A method and related equipment for warning the risk of myelosuppression caused by treatment

Abstract

The present invention provides a method and related equipment for early warning of myelosuppression risk caused by treatment, which are applied to the field of data processing technology. The present application processes the multimodal data information of a target user to generate a target physiological state prediction factor; processes the target user's serum marker characteristic information to generate a myelosuppression influencing factor; processes the target user's lifestyle information to generate a myelosuppression risk adjustment factor; processes the myelosuppression influencing factor and the myelosuppression risk adjustment factor to generate dynamic risk parameter information for the target user; and processes the target physiological state prediction factor and the target user's dynamic risk parameter information based on a target high-dose methotrexate myelosuppression risk prediction model to generate a myelosuppression risk value and a bone marrow transplant risk level.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method for early warning of myelosuppression risk caused by treatment and related equipment. Background Art

[0002] In the field of cancer treatment, high-dose methotrexate (HDMTX) is a commonly used first-line chemotherapy drug for diseases such as acute lymphoblastic leukemia, non-Hodgkin's lymphoma, and osteosarcoma. However, the bone marrow suppression it causes cannot be ignored. Literature reports suggest that the incidence of HDMTX-induced bone marrow suppression can be as high as 32%.

[0003] Early identification and prediction of individuals at high risk of adverse events after using such drugs is crucial to ensuring the safety of users and their subsequent condition. However, current prediction models for adverse events related to this drug have significant flaws. Most of these models are based on data from a single source, and the number of samples is extremely limited, which makes the model prone to overfitting and greatly weakens its applicability in different scenarios. For example, common research samples are only dozens to hundreds of cases. Although a few larger-scale studies have a certain sample size, the data sources are limited to the same region and cannot fully cover individual characteristics in different environments. In addition, existing prediction models mainly focus on the prediction of a specific adverse condition (similar to delayed clearance), and there is little research on the problem of bone marrow suppression, which has a higher incidence rate, and it is difficult to meet the urgent need for bone marrow suppression risk assessment in practical applications.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the Invention

[0005] The purpose of this application is to provide a method and related equipment for early warning of the risk of myelosuppression caused by treatment, which, at least to a certain extent, overcomes the problems existing in the prior art by obtaining multiple types of information about the target user, such as multimodality, serum markers, lifestyle habits, medication information, models, and samples. Then, during data processing, multimodal data is converted to generate target physiological state prediction factors, serum information is processed to obtain myelosuppression influencing factors, and lifestyle information is used to obtain risk adjustment factors and combine them to generate dynamic risk parameters. When constructing the model, the sampling ratio is determined from the training samples, and the training and verification are performed to obtain the target model. Finally, the target model is used to integrate relevant factors, and the myelosuppression risk value and level are calculated. Effective early warning is achieved by comprehensively considering the parameters of each factor.

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

[0007] According to one aspect of the present application, a method for early warning of myelosuppression risk caused by treatment is provided, comprising: obtaining multimodal data information of a target user, serum marker characteristic information of the target user, life habit information of the target user, medication information of the target user, a preset high-dose methotrexate myelosuppression risk prediction model and a training sample set, wherein the multimodal data information of the target user includes the target user's age, weight, blood creatinine, BMI, height, red blood cell count, lymphocyte count, potassium, urine white blood cells, high-density lipoprotein, magnesium and uric acid; processing the multimodal data information of the target user to generate a target physiological state prediction factor; and processing the serum marker characteristic information of the target user to generate a target physiological state prediction factor. The characteristic information of the target user is processed to generate a bone marrow suppression influencing factor; the living habit information of the target user is processed to generate a bone marrow suppression risk adjustment factor; the bone marrow suppression influencing factor and the bone marrow suppression risk adjustment factor are processed to generate dynamic risk parameter information of the target user; based on the training sample set, the preset high-dose methotrexate bone marrow suppression risk prediction model is processed to generate a target high-dose methotrexate bone marrow suppression risk prediction model; based on the target high-dose methotrexate bone marrow suppression risk prediction model, the target physiological state prediction factor and the dynamic risk parameter information of the target user are processed to generate a bone marrow suppression risk value and a bone marrow transplantation risk level.

[0008] Another aspect of the present application is a device for early warning of myelosuppression risk caused by treatment, characterized in that it includes: an acquisition module for acquiring multimodal data information of a target user, serum marker characteristic information of the target user, life habit information of the target user, medication information of the target user, a preset high-dose methotrexate myelosuppression risk prediction model and a training sample set, wherein the multimodal data information of the target user includes the target user's age, weight, blood creatinine, BMI, height, red blood cell count, lymphocyte count, potassium, urine white blood cells, high-density lipoprotein, magnesium and uric acid; a processing module for processing the multimodal data information of the target user to generate a target physiological state prediction factor; and a processing module for processing the multimodal data information of the target user to generate a target physiological state prediction factor. The target user's serum marker characteristic information is processed to generate a bone marrow suppression influencing factor; the target user's living habit information is processed to generate a bone marrow suppression risk adjustment factor; the bone marrow suppression influencing factor and the bone marrow suppression risk adjustment factor are processed to generate the target user's dynamic risk parameter information; based on the training sample set, the preset high-dose methotrexate bone marrow suppression risk prediction model is processed to generate a target high-dose methotrexate bone marrow suppression risk prediction model; based on the target high-dose methotrexate bone marrow suppression risk prediction model, the target physiological state prediction factor and the target user's dynamic risk parameter information are processed to generate a bone marrow suppression risk value and a bone marrow transplantation risk level.

[0009] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a second processor, the computer program implements the above-mentioned method for early warning of the risk of myelosuppression caused by treatment.

[0010] The present application provides a method and related equipment for warning the risk of myelosuppression caused by treatment. The server obtains multiple types of information about the target user, such as multimodality, serum markers, lifestyle habits, medication information, models, and samples. Then, in data processing, multimodal data is converted to generate target physiological state prediction factors, serum information is processed to obtain myelosuppression influencing factors, and lifestyle information is used to obtain risk adjustment factors and combine them to generate dynamic risk parameters. When constructing the model, the sampling ratio is determined from the training samples, and the training and verification are performed to obtain the target model. Finally, the target model is used to fuse the relevant factors, and the myelosuppression risk value and level are calculated. Effective warning is achieved by comprehensive consideration of the parameters of each factor.

[0011] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flowchart showing a method for early warning of myelosuppression risk caused by treatment provided by one embodiment of the present application is shown;

[0013] Figure 2 A schematic structural diagram of a device for early warning of risk of myelosuppression caused by treatment provided in one embodiment of the present application is shown. DETAILED DESCRIPTION

[0014] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0015] The following combination Figure 1 The following describes a method for early warning of myelosuppression risk caused by treatment according to an exemplary embodiment of the present application. In one embodiment, the present application also provides a method for early warning of myelosuppression risk caused by treatment and related equipment. Figure 1 The flowchart of a method for early warning of the risk of myelosuppression caused by treatment according to an embodiment of the present application is schematically shown. Figure 1 As shown, the method is applied to the server and includes:

[0016] S101, obtaining multimodal data information of a target user, serum marker characteristic information of the target user, lifestyle information of the target user, medication information of the target user, a preset high-dose methotrexate bone marrow suppression risk prediction model, and a training sample set.

[0017] In one embodiment, it is assumed that the target user is a 55-year-old male. Age is an important physiological characteristic. In the risk assessment of bone marrow suppression, different age groups have different tolerance to drugs. With age, the various functions of the body gradually decline, and the hematopoietic function of the bone marrow is also affected to a certain extent. The tolerance to large doses of methotrexate decreases, thereby increasing the risk of bone marrow suppression. The user weighs 70 kilograms. The relative relationship between weight and drug dosage is of great significance in assessing the risk of bone marrow suppression. The drug dosage is calculated to a certain extent based on weight. If the weight is too light or too heavy, it will affect the metabolic process and concentration distribution of the drug in the body, thereby affecting the probability of bone marrow suppression. For example, patients who are underweight face a higher risk of bone marrow suppression due to relatively high drug doses.

[0018] The user's blood creatinine level was 100 μmol / L. Serum creatinine is a key indicator of renal function, and normal renal function is directly related to drug excretion. Elevated serum creatinine indicates a degree of renal impairment, which slows methotrexate excretion and prolongs the duration of drug concentration in the body, increasing toxicity to the bone marrow and the risk of bone marrow suppression. The user's BMI was calculated to be 25. BMI comprehensively reflects the relationship between height and weight and can be used to assess a patient's nutritional status and body fat content. A higher BMI is associated with chronic conditions such as cardiovascular disease and diabetes, which indirectly affect bone marrow function and drug response. For example, obesity (higher BMI) leads to a chronic inflammatory state, which affects the bone marrow microenvironment and increases the risk of bone marrow suppression. The user's height is 175 cm. Height, when combined with other indicators such as weight, helps more accurately assess overall physical condition and physiological function. Height is included in the calculation as an auxiliary variable to more comprehensively reflect individual differences in physiological characteristics.

[0019] The red blood cell count was 4.5 × 10 12 / L. Red blood cells are primarily responsible for transporting oxygen in the body, and changes in their number reflect the state of hematopoietic function. If the red blood cell count is below the normal range, it indicates that the bone marrow hematopoietic function is suppressed or there are other blood system diseases, which also increases the risk of bone marrow suppression after high-dose methotrexate treatment. The lymphocyte count is 1.5×10 9Lymphocytes play a vital role in the immune system, and changes in their number reflect the immune system's functional status and the body's response to disease or medication. In cases of myelosuppression, lymphocyte counts can fluctuate, and certain factors that affect lymphocyte function are also associated with the pathogenesis of myelosuppression. The blood potassium concentration was 4.0 mmol / L. Potassium ions are one of the key electrolytes that maintain cellular physiological function. Its balance in the body is crucial for neuromuscular excitability, cardiac function, and other functions. Abnormal blood potassium levels affect cellular metabolism, including that of bone marrow cells, and thus indirectly impact the risk of myelosuppression. The urine white blood cell count was 5 cells / uL. The presence of urine white blood cells indicates infection or inflammation in the urinary tract, reflecting the body's overall immune status or underlying disease. Inflammation, through regulatory effects of the immune system, affects the bone marrow microenvironment and hematopoietic function, thereby modifying the risk of myelosuppression to a certain extent. The high-density lipoprotein level was 1.2 mmol / L.

[0020] High-density lipoprotein (HDL) is primarily involved in reverse cholesterol transport in the body and is crucial for cardiovascular health. By influencing overall metabolism and blood circulation, changes in HDL levels indirectly affect the risk of myelosuppression. The blood magnesium concentration is 0.8 mmol / L. Magnesium ions play a vital role in many enzymatic reactions and cellular physiological processes, including DNA and RNA synthesis and maintaining cell membrane stability. Bone marrow cell proliferation and differentiation also rely on the normal functioning of these enzymatic reactions and cellular physiological processes. Therefore, abnormal blood magnesium concentrations can affect bone marrow hematopoietic function and, in turn, be associated with the risk of myelosuppression. The uric acid level is 400 μmol / L. Uric acid is the end product of purine metabolism in the body. Hyperuricemia is associated with gout, kidney disease, and other conditions. These conditions affect the body's metabolic function and internal environment, potentially impacting bone marrow hematopoietic function and increasing the risk of myelosuppression.

[0021] The target user's alanine aminotransferase (ALT) is 40U / L, lactate dehydrogenase (LDH) is 200U / L, and aspartate aminotransferase (AST) is 35U / L. These serum markers usually reflect the functional status of the liver and other tissues. ALT and AST are mainly present in hepatocytes, and their increase indicates hepatocyte damage, such as drug-induced liver damage or other liver diseases. During treatment with high-dose methotrexate, abnormal liver function affects the metabolism and detoxification process of the drug, resulting in abnormally high drug concentrations in the body, thereby increasing the risk of bone marrow suppression. LDH is present in many tissues, and its increase is related to damage or metabolic abnormalities of tissue cells. In the case of bone marrow suppression, damage to bone marrow cells or abnormal hematopoietic function leads to increased LDH release, so LDH levels can be used as a potential reference indicator for bone marrow suppression.

[0022] The target user's daily diet consists of relatively low intake of vegetables and fruits, averaging approximately 1,000 grams of vegetables and 500 grams of fruit per week. However, their intake of meat and fat is high, averaging approximately 1,500 grams of meat and 200 grams of fat per week. This unbalanced diet leads to nutritional deficiencies (such as vitamins and minerals) and obesity. Vitamin and mineral deficiencies affect the normal metabolism and hematopoietic function of bone marrow cells, while the chronic inflammatory state associated with obesity disrupts the bone marrow microenvironment, increasing the risk of bone marrow suppression. The user typically engages in light exercise, such as walking three times a week for approximately 30 minutes each time, and rarely engages in high-intensity aerobic exercise or strength training. Moderate exercise helps promote blood circulation and metabolism, enhances immunity, and improves drug tolerance. Lack of adequate exercise can reduce body function, impair blood supply and nutrient delivery to the bone marrow, and thus increase the risk of bone marrow suppression. The user also consumes alcohol, averaging approximately 200 ml of alcohol per week, primarily baijiu. Alcohol and its metabolites are toxic to organs such as the liver and bone marrow. Long-term drinking can cause liver damage, affect drug metabolism, and directly damage bone marrow cells, increasing the risk of myelosuppression. This user smokes about 10 cigarettes a day. Smoking is a clear-cut negative lifestyle habit. Tobacco's harmful substances (such as nicotine and tar) can reach the bone marrow through the bloodstream, damaging bone marrow cells and inhibiting the bone marrow's hematopoietic function, significantly increasing the risk of myelosuppression.

[0023] In addition to high-dose methotrexate treatment, the target patient also took a blood pressure medication (such as nifedipine) and an antibiotic (such as amoxicillin). These combined medications can interact with methotrexate, affecting its metabolism, distribution, or efficacy. For example, some medications inhibit methotrexate excretion, leading to elevated concentrations in the body and increasing the risk of bone marrow suppression. Alternatively, some medications can affect the normal function of the bone marrow, compounding the bone marrow suppression effect of methotrexate and exacerbating bone marrow damage.

[0024] The pre-configured high-dose methotrexate myelosuppression risk prediction model is constructed using the LightGBM algorithm. The model can be viewed as an integrated model consisting of multiple decision trees, processing and learning input data in a layered and progressive manner. During training, each decision tree learns and optimizes based on the residuals of the previous tree, continuously improving the model's predictive capabilities. This hierarchical structure enables the model to capture complex nonlinear relationships in the data and analyze and model factors associated with myelosuppression risk from different perspectives and levels.

[0025] The model consists of an input layer, hidden layers, and an output layer. Its specific structure is as follows: The input layer receives preprocessed multimodal data, serum marker characteristics, lifestyle information, and medication information from target users as input features. These features undergo preprocessing, such as standardization and normalization, to ensure consistency in their numerical range and distribution, enabling the model to better learn and process them. For example, numerical features such as age and weight are normalized to fall within the interval [0, 1]. Categorical features such as gender and smoking status are converted into binary vectors using one-hot encoding. The hidden layer consists of multiple decision trees, whose structural parameters, such as depth and number of nodes, are automatically adjusted and optimized during training based on data characteristics and model performance. Each decision tree splits and grows based on the input features, selecting the optimal features and split points to partition the data into different subsets to minimize the loss function. Optimization techniques, such as the histogram-based algorithm, are employed in constructing the decision trees to reduce data storage and computational complexity, thereby improving training efficiency. The output layer outputs the myelosuppression risk value and bone marrow transplant risk level. The risk value is a continuous value, derived through internal model calculations and mapping. Its range is set within [0, 1] or other reasonable intervals based on specific training and adjustments, and represents the likelihood of myelosuppression. The risk level is categorized into different levels (such as low risk, medium risk, and high risk) based on pre-set thresholds, allowing clinicians to more intuitively understand and apply the model's predictions.

[0026] The model's learning rate (LearningRate) was set to 0.05. The learning rate controls the update step size of the model parameters in each iteration. A smaller learning rate helps the model converge more stably, but requires more training time. A larger learning rate causes the model to skip the optimal solution during training, but trains faster. Through multiple experiments and fine-tuning, 0.05 was determined to be a suitable learning rate for this model, ensuring that the model converges while completing the training process relatively quickly. The number of iterations (Number of Iterations) was determined to be 500 after training and validation. The number of iterations determines the number of rounds the model learns from the data. As the number of iterations increases, the model's performance on the training set generally improves, but overfitting may occur. By monitoring the model's performance metrics on the validation set (such as AUC and accuracy), it was found that when the number of iterations reached 500, the model's performance on the validation set reached a relatively stable state. Further increases in the number of iterations did not significantly improve model performance and increased the risk of overfitting. Therefore, the number of iterations was set to 500.

[0027] The final model contains 200 decision trees. The number of decision trees is closely related to the model's complexity and generalization ability. A larger number of decision trees improves the model's expressiveness, but also increases the computational burden and the risk of overfitting. During training, by gradually increasing the number of decision trees and observing the model's performance on the validation set, we found that when the number of decision trees reached 200, the model achieved a good balance between prediction accuracy and generalization ability. The average tree depth was 6. The tree depth limits the growth complexity of the decision tree. Shallow trees cannot fully capture the complex relationships in the data, while too deep trees are prone to overfitting. Experimentation and evaluation of different tree depths found that an average tree depth of 6 yielded ideal performance on both the training and validation sets, effectively learning the characteristic patterns in the data while maintaining a certain level of generalization ability. The number of leaf nodes per tree ranged from 30 to 50. The number of leaf nodes is related to the tree's complexity and the model's generalization ability. An appropriate number of leaf nodes ensures that the model can effectively classify and predict the data without overfitting. During the model training process, the number of leaf nodes was dynamically adjusted and optimized based on the distribution and characteristics of the data, so that the number of leaf nodes in each tree was maintained between 30 and 50 to obtain better model performance.

[0028] The training sample set comprises a large collection of information on patients receiving high-dose methotrexate, including the aforementioned multimodal data, serum marker profiles, lifestyle information, medication history, and whether or not the patients experienced myelosuppression. These samples represent patients from diverse regions, age groups, and disease types. By learning and training on these samples, the model can identify potential relationships between different factors and the risk of myelosuppression. This allows the model to accurately predict the risk value and level of myelosuppression based on the input of new target users. For example, the training sample set includes data on 1,000 patients with hematologic malignancies receiving high-dose methotrexate, 300 of whom developed myelosuppression. Through analysis of this data and model training, the model's parameters and structure are continuously optimized, improving its predictive performance. During model training, the SHAP (SHapley Additive Explanations) method is used to calculate the contribution of each feature to the model's predictions, thereby assessing feature importance. A higher SHAP value indicates a more critical role for the feature in the model's predictions. By analyzing the SHAP values, we can understand the relative importance of different factors (such as age, serum markers, lifestyle habits, etc.) in predicting the risk of myelosuppression, providing a basis for further model interpretation and clinical application.

[0029] S102: Process the multimodal data information of the target user to generate a target physiological state prediction factor.

[0030] In one embodiment, multimodal data of a target user is processed to generate physiological status classification information and physiological characteristic influencing factors. A 55-year-old is in the transition period from middle age to old age. Physical function declines compared to young adults, impacting drug metabolism and tolerance to a certain extent, placing him or her in the moderate-risk physiological age category. The BMI, calculated based on weight and height, is within the normal range, indicating relatively balanced nutritional status and a good basis for drug tolerance, thus classifying him or her as nutritionally balanced. A red blood cell count is within the normal range, indicating that hematopoietic function is currently largely normal, placing him or her in the normal hematopoietic function category. A normal lymphocyte count reflects a stable immune system, placing him or her in the normal immune function category. The age factor is weighted at 0.2 to account for its potential impact on overall body function and drug metabolism. The BMI is weighted at 0.1 because it is within the normal range and has a relatively small impact on overall risk. The red blood cell count and lymphocyte count are each weighted at 0.15. Although within the normal range, they are still indicators of important physiological functions and provide a certain indication of the risk of myelosuppression. Weights for indicators such as serum potassium, urine white blood cells, high-density lipoprotein, serum magnesium, and uric acid were each set at 0.05. A urine white blood cell count slightly above the lower limit of the normal range suggests mild inflammation, necessitating appropriate attention to its potential impact on the bone marrow microenvironment. Other indicators were generally normal, and their impact on overall risk was relatively minimal.

[0031] Physiological status classification information and the target user's physiological characteristics influencing factors are processed to generate risk abnormality signature information. For the moderate-risk physiological age category, with a weight of 0.2, the overall baseline risk value increases by 20 points (assuming a full score of 100; this is a relative quantification). A nutritionally balanced BMI, with a weight of 0.1, has a minimal impact on risk and therefore does not increase the risk score. A normal red blood cell count (for hematopoietic function) and a normal lymphocyte count (for immune function), with a weight of 0.15 each, add little additional risk. Normal serum potassium, magnesium, and high-density lipoprotein (HDL) levels do not increase risk. However, a slightly elevated urine white blood cell count, with a weight of 0.05, increases risk by 3 points. A uric acid level of 400 μmol / L is considered high-normal, and with a weight of 0.05, increases risk by 2 points. Combining these factors, the generated risk abnormality signature information indicates that this user has a moderate risk of bone marrow suppression based on their baseline physiological status, with a total score of 25 points, primarily contributed by age and indicators such as urine white blood cells and uric acid.

[0032] The risk abnormality characteristic information is processed to generate a target physiological state prediction factor. The above risk abnormality characteristic information is further analyzed and standardized. Using a linear conversion formula, the converted target physiological state prediction factor is set to P, P = 0.3 + 0.003 × 25 = 0.375 (here 0.3 is the basic offset and 0.003 is the adjustment coefficient). This target physiological state prediction factor P will serve as one of the important inputs for the subsequent comprehensive assessment of bone marrow suppression risk, and will participate in the calculation of the final bone marrow suppression risk value and risk level together with other factors (such as serum marker influencing factors, lifestyle risk adjustment factors, etc.). The higher the value, the higher the risk of bone marrow suppression based on physiological status.

[0033] S103: Process the target user's serum marker characteristic information to generate a bone marrow suppression influencing factor.

[0034] In one embodiment, the serum marker characteristic information of the target user is processed to generate target serum marker characteristic information, wherein the target serum marker characteristic information includes alanine aminotransferase, lactate dehydrogenase and aspartate aminotransferase. Assume that the alanine aminotransferase (ALT) of this 55-year-old male is 45U / L, lactate dehydrogenase (LDH) is 220U / L, and aspartate aminotransferase (AST) is 40U / L. Compared with the normal reference range, ALT and AST are slightly higher than the upper limit of normal (generally ALT normal range is 0-40U / L, AST normal range is 0-37U / L), indicating that there is mild damage or inflammatory response in liver cells. LDH is also higher than the normal range (generally LDH normal range is 109-245U / L), reflecting a certain degree of damage or metabolic abnormality in tissue cells, and these three items are determined as target serum marker characteristic information.

[0035] The target serum marker characteristic information is mapped to generate a continuous low-dimensional vector space. Using dimensionality reduction algorithms such as principal component analysis (PCA), the target serum marker characteristic information (ALT, LDH, and AST values) is mapped into a three-dimensional, continuous low-dimensional vector space. In this space, each dimension represents a comprehensive feature, and its coordinate value reflects the relative magnitude and changing trend of these features.

[0036] Semantic analysis is performed on the target serum marker feature information to generate semantic information for each feature. For elevated ALT, the semantic information is "mild liver cell damage, affecting drug metabolism and increasing the risk of myelosuppression"; for elevated LDH, the semantic information is "tissue cell damage or metabolic abnormalities, associated with bone marrow cell status, potentially affecting myelosuppression"; and for elevated AST, the semantic information is "abnormal liver cell function, disrupting biochemical balance and affecting the bone marrow microenvironment." The semantic information of each feature and the continuous low-dimensional vector space are processed to generate a myelosuppression impact factor. This factor is calculated using a pre-trained machine learning model or a specific mathematical algorithm, taking into account the semantic information of each feature and its positional relationship in the continuous low-dimensional vector space. Assume that the model calculates a myelosuppression impact factor of 0.45 based on this information (the value ranges from 0 to 1, with 0 indicating little to no impact and 1 indicating a significant impact). This factor will serve as a key parameter in the subsequent myelosuppression risk calculation and will be used in conjunction with other factors to more accurately assess the risk of myelosuppression in this 55-year-old man.

[0037] For example, alanine aminotransferase (ALT) is 45 U / L, which accounts for 0.1 in the calculation after standardization and weight distribution; lactate dehydrogenase (LDH) is 220 U / L, accounting for 0.2; aspartate aminotransferase (AST) is 40 U / L, accounting for 0.1.

[0038] The bone marrow suppression impact factor is calculated using a pre-set weighted summation formula: Bone marrow suppression impact factor = (0.1 × ALT normalized value) + (0.1 × LDH normalized value) + (0.1 × AST normalized value) = (0.04 + 0.1 + 0.03 = 0.17). The calculated bone marrow suppression impact factor is 0.17, indicating that based on the comprehensive assessment of these serum markers, the user has a certain degree of bone marrow suppression risk. This factor, along with other factors, is further used in the final bone marrow suppression risk assessment calculation.

[0039] S104: Process the target user's lifestyle information to generate a myelosuppression risk adjustment factor.

[0040] In one embodiment, feature extraction is performed on the target user's lifestyle information to generate dietary intake features, exercise type features, drinking type features, and smoking type features. These dietary intake features, exercise type features, drinking type features, and smoking type features are then processed to generate dietary risk values, drinking risk values, smoking risk values, and dynamic feature weights. The dynamic feature weights are generated based on the lifestyle information of different target users over a preset time period. For this 55-year-old male, his daily diet consists of relatively low intake of vegetables and fruits, averaging approximately 1000 grams of vegetables and 500 grams of fruit per week. His intake of meat and fats is high, averaging approximately 1500 grams of meat and 200 grams of fat per week. This unbalanced diet leads to nutritional deficiencies (such as vitamins and minerals) and obesity. He typically engages in light exercise, such as walking three times a week for approximately 30 minutes each time, and rarely engages in high-intensity aerobic exercise or strength training. He has a drinking habit, averaging approximately 200 milliliters of alcohol per week, primarily drinking white wine, and smoking approximately 10 cigarettes per day.

[0041] Due to an unbalanced diet, a lack of important nutrients, and the risk of obesity, a dietary risk value of 0.4 was calculated based on a pre-set dietary risk assessment model. For example, insufficient intake of vitamins C, K, and dietary fiber affects the normal metabolism and hematopoietic function of bone marrow cells, while excessive intake of fats and meat triggers chronic inflammation and increases the risk of bone marrow suppression. Taking into account the high alcohol content of baijiu (white wine), as well as the frequency and amount of drinking, the drinking risk value is 0.3 according to the drinking risk assessment rules. Alcohol and its metabolites are toxic to organs such as the liver and bone marrow. Long-term drinking can cause liver damage, affect drug metabolism, and directly damage bone marrow cells. Based on research on the health risks of smoking and related assessment models, the smoking risk value for smoking 10 cigarettes per day is 0.35. Tobacco's harmful substances (such as nicotine and tar) can reach the bone marrow through the bloodstream, damaging bone marrow cells and suppressing its hematopoietic function.

[0042] By analyzing and statistically modeling the lifestyle data of a large number of different users over the past year (preset time period), we found that for 55-year-old men, the relative importance of diet, exercise, alcohol consumption, and smoking in terms of myelosuppression risk was 0.3, 0.2, 0.25, and 0.25, respectively. These weights vary based on factors such as age, gender, and medical background, and are continuously updated and optimized as new data accumulates. The risk adjustment factor (RAF) for myelosuppression is generated by processing the dietary risk value, exercise risk value, alcohol consumption risk value, smoking risk value, and dynamic feature weights. This RAF is calculated using a weighted summation approach. The formula is: RAF = Dietary risk value × Dietary weight + Exercise risk value × Exercise weight + Alcohol consumption risk value × Alcohol consumption weight + Smoking risk value × Smoking weight. Substituting these values ​​into the formula yields: RAF = 0.4 × 0.3 + 0.2 × 0.3 + 0.3 × 0.25 + 0.35 × 0.25 = 0.31. This bone marrow suppression risk adjustment factor will be used together with other factors (such as bone marrow suppression influencing factors, target physiological state predictors, etc.) to calculate the subsequent bone marrow suppression risk value and risk level to more comprehensively assess the risk of bone marrow suppression in this man.

[0043] S105: Process the myelosuppression influencing factor and the myelosuppression risk adjustment factor to generate dynamic risk parameter information of the target user.

[0044] In one embodiment, the bone marrow suppression impact factor and the bone marrow suppression risk adjustment factor are processed to generate a bone marrow hematopoietic stem cell morphological change value. Assume that the bone marrow suppression impact factor of the 55-year-old male was previously calculated to be 0.17 (as in the example above) and the bone marrow suppression risk adjustment factor was 0.31. First, the bone marrow suppression impact factor and the bone marrow suppression risk adjustment factor are processed to generate a bone marrow hematopoietic stem cell morphological change value. Using a pre-trained mathematical model based on a large amount of clinical data and experimental results, assuming the model is a linear weighted model, the two factors are weighted and summed and then transformed using a specific function. Let the bone marrow hematopoietic stem cell morphological change value be M, and the calculation formula is M = 0.4 × bone marrow suppression impact factor + 0.6 × bone marrow suppression risk adjustment factor + 0.1 (the coefficients and constants here are based on model training and empirical settings), then M = 0.4 × 0.17 + 0.6 × 0.31 + 0.1 = 0.252. This value represents the degree of morphological change of bone marrow hematopoietic stem cells under the influence of the current comprehensive factors. A larger value indicates a more significant morphological change and a higher risk of bone marrow suppression.

[0045] The morphological changes of bone marrow hematopoietic stem cells are processed to generate characteristic parameter information and dynamic bone marrow hematopoietic stem cell structural information for different testing cycles. Assume that the testing cycle is set to once a week for four consecutive weeks. In the first week, based on the M value and other relevant physiological data, a complex biological model calculates characteristic parameter information, such as a 10% decrease in cell proliferation rate, an 8% increase in cell apoptosis rate, and a 12% decrease in cell metabolic activity. Simultaneously, advanced imaging technology and image analysis algorithms are used to obtain dynamic bone marrow hematopoietic stem cell structural information, such as irregular nuclear morphology, a relatively reduced proportion of cytoplasm, and certain abnormalities in the number and function of mitochondria. Over the following weeks, as time passes and the patient's physical condition changes, these characteristic parameter information and structural information will continue to update and change.

[0046] The dynamic risk parameter information for the target user is processed by analyzing characteristic parameters from different detection cycles and dynamic bone marrow hematopoietic stem cell structure information. This is done using a time series analysis method and a recurrent neural network (RNN) model within a machine learning algorithm. The resulting dynamic risk parameter information is converted into a risk score ranging from 0 to 100. This man's dynamic risk parameter is 45. This score reflects the dynamic changes in the user's bone marrow suppression risk over time, providing doctors and researchers with a more comprehensive understanding of the patient's risk status and enabling timely intervention.

[0047] S106 , processing the preset high-dose methotrexate bone marrow suppression risk prediction model based on the training sample set to generate a target high-dose methotrexate bone marrow suppression risk prediction model.

[0048] In one embodiment, any number of data features in the training sample set are obtained, and a sampling ratio is generated based on the number of each data feature in the training sample set. The training sample set is processed based on the sampling ratio to generate a preset number of sampling features. From the training sample set of 1,000 patients, data features (such as age, weight, various blood indicators, medication status, etc.) of 200 patients are randomly selected for analysis. According to statistics, it is found that the age feature has 200 different values, the weight feature has 180 different values, and the serum marker feature (such as alanine aminotransferase, etc.) has 150 different values. Calculate the sampling ratio: for example, for the age feature, the sampling ratio = 200 (the number of age features in the sample extracted) / 1,000 (the total number of age features in the training sample set) = 0.2; for the weight feature, the sampling ratio = 180 / 1,000 = 0.18; and so on, to calculate the sampling ratio of each data feature.

[0049] Based on any data feature and each sampled feature, multiple data sets are generated. Each data set contains a preset number of data samples, at least one of which includes identification information. Based on the calculated sampling ratio, the training sample set is sampled with replacement. For example, for the age feature, 200 (with duplicates) are randomly selected from 1000 age data points at a sampling ratio of 0.2 as sampling features; for the weight feature, 180 are selected, and so on. Ultimately, a preset number of 500 sampling features are generated (the number 500 is set based on actual needs and model training experience). Then, any extracted data feature (such as a patient's complete information) is combined with each sampling feature for processing. For example, a patient's age, weight, serum markers, and other information are matched one by one with these 500 sampling features to form data sets. Each data set contains 10 data samples (also set based on model training requirements), and at least one data sample includes identification information indicating whether the patient has myelosuppression. This generates multiple data sets for subsequent model training.

[0050] The preset high-dose methotrexate bone marrow suppression risk prediction model is trained based on the data samples in the multiple data sets to generate a trained high-dose methotrexate bone marrow suppression risk prediction model. The trained high-dose methotrexate bone marrow suppression risk prediction model is processed based on the validation sample set to generate a validation result. If the data sample containing the identification information in the validation result is a risk factor that characterizes the risk of bone marrow suppression, the trained high-dose methotrexate bone marrow suppression risk prediction model is used as the target high-dose methotrexate bone marrow suppression risk prediction model. These data sets are used to train the preset high-dose methotrexate bone marrow suppression risk prediction model. During the training process, the model continuously learns the relationship between each feature in the data set and the bone marrow suppression result, and adjusts the node splitting rules and weight parameters of the decision tree. After training is completed, the trained model is processed using an independent validation sample set (assuming that it contains 200 patient information). The model predicts the patient data in the validation sample set and compares it with the actual identification information (whether bone marrow suppression occurs). For example, the model predicts that 80 patients will develop bone marrow suppression, while 75 patients in the actual validation sample set will develop bone marrow suppression. The verification results are evaluated by calculating indicators such as precision and recall.

[0051] In the verification results, patients correctly predicted by the model to have bone marrow suppression (i.e., data samples containing identification information and accurate model predictions) do indeed represent risk factors affecting bone marrow suppression (such as these patients having characteristics such as older age, abnormal serum markers, and unhealthy lifestyle habits). Therefore, this trained high-dose methotrexate bone marrow suppression risk prediction model is used as the target high-dose methotrexate bone marrow suppression risk prediction model to predict bone marrow suppression risk for target users such as 55-year-old men.

[0052] S107 , processing the target physiological state prediction factor and the dynamic risk parameter information of the target user based on the target high-dose methotrexate bone marrow suppression risk prediction model to generate a bone marrow suppression risk value and a bone marrow transplantation risk level.

[0053] In one embodiment, the target physiological state prediction factor and the dynamic risk parameter information of the target user are fused based on the target high-dose methotrexate bone marrow suppression risk prediction model to generate a fusion feature sequence. For a 55-year-old male target user, the data are as follows: the physiological state prediction factor PF = 0.375, which is generated by processing multimodal data information, comprehensively considering the influence of factors such as age, BMI, red blood cell count, and lymphocyte count on physiological state, reflecting that there is a certain risk tendency of bone marrow suppression in basic physiology. The bone marrow suppression influence factor MF = 0.35 is jointly determined by the serum marker characteristic information (alanine aminotransferase ALT = 45U / L, lactate dehydrogenase LDH = 220U / L, aspartate aminotransferase AST = 40U / L), indicating that there is a certain risk of bone marrow suppression under the action of these factors.

[0054] The bone marrow suppression risk adjustment factor AF = 0.31, which is calculated based on his lifestyle information (unbalanced diet, light exercise, drinking about 200 ml of liquor per week, and smoking about 10 cigarettes per day), reflecting the impact of bad lifestyle habits on the risk of bone marrow suppression. The medication information factor DF = 0.2 is determined by the interaction between the simultaneous use of antihypertensive drugs (nifedipine) and antibiotics (amoxicillin) and high-dose methotrexate. The comorbidity factor CF = 0.1, the man has no other obvious comorbidities. The medical history factor HF = 0.05, there is no history of bone marrow suppression or related blood system diseases. The model parameters are set as follows:

[0055] Weight coefficients: a1=0.2, a2=0.2, a3=0.15, a4=0.15, a5=0.1, a6=0.2.

[0056] Influence intensity coefficient: b1=0.4, b2=0.5, b3=0.3, b4=0.4, b5=0.2, b6=0.3.

[0057] Offset constants: c1 = 0.1, c2 = 0.2, c3 = 0.1, c4 = 0.15, c5 = 0.05, c6 = 0.1.

[0058] Each factor is input into the multi-layer perceptron (MLP) as a different feature channel. PF =ReLU(W PF ×PF+b PF) and other similar transformations (the same applies to other factors), and then concatenate them in the second layer of MLP to form a fusion feature sequence F = [y PF ,y MF ,y AF ,y DF ,y CF ,y HF The fused feature sequence is processed based on the target high-dose methotrexate myelosuppression risk prediction model to generate target classification header information. Features are further extracted and integrated through the fully connected layer and activation function layer, such as z = ReLU(W1×F+b1). The target classification header information H is then generated through the output layer W2×z+b2, which contains information such as the category and confidence level of myelosuppression risk.

[0059] The target classification header information is processed to generate a bone marrow suppression risk value and a bone marrow transplant risk level. The method includes a calculation formula for obtaining the bone marrow suppression risk value, which is: Risk value =(a1×

[0060]

[0061] Among them, PF represents physiological status predictor, MF represents bone marrow suppression influencing factor, AF represents bone marrow suppression risk adjustment factor, DF represents medication information factor, CF represents comorbidity factor, HF represents medical history factor, a1, a2, a3, a4, a5, a6 are weight coefficients of each factor in the overall risk assessment, the sum of which is 1; b1, b2, b3, b4, b5, b6 are the influence intensity coefficients of each factor, which are used to adjust the degree of influence of each factor on the risk value; c1, c2, c3, c4, c5, c6 are the offset constants of each factor. Substitute the above example data into the formula, specifically, Risk value =(0.2×0.598)+(0.2×0.530)+(0.15×0.629)+(0.15×0.593)+(0.1×0.934)+(0.2×0.891)=0.1196+0.106+0.09435+0.8895+0.934+0.1782=0.6795. According to the pre-defined risk classification rules (0-0.4 is low risk, 0.4-0.7 is medium risk, and 0.7 or above is high risk), this man's risk of myelosuppression is medium. This indicates that during high-dose methotrexate treatment, close monitoring of blood tests and other indicators is necessary. Based on these results, doctors can consider adjusting treatment plans or taking preventive measures, such as administering hematopoietic drugs, to reduce the incidence and severity of myelosuppression.

[0062] The server acquires a wide range of information covering the target user, including multimodal data such as age and weight, serum markers including specific enzyme indicators, lifestyle information covering diet, exercise, tobacco and alcohol consumption, medication details, as well as pre-set models and training samples. In the data processing pipeline, the multimodal data is first converted into physiological state classifications and characteristic influencing factors, and then the target physiological state prediction factor is generated. Serum markers are screened, analyzed, mapped, and semantically processed to derive myelosuppression influencing factors. Lifestyle information is extracted and calculated to determine the myelosuppression risk adjustment factor. The two are combined to generate dynamic risk parameter information. For model construction, data features are extracted from the training samples to determine the sampling ratio, generating sampling features and data sets. The pre-set model is trained and validated with a validation set, and the model that meets the requirements becomes the target model. Finally, the target model integrates relevant factors to generate sequence and classification header information. The myelosuppression risk value and level are calculated using a formula. The factor weights, strength coefficients, and offset constants in the formula work together to comprehensively consider multiple factors and effectively provide early warning of myelosuppression risk.

[0063] In one embodiment, Figure 2 As shown, the present application also provides a device for early warning of the risk of myelosuppression caused by treatment, comprising:

[0064] An acquisition module 201 is configured to acquire multimodal data information of a target user, serum marker characteristic information of the target user, lifestyle information of the target user, medication information of the target user, a preset high-dose methotrexate bone marrow suppression risk prediction model, and a training sample set, wherein the multimodal data information of the target user includes the target user's age, weight, serum creatinine, BMI, height, red blood cell count, lymphocyte count, potassium, urine white blood cells, high-density lipoprotein, magnesium, and uric acid;

[0065] Processing module 202 is used to process the multimodal data information of the target user to generate a target physiological state prediction factor; process the serum marker characteristic information of the target user to generate a bone marrow suppression influencing factor; process the living habit information of the target user to generate a bone marrow suppression risk adjustment factor; process the bone marrow suppression influencing factor and the bone marrow suppression risk adjustment factor to generate dynamic risk parameter information of the target user; process the preset high-dose methotrexate bone marrow suppression risk prediction model based on the training sample set to generate a target high-dose methotrexate bone marrow suppression risk prediction model; process the target physiological state prediction factor and the target user's dynamic risk parameter information based on the target high-dose methotrexate bone marrow suppression risk prediction model to generate a bone marrow suppression risk value and a bone marrow transplantation risk level.

[0066] Each embodiment of this application is described in a related manner. Similar portions between embodiments can be referenced to each other. Each embodiment focuses on the differences from other embodiments. In particular, the embodiments of the method, electronic device, electronic device, and readable storage medium for assessing the risk of myelosuppression due to treatment are generally similar to the aforementioned embodiments of the method for assessing the risk of myelosuppression due to treatment, so the description is relatively simple. For relevant portions, reference can be made to the description of the aforementioned embodiments of the method for assessing the risk of myelosuppression due to treatment.

Claims

1. A method for early warning of the risk of myelosuppression caused by treatment, characterized in that: include: Obtaining multimodal data information of a target user, serum marker characteristic information of the target user, lifestyle information of the target user, medication information of the target user, a preset high-dose methotrexate bone marrow suppression risk prediction model, and a training sample set, wherein the multimodal data information of the target user includes the target user's age, weight, blood creatinine, BMI, height, red blood cell count, lymphocyte count, potassium, urine white blood cells, high-density lipoprotein, magnesium, and uric acid; Processing the multimodal data information of the target user to generate a target physiological state prediction factor; Processing the target user's serum marker characteristic information to generate a bone marrow suppression influencing factor; Processing the target user's lifestyle information to generate a myelosuppression risk adjustment factor; Processing the myelosuppression influencing factor and the myelosuppression risk adjustment factor to generate dynamic risk parameter information of the target user; Processing the preset high-dose methotrexate bone marrow suppression risk prediction model based on the training sample set to generate a target high-dose methotrexate bone marrow suppression risk prediction model; The target physiological state prediction factor and the dynamic risk parameter information of the target user are processed based on the target high-dose methotrexate bone marrow suppression risk prediction model to generate a bone marrow suppression risk value and a bone marrow transplant risk level, including: fusing the target physiological state prediction factor and the dynamic risk parameter information of the target user based on the target high-dose methotrexate bone marrow suppression risk prediction model to generate a fusion feature sequence; processing the fusion feature sequence based on the target high-dose methotrexate bone marrow suppression risk prediction model to generate target classification header information; processing the target classification header information to generate a bone marrow suppression risk value and a bone marrow transplant risk level; the method includes a calculation formula for obtaining the bone marrow suppression risk value, The calculation formula is: ； Among them, PF represents physiological status predictor, MF represents bone marrow suppression influencing factor, AF represents bone marrow suppression risk adjustment factor, DF represents medication information factor, CF represents comorbidity factor, and HF represents medical history factor. is the weight coefficient of each factor in the overall risk assessment, the sum of which is 1; is the influence intensity coefficient of each factor, Used to adjust the impact of each factor on the risk value; it is the offset constant of each factor.

2. The method according to claim 1, wherein Processing the multimodal data information of the target user to generate a target physiological state prediction factor includes: Processing the multimodal data information of the target user to generate physiological state classification information and physiological characteristic influencing factors of the target user; Processing the physiological status classification information and the physiological characteristic influencing factors of the target user to generate risk abnormality characteristic information; The risk abnormality characteristic information is processed to generate a target physiological state prediction factor.

3. The method according to claim 1, wherein Processing the target user's serum marker characteristic information to generate a bone marrow suppression influencing factor includes: Processing the target user's serum marker characteristic information to generate target serum marker characteristic information, wherein the target serum marker characteristic information includes alanine aminotransferase, lactate dehydrogenase, and aspartate aminotransferase; Mapping the characteristic information of target serum markers to generate a continuous low-dimensional vector space; Perform semantic analysis on the target serum marker feature information to generate semantic information for each feature information; The semantic information and continuous low-dimensional vector space of each feature information are processed to generate the bone marrow suppression influencing factor.

4. The method according to claim 3, wherein Processing the target user's lifestyle information to generate a myelosuppression risk adjustment factor includes: Performing feature extraction processing on the target user's lifestyle information to generate dietary intake features, exercise type features, drinking type features, and smoking type features; Processing the dietary intake characteristics, the exercise type characteristics, the drinking type characteristics, and the smoking type characteristics to generate a dietary risk value, a drinking risk value, a smoking risk value, and a dynamic feature weight value, wherein the dynamic feature weight value is generated based on the living habit information of different target users within a preset time period; The diet risk value, the exercise risk value, the drinking risk value, the smoking risk value and the dynamic feature weight value are processed to generate a bone marrow suppression risk adjustment factor.

5. The method according to claim 4, wherein The myelosuppression influencing factor and the myelosuppression risk adjustment factor are processed to generate dynamic risk parameter information of the target user, including: Processing the bone marrow suppression influencing factors and the bone marrow suppression risk adjustment factors to generate a bone marrow hematopoietic stem cell morphology change value; Processing the bone marrow hematopoietic stem cell morphological change values ​​to generate characteristic parameter information of different detection cycles and dynamic bone marrow hematopoietic stem cell structural information; The characteristic parameter information of different detection cycles and the structural information of dynamic bone marrow hematopoietic stem cells are processed to generate dynamic risk parameter information of the target user.

6. The method according to claim 1, wherein The preset high-dose methotrexate bone marrow suppression risk prediction model is processed based on the training sample set to generate a target high-dose methotrexate bone marrow suppression risk prediction model, including: Get any number of data features in the training sample set; Generate a sampling ratio based on the number of each data feature in the training sample set; Processing the training sample set based on the sampling ratio to generate a preset number of sampling features; Processing any data feature and each sampling feature to generate multiple data groups, wherein each data group includes a preset number of data samples, and at least one data sample includes identification information; Training the preset high-dose methotrexate bone marrow suppression risk prediction model based on data samples in multiple data groups to generate a trained high-dose methotrexate bone marrow suppression risk prediction model; processing the trained high-dose methotrexate bone marrow suppression risk prediction model based on the validation sample set to generate a validation result; If the data sample containing identification information in the verification result is a risk factor representing an impact on bone marrow suppression, the trained high-dose methotrexate bone marrow suppression risk prediction model is used as the target high-dose methotrexate bone marrow suppression risk prediction model.

7. A device for early warning of the risk of myelosuppression caused by treatment, characterized in that: For implementing the method of claim 1, the apparatus comprises: an acquisition module, configured to acquire multimodal data information of a target user, serum marker characteristic information of the target user, lifestyle information of the target user, medication information of the target user, a preset high-dose methotrexate bone marrow suppression risk prediction model, and a training sample set, wherein the multimodal data information of the target user includes the target user's age, weight, serum creatinine, BMI, height, red blood cell count, lymphocyte count, potassium, urine white blood cells, high-density lipoprotein, magnesium, and uric acid; A processing module is used to process the multimodal data information of the target user to generate a target physiological state prediction factor; process the serum marker characteristic information of the target user to generate a bone marrow suppression influencing factor; process the living habit information of the target user to generate a bone marrow suppression risk adjustment factor; process the bone marrow suppression influencing factor and the bone marrow suppression risk adjustment factor to generate dynamic risk parameter information of the target user; process the preset high-dose methotrexate bone marrow suppression risk prediction model based on the training sample set to generate a target high-dose methotrexate bone marrow suppression risk prediction model; process the target physiological state prediction factor and the dynamic risk parameter information of the target user based on the target high-dose methotrexate bone marrow suppression risk prediction model to generate a bone marrow suppression risk value and a bone marrow transplantation risk level.

8. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; Wherein, the first processor is configured to execute the method for early warning of the risk of myelosuppression caused by treatment 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, the method for early warning of the risk of myelosuppression caused by treatment according to any one of claims 1 to 6 is implemented.

Images