Early warning method and system for pheochromocytoma and paraganglioma

Abstract

The invention provides an early warning method and system for pheochromocytoma and paraganglioma, and is applied to the technical field of data processing. The method comprises the following steps: processing clinical parameters, biochemical marker data and CT image feature data in a training sample set in combination with a Logistic regression analysis method to generate an initial neuroendocrine tumor early warning model; processing the initial neuroendocrine tumor early warning model based on K-fold cross validation and the validation sample set to generate a target neuroendocrine tumor early warning model; processing the biochemical marker data based on 24-hour urine catecholamine pairs in clinical parameters to generate standardized biochemical detection index information; processing the CT image feature data to generate imaging comprehensive score information; and processing the standardized biochemical detection index information and the imaging comprehensive score information based on the target neuroendocrine tumor early warning model to generate early warning information of pheochromocytoma and paraganglioma.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a pheochromocytoma and paraganglioma early warning method and system. BACKGROUND

[0002] Traditional early warning for PPGLs has many limitations, and existing early warning model construction methods have limitations. On the one hand, when training the model, a single analysis method is often used, such as simple statistical analysis, which is difficult to deeply mine the complex nonlinear relationship between data. Taking the Logistic regression analysis method as an example, if the input features cannot be reasonably selected and processed, overfitting or underfitting problems are likely to occur, affecting the prediction accuracy of the model on the probability of PPGLs occurrence, tumor properties and influence range. On the other hand, in the model verification link, there is a lack of effective cross-validation and external data verification mechanism, making it difficult to fully guarantee the stability and generalization ability of the model, and the model cannot be accurately applied in different patient groups and medical environments. Due to insufficient in-depth and comprehensive analysis of data, the risk assessment value cannot accurately reflect the true risk degree of patients suffering from PPGLs. When comparing the risk assessment value with the standard range, there is a lack of effective strategy to dynamically adjust the model parameters, making it difficult to accurately warn according to the individual differences of patients. In addition, the process of converting the early warning bias correction vector into intuitive early warning information is not perfect, which is not conducive to the formulation of early intervention decisions.

[0003] The existing PPGLs early warning technology has poor adaptability to primary medical care. Primary hospitals usually lack high-end detection equipment and professional technical personnel, making it difficult to obtain high-quality clinical parameter, biochemical marker data and CT image feature data. The existing early warning method has high requirements for data quality and relies on complex equipment and professional analysis software, making it difficult for primary hospitals to apply these technologies, and unable to provide effective early warning services for patients in a timely manner, further exacerbating the problem of uneven distribution of medical resources.

[0004] It should be noted that the information disclosed in the above background section is only used to strengthen the understanding of the background of the present disclosure, and therefore can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide a pheochromocytoma and paraganglioma early warning method and system, at least to some extent to overcome the problems existing in the prior art, by collecting patient clinical, biochemical and CT image data to construct a training sample set. After data cleaning, feature extraction and normalization, combined with Logistic regression analysis method to generate an initial early warning model, and then using K-fold cross-validation and validation sample set to optimize the target model. On data processing, compared with the measurement value and the normal upper limit to generate standardized biochemical indicators, and analyzing CT images to obtain tumor information to obtain an imaging comprehensive score. Based on the target model, integrate the two types of information to obtain a risk assessment value, compare the standard range to adjust the model parameters to generate early warning information, integrate multiple parameters and verify, improve the diagnostic accuracy, reduce the dependence on high-end equipment, reduce the medical burden of patients, and provide strong support for clinical decision-making.

[0006] Other characteristics and advantages of the present application will become apparent from the following detailed description, or will be learned by practice of the application.

[0007] According to one aspect of the present application, an early warning method for pheochromocytoma and paraganglioma is provided, comprising: obtaining clinical parameters, biochemical marker data, CT image feature data of pheochromocytoma and paraganglioma patients, and a training sample set, wherein the clinical parameters include 24-hour urinary catecholamines, the biochemical marker data includes methoxyepinephrine, norepinephrine and methoxytyramine in blood; based on the clinical parameters, biochemical marker data and CT image feature data in the training sample set, combined with Logistic regression analysis method, an initial neuroendocrine tumor early warning model is generated; based on K-fold cross-validation and validation sample set, the initial neuroendocrine tumor early warning model is processed to generate a target neuroendocrine tumor early warning model; based on the 24-hour urinary catecholamines in the clinical parameters, the biochemical marker data is processed to generate standardized biochemical detection index information; the CT image feature data is processed to generate imaging comprehensive score information; based on the target neuroendocrine tumor early warning model, the standardized biochemical detection index information and the imaging comprehensive score information are processed to generate early warning information for pheochromocytoma and paraganglioma.

[0008] In another aspect of the present application, a pheochromocytoma and paraganglioma early warning device is characterized in that it comprises: an acquisition module for acquiring clinical parameters, biochemical marker data, CT image feature data, and a training sample set of a pheochromocytoma and paraganglioma patient, wherein the clinical parameters include 24-hour urinary catecholamines, and the biochemical marker data includes methoxyepinephrine, norepinephrine, and methoxytyramine in blood; a processing module for processing the clinical parameters, the biochemical marker data, and the CT image feature data in the training sample set based on a Logistic regression analysis method to generate an initial neuroendocrine tumor early warning model; processing the initial neuroendocrine tumor early warning model based on K-fold cross-validation and a verification sample set to generate a target neuroendocrine tumor early warning model; processing the biochemical marker data based on the 24-hour urinary catecholamines in the clinical parameters to generate standardized biochemical detection index information; processing the CT image feature data to generate imaging comprehensive score information; and processing the standardized biochemical detection index information and the imaging comprehensive score information based on the target neuroendocrine tumor early warning model to generate early warning information of the pheochromocytoma and paraganglioma.

[0009] According to still another aspect of the present application, an electronic device is characterized in that it comprises: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the executable instructions to implement the above-mentioned pheochromocytoma and paraganglioma early warning method.

[0010] According to still another aspect of the present application, a computer-readable storage medium is provided, which stores a computer program, and the computer program is executed by a second processor to implement the above-mentioned pheochromocytoma and paraganglioma early warning method.

[0011] The pheochromocytoma and paraganglioma early warning method and system provided by the present application constructs a training sample set by collecting clinical, biochemical, and CT image data of a patient. After data cleaning, feature extraction, and normalization, an initial early warning model is generated in combination with a Logistic regression analysis method, and a target model is obtained by using K-fold cross-validation and a verification sample set. In terms of data processing, standardized biochemical indicators are generated by comparing measured values with normal upper limits, and imaging comprehensive scores are obtained by analyzing CT images to acquire multiple information of a tumor. Based on the target model, two types of information are integrated to obtain a risk assessment value, and early warning information is generated by comparing standard ranges to adjust model parameters. The integration of multiple parameters and verification improves the accuracy of diagnosis, reduces the dependence on high-end equipment, reduces the medical burden of patients, and provides strong support for clinical decision-making.

[0012] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 A flow chart of a pheochromocytoma and paraganglioma early warning method provided by an embodiment of the present application is shown.

[0014] Figure 2 A structural schematic diagram of a pheochromocytoma and paraganglioma early warning device provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0015] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings, which should be understood as merely illustrative and explanatory, and not limiting the present application.

[0016] The pheochromocytoma and paraganglioma early warning method according to the exemplary embodiments of the present application is described below in conjunction with Figure 1 It should be noted that the following application scenarios are merely shown for the purpose of facilitating the understanding of the spirit and principles of the present application, and the embodiments of the present application are not limited in this respect. On the contrary, the embodiments of the present application are applicable to any applicable scenario.

[0017] In one embodiment, the present application further provides a pheochromocytoma and paraganglioma early warning method and system. Figure 1 A flow schematic diagram of a pheochromocytoma and paraganglioma early warning method according to an embodiment of the present application is schematically shown. As shown in Figure 1 The method is applied to a server and includes the following steps.

[0018] S101, obtaining clinical parameters, biochemical marker data, CT image feature data, and a training sample set of a pheochromocytoma and paraganglioma patient.

[0019] In one embodiment, for a PPGLs patient, the clinical parameter collection covers multiple aspects of information. Through consulting medical records, communicating with the patient, etc., the patient's age is recorded, and the age information can assist in judging the probability of disease occurrence. The incidence risk of different age groups is different. The gender is recorded because the gender may be different in the pathogenesis and performance of PPGLs. The clinical symptoms such as paroxysmal or persistent hypertension, headache, palpitation, and sweating are recorded in detail. These symptoms are important external manifestations of the disease. Blood pressure and heart rate are measured and recorded. PPGLs patients often have abnormal fluctuations in blood pressure and heart rate, and the numerical changes are of great significance to diagnosis. 24-hour urine catecholamine (24huCAs) is collected, which is a key indicator reflecting catecholamine metabolism in the body. The 24-hour collection method can more comprehensively and accurately evaluate the metabolic level. These clinical parameters reflect the patient's physical condition from different angles and provide a basis for subsequent diagnosis.

[0020] The biochemical marker data mainly focuses on methoxyepinephrine, norepinephrine and methoxytyramine. Using professional blood testing equipment and methods, the patient's blood sample is collected, the serum is separated, and then the content of these biochemical markers in the blood is measured by using precise detection technologies such as high performance liquid chromatography or enzyme-linked immunosorbent assay. These substances, as the metabolites of catecholamines, their content changes are closely related to the occurrence and development of PPGLs, and can provide important biochemical basis for diagnosis and reveal disease information from the molecular level.

[0021] The patient is examined by CT scanning equipment, and the scanning range needs to cover the areas where tumors may exist, such as adrenal glands and surrounding tissues, retroperitoneal and other parts. During the scanning process, appropriate scanning parameters such as slice thickness, slice spacing, voltage, current, etc. are adjusted to obtain high-quality CT images. Then, professional imaging doctors or trained technicians analyze the images and extract key image features such as tumor diameter, location, enhancement grade, necrosis characteristics, etc. Tumor diameter reflects the size of the tumor and is an important indicator for judging the development of the tumor; location information helps to determine the source of the tumor and the possible range of invasion; enhancement grade and necrosis characteristics can reflect the blood supply and internal structure of the tumor, which is of great significance for judging the nature of the tumor.

[0022] From the hospital's case database, select patients diagnosed with PPGLs, and select a certain number of patients who have been ruled out as PPGLs as controls. Collect the clinical parameters, biochemical marker data and CT image feature data of these cases, and organize and label them according to unified standards and formats. For patients diagnosed, clearly mark their disease type, tumor nature (benign or malignant) and other information; for control patients, mark as normal or other disease diagnosis results. By constructing such a training sample set, provide rich and accurate data basis for subsequent model training, so that the model can learn the differences between PPGLs patients and non-patients in various data, so as to realize accurate diagnosis and prediction.

[0023] S102, based on the clinical parameters, biochemical marker data, CT image feature data in the training sample set, combined with Logistic regression analysis method for processing, generate an initial neuroendocrine tumor early warning model.

[0024] In an embodiment, various data are extracted from the training sample set, and data such as age, gender, blood pressure, and the like in the clinical parameters are checked to remove obvious errors (such as negative age) and missing values (such as missing blood pressure data); for the biochemical marker data, data with excessively large detection errors are removed; in the CT image feature data, image data that cannot be accurately analyzed due to poor scanning quality is excluded. By analyzing the age, gender, symptoms (such as whether there is hypertension, headache, etc.), blood pressure, and heart rate values of the patient, the clinical features are formed. For example, if the patient is 45 years old, male, has paroxysmal hypertension and headache symptoms, the blood pressure fluctuates between 160 / 100 mmHg and 180 / 110 mmHg, and the heart rate is between 90 and 110 times per minute, these information constitutes the clinical features.

[0025] The measured values of 24-hour urine catecholamine, methoxyepinephrine, normethanephrine, and methoxytyramine in the blood are compared with the upper limit of normal to obtain the fold change. Assuming that the upper limit of normal for 24-hour urine catecholamine is 100 μg / 24h, and the measured value of a patient is 200 μg / 24h, the fold change is 2 (UFC=2); the upper limit of normal for methoxyepinephrine is 0.5 nmol / L, and the measured value of the patient is 1.5 nmol / L, the fold change is 3 (BFC methoxyepinephrine=3). The CT image is observed, the enhancement grade is determined according to the enhancement degree of the tumor in the image (such as mild, moderate, and obvious enhancement); the tumor necrosis feature is determined by checking whether there is a tumor necrosis area and the size of the necrosis area; the tumor diameter feature is obtained by measuring the maximum diameter of the tumor; and the specific position of the tumor in the body is determined, such as being located on the left side of the adrenal gland, and the like, which form the image enhancement grade, tumor necrosis feature, tumor diameter, and tumor position feature, respectively.

[0026] The above initial image features are normalized to convert data of different dimensions and different value ranges to a unified range for subsequent analysis. For example, the age in the clinical features is assumed to range from 18 to 70 years old in the training sample set, and a linear normalization formula is used The age value is mapped to between 0 and 1; the fold change and the fold change ratio feature are normalized according to their own data distribution; the image enhancement grade can be assigned values of 1, 2, and 3 for mild, moderate, and severe, respectively, and then normalized; the tumor necrosis feature, tumor diameter, and tumor position feature are also normalized according to their respective data characteristics. After processing, the preprocessed feature data containing the clinical feature data, biochemical feature data, and CT image feature data are obtained.

[0027] The preprocessed feature data is input into a logistic regression model. The model determines the relationship between each feature and the occurrence of pheochromocytoma and paraganglioma through learning from a large amount of data. For example, if the clinical feature of hypertension symptoms is highly correlated with the probability of disease, the model will give it a higher weight; if the fold change ratio of a certain biochemical marker plays an important role in determining the nature of the tumor (benign or malignant), it will also reflect the corresponding weight in the model. The prediction parameter vector of the early warning model is calculated, and the parameters in the vector correspond to the prediction strategy and related parameter information of the occurrence probability of pheochromocytoma and paraganglioma, tumor nature related features (such as the degree of malignant tendency), and the range of influence (such as the range of peripheral tissues that the tumor may invade). Assuming that a parameter in the prediction parameter vector represents the probability coefficient of the tumor being malignant, if the parameter value is high, combined with the comprehensive judgment of other parameters, it suggests that the tumor is more likely to be malignant.

[0028] The early warning model for neuroendocrine tumors is a multilayer perceptron (MLP), which is a type of feedforward neural network consisting of an input layer, multiple hidden layers, and an output layer. It is suitable for handling complex nonlinear classification and regression problems and can effectively learn the complex relationships between clinical parameters, biochemical marker data, and CT image feature data to predict information related to neuroendocrine tumors (such as pheochromocytoma and paraganglioma, PPGLs). Model hierarchy and structure Input layer: The number of nodes in the input layer is determined by the number of input features. The input features include clinical feature data (such as age, gender, blood pressure, heart rate, 24-hour urinary catecholamine-related fold change information) after data cleaning, feature extraction, and normalization processing, biochemical feature data (fold change and fold change ratio information of methoxyepinephrine, norepinephrine, and methoxytyramine in blood), and CT image feature data (image enhancement grade, tumor necrosis characteristics, tumor diameter, tumor location characteristics). Assuming that there are 20 features after processing, the input layer has 20 nodes. These nodes receive preprocessed feature data and pass it to the next layer. Two hidden layers are set. The first hidden layer contains 10 neurons, and the second hidden layer contains 5 neurons. The neurons in the hidden layer are connected to the input layer and the next layer through weights. The neurons perform weighted summation on the input data and perform nonlinear transformation through an activation function. Here, ReLU (Rectified Linear Unit) is chosen as the activation function, with the formula f(x) = max(0, x). The number of nodes in the output layer is determined by the prediction target. Since the occurrence probability of PPGLs, tumor property-related features (simplified as benign or malignant, represented by 0 and 1), and the scope of influence (represented by tumor size and location information, encoded as a numerical range) need to be predicted, the output layer is set to have 3 nodes. The output layer nodes are connected to the hidden layer through weights, and the output results are processed through the Sigmoid function to map the output values to the range of 0-1, representing the probability or state of the corresponding prediction target. The Sigmoid function formula is

[0029] The weights are the parameters that connect different layers of neurons in the model, determining the strength of signal transmission between neurons. During training, the weights are constantly adjusted to optimize the performance of the model. For example, the weight matrix W1 from the input layer to the first hidden layer is a 20x10 matrix, with each element representing a connected weight value; the weight matrix W2 from the first hidden layer to the second hidden layer is a 10x5 matrix; the weight matrix W3 from the second hidden layer to the output layer is a 5x3 matrix. The initial values of these weights are usually randomly initialized, such as randomly taking values in the interval [-0.5, 0.5], and then updated during the training process through the backpropagation algorithm. The bias (also known as the bias) is an additional parameter for each neuron, used to adjust the activation threshold of the neuron. There is a bias parameter in each neuron of the hidden layers and the output layer. For example, the 10 neurons in the first hidden layer have 10 bias parameters b 11 ,b 12 ...,b 110 ; the 5 neurons in the second hidden layer have 5 bias parameters b 21 ,b 22 ,...,b 25 ; the 3 neurons in the output layer have 3 bias parameters b 31 ,b 32 ,b 33 . The initial values of the bias parameters can also be randomly set, such as taking values in the interval [-0.5, 0.5], and updated together with the weight parameters during the training process. During the training process, the preprocessed feature data is input into this multi-layer perceptron model, the prediction result is calculated through forward propagation, and then the difference between the prediction result and the true label is calculated using the loss function (such as the cross-entropy loss function). Then, the weight and bias parameters are updated through the backpropagation algorithm, and the training is iterated until the model achieves good performance on the training samples, thereby generating an initial model that can be used for preliminary early warning of neuroendocrine tumors.

[0030] S103, based on K-fold cross-validation and the validation sample set, processing the initial neuroendocrine tumor early warning model to generate a target neuroendocrine tumor early warning model.

[0031] In an embodiment, the initial neuroendocrine tumor early warning model is processed based on K-fold validation set to generate simulated diagnosis results. Assuming that 1000 samples of PPGLs patients and non-patients are collected as a training sample set, 5-fold cross-validation (K=5) is adopted. The 1000 samples are randomly and uniformly divided into 5 subsets, each containing 200 samples. Each time, one of the subsets is selected as the validation set, and the remaining 4 subsets are combined as the training set. This is done 5 times, resulting in 5 different training set and validation set combinations. Taking the validation set of the first division as an example, the clinical parameters (such as patient age 40 years old, female, with persistent hypertension, headache symptoms, blood pressure 170 / 105 mmHg, heart rate 95 beats per minute, 24-hour urine catecholamine measurement value is 3 times the normal upper limit), biochemical marker data (methoxyepinephrine measurement value is 2.5 times the normal upper limit, norepinephrine measurement value is 2 times the normal upper limit, methoxytyramine measurement value is 1.8 times the normal upper limit, the fold change ratio is calculated to obtain the corresponding value) and CT image feature data (tumor diameter 3 cm, located on the right side of the adrenal gland, moderate enhancement, with partial necrotic area) of each sample in the validation set are input into the initial neuroendocrine tumor early warning model (such as the previously constructed multi-layer perception model). The model calculates and outputs the probability of each sample having PPGLs, tumor properties (probability of being predicted as malignant), and impact range (predicted possible impact on surrounding tissues according to tumor size and location), etc. Simulated diagnosis results. This operation is performed on the 200 samples in the validation set to obtain 200 simulated diagnosis results. Repeat the above steps to perform simulated diagnosis on the validation sets of the other 4 divisions, and obtain 5 groups of simulated diagnosis results.

[0032] The simulated diagnosis results are compared with the validation sample set to generate validation results. The simulated diagnosis results of the first validation set are compared with the true situation of the validation sample set. For example, a sample in the validation set is actually diagnosed as PPGLs and is a malignant tumor, located near the surrounding blood vessels. The model predicts that the sample has a probability of 0.8 of having PPGLs (correct prediction), a probability of 0.7 of being a malignant tumor (correct prediction), but in the impact range prediction, only a small part of the surrounding tissue is predicted to be affected, and the blood vessels near the tumor are not accurately predicted (incorrect prediction). This comparison is performed on each of the 200 samples in the validation set.

[0033] The accuracy rate is calculated. If the judgment of 150 samples in the validation set is correct (including the number of samples predicted to be ill and actually ill, and the number of samples predicted to be healthy and actually healthy), the accuracy rate is 150 / 200 = 75%; the precision rate of tumor property prediction is calculated. If 80% of the samples predicted to be malignant are actually malignant, the precision rate of tumor property prediction is 80%; the recall rate of the impact range prediction is calculated. If the proportion of the actually affected peripheral tissue area that is accurately predicted by the model is 60%, the recall rate of the impact range prediction is 60%. In the same way, the other four groups of simulated diagnosis results are compared with the corresponding validation set, and the indicators are calculated. The average accuracy rate, average precision rate, and average recall rate are obtained by comprehensively analyzing the results of the five groups, and the performance of the model in different aspects is evaluated.

[0034] According to the calculated validation result indicators, the performance of the initial model is evaluated. If the average accuracy rate is low, such as only 60%, it means that the overall prediction accuracy of the model needs to be improved; if the precision rate of tumor property prediction is low, it means that the model has deficiencies in judging the benign and malignant tumors; and if the recall rate of the impact range prediction is low, it means that the model has weak prediction ability for the impact range of the tumor. Analyze these indicators to find out the aspects where the model performs poorly. Based on the evaluation results, adjust the model parameters. If it is found that the weight setting of some features of the model is unreasonable, resulting in poor performance, for example, in the Logistic regression model, it is found that the weight of the 24-hour urinary catecholamine-related fold change feature is too high, which makes the feature have too much influence on the result. The weight of this feature can be appropriately reduced; in the multilayer perception model, if the connection weight of a certain hidden layer neuron causes deviation in information transmission, the weights can be adjusted using the back propagation algorithm. The bias parameter is also adjusted according to the evaluation results to optimize the activation threshold of the neuron. After multiple adjustments and re-verification, until the indicators of the model on the validation set reach the expected standard, such as the accuracy rate is improved to more than 80%, the precision rate of tumor property prediction and the recall rate of impact range prediction are significantly improved, and the model obtained at this time is the target neuroendocrine tumor early warning model. This model has higher accuracy and reliability in diagnosing PPGLs.

[0035] In S104, the 24-hour urinary catecholamine in the clinical parameters is used to process the biochemical marker data to generate standardized biochemical detection index information.

[0036] In one implementation, the measurement value of 24-hour urine catecholamine is compared with the normal upper limit value of urine catecholamine to generate fold change information of urine catecholamine. The measurement value of 24-hour urine catecholamine (24huCAs) of a patient is obtained. For example, assuming that the normal upper limit value of urine catecholamine is 100 μg / 24h, the 24-hour urine catecholamine measurement value of a certain PPGLs patient is 300 μg / 24h. The measurement value is compared with the normal upper limit value, and the fold change information of urine catecholamine of the patient is calculated by the formula: fold change = measurement value ÷ normal upper limit value, which is 300 ÷ 100 = 3. This indicates that the 24-hour urine catecholamine level of the patient is 3 times the normal upper limit, and the fold change information quantitatively reflects the abnormality of the patient's index, providing preliminary data support for subsequent diagnosis.

[0037] The methoxyepinephrine, norepinephrine, and methoxytyramine data are also processed in the manner of comparing the measurement value with the normal upper limit value. Assuming that the normal upper limit value of methoxyepinephrine is 0.5 nmol / L, the measurement value of a certain patient is 1.2 nmol / L, the fold change is 1.2 ÷ 0.5 = 2.4; the normal upper limit value of norepinephrine is 0.8 nmol / L, the measurement value of the patient is 2.0 nmol / L, the fold change is 2.0 ÷ 0.8 = 2.5; the normal upper limit value of methoxytyramine is 1.0 nmol / L, the measurement value of the patient is 1.5 nmol / L, the fold change is 1.5 ÷ 1.0 = 1.5. These fold change data respectively reflect the deviation of the three biochemical markers in the patient's body relative to the normal level, which helps to analyze the correlation between the patient's physiological condition and PPGLs from the perspective of different metabolic products.

[0038] The fold change information of urine catecholamine obtained above and the fold change information of methoxyepinephrine, norepinephrine, and methoxytyramine are processed to calculate the ratio between them to obtain fold change ratio information. For example, taking the ratio of the fold change of urine catecholamine to the fold change of methoxyepinephrine as an example, in the above example, the ratio is 3 ÷ 2.4 = 1.25; the ratio of the fold change of urine catecholamine to the fold change of norepinephrine is 3 ÷ 2.5 = 1.2; the ratio of the fold change of urine catecholamine to the fold change of methoxytyramine is 3 ÷ 1.5 = 2. These ratios can reflect the relative change relationship between different biochemical markers. Compared with a single fold change, the ratio information can provide more abundant biochemical metabolic correlation information to assist in judging the disease state. For example, if it is found in a large sample study that when the ratio of the fold change of urine catecholamine to the fold change of methoxyepinephrine is in a certain specific range, it has a correlation with a certain subtype or disease severity of PPGLs, then the ratio information has important reference value in the diagnosis process.

[0039] The fold change ratio information, the fold change information of urinary catecholamine, and the fold change information of each of methoxyepinephrine, norepinephrine, and methoxytyramine measurement values are processed to generate standardized biochemical test index information. A comprehensive index system is constructed, different fold changes and fold change ratios are given corresponding weights, and then weighted summation is performed. The weight of urinary catecholamine fold change is 0.4, the weight of methoxyepinephrine fold change is 0.2, the weight of norepinephrine fold change is 0.2, the weight of methoxytyramine fold change is 0.1, the weight of the fold change ratio of urinary catecholamine and methoxyepinephrine is 0.05, the weight of the fold change ratio of urinary catecholamine and norepinephrine is 0.03, and the weight of the fold change ratio of urinary catecholamine and methoxytyramine is 0.02. The standardized biochemical test index is calculated according to the data in the above example: standardized index = 3 x 0.4 + 2.4 x 0.2 + 2.5 x 0.2 + 1.5 x 0.1 + 1.25 x 0.05 + 1.2 x 0.03 + 2 x 0.02 = 1.2 + 0.48 + 0.5 + 0.15 + 0.0625 + 0.036 + 0.04 = 2.4685. This standardized biochemical test index integrates multiple related biochemical data into a comprehensive numerical value, which can more comprehensively and standardizedly reflect the abnormality of the patient's biochemical index in the PPGLs diagnosis model, and provides a key quantitative basis for subsequent model-based diagnosis analysis.

[0040] S105, processing the CT image feature data to generate imaging comprehensive score information.

[0041] In one implementation, the CT image feature data is extracted and classified to generate tumor morphology feature information, enhancement grade feature information, necrosis feature information, and location feature information. Taking the CT image diagnosis of a suspected PPGLs patient as an example, the doctor first imports the acquired CT image data into professional analysis software, such as the GE Medical AW (Advantage Workstation) post-processing workstation software or the Siemens syngo.via software platform. When observing the tumor morphology, the software's morphological measurement tool is used to outline the tumor's contour. Through the software's automatic calculation function, specific values such as the tumor's circumference, area, long diameter, and short diameter can be obtained. Assuming that the long diameter of the tumor is measured to be 3 cm and the short diameter is 2.5 cm, and the ratio of the long diameter to the short diameter is close to 1, combined with the pre-set morphological analysis algorithm in the software, it is determined that the tumor is round. At the same time, with the help of the software's edge detection function, the change in the tumor edge pixels can be clearly observed. If the edge pixels transition smoothly without burrs or lobulated appearance, it can be concluded that the tumor edge is smooth. This morphological feature is of great significance in determining the nature of the tumor. Generally speaking, tumors with smooth edges are more common in benign lesions, but they can also occur in PPGLs and need to be further judged in combination with other features.

[0042] In determining the enhancement grade feature information, the software can perform special processing on the CT enhanced scan image. By adjusting the window width and window level of the image, the tumor's enhancement situation is more clearly displayed. The software can also measure the CT values of different regions of the tumor and compare them with the CT values of the surrounding normal tissue. When the tumor is significantly enhanced, i.e., the CT value of the tumor is significantly higher than that of the surrounding normal tissue, according to the built-in enhancement grade determination standard in the software, if the enhancement amplitude exceeds a certain threshold (e.g., the CT value after enhancement increases by more than 50 HU compared to the plain scan), combined with the doctor's clinical experience, it is determined that the tumor is highly enhanced. High enhancement usually means that the tumor has abundant blood supply, which is related to the biological characteristics of PPGLs, as PPGLs are neuroendocrine tumors that often have a relatively rich blood supply to maintain their growth and function.

[0043] For the generation of necrosis feature information, the software identifies low-density areas by analyzing the density values in the image. In CT images, necrotic areas usually appear as low-density. The doctor uses the software's density measurement tool to measure the CT value range of the low-density area and observes its size and shape. If a region is found within the tumor with a CT value significantly lower than the tumor's substance and the region has an irregular shape and fuzzy boundary, it can be determined that there is an irregular low-density necrotic area inside the tumor. The presence of tumor necrosis may indicate that the tumor is growing rapidly, has relatively insufficient blood supply, or has undergone ischemic and hypoxic changes, which can assist in determining the malignancy and development stage of the tumor to some extent.

[0044] Finally, with the help of anatomical positioning markers in the images, the software can accurately determine the location characteristics of the tumor. During the CT image acquisition process, precise positioning markers will be placed on the patient's anatomical structures. The software can read these marker information and, combined with the three-dimensional reconstruction function of the image, observe the location of the tumor from multiple perspectives such as coronal, sagittal and axial. This location information is crucial for subsequent treatment planning. For example, in surgical planning, doctors can accurately choose the surgical approach according to the location of the tumor to avoid damaging important structures around it.

[0045] By analyzing and processing the tumor shape feature information, enhancement level feature information, necrosis feature information, and location feature information, tumor growth pattern information, metabolic activity feature information, and tissue abnormality feature information are generated. When observing a smooth tumor edge, this feature reflects the relatively weak invasiveness of the tumor to the surrounding tissue during its growth to some extent. Because if the tumor cells have strong invasiveness during proliferation, they will break through the barrier of the surrounding tissue, resulting in irregular shapes such as burrs and lobes on the edge. A smooth edge means that the growth of tumor cells is relatively orderly and there is no extensive infiltration into the surrounding tissue. At the same time, if the necrotic area is small and concentrated in the center, there is a specific pathophysiological mechanism behind it. During tumor growth, the demand for nutrients and oxygen of cells inside the tumor increases. When the tumor volume increases to a certain extent, the cells in the central part are far from the blood vessels and have insufficient blood supply, which can easily lead to ischemia and hypoxia, and then necrosis. If the necrotic area is small and concentrated in the center, it means that although the blood supply in the central part is insufficient, the overall blood supply can still maintain the growth of the tumor, and the tumor cells have not extensively broken through the surrounding tissue to obtain more nutrients, so it is more inclined to be a swelling growth.

[0046] If the tumor is a swelling growth, it is more inclined to complete resection in terms of treatment selection because its invasion to the surrounding tissue is relatively light and the possibility of complete resection is greater. When analyzing the enhancement level characteristics of the tumor, take another suspected PPGLs patient as an example. When the CT image shows that the tumor is highly enhanced, it is closely related to the blood supply and cell metabolism of the tumor. During the CT enhancement scan, the contrast agent injected will be distributed to the tissues throughout the body with the blood circulation. The more abundant the blood supply of the tumor tissue, the more contrast agent enters the tumor cells, which appears as obvious enhancement on the image. High enhancement means that there is a large amount of contrast agent accumulation in the tumor, which directly reflects that the tumor has abundant blood supply. Blood not only provides nutrients for tumor cells but also carries away metabolic waste. Abundant blood supply provides material basis for the vigorous metabolism of tumor cells, enabling them to carry out various metabolic activities such as protein synthesis and nucleic acid replication. Therefore, tumors with high enhancement often have active cell metabolism.

[0047] By professional medical image analysis software, the CT value of the tumor after enhancement increased by 80HU (higher than the threshold of 50HU for general judgment of high enhancement), and the enhanced area was uniform, which indicated that the blood supply of the tumor was relatively rich in each part. Combined with relevant research data, when the enhanced CT value of the tumor increases by more than 50HU, the metabolic enzyme activity in the tumor cells is significantly increased, such as the up-regulation of glucose transporter 1 (GLUT1) expression, which promotes the uptake and metabolism of glucose and provides energy for the proliferation of tumor cells. Metabolic activity characteristic information is generated, such as "the tumor has high metabolic activity, cell proliferation is active, and its growth rate and invasion potential should be closely monitored". This information is helpful for doctors to evaluate the malignancy and development trend of the tumor, because tumors with high metabolic activity tend to grow faster and are more likely to metastasize.

[0048] When evaluating tissue abnormalities, it is necessary to consider tumor morphology, enhancement grade, and necrosis characteristics comprehensively. When the tumor morphology is irregular, it indicates that the growth of tumor cells has lost the normal regulatory mechanism and presents a disordered growth state. Irregular morphology may be due to the heterogeneity of tumor cells, with different regions of tumor cells having different proliferation and invasion abilities, resulting in uneven tumor edges. Inhomogeneous enhancement further indicates that there are differences in blood supply and cell metabolism within the tumor. This may be because there are subpopulations of cells with different degrees of differentiation within the tumor, or abnormal angiogenesis within the tumor tissue, with rich blood supply in some areas and insufficient blood supply in some areas. This difference in blood supply and metabolism is manifested as inhomogeneous enhancement on CT enhancement images.

[0049] At the same time, a larger necrotic area indicates that there is more cell death within the tumor, which is due to the rapid growth of the tumor, insufficient blood supply, or the toxic effects of metabolic products on tumor cells. Combining these characteristics, it indicates that the tumor tissue is significantly different from normal tissue in structure and function, and the degree of tissue abnormality is high. For example, in the CT image of a certain case, the tumor presents a lobulated shape, with obvious edge spiculation, and after enhancement scanning, it can be seen that part of the tumor is enhanced obviously, part of the tumor is enhanced weakly, and the necrotic area accounts for 30% of the tumor volume. According to these characteristics, tissue abnormality characteristic information is generated, such as "the tumor tissue has a high degree of abnormality, with obvious cell atypia and structural disorder, high risk of malignancy, and further examination is needed". This information is of great significance for guiding subsequent diagnosis and treatment decisions.

[0050] In this case, the original CT image feature data is thoroughly mined using professional medical image analysis software such as Siemens' syngo.via, GE's AW, etc. First, attention is paid to the key data of tumor edge blur. In CT images, the tumor edge is one of the important bases for judging its nature. The boundary between normal tissue and tumor tissue is usually clear in benign lesions, while malignant tumors, due to their invasive growth characteristics, will infiltrate the surrounding tissue, resulting in blurred edges. Using the edge detection tool of the image analysis software, the pixel changes of the tumor edge can be accurately identified and marked. For example, through the gradient algorithm of the software, the gray scale change rate of the tumor edge pixels is calculated, and when the change rate exceeds a certain threshold (the threshold is determined based on a large number of clinical case data and research), the software automatically marks this part of the edge as a blurred area. At the same time, in order to more accurately quantify the degree of edge blur, the width, irregularity and other parameters of the blurred area can be measured. Assuming that the average width of the tumor edge blurred area is 2mm and the irregularity index is 0.6 (the higher the value, the more irregular), these data are recorded in detail.

[0051] Heterogeneous enhancement is also one of the important features of malignant tumors, which reflects the differences in blood supply and cell metabolism within the tumor. In the enhanced CT image, the density measurement and analysis function of the software is used to quantify the enhancement degree of different regions of the tumor. Normal tissues have a relatively uniform enhancement degree after enhancement scanning, while malignant tumors, due to abnormal angiogenesis, different cell differentiation levels, etc., will appear heterogeneous enhancement. The software can automatically divide the tumor area and measure the CT value of each sub-region. For example, in this tumor image, it is found that the CT value of some regions increases by 60HU after enhancement, while another part of the region only increases by 20HU, with a large difference between the two, indicating heterogeneous enhancement. Mark these regions with obvious enhancement differences, and record the location, area ratio, etc. of different enhancement regions. For example, the region with high enhancement degree accounts for 30% of the total tumor area, the region with low enhancement degree accounts for 20%, and the remaining is the moderate enhancement region. In addition to tumor edge blur and heterogeneous enhancement, there may be other features related to malignant PPGLs, such as the presence of microcalcification within the tumor, signs of invasion of the surrounding tissue, etc., which are also marked. For example, through the software's calcification detection function, it is found that there are multiple microcalcifications with a diameter of less than 2mm inside the tumor, and the location and number of these calcifications are also accurately recorded.

[0052] After the key data is marked, the marked data is extracted from the original CT image feature data using the filtering function of the software to form a potential PPGLs related data screening result. The screening result is saved in a specific data format for further analysis. For example, the marked tumor edge blur area, uneven enhancement area, and microcalcification related data are integrated into a new data set, including the position coordinates, morphological parameters (such as area, perimeter, diameter, etc.), CT value changes, and other detailed information of each marked area. These screening results are of great significance for subsequent diagnosis and treatment. On the one hand, doctors can more intuitively observe the malignant characteristics of the tumor to further confirm the diagnostic possibility of PPGLs. For example, if the tumor edge is blurred and the range of invasion of the surrounding tissue is large, a wider surgical resection range or a comprehensive treatment plan needs to be considered.

[0053] S106, based on the target neuroendocrine tumor early warning model, the standardized biochemical detection index information and the imaging comprehensive score information are processed to generate early warning information of pheochromocytoma and paraganglioma.

[0054] In an embodiment, the input data of the model includes standardized biochemical detection index information and imaging comprehensive score information, wherein the standardized biochemical detection index information includes the measurement values of 24-hour urine catecholamine (24h uCAs), methoxyepinephrine (MN) in blood, noradrenaline (NMN), and methoxytyramine (MT) and the fold change (UFC, BFC) and ratio compared with the upper limit of normal (ULN). After comparing the above indicators with the ULN, the comprehensive standardized indicators are generated by weighted calculation (such as using 2 ULN as the critical value to calibrate the abnormal degree). The imaging comprehensive score information includes the enhancement level (such as mild / moderate / obvious enhancement) in the CT enhanced image, the necrosis performance (such as ring sign, necrosis area proportion), tumor diameter, location, and other characteristics. The CT characteristics are quantitatively scored (such as the combination score of enhancement level and necrosis characteristics) to form the imaging comprehensive score.

[0055] The target model is a multi-factor prediction model constructed by Logistic regression analysis method, which is verified by K-fold cross-validation and external data (such as a hospital data set). The standardized biochemical indicators and imaging scores are used as input variables, and the linear prediction value is calculated by the regression equation to convert into the probability of PPGLs occurrence (such as the probability value ≥0.5 is determined as high risk). The stability is ensured by K-fold cross-validation internally, and the generalization ability is verified by independent data set externally to ensure the accuracy of the model in different scenarios.

[0056] The early warning information generation path is as follows: risk assessment value calculation, target model weighted operation on input data, and output of probability value between 0 and 1 (such as 0.7 indicating a 70% probability of disease). The risk assessment value is compared with the risk standard range preset by the model (such as a probability threshold based on a large sample), the deviation is corrected by adjusting the regression parameters (such as the weight coefficient), and the final early warning result is generated. Through tools such as nomograms, abstract probabilities are converted into visual scores, and clinicians can quickly query the risk of disease according to symptoms, biochemical and imaging parameters.

[0057] Multi-parameter integration overcomes the limitations of single indicators, improves early diagnosis rate, reduces missed and misdiagnosed cases, and improves accuracy. Relying on conventional CT and biochemical detection reduces the dependence on functional imaging (such as PET-CT) and is suitable for resource-limited areas. Shorten the diagnosis process, reduce unnecessary examinations, and reduce patient waiting time and medical costs. Quantitative risk values provide the basis for clinical intervention (such as surgical timing and follow-up frequency) and support personalized diagnosis and treatment.

[0058] In this application, the server collects patient clinical parameters, biochemical marker data and CT image feature data to construct a training sample set. Then the data is cleaned, feature extracted and normalized, combined with Logistic regression analysis method to generate an initial neuroendocrine tumor early warning model. Then use K-fold cross-validation and validation sample set to optimize the initial model to get the target model. In terms of data processing, by comparing the 24-hour urinary catecholamine and related metabolite measurement values in blood with the upper limit of normal, the fold change and ratio information are obtained, and then the standardized biochemical detection index is generated. At the same time, the CT image feature data is extracted, classified and analyzed to obtain information such as tumor growth pattern and metabolic activity, and generate an imaging comprehensive score. Based on the target model, the standardized biochemical detection index information and the imaging comprehensive score are combined to obtain the risk assessment value. By comparing with the expected risk standard range of the model, the model parameters are adjusted to generate early warning information. By integrating multiple parameters and cross-validation, the diagnostic accuracy is improved, the dependence on high-end equipment is reduced, and strong support is provided for clinical decision-making.

[0059] In one embodiment, as shown in Figure 2 The application also provides a pheochromocytoma and paraganglioma early warning device, comprising:

[0060] The acquisition module 201 is used to acquire the clinical parameters, biochemical marker data, CT image feature data and training sample set of pheochromocytoma and paraganglioma patients, wherein the clinical parameters include 24-hour urinary catecholamine, and the biochemical marker data includes methoxyepinephrine, norepinephrine and methoxytyramine in blood.

[0061] The processing module 202 is configured to process the clinical parameters, the biochemical marker data, and the CT image feature data in the training sample set based on a Logistic regression analysis method to generate an initial neuroendocrine tumor early warning model; process the initial neuroendocrine tumor early warning model based on K-fold cross-validation and a verification sample set to generate a target neuroendocrine tumor early warning model; process the biochemical marker data based on 24-hour urinary catecholamine in the clinical parameters to generate standardized biochemical detection index information; process the CT image feature data to generate imaging comprehensive score information; and process the standardized biochemical detection index information and the imaging comprehensive score information based on the target neuroendocrine tumor early warning model to generate early warning information of pheochromocytoma and paraganglioma.

[0062] The computer-readable storage medium provided by the above-mentioned embodiments of the present application has the same beneficial effects as the method adopted, run, or implemented by the application program stored therein, based on the same inventive concept as the early warning method of pheochromocytoma and paraganglioma provided by the embodiments of the present application.

[0063] Each of the embodiments in the present application is described in a related manner, and the same or similar parts between the embodiments can be referred to each other. Each of the embodiments mainly describes the difference from other embodiments. In particular, the early warning method of pheochromocytoma and paraganglioma, the electronic device, the electronic equipment, and the readable storage medium are basically similar to the above-mentioned embodiments of the early warning method of pheochromocytoma and paraganglioma, and thus the description is relatively simple, and the relevant parts can be referred to the above-mentioned embodiments of the early warning method of pheochromocytoma and paraganglioma.

Claims

1. A method for early warning of pheochromocytoma and paraganglioma, characterized in that, include: Acquire clinical parameters, biochemical marker data, CT image feature data, and training sample set of patients with pheochromocytoma and paraganglioma. The clinical parameters include PPGL-related symptoms, and the biochemical marker data include blood levels of methoxyepinephrine, nomethoxyepinephrine, and methoxytyramine. Based on the clinical parameters, biochemical biomarker data, and CT image feature data in the training sample set, combined with Logistic regression analysis, an initial neuroendocrine tumor early warning model was generated. The initial neuroendocrine tumor early warning model is processed based on K-fold cross-validation and the validation sample set to generate the target neuroendocrine tumor early warning model; Based on the 24-hour urinary catecholamine levels in clinical parameters, biochemical biomarker data are processed to generate standardized biochemical detection index information; CT image feature data are processed to generate comprehensive imaging score information; Based on the target neuroendocrine tumor early warning model, standardized biochemical test indicators and imaging comprehensive score information are processed to generate early warning information for pheochromocytoma and paraganglioma.

2. The method as described in claim 1, characterized in that, Based on clinical parameters, biochemical biomarker data, and CT image feature data from the training sample set, combined with logistic regression analysis, an initial neuroendocrine tumor early warning model was generated, including: The clinical parameters, biochemical markers, and CT image features in the training sample set were cleaned and feature extracted to generate initial image features. These initial image features included clinical features, fold change and fold change ratio features compared with the upper limit of normal, image enhancement level, tumor necrosis features, tumor diameter, and tumor location features. The initial image features are normalized to generate preprocessed feature data, which includes clinical feature data, biochemical feature data, and CT image feature data. The preprocessed feature data is processed using Logistic regression analysis to generate prediction parameters for the early warning model. The prediction parameter vector of the early warning model is used to characterize the prediction strategy and related parameter information for the probability of occurrence, tumor nature-related characteristics, and scope of influence of pheochromocytoma and paraganglioma. The initial neuroendocrine tumor early warning model is optimized and trained based on the predicted parameter vector of the early warning model to generate an initial neuroendocrine tumor early warning model.

3. The method as described in claim 2, characterized in that, The initial neuroendocrine tumor early warning model is processed based on K-fold cross-validation and the validation sample set to generate the target neuroendocrine tumor early warning model, including: The initial neuroendocrine tumor early warning model is processed based on the validation set of K-fold partitioning to generate simulated diagnostic results; The simulated diagnostic results are compared with the validation sample set to generate validation results; Based on the validation results, the initial neuroendocrine tumor early warning model was evaluated and adjusted to generate the target neuroendocrine tumor early warning model.

4. The method as described in claim 1, characterized in that, Based on the 24-hour urinary catecholamine levels in clinical parameters, biochemical biomarker data are processed to generate standardized biochemical detection indicator information, including: The measured values ​​of urinary catecholamines over 24 hours are compared with the upper limit of normal urinary catecholamine levels to generate information on the multiple changes in urinary catecholamine levels. The measured values ​​of methoxyepinephrine, nomethoxyepinephrine, and methoxytyramine in the blood were compared with their respective upper limits of normal to generate information on their fold changes. The fold change information of urinary catecholamines and the fold change information of the measured values ​​of methoxyadrenaline, nomethoxyadrenaline and methoxytyramine are processed to generate fold change ratio information. The information on fold change ratios, fold change information on urinary catecholamines, and fold change information on the individual measurements of methoxyadrenaline, nomethoxyadrenaline, and methoxytyramine are processed to generate standardized biochemical detection index information.

5. The method as described in claim 1, characterized in that, The CT image feature data is processed to generate comprehensive radiological scoring information, including: The CT image feature data is extracted and classified to generate tumor morphology feature information, enhancement level feature information, necrosis feature information, and location feature information. The tumor morphological characteristics, enhancement level characteristics, necrosis characteristics, and location characteristics are analyzed and processed to generate tumor growth pattern information, metabolic activity characteristics, and tissue abnormality characteristics. The tumor growth pattern information, metabolic activity characteristics information, tissue abnormality characteristics information, and location characteristics information are analyzed and processed to generate potential characteristics information and tumor nature characteristics information of pheochromocytoma and paraganglioma; Based on the potential characteristics and tumor nature characteristics of pheochromocytoma and paraganglioma, key data in the original CT image feature data are marked and screened to generate relevant data screening results for potential pheochromocytoma and paraganglioma. The relevant data screening results of potential pheochromocytoma and paraganglioma were integrated and quantified to generate pheochromocytoma and paraganglioma characteristics. The characteristics of pheochromocytoma and paraganglioma are processed to generate comprehensive imaging scoring information.

6. The method as described in claim 1, characterized in that, Based on the target neuroendocrine tumor early warning model, standardized biochemical test indicators and comprehensive imaging scores are processed to generate early warning information for pheochromocytoma and paraganglioma, including: Based on the target neuroendocrine tumor early warning model, standardized biochemical test index information and imaging comprehensive score information are analyzed and processed to generate risk assessment values ​​for pheochromocytoma and paraganglioma. The warning decision parameter vector inside the target neuroendocrine tumor warning model is processed, and the risk assessment value is compared with the risk standard range expected by the model to generate a corresponding warning deviation correction vector. The warning deviation correction vector is analyzed and transformed to generate warning information for pheochromocytoma and paraganglioma.

7. An early warning device for pheochromocytoma and paraganglioma, characterized in that, The device includes: The acquisition module is used to acquire clinical parameters, biochemical marker data, CT image feature data, and training sample sets of patients with pheochromocytoma and paraganglioma. The clinical parameters include 24-hour urinary catecholamines, and the biochemical marker data include blood methoxyepinephrine, nomethoxyepinephrine, and methoxytyramine. The processing module is used to process clinical parameters, biochemical marker data, and CT image feature data from the training sample set using Logistic regression analysis to generate an initial neuroendocrine tumor early warning model; to process the initial neuroendocrine tumor early warning model based on K-fold cross-validation and a validation sample set to generate a target neuroendocrine tumor early warning model; to process biochemical marker data based on 24-hour urinary catecholamines from the clinical parameters to generate standardized biochemical detection index information; to process CT image feature data to generate comprehensive imaging score information; and to process the standardized biochemical detection index information and comprehensive imaging score information based on the target neuroendocrine tumor early warning model to generate early warning information for pheochromocytoma and paraganglioma.

8. An electronic device, characterized in that, include: First processor; and memory for storing executable instructions of the first processor; The first processor is configured to execute the early warning method for pheochromocytoma and paraganglioma according to any one of claims 1 to 6 by executing the executable instructions.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the second processor, it implements the early warning method for pheochromocytoma and paraganglioma as described in any one of claims 1 to 6.

Images