning algorithms to construct multiple machine learning prediction models that predict whether a patient will experience myocardial physiological uptake based on their independent risk factors.

[0012] Step S5: Calculate the AUC value (area under the ROC curve) of the logistic regression prediction model and each machine learning prediction model through ROC curve analysis to evaluate the model's discriminative power, and record the model with the largest AUC value as the undetermined prediction model.

[0013] Step S6: Evaluate the accuracy and clinical applicability of the undetermined prediction model through calibration curve analysis and decision curve analysis (DCA). If the preset accuracy requirements and clinical applicability requirements are met simultaneously, the undetermined prediction model is selected as the best prediction model; otherwise, repeat steps S1 to S6.

[0014] Step S7, Perform on new patients 18 Before F-FDG PET / CT tumor imaging, the optimal prediction model is used to predict whether there is a risk of myocardial physiological uptake.

[0015] Therefore, the present invention collects and processes... 18Candidate clinical data of malignant tumor patients obtained from F-FDG PET / CT tumor imaging were used to identify independent risk factors for myocardial physiological uptake using univariate and multivariate logistic regression analyses. Logistic regression prediction models and multiple machine learning prediction models were established, and the undetermined prediction model with the largest AUC value was selected through ROC curve analysis. Thus, on the one hand, the excellent linear prediction performance of logistic regression analysis was used to identify independent risk factors, and on the other hand, multiple machine learning algorithms were used to explore the deep nonlinear relationships between various independent risk factors, thereby overcoming the prediction bottleneck that may exist in the logistic regression prediction model (i.e., when the machine learning prediction model is an undetermined prediction model).

[0016] Furthermore, this invention evaluates and determines whether the undetermined prediction model is the optimal prediction model through calibration curve analysis and decision curve analysis, enabling it to be used on new patients. 18 Before F-FDG PET / CT tumor imaging, the best prediction model is used to predict whether there is a risk of myocardial physiological uptake, which has the advantages of fast prediction speed, high accuracy and high reliability.

[0017] Among them, the malignant tumor patients mentioned in step S1 are patients recorded in the hospital's medical record system;

[0018] The pre-defined inclusion and exclusion criteria include: Inclusion criterion one: age 18 years or older; Inclusion criterion two: complete clinical data; Exclusion criterion one: having diabetes; Exclusion criterion two: a clear history of heart disease; Exclusion criterion three: undergoing... 18 Pre-set criteria for F-FDG PET / CT tumor imaging examination: recent use of drugs affecting glucose metabolism;

[0019] The target patients are the remaining patients who meet all the inclusion criteria after excluding patients who meet any one of the exclusion criteria among the malignant tumor patients.

[0020] Step S3 includes:

[0021] Step S3-1: Using the aforementioned basic variable dataset, perform univariate logistic regression analysis on each candidate variable in the candidate clinical data, and select candidate variables with a P value below 0.05 as undetermined variables, indicating that they have a significant impact on whether myocardial physiological uptake occurs.

[0022] Step S3-2: Using the aforementioned basic variable dataset, perform multifactor logistic regression analysis on all the undetermined variables, and use the back-wald method to filter out the undetermined variables. 18 Independent risk factor for myocardial physiological uptake in patients with malignant tumors during F-FDG PET / CT tumor imaging;

[0023] Step S3-3: Based on the analysis results of step S3-2, construct the logistic regression prediction model.

[0024] The candidate clinical data mentioned in step S2 includes the following nine candidate variables: gender, age, height, weight, BMI, fasting blood glucose, injection dose, fasting time, and maximum standard uptake value (SUV) of the tumor lesion. max ); where gender is a categorical variable including both male and female, and fasting time is a categorical variable including both yes and no states, representing the duration of fasting. 18 For F-FDG PET / CT tumor imaging examinations, whether fasting for more than 16 hours was required was used. Categorical variables were expressed as frequency and percentage (%). Age, height, weight, BMI, fasting blood glucose, injection dose, and maximum standard uptake value (SUV) of the tumor lesion were also considered. max All of these are continuous variables, and continuous variables are expressed as mean ± standard deviation;

[0025] Experiments revealed that the undetermined variables in step S3-1 and the independent risk factors in step S3-2 were: fasting blood glucose, injection dose, fasting time, and maximum standard uptake value (SUV) of the tumor lesion. max ).

[0026] Therefore, this invention uses fasting blood glucose, injection dosage, fasting time variables, and the maximum standard uptake value (SUV) of the tumor lesion as the basis for its formulation. max As an independent risk factor for predicting the occurrence of myocardial physiological uptake, the maximum standard uptake value (SUV) of the tumor lesion was used. max The 'Tumor Sink Effect' (competitive uptake of tracers by hypermetabolic tumor lesions) was precisely quantified. For the first time, this complex nonlinear pathological mechanism was transformed into a computable mathematical weight, fundamentally correcting the bias of traditional prediction models. This greatly improved the specificity and clinical fit of myocardial uptake risk prediction and avoided the problems existing in current clinical prediction methods, namely: the limitation of prediction is due to the fact that the patient's basic physiological indicators are used as independent risk factors, ignoring the systematic perturbation of the global tracer distribution by malignant tumors as hypermetabolic entities.

[0027] The logistic regression prediction model described in step S3 is as follows:

[0028] RV=37.352 - 5.484×FBG - 0.663×Dose - 2.618×FT - 0.095×SUV max ;

[0029] In the formula, RV represents the risk value for myocardial physiological uptake; FBG is fasting blood glucose; Dose is the injection dose; FT is the fasting time variable, and FT is the time during which fasting occurs. 18 For F-FDG PET / CT tumor imaging, a value of 1 is assigned if fasting for more than 16 hours is required; otherwise, a value of 0 is assigned. max This represents the maximum standard uptake value for tumor lesions.

[0030] The various machine learning algorithms mentioned in step S4 include: Decision Tree (DT), K-Nearest Neighbors (KNN), Naive Bayes Model (NBM), Random Forest (RF), Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost).

[0031] Preferably, the risk prediction method further includes: continuously acquiring independent risk factors for new patients with malignant tumors and their status during the course of treatment. 18 The results of determining whether myocardial physiological uptake occurs during F-FDG PET / CT tumor imaging are used to train and iterate the optimal prediction model.

[0032] Preferably, the risk prediction method further includes:

[0033] Step S8, see Figure 2 If the best prediction model is a logistic regression prediction model, then its nomogram is generated to display the complex model prediction results in an intuitive way, making them easier to understand and apply.

[0034] See Figure 3 If the best prediction model is one of the machine learning prediction models, then the SHAP method is used to perform interpretability analysis on it to make up for the "black box" defect of machine learning algorithms, reveal the internal decision-making mechanism of the model, and make the model have the transparency of traditional statistics, providing quantitative reference for clinical research.

[0035] In this invention, the R software package (version 4.2.1) is preferably used to perform rapid statistical analysis of the data: the single-factor logistic regression analysis and multi-factor logistic regression analysis in step S3 use the glm {stats} package; the ROC curve analysis in step S5 uses the pROC package; the calibration curve analysis in step S6 uses the val.prob function and calibrate in the rms package; the decision curve analysis (DCA) in step S6 uses the rmda package; and the nomogram in step S8 uses the rms package.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] First, this invention involves collecting and processing... 18 Candidate clinical data of malignant tumor patients obtained from F-FDG PET / CT tumor imaging were used to identify independent risk factors for myocardial physiological uptake using univariate and multivariate logistic regression analyses. Logistic regression prediction models and multiple machine learning prediction models were established, and the undetermined prediction model with the largest AUC value was selected through ROC curve analysis. Thus, on the one hand, the excellent linear prediction performance of logistic regression analysis was used to identify independent risk factors, and on the other hand, multiple machine learning algorithms were used to explore the deep nonlinear relationships between various independent risk factors, thereby overcoming the prediction bottleneck that may exist in the logistic regression prediction model (i.e., when the machine learning prediction model is an undetermined prediction model).

[0038] Furthermore, this invention evaluates and determines whether the undetermined prediction model is the optimal prediction model through calibration curve analysis and decision curve analysis, enabling it to be used on new patients. 18 Before F-FDG PET / CT tumor imaging, the best prediction model is used to predict whether there is a risk of myocardial physiological uptake, which has the advantages of fast prediction speed, high accuracy and high reliability.

[0039] Second, this invention uses fasting blood glucose, injection dosage, fasting time variables, and the maximum standard uptake value (SUV) of tumor lesions as parameters. max As an independent risk factor for predicting the occurrence of myocardial physiological uptake, the maximum standard uptake value (SUV) of the tumor lesion was used. maxThe 'Tumor Sink Effect' (competitive uptake of tracers by hypermetabolic tumor lesions) was precisely quantified. For the first time, this complex nonlinear pathological mechanism was transformed into a computable mathematical weight, fundamentally correcting the bias of traditional prediction models. This greatly improved the specificity and clinical fit of myocardial uptake risk prediction and avoided the problems existing in current clinical prediction methods, namely: the limitation of prediction is due to the fact that the patient's basic physiological indicators are used as independent risk factors, ignoring the systematic perturbation of the global tracer distribution by malignant tumors as hypermetabolic entities. Attached Figure Description

[0040] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments:

[0041] Figure 1 This is a flowchart of the risk prediction method of the present invention;

[0042] Figure 2 Example of a Nomograph for the best prediction model;

[0043] Figure 3 This is an example of the interpretability analysis results of the SHAP method for the best prediction model;

[0044] Figure 4 The graph shows a comparison of ROC curves obtained from ROC curve analysis of seven models. The left and right graphs correspond to the training set and validation set, respectively.

[0045] Figure 5 The left and right plots represent the calibration curves for the undetermined prediction model, with the left and right plots corresponding to the training and validation sets in clinical settings, respectively.

[0046] Figure 6 The left and right plots represent the decision curves (DCA) of the undetermined prediction model, with the left and right plots corresponding to the training and validation sets, respectively. Detailed Implementation

[0047] The present invention will now be described in detail with reference to the embodiments and accompanying drawings to help those skilled in the art better understand the inventive concept of the present invention. However, the scope of protection of the claims of the present invention is not limited to the following embodiments. For those skilled in the art, all other embodiments obtained without creative effort without departing from the inventive concept of the present invention are within the scope of protection of the present invention.

[0048] like Figure 1 As shown, the present invention discloses a kind of 18 Methods for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging include:

[0049] Step S1: Based on the preset arrangement standards, from the... 18 Target patients were screened from patients with malignant tumors undergoing F-FDG PET / CT tumor imaging examination;

[0050] Among them, the malignant tumor patients mentioned in step S1 are patients recorded in the hospital's medical record system;

[0051] The pre-defined inclusion and exclusion criteria include: Inclusion criterion one: age 18 years or older; Inclusion criterion two: complete clinical data; Exclusion criterion one: having diabetes; Exclusion criterion two: a clear history of heart disease; Exclusion criterion three: undergoing... 18 Pre-set criteria for F-FDG PET / CT tumor imaging examination: recent use of drugs affecting glucose metabolism;

[0052] The target patients are the remaining patients who meet all the inclusion criteria after excluding patients who meet any one of the exclusion criteria among the malignant tumor patients.

[0053] Step S2: Obtain the basic variable dataset, which includes candidate clinical data for each target patient, and the target patient's clinical data during the treatment process. 18 The assessment of whether myocardial physiological uptake occurs during F-FDG PET / CT tumor imaging; generally, the basic variable dataset is divided into training and test sets in a 7:3 ratio for subsequent model training, which will not be elaborated here.

[0054] The candidate clinical data mentioned in step S2 includes the following nine candidate variables: gender, age, height, weight, BMI, fasting blood glucose, injection dose, fasting time, and maximum standard uptake value (SUV) of the tumor lesion. max ); where gender is a categorical variable including both male and female, and fasting time is a categorical variable including both yes and no states, representing the duration of fasting. 18 For F-FDG PET / CT tumor imaging examinations, whether fasting for more than 16 hours was required was used. Categorical variables were expressed as frequency and percentage (%). Age, height, weight, BMI, fasting blood glucose, injection dose, and maximum standard uptake value (SUV) of the tumor lesion were also considered. max All of them are continuous variables, and continuous variables are expressed as mean ± standard deviation.

[0055] Step S3: Perform univariate logistic regression analysis and multivariate logistic regression analysis (LR) on the aforementioned basic variable dataset to determine from the candidate clinical data: 18Independent risk factors for myocardial physiological uptake in patients with malignant tumors during F-FDG PET / CT tumor imaging were identified; and a logistic regression prediction model was obtained to predict whether myocardial physiological uptake would occur based on the independent risk factors of the patients.

[0056] Step S3 includes:

[0057] Step S3-1: Using the aforementioned basic variable dataset, perform univariate logistic regression analysis on each candidate variable in the candidate clinical data, and select candidate variables with a P value below 0.05 as undetermined variables, indicating that they have a significant impact on whether myocardial physiological uptake occurs.

[0058] Step S3-2: Using the aforementioned basic variable dataset, perform multifactor logistic regression analysis on all the undetermined variables, and use the back-wald method to filter out the undetermined variables. 18 Independent risk factor for myocardial physiological uptake in patients with malignant tumors during F-FDG PET / CT tumor imaging;

[0059] Step S3-3: Based on the analysis results of step S3-2, construct the logistic regression prediction model.

[0060] Experiments revealed that the undetermined variables in step S3-1 and the independent risk factors in step S3-2 were: fasting blood glucose, injection dose, fasting time, and maximum standard uptake value (SUV) of the tumor lesion. max ).

[0061] Therefore, this invention uses fasting blood glucose, injection dosage, fasting time variables, and the maximum standard uptake value (SUV) of the tumor lesion as the basis for its formulation. max As an independent risk factor for predicting the occurrence of myocardial physiological uptake, the maximum standard uptake value (SUV) of the tumor lesion was used. max The 'Tumor Sink Effect' (competitive uptake of tracers by hypermetabolic tumor lesions) was precisely quantified. For the first time, this complex nonlinear pathological mechanism was transformed into a computable mathematical weight, fundamentally correcting the bias of traditional prediction models. This greatly improved the specificity and clinical fit of myocardial uptake risk prediction and avoided the problems existing in current clinical prediction methods, namely: the limitation of prediction is due to the fact that the patient's basic physiological indicators are used as independent risk factors, ignoring the systematic perturbation of the global tracer distribution by malignant tumors as hypermetabolic entities.

[0062] The experimental results of steps S3-1 and S3-2 are shown in Table 1 below. The results show the variables of fasting blood glucose, injection dose, fasting time, and maximum standard uptake value (SUV) of the tumor lesion. max All of them are18 An independent risk factor for myocardial physiological uptake in patients with malignant tumors in F-FDG PET / CT tumor imaging.

[0063] Table 1

[0064]

[0065] The logistic regression prediction model described in step S3 is as follows:

[0066] RV=37.352 - 5.484×FBG - 0.663×Dose - 2.618×FT - 0.095×SUV max ;

[0067] In the formula, RV represents the risk value for myocardial physiological uptake; FBG is fasting blood glucose; Dose is the injection dose; FT is the fasting time variable, and FT is the time during which fasting occurs. 18 For F-FDG PET / CT tumor imaging, a value of 1 is assigned if fasting for more than 16 hours is required; otherwise, a value of 0 is assigned. max This represents the maximum standard uptake value for tumor lesions.

[0068] Step S4: Based on the aforementioned basic variable dataset, train the models using various machine learning algorithms to construct multiple machine learning prediction models that predict whether a patient will experience myocardial physiological uptake based on their independent risk factors.

[0069] The various machine learning algorithms mentioned in step S4 include: Decision Tree (DT), K-Nearest Neighbors (KNN), Naive Bayes Model (NBM), Random Forest (RF), Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost).

[0070] Step S5: Calculate the AUC value (area under the ROC curve) of the logistic regression prediction model and each machine learning prediction model through ROC curve analysis to evaluate the model's discriminative power, and record the model with the largest AUC value as the undetermined prediction model.

[0071] Step S6: Evaluate the accuracy and clinical applicability of the undetermined prediction model through calibration curve analysis and decision curve analysis (DCA). If the preset accuracy requirements and clinical applicability requirements are met simultaneously, the undetermined prediction model is selected as the best prediction model; otherwise, repeat steps S1 to S6.

[0072] Step S7, Perform on new patients 18 Before F-FDG PET / CT tumor imaging, the optimal prediction model is used to predict whether there is a risk of myocardial physiological uptake.

[0073] Therefore, the present invention collects and processes... 18 Candidate clinical data of malignant tumor patients obtained from F-FDG PET / CT tumor imaging were used to identify independent risk factors for myocardial physiological uptake using univariate and multivariate logistic regression analyses. Logistic regression prediction models and multiple machine learning prediction models were established, and the undetermined prediction model with the largest AUC value was selected through ROC curve analysis. Thus, on the one hand, the excellent linear prediction performance of logistic regression analysis was used to identify independent risk factors, and on the other hand, multiple machine learning algorithms were used to explore the deep nonlinear relationships between various independent risk factors, thereby overcoming the prediction bottleneck that may exist in the logistic regression prediction model (i.e., when the machine learning prediction model is an undetermined prediction model).

[0074] Furthermore, this invention evaluates and determines whether the undetermined prediction model is the optimal prediction model through calibration curve analysis and decision curve analysis, enabling it to be used on new patients. 18 Before F-FDG PET / CT tumor imaging, the best prediction model is used to predict whether there is a risk of myocardial physiological uptake, which has the advantages of fast prediction speed, high accuracy and high reliability.

[0075] The above are the basic embodiments of the present invention, and further optimizations, improvements and limitations can be made based on these basic embodiments:

[0076] Preferably, the risk prediction method further includes: continuously acquiring independent risk factors for new patients with malignant tumors and their status during the course of treatment. 18 The results of determining whether myocardial physiological uptake occurs during F-FDG PET / CT tumor imaging are used to train and iterate the optimal prediction model.

[0077] Preferably, the risk prediction method further includes:

[0078] Step S8, see Figure 2If the best prediction model is a logistic regression prediction model, then its nomogram is generated to display the complex model prediction results in an intuitive way, making them easier to understand and apply.

[0079] See Figure 3 If the best prediction model is one of the machine learning prediction models, then the SHAP method is used to perform interpretability analysis on it to make up for the "black box" defect of machine learning algorithms, reveal the internal decision-making mechanism of the model, and make the model have the transparency of traditional statistics, providing quantitative reference for clinical research.

[0080] In this invention, the R software package (version 4.2.1) is preferably used to perform rapid statistical analysis of the data: the single-factor logistic regression analysis and multi-factor logistic regression analysis in step S3 use the glm {stats} package; the ROC curve analysis in step S5 uses the pROC package; the calibration curve analysis in step S6 uses the val.prob function and calibrate in the rms package; the decision curve analysis (DCA) in step S6 uses the rmda package; and the nomogram in step S8 uses the rms package.

[0081] The performance of the best prediction model obtained in this invention is evaluated below using example data:

[0082] like Figure 4 The figure shows the comparison of ROC curves obtained from step S3 (logistic regression prediction model) and step S4 (six machine learning prediction models, corresponding to six machine learning algorithms: decision tree, K-nearest neighbors, Naive Bayes model, random forest, support vector machine, and extreme gradient boosting). As can be seen from the figure, the model with the largest AUC value is the model obtained by extreme gradient boosting (XGBoost), which is denoted as the undetermined prediction model, with an AUC value of 0.963 (95% CI: 0.921~1.000).

[0083] like Figure 5 As shown in the figure above, the consistency between the calibration curve and the ideal curve is good, indicating that the model has good accuracy and meets the preset accuracy requirements.

[0084] like Figure 6 As shown in the figure, the decision curve (DCA) of the above-mentioned undetermined prediction model is as follows. It can be seen from the figure that the curve is higher than the reference line in the threshold range of 0 to 1, indicating that the net benefit value of the model is high. This suggests that the nomotu prediction model has good clinical effectiveness and meets the preset clinical applicability requirements. Therefore, the above-mentioned undetermined prediction model is the best prediction model.

[0085] This invention is not limited to the specific embodiments described above. Based on the above content and in accordance with common technical knowledge and conventional methods in the field, without departing from the basic technical concept of this invention, this invention can also make other equivalent modifications, substitutions or alterations, all of which fall within the protection scope of this invention.

Claims

1. A kind 18 A method for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging, characterized in that... include: Step S1: Based on preset sorting and arrangement standards, from the... 18 Target patients were identified from patients with malignant tumors examined by F-FDG PET / CT tumor imaging. Step S2: Obtain the basic variable dataset, which includes candidate clinical data for each target patient, and the target patient's clinical data during the treatment process. 18 The result of determining whether myocardial physiological uptake occurs during F-FDG PET / CT tumor imaging; Step S3: Perform univariate logistic regression analysis and multivariate logistic regression analysis using the aforementioned basic variable dataset to determine from the candidate clinical data: 18 Independent risk factor for myocardial physiological uptake in patients with malignant tumors during F-FDG PET / CT tumor imaging; Furthermore, a logistic regression prediction model was obtained to predict whether myocardial physiological uptake would occur based on the patient's independent risk factors. Step S4: Based on the aforementioned basic variable dataset, train the models using various machine learning algorithms to construct multiple machine learning prediction models that predict whether a patient will experience myocardial physiological uptake based on their independent risk factors. Step S5: Calculate the AUC value of the logistic regression prediction model and each machine learning prediction model through ROC curve analysis, and record the model with the largest AUC value as the undetermined prediction model. Step S6: Evaluate the accuracy and clinical applicability of the undetermined prediction model through calibration curve analysis and decision curve analysis respectively. If the preset accuracy requirements and clinical applicability requirements are met at the same time, the undetermined prediction model is taken as the best prediction model. Otherwise, repeat steps S1 to S6; Step S7, Perform on new patients 18 Before F-FDG PET / CT tumor imaging, the optimal prediction model is used to predict whether there is a risk of myocardial physiological uptake.

2. As described in claim 1 18 A method for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging, characterized by: The malignant tumor patients mentioned in step S1 are those recorded in the hospital's medical record system; The pre-defined inclusion and exclusion criteria include: Inclusion criterion one: age 18 years or older; Inclusion criterion two: complete clinical data; Exclusion criterion one: having diabetes; Exclusion criterion two: a clear history of heart disease; Exclusion criterion three: undergoing... 18 Pre-set criteria for F-FDG PET / CT tumor imaging examination: recent use of drugs affecting glucose metabolism; The target patients are the remaining patients who meet all the inclusion criteria after excluding patients who meet any one of the exclusion criteria among the malignant tumor patients.

3. As described in claim 1 18 A method for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging, characterized by: Step S3 includes: Step S3-1: Using the aforementioned basic variable dataset, perform univariate logistic regression analysis on each candidate variable in the candidate clinical data, and select candidate variables with a P-value below 0.05 as undetermined variables; Step S3-2: Using the aforementioned basic variable dataset, perform multifactor logistic regression analysis on all the undetermined variables, and use the backward stepwise method to filter out the undetermined variables. 18 Independent risk factor for myocardial physiological uptake in patients with malignant tumors during F-FDG PET / CT tumor imaging; Step S3-3: Based on the analysis results of step S3-2, construct the logistic regression prediction model.

4. As described in claim 3 18 A method for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging, characterized by: The candidate clinical data mentioned in step S2 includes the following nine candidate variables: gender, age, height, weight, BMI, fasting blood glucose, injection dose, fasting time, and maximum standard uptake value of tumor lesions; among them, gender is a categorical variable including male and female, and fasting time is a categorical variable including whether or not the fasting period is active. 18 Is fasting for more than 16 hours required before an F-FDG PET / CT tumor imaging examination? The undetermined variables in step S3-1 and the independent risk factors in step S3-2 are: fasting blood glucose, injection dose, fasting time variable, and maximum standard uptake value of tumor lesions.

5. As described in claim 1 18 A method for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging, characterized by: The logistic regression prediction model mentioned in step S3 is: RV=37.352 - 5.484×FBG - 0.663×Dose - 2.618×FT - 0.095×SUV max ; In the formula, RV represents the risk value for myocardial physiological uptake; FBG is fasting blood glucose; Dose is the injection dose; FT is the fasting time variable, and FT is the time during which fasting occurs. 18 For F-FDG PET / CT tumor imaging, a value of 1 is assigned if fasting for more than 16 hours is required; otherwise, a value of 0 is assigned. max This represents the maximum standard uptake value for tumor lesions.

6. As described in claim 1 18 A method for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging, characterized by: The various machine learning algorithms mentioned in step S4 include: decision tree, K-nearest neighbor, Naive Bayes model, random forest, support vector machine, and extreme gradient boosting.

7. As described in claim 1 18 A method for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging, characterized by: The risk prediction method also includes: continuously acquiring independent risk factors for new patients with malignant tumors and their status during treatment. 18 The results of determining whether myocardial physiological uptake occurs during F-FDG PET / CT tumor imaging are used to train and iterate the optimal prediction model.

8. As described in claim 1 18 A method for predicting the risk of myocardial physiological uptake in F-FDG PET / CT tumor imaging, characterized by: The risk prediction method also includes: Step S8: If the optimal prediction model is a logistic regression prediction model, then generate its Nomograph; If the best prediction model is one of the machine learning prediction models, then its interpretability is analyzed using the SHAP method.

Images