Intelligent evaluation tool for dystrophy risk of nasopharyngeal carcinoma patient

Abstract

The invention provides a tool for intelligently evaluating the dystrophy risk of a patient with nasopharyngeal carcinoma, and the tool for evaluating the specific dystrophy of nasopharyngeal carcinoma is constructed by integrating indexes such as social demographic data, physique measurement, tumor staging, chemoradiotherapy conditions, recent adverse reactions and hematology indexes, so that high-precision risk evaluation is realized. Through single-factor and multi-factor analysis, the variable weight is optimized, and the discrimination capability and stability of the model are improved. An intelligent applet based on artificial intelligence is developed, patient data are extracted, nutrition risk calculation is completed within seconds, evaluation time is shortened, and clinical working efficiency is improved. The system also has the functions of dynamic monitoring and intelligent early warning, automatically updates the evaluation result along with the treatment progress of the patient, and provides individualized intervention suggestions. The defects that a traditional nutrition evaluation tool is high in subjectivity, low in efficiency, lack of tumor specificity and low in informatization degree are overcome, and an efficient, accurate and automatic intelligent evaluation solution is provided for clinic.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent evaluation, in particular to an intelligent evaluation tool for malnutrition risk of nasopharyngeal carcinoma patients. BACKGROUND

[0002] With the development of comprehensive treatment of tumors, the survival period of patients is significantly prolonged, but malnutrition is still one of the key factors affecting treatment compliance, immune function and overall prognosis. The invention shows that the incidence of malnutrition in patients with malignant tumors can reach 40%-80%, which not only increases the toxic side effects of chemotherapy, prolongs the hospitalization time, but also reduces the quality of life and survival rate. In patients with nasopharyngeal carcinoma (NPC), the problem of malnutrition is particularly prominent. Nasopharyngeal carcinoma mainly occurs in Southeast Asia and southern China, and its treatment is mainly radiotherapy and chemotherapy, but radiotherapy and chemotherapy often cause adverse reactions such as dry mouth, dysphagia, taste disorder, nausea and mucositis, which significantly affect the patient's food intake and nutrient absorption, leading to a high incidence of malnutrition of about 60%.

[0003] The existing malnutrition evaluation tools all have limitations: (1) PG-SGA (patient self-evaluation subjective overall assessment) is one of the recommended tools for nutritional evaluation of tumor patients, covering seven dimensions such as body weight, food intake, symptoms, activity function, disease-nutrition relationship, metabolic demand and physical examination. However, its operation is complex, time-consuming, and has more subjective components, which limits its promotion and application in clinical practice. (2) MNA is mainly aimed at the elderly population; although it has high sensitivity in adult tumor patients, its specificity is moderate, and its effectiveness is lower in Asian tumor population; at the same time, the peak incidence of NPC is mostly seen in 40-59 years old, and the age structure does not completely match the target population of MNA. (3) GLIM is a diagnostic framework proposed in 2018 and is being verified in various clinical scenarios. However, some inventions using PG-SGA as a reference show that the consistency of GLIM and PG-SGA in tumors is low, and the sensitivity is greatly affected by the screening tool used. There is no uniform consensus on the diagnostic threshold of muscle mass decline in GLIM patients, resulting in poor consistency of evaluation results, and GLIM is difficult to fully reflect the nutritional metabolism characteristics of tumor patients. In addition, in some areas, BMI or serum albumin is still used as a single indicator for judgment, which is obviously insufficient in accuracy.

[0004] In summary, the existing nutritional evaluation technology has the following main problems: (1) lack of tumor type specificity; (2) complex and subjective evaluation; (3) imperfect variable system; (4) limited diagnostic accuracy; (5) insufficient clinical accessibility.

[0005] Existing technical solutions: (1) PG-SGA (patient self-evaluation subjective overall assessment) The PG-SGA is a recommended tool for the assessment of malnutrition in cancer patients, which combines patient self-assessment and medical staff physical examination, including seven parts: weight, food intake, symptoms, activity function, disease metabolic demand, etc. It has high sensitivity and predictive ability, but the content is complex, time-consuming, and the clinical operation is complicated, which is not suitable for rapid assessment in clinical practice, and the assessment results are still affected by subjective factors.

[0006] (2) MNA (Mini Nutritional Assessment) MNA is mainly used for the elderly population, covering physical measurement, dietary intake, living conditions and self-perception. Its advantages are simple to use, can identify nutritional risk early, and is suitable for community and hospital environment. However, MNA is designed for the elderly population, and its diagnostic performance in young and middle-aged cancer patients is low, with insufficient specificity (about 0.70-0.72), and some items rely on patients' subjective answers, lacking consideration of tumor metabolic characteristics and treatment factors, limiting its applicability in cancer patients.

[0007] (3) GLIM (Global Leadership in Clinical Nutrition Initiative Criteria) GLIM criteria were proposed by the International Society of Clinical Nutrition in 2018, combining phenotypic criteria (weight loss, BMI reduction, muscle loss) and etiological criteria (food intake reduction, inflammation). Its advantages are high degree of structuring, international applicability, and can be applied to various cancer patients. However, GLIM has low consistency with PG-SGA in cancer, and its sensitivity is greatly affected by the screening tool used. There is no consensus on the diagnostic threshold of muscle mass loss in GLIM, resulting in poor consistency of assessment results, and GLIM cannot fully reflect the nutritional and metabolic characteristics of cancer patients, limiting its application in the field of cancer nutrition.

[0008] Based on risk factors, constructing disease diagnosis or assessment tools has become one of the important directions of modern precision medicine, and has been widely used in various disease fields. For example, the Framingham cardiovascular risk score model integrates age, gender, blood pressure, cholesterol, and smoking as risk factors to quantitatively predict the occurrence of cardiovascular events, which has been verified and promoted worldwide. In the field of diabetes, the inventors established a risk prediction tool based on body mass index, fasting blood glucose, and family history as the main variables for early screening and health intervention. In the field of oncology, breast cancer recurrence risk score, lung cancer recurrence model, and gastric cancer postoperative complication prediction model are all based on clinical and laboratory risk factors. These achievements show that by quantifying risk factors and integrating multi-dimensional data, early identification, risk stratification and precision intervention of diseases can be effectively achieved.

[0009] The present application draws on mature technologies in other medical fields, and aims at the clinical pain point of malnutrition assessment of nasopharyngeal carcinoma patients, constructs an evaluation tool based on multi-dimensional risk factors, and innovatively develops an intelligent applet system. Doctors or nurses only need to input the basic information of the patient, and the patient's nutritional risk level and intervention suggestions can be obtained within a few seconds, which provides an efficient, objective and generalizable technical path for clinical nutrition management. Therefore, the present application proposes an intelligent malnutrition risk assessment tool for nasopharyngeal carcinoma patients. SUMMARY

[0010] In view of the deficiencies of the existing nutrition assessment methods, the purpose of the present application is to provide an intelligent malnutrition assessment tool for nasopharyngeal carcinoma patients based on risk factors, so as to realize rapid, objective and dynamic nutrition risk identification and management. (1) The present application constructs a specific malnutrition assessment tool for nasopharyngeal carcinoma by comprehensively considering social demographic data, physical measurements, tumor stage, radiotherapy and chemotherapy, recent adverse reactions and hematological indicators, so as to realize high-precision risk assessment. (2) Through single factor and multi-factor analysis, the multi-source clinical data are trained and verified, the variable weight is optimized, and the discrimination ability and stability of the model are improved. At the same time, an intelligent applet based on artificial intelligence is developed, which can extract patient data in real time and complete nutrition risk calculation within a few seconds, greatly shortening the evaluation time and improving the clinical work efficiency. (3) The system also has dynamic monitoring and intelligent early warning functions, which can automatically update the evaluation results as the patient's treatment progresses, and provide individualized intervention suggestions. Through the above innovations, the present application overcomes the defects of traditional nutrition assessment tools such as strong subjectivity, low efficiency, lack of tumor specificity and low informatization, and provides an efficient, accurate and automated intelligent assessment solution for clinical use.

[0011] Through literature review, it is found that there is no gold standard for evaluating malnutrition, and there is no specific evaluation tool for nasopharyngeal carcinoma patients. The commonly used malnutrition assessment tools in clinical practice, such as patient self-assessment subjective overall assessment (PG-SGA), global clinical nutrition leadership initiative (GLIM) and mini nutritional assessment scale (MNA), all lack the integration of tumor treatment related variables (such as radiotherapy and chemotherapy dose and cycle) and biochemical indicators, resulting in insufficient sensitivity and specificity of the evaluation results. In addition, although PG-SGA is recommended for the evaluation of malnutrition in cancer patients, it is time-consuming and depends on subjective judgment, which makes it difficult to be widely used in clinical practice. MNA is an evaluation tool for malnutrition in the elderly. GLIM is the latest malnutrition assessment tool, but it has not been corrected for the characteristics of Asian population, and there is no uniform consensus on the diagnostic threshold of muscle mass decline in GLIM patients, resulting in poor consistency of the evaluation results, and GLIM is difficult to fully reflect the nutritional metabolism characteristics of tumor patients.

[0012] In view of the above problems, the technical goal of the present application is: The application integrates social demographic data, physical measurement, tumor staging, radiotherapy and chemotherapy, recent adverse reactions and hematological indicators and other multi-dimensional indicators, and invents a specific evaluation tool for malnutrition of nasopharyngeal carcinoma patients based on key risk factors, and designs it into an intelligent computing program. It has been verified that the tool has good discrimination, calibration and short evaluation time, can accurately and efficiently evaluate the nutritional status of nasopharyngeal carcinoma patients, and can be popularized in clinical use. The application solves the problems of the existing evaluation tools, such as not including tumor-related indicators, complicated operation, long time consumption, relying on subjective judgment, lack of diagnostic threshold for diagnostic indicators, low sensitivity and specificity, and lack of specificity for nasopharyngeal carcinoma patients.

[0013] In order to achieve the above-mentioned application purposes, the application provides the following technical solutions: a nasopharyngeal carcinoma patient malnutrition risk intelligent evaluation tool, which is characterized by comprising a malnutrition prediction model constructed based on Logistic regression analysis, a nutritional status score calculation equation corresponding to the model, an optimal risk judgment threshold, and an intelligent calculation module developed based on the equation; the prediction model is constructed by screening independent risk factors of nasopharyngeal carcinoma patients, and the independent risk factors include age increase, tumor stage IV, chemotherapy cycle ≥1, radiotherapy frequency ≥1, low BMI, appetite reduction in the past week, and prealbumin ≤250 mg / L.

[0014] As a preferred technical solution of the application, the nutritional status score calculation equation is: nutritional score = 16.372 + 1.331 x (chemotherapy ≥1) + 1.017 x (radiotherapy ≥1) + 0.052 x age + 1.124 x (stage IV) + 2.287 x (appetite reduction) - 0.570 x BMI - 0.950 x (prealbumin > 250); wherein, "chemotherapy ≥1", "radiotherapy ≥1", "stage IV", "appetite reduction" and "prealbumin > 250" are binary variables, taking 1 when meeting the conditions and 0 when not meeting the conditions; the unit of age is year, and the unit of BMI is kg / m².

[0015] As a preferred technical solution of the application, the optimal risk judgment threshold is 6.6586, when the nutritional score of the patient is ≥6.6586, it is determined that there is a risk of malnutrition; the threshold is determined by ROC curve analysis method, and the maximum Youden index is taken as the cutoff value selection standard, corresponding to the sensitivity 0.91 and the specificity 0.87.

[0016] As a preferred technical scheme of the present application, the construction process of the prediction model comprises: adopting a prospective cross-sectional diagnostic invention design, enrolling 668 newly diagnosed nasopharyngeal carcinoma patients in the first stage, collecting 22 potential variables, taking PG-SGA≥2 points as the malnutrition judgment standard, screening out 7 independent risk factors and parameterizing through single factor and multi-factor Logistic regression analysis, and obtaining the nutritional status score calculation equation; in the second stage, the model performance is verified from the aspects of discrimination and calibration, the discrimination is evaluated through the ROC curve and AUC value, and the calibration is verified through the calibration curve and Hosmer-Lemeshow goodness-of-fit test.

[0017] As a preferred technical scheme of the present application, the AUC value of the prediction model in the training set is 0.918, the calibration curve is close to the 45° ideal reference line, the Hosmer-Lemeshow goodness-of-fit test p>0.05, indicating that the model has good discrimination ability and calibration performance.

[0018] As a preferred technical scheme of the present application, the intelligent computing module is an intelligent applet embedded in the hospital information system, and the use steps thereof comprise: checking or leaving blank the binary variable option, the binary variable option comprising "chemotherapy cycle≥1", "radiotherapy frequency≥1", "tumor stage IV", "appetite loss in the past week", "prealbumin>250mg / L"; inputting the age and BMI values of the patient; clicking the calculation button to automatically analyze the nutritional status score of the patient.

[0019] As a preferred technical scheme of the present application, the intelligent computing module can realize rapid, objective and standardized evaluation of the malnutrition risk of nasopharyngeal carcinoma patients, and overcome the problems of long time consumption, strong subjectivity and complex operation of traditional evaluation tools.

[0020] An evaluation method of a nasopharyngeal carcinoma patient malnutrition risk intelligent evaluation tool, comprising the following steps: S1, collecting the age, BMI, tumor stage, chemotherapy cycle number, radiotherapy frequency, appetite status in the past week and prealbumin level data of the nasopharyngeal carcinoma patient; S2, inputting the collected data into the intelligent computing module, or completing data input through the option checking and value input of the intelligent computing module; S3, the intelligent computing module automatically calculates the nutritional score of the patient according to the nutritional status score calculation equation; S4, determining whether the patient has a malnutrition risk according to the comparison result of the nutritional score and the optimal risk judgment threshold.

[0021] Compared with the prior art, the present application has the following advantages: (1) Improve the accuracy and pertinence of nasopharyngeal carcinoma malnutrition evaluation The existing nutrition assessment tools are mostly general scales, which fail to combine the disease characteristics of nasopharyngeal carcinoma and the radiotherapy and chemotherapy related indexes. The present application takes the specific risk factors of nasopharyngeal carcinoma patients as core variables (including radiotherapy and chemotherapy cycles, tumor stage, BMI, age, prealbumin and appetite status, etc.), and establishes a prediction model through Logistic regression, which significantly improves the specificity and accuracy of the malnutrition assessment of nasopharyngeal carcinoma patients, with the AUC value being above 0.85, which is superior to the existing general tools.

[0022] (2) Significantly shorten the evaluation time, realize automation and standardization The traditional nutrition assessment needs to be filled in the questionnaire manually and judged subjectively, which is time-consuming and has poor consistency. The present application develops an intelligent calculation module based on mathematical equations, which can automatically analyze and evaluate the results only by inputting the clinical data of the patients, the whole process only takes a few seconds, avoids manual errors, and realizes the rapidity, objectivity and standardization of the evaluation process.

[0023] (3) Good clinical generalizability and system compatibility The parameters used in the present application are all clinically routine obtainable indexes, which can be seamlessly connected with the hospital electronic medical record system (EMR) and the laboratory information system (LIS), realize automatic data extraction and real-time calculation, and are convenient for popularization and application in different medical institutions.

[0024] (4) Dynamic monitoring and risk warning function can be realized The system of the present application can dynamically update the nutrition assessment results according to the treatment progress and index changes of the patients, and intelligently warn and prompt the patients with increased risk of malnutrition, which assists the medical staff to intervene in time, helps to reduce the incidence of malnutrition and improve the prognosis of the patients. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The ROC curve area diagram of the evaluation tool provided by the present application is shown; Figure 2 The calibration curve diagram of the evaluation tool provided by the present application is shown; Figure 3 The interface diagram of the nasopharyngeal carcinoma patient malnutrition tool applet provided by the present application is shown; Figure 4 The interface diagram of the English version of the nasopharyngeal carcinoma patient malnutrition tool applet provided by the present application is shown. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments.

[0027] Therefore, the following detailed description of the embodiments of the application is not intended to limit the scope of the application as claimed, but merely represents some embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the application. It should be noted that the embodiments in the application and the features and technical solutions in the embodiments can be combined with each other without conflict, and similar reference numbers and letters represent similar items in the following drawings, so that once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0028] Embodiment 1: The application proposes an intelligent assessment tool for malnutrition risk of nasopharyngeal carcinoma patients. The tool is established by analyzing large sample clinical data, combining logistic regression and artificial intelligence algorithm, and is suitable for nasopharyngeal carcinoma patients. The tool realizes rapid, objective and standardized malnutrition assessment through intelligent program, and overcomes the problems of long time consumption, strong subjectivity and complex operation of the existing PG-SGA, GLIM, MNA and other traditional tools.

[0029] Invention method and results: (1) Invention design: The model construction of the application is based on a prospective cross-sectional diagnostic invention design. The design can evaluate the nutritional status of patients and measure potential risk factors at the same time point, ensure the timeliness and representativeness of data, and quickly identify malnutrition-related variables. The whole invention is divided into two stages: The first stage: risk factor identification and tool development: 668 newly diagnosed nasopharyngeal carcinoma patients were enrolled in the First Affiliated Hospital of Guangxi Medical University, and continuous sampling method was used. 22 potential variables were collected through questionnaire survey, physical measurement, laboratory detection and clinical records, including social demographic characteristics, body measurement data (BMI, calf circumference, grip strength), lifestyle (smoking, drinking), tumor stage, treatment parameters (number of chemotherapy and radiotherapy), adverse reactions (oral mucositis, dysphagia, anorexia, etc.) and biochemical indicators (albumin, prealbumin, chloride, etc.). The nutritional status was evaluated by PG-SGA as the standard assessment tool. PG-SGA≥2 points was used as the malnutrition judgment standard, and single factor and multiple factor logistic regression analysis was used to screen independent risk factors. Finally, 7 independent risk factors were established: (1) increasing age; (2) tumor stage IV; (3) chemotherapy cycle≥1; (4) radiotherapy frequency≥1; (5) low BMI; (6) anorexia in the past week; (7) prealbumin≤250mg / L.

[0030] The above risk factor parameters are parameterized to obtain two kinds of nutrition prediction equations: Final Nutritional Assessment Tool (FNA) for nasopharyngeal carcinoma patients: Nutrition score = 16.372 + 1.331 x (chemotherapy≥1) + 1.017 x (radiotherapy≥1) + 0.052 x age + 1.124 x (stage IV) + 2.287 x (loss of appetite) - 0.570 x BMI - 0.950 x (prealbumin>250) To determine the optimal risk determination threshold, the present application uses the receiver operating characteristic curve (ROC) analysis method, and takes the maximum Youden index as the cutoff value selection standard. (1) Taking PG-SGA≥2 as the gold standard for determining "malnutrition"; (2) Using MedCalc software to draw the ROC curve of FNA; (3) Calculate the sensitivity and specificity corresponding to each possible score; (4) Determine the maximum Youden index as the best cutoff point. The cutoff value determination result: the best cutoff value of FNA is 6.6586, corresponding to a sensitivity of 0.91 and a specificity of 0.87; Second stage: verification of the tool To verify the predictive performance of the newly developed nasopharyngeal carcinoma patient malnutrition assessment tool (FNA), the present application comprehensively evaluates from the aspects of discrimination and calibration.

[0031] Discrimination is used to measure the ability of the model to distinguish between well-nourished and malnourished patients. The model prediction score is matched with the true nutritional status based on PG-SGA (≥2 for malnutrition), and the receiver operating characteristic curve (ROC) is drawn using MedCalc software (version 20.100, Belgium). The area under the curve (AUC) is calculated. The closer the AUC value is to 1, the stronger the model discrimination ability.

[0032] Calibration is used to evaluate the consistency between the model's predicted malnutrition probability and the actual incidence. The rms package in R language (version 4.2.1) is used to draw the calibration plot. The specific method is as follows: according to the model's predicted nutritional risk probability, the samples are divided into 10 quantile groups, the actual incidence and average prediction probability of each group are calculated, and the curve is drawn with the prediction value as the horizontal coordinate and the observation value as the vertical coordinate. The 45° diagonal line is used as the ideal reference line. When the curve is close to the 45° line, it indicates that the tool has good calibration performance. In addition, through the Hosmer-Lemeshow goodness-of-fit test (p>0.05), it is further verified that there is no significant difference between the predicted probability and the actual probability, and the model fitting effect is good.

[0033] The results showed that: in terms of discrimination, the newly developed nutritional assessment tool (FNA) showed excellent discriminant ability in different data sets. In the training set, the AUC value of FNA was 0.918; in terms of calibration, the predicted probability of FNA was highly consistent with the actual observed probability, and the calibration curve was close to the ideal 45° diagonal line, indicating that the model had good prediction consistency.

[0034] The third stage of nasopharyngeal carcinoma patient malnutrition risk intelligent program design In order to facilitate the calculation of the nutritional status score of nasopharyngeal carcinoma patients, the present application designs a nutritional status score calculator (see Figure 3 ) based on mathematical equations. The following is the use instruction of the calculation program: If the patient has completed at least one cycle of chemotherapy, check the "chemotherapy cycle ≥1"; if the patient has received at least one time of radiotherapy, check the "radiotherapy times ≥1"; if the patient's tumor stage is stage IV, check "tumor stage IV"; if the patient has anorexia in the past week, check "anorexia"; if the patient's prealbumin level is >250 mg / L, check "prealbumin (PA) >250 mg / L". If the above conditions are not met, the corresponding options are left blank. Then, input the patient's age and body mass index (BMI) in the designated column, and finally click the "calculate" button to automatically analyze the patient's nutritional status score.

[0035] For example, a 51-year-old male nasopharyngeal carcinoma patient, tumor stage III, BMI 23.7 kg / m². The patient has completed two cycles of chemotherapy and has received five times of radiotherapy. In the past week, he has anorexia, and his prealbumin level is 208.7 mg / L. Using the calculator to calculate, the nutritional status score is 10.15. The use steps of the nutritional status score calculator are as shown in Figure 3 .

[0036] Figure 4 The English version of the nasopharyngeal carcinoma patient malnutrition risk intelligent program is linked, and the user clicks to open, inputs or checks the patient variables (such as age, BMI, radiotherapy and chemotherapy, prealbumin value, etc.) to automatically analyze the nutritional status score within a few seconds.

[0037] (1) PG-SGA (patient self-assessment subjective overall assessment) PG-SGA is often used as a recommended tool for nutritional assessment of tumor patients, combining patient self-assessment and physical examination by medical staff, including body weight, eating habits, symptoms, activity function, disease metabolic demand, etc. Seven parts. It has high sensitivity and prediction ability, but the content of PG-SGA is too much, time-consuming, and the clinical operation is complicated, which is not suitable for rapid assessment in clinical practice, and the assessment results are still affected by subjective factors.

[0038] (2) MNA (Mini Nutritional Assessment) MNA is mainly used for the elderly population, covering four aspects of physical measurement, dietary intake, living conditions and self-perception. Its advantages are simple to use, can identify nutritional risk early, and is suitable for community and hospital environment. However, MNA is designed for the elderly population, and its diagnostic performance in young and middle-aged cancer patients is low, with insufficient specificity (about 0.70-0.72), and some items rely on the subjective answers of patients, lacking consideration of tumor metabolic characteristics and treatment factors, limiting its applicability in cancer patients.

[0039] (3) GLIM (Global Leadership in Clinical Nutrition Initiative Criteria) The GLIM criteria were proposed by the International Society for Clinical Nutrition in 2018, combining phenotypic criteria (weight loss, BMI reduction, muscle loss) with etiological criteria (food intake reduction, inflammation). Its advantages are high degree of structuring, international applicability, and can be applied to various cancer patients. However, the consistency of GLIM and PG-SGA in tumors is low, and the sensitivity is greatly affected by the screening tool used. There is no consensus on the diagnostic threshold of muscle mass loss in GLIM, resulting in poor consistency of the evaluation results, and GLIM cannot fully reflect the nutritional metabolism characteristics of cancer patients, limiting its application in the field of tumor nutrition.

[0040] Based on risk factors, the construction of disease diagnosis or evaluation tools has become one of the important invention directions of modern precision medicine, and has been widely used in many disease fields. For example, the Framingham cardiovascular risk score model integrates age, gender, blood pressure, cholesterol and smoking, etc. risk factors to quantitatively predict the occurrence of cardiovascular events, which has been verified and promoted worldwide. In the field of diabetes, the inventor established a risk prediction tool based on body mass index, fasting blood glucose and family history as the main variables for early screening and health intervention. In the field of oncology, breast cancer recurrence risk score, lung cancer recurrence model and gastric cancer postoperative complication prediction model are all based on clinical and laboratory risk factors. These achievements show that by quantifying risk factors and integrating multi-dimensional data, early identification, risk stratification and precision intervention of diseases can be effectively achieved.

[0041] This invention draws on mature technologies in other medical fields, and constructs an evaluation tool based on multi-dimensional risk factors for the clinical pain points of malnutrition assessment in nasopharyngeal carcinoma patients, and innovatively develops an intelligent applet system. Doctors or nurses only need to input the patient's basic information, and the patient's nutritional risk level and intervention suggestions can be obtained within a few seconds, providing an efficient, objective and promotable technical path for clinical nutrition management.

[0042] (1) Construction of a nasopharyngeal carcinoma malnutrition prediction model based on clinical risk factors Multidimensional clinical data of NPC patients (including radiotherapy and chemotherapy cycles, tumor stage, BMI, age, anorexia, and prealbumin levels) were collected by the system. Single-factor and multi-factor logistic regression analysis was used to screen significant risk factors, establish a specific nutritional risk prediction model for NPC, and achieve quantitative risk assessment.

[0043] (2) Innovative design and modeling of FNA tools Based on the model, the “Final Nutritional Assessment (FNA)” was developed, which converted each risk factor into a weighted scoring system through a regression equation and formed a nutritional status prediction equation, enabling automatic calculation and visual presentation of individual nutritional risk.

[0044] (3) Scientificity and reliability of model performance verification ROC curve and area under curve (AUC) were used to evaluate the model's discriminative ability, and calibration curve and Hosmer-Lemeshow test were used to verify the model's fitting degree. The results showed that the AUC of FNA in the training set, internal validation set, and external validation set was 0.918, 0.906, and 0.849, respectively, and the Hosmer-Lemeshow test p>0.05, indicating that the model had high accuracy and stability.

[0045] (4) Intelligent calculation module based on the equation Based on the FNA mathematical model, the invention developed an intelligent nutritional assessment applet that can be embedded in the hospital information system. By inputting or checking the patient's variables (such as age, BMI, radiotherapy and chemotherapy, prealbumin value, etc.), the nutritional status score can be automatically analyzed within seconds, achieving rapid, objective, and standardized identification of nutritional risk.

[0046] In the preferred embodiment, the present application uses a logistic regression model to model the multi-dimensional risk factors of nasopharyngeal carcinoma patients to construct a malnutrition prediction equation. However, without changing the technical purpose and implementation principle of the present application, other algorithms can also be used as equivalent alternatives to achieve the same malnutrition risk assessment function. For example, ridge regression or LASSO regression can be selected to control the multicollinearity between variables by introducing a regularization term; integrated algorithms such as random forest or extreme gradient boosting (XGBoost) can also be used to model the clinical multi-source data non-linearly; or algorithms such as support vector machine (SVM), Bayesian logistic regression, and shallow neural network can be used to train and classify the variables of patients' radiotherapy cycles, tumor stages, body mass index (BMI), prealbumin, appetite status, and age, thereby outputting the malnutrition risk probability. The above different algorithms differ in calculation path and parameter processing, but achieve the same function as the logistic regression model, which is based on the statistical learning principle to allocate weights to input variables and estimate risks, and can complete the determination and stratified evaluation of malnutrition in nasopharyngeal carcinoma patients. Therefore, these algorithms all belong to equivalent alternatives to the logistic regression model, do not change the core technical idea and implementation goal of the present application, and should be included in the protection scope of the present application.

[0047] The above embodiments are only used to illustrate the technical solutions described in the present application and do not limit the present application. Although the present application has been described in detail with reference to the above embodiments, the present application is not limited to the above specific embodiments. Therefore, any modification or equivalent replacement of the present application; all technical solutions and improvements that do not deviate from the spirit and scope of the present application are included in the scope of the claims of the present application.

Claims

1. A malnutrition risk intelligent assessment tool for nasopharyngeal carcinoma patients, characterized in that, The application relates to a malnutrition prediction model based on Logistic regression analysis, a nutrition state score calculation equation corresponding to the model, an optimal risk judgment threshold, and an intelligent calculation module developed based on the equation; the prediction model is constructed by screening independent risk factors of nasopharyngeal carcinoma patients, and the independent risk factors include age increase, tumor stage IV, chemotherapy cycle >=1, radiotherapy times >=1, low BMI, appetite loss in the past week, and prealbumin <=250 mg / L.

2. The nasopharyngeal carcinoma patient malnutrition risk intelligent assessment tool according to claim 1, characterized in that, The nutrition state score calculation equation is: nutrition score = 16.372 + 1.331* (chemotherapy >=1) + 1.017* (radiotherapy >=1) + 0.052* age + 1.124* (stage IV) + 2.287* (appetite loss) - 0.570* BMI - 0.950* (prealbumin > 250); wherein "chemotherapy >=1", "radiotherapy >=1", "stage IV", "appetite loss", and "prealbumin > 250" are binary variables, 1 is taken when the conditions are met, and 0 is taken when the conditions are not met; the unit of age is year, and the unit of BMI is kg / m2.

3. The nasopharyngeal carcinoma patient malnutrition risk intelligent assessment tool of claim 1, wherein, The optimal risk judgment threshold is 6.6586, when the nutrition score of a patient is greater than or equal to 6.6586, it is judged that there is a malnutrition risk; the threshold is determined by the ROC curve analysis method, and the maximum Youden index is taken as the selection standard of the cutoff value, corresponding to the sensitivity 0.91 and the specificity 0.

87.

4. The nasopharyngeal carcinoma patient malnutrition risk intelligent assessment tool of claim 1, wherein, The construction process of the prediction model comprises the following steps: adopting a prospective cross-sectional diagnostic invention design, enrolling 668 newly diagnosed nasopharyngeal carcinoma patients in the first stage, collecting 22 potential variables, taking PG-SGA >=2 points as the malnutrition judgment standard, screening out 7 independent risk factors through single factor and multi-factor Logistic regression analysis and parameterization, and obtaining the nutrition state score calculation equation; in the second stage, the model performance is verified from the aspects of discrimination and calibration, the discrimination is evaluated by the ROC curve and the AUC value, and the calibration is verified by the calibration curve and the Hosmer-Lemeshow goodness-of-fit test.

5. The nasopharyngeal carcinoma patient malnutrition risk intelligent assessment tool of claim 4, wherein, The AUC value of the prediction model in the training set is 0.918, the calibration curve is close to the 45-degree ideal reference line, the Hosmer-Lemeshow goodness-of-fit test p>0.05, and it is indicated that the model has good discrimination ability and calibration performance.

6. The nasopharyngeal carcinoma patient malnutrition risk intelligent assessment tool of claim 1, wherein, The intelligent calculation module is an intelligent applet embedded in a hospital information system, and the use steps thereof comprise the following steps: checking or leaving blank the binary variable options, the binary variable options comprise "chemotherapy cycle >=1", "radiotherapy times >=1", "tumor stage IV", "appetite loss in the past week", and "prealbumin > 250 mg / L"; inputting the age and BMI values of the patient; and clicking the calculation button to automatically analyze the nutrition state score of the patient.

7. The nasopharyngeal carcinoma patient malnutrition risk intelligent assessment tool of claim 1, wherein, The intelligent calculation module can realize rapid, objective and standardized evaluation of the malnutrition risk of nasopharyngeal carcinoma patients, and overcomes the problems of long time consumption, strong subjectivity and complex operation of traditional evaluation tools.

8. An evaluation method of an intelligent assessment tool for malnutrition risk in nasopharyngeal carcinoma patients, characterized in that, The application right claim 1-7 any one of the evaluation tool, including the following steps: S1, the age of nasopharyngeal carcinoma patient, BMI, tumor stage, the number of chemotherapy cycle, the number of radiotherapy, the appetite condition of nearly a week and the proalbumin level data are collected;S2, the data collected are input into the intelligent computing module, or the data input is completed by the option check of the intelligent computing module;S3, the intelligent computing module automatically calculates the patient's nutrition score according to the nutrition state score calculation equation;S4, according to the comparison result of nutrition score and the best risk judgment threshold, whether the patient has malnutrition risk is judged.

Images