Systems and methods for determining early cancer risk impact factors

Abstract

The invention discloses a system and a method for determining early cancer risk influence factors. The method comprises the following steps: taking blood routine, blood biochemical and tumor marker data of a cancer group; determining a peripheral blood-derived inflammatory marker for the group of cancers; comparing the tumor marker data and the peripheral blood-derived inflammatory marker with corresponding data of a corresponding normal group, and determining a key abnormal factor with statistical significance; constructing a Lasso regression model based on all the key abnormal factors, and screening out a first-level influence factor from all the key abnormal factors according to the selected target parameters; a single-factor logistic regression analysis and parameter estimation method is utilized to screen out second-level influence factors from all the first-level influence factors; and a multi-factor logistic regression analysis and parameter estimation method is utilized to determine a significant impact factor from all the secondary impact factors. According to the method, the obvious influence factors on the early-stage cancer can be more effectively determined, and the accuracy and the reliability of cancer risk early warning are improved.

Description

Technical Field

[0001] The present application belongs to the field of biomedical data processing technology, and specifically relates to a system and method for determining factors affecting early cancer risk. Background Art

[0002] Gastric cancer is a malignant tumor located in the gastric mucosal epithelium. It is one of the most common malignant tumors of the digestive tract, and the incidence rate in men is much higher than that in women. Since the early symptoms of gastric cancer are not obvious, most patients are diagnosed in the late stage. In addition to the high recurrence and metastasis rates, the 5-year overall survival rate of patients does not exceed 30%. At present, gastroscopy is still the most effective guidance and recommended method for gastric cancer. However, due to the uncomfortable experience and high economic burden, the widespread use of gastroscopy in the early screening of gastric cancer is still difficult. In addition, the most common tumor markers such as CEA, CA72-4 or CA19-9 are not ideal biomarkers for diagnosing gastric cancer due to their low sensitivity and specificity. Therefore, it is crucial to determine non-invasive, highly sensitive, specific and easily accessible biological indicators.

[0003] Studies in recent years have shown that cancer patients have systemic immune responses, and the role of inflammatory factors and the immune system is related to tumor progression. When the body has an inflammatory response, blood inflammatory indicators such as neutrophils (NE), lymphocytes (LY), monocytes (MO), and platelet counts (Plt) become abnormal. NE releases reactive oxygen species in the process of killing tumor cells, promoting the secretion of vascular endothelial growth factor, which facilitates the metastasis and growth of tumor cells; the reduction or dysfunction of LY leads to a decrease in the body's immune system's ability to monitor tumor cells and immune escape; MO participates in inflammatory responses and immune regulation, providing a microenvironment for tumor recurrence and metastasis; Plt protects tumor cells from NK cell tumor lysis and immune surveillance, especially in patients with advanced tumors, who are often accompanied by increased Plt. In addition, serum albumin (Alb) can reflect the body's nutritional status, and low albumin levels indicate malnutrition and systemic inflammation. There are many types of inflammatory markers derived from peripheral blood counts, including neutrophil count to lymphocyte count ratio (NLR), platelet count to lymphocyte count ratio (PLR), systemic immune inflammatory index (SII) and prognostic nutritional index (PNI), etc. However, there are no reports on screening valuable risk factors for gastric cancer in patients based on multiple peripheral blood-derived inflammatory indicators. Summary of the invention

[0004] In order to overcome the defects of the prior art, the technical problem to be solved by the present application is to provide a system and method for determining the influencing factors of early cancer risk, which can more effectively determine the influencing factors of early cancer.

[0005] In order to achieve the above technical objectives, this application adopts the following technical means:

[0006] In a first aspect, the present application provides a system for determining early-stage cancer risk influencing factors, including:

[0007] A data acquisition module is configured to obtain routine blood data, blood biochemical data and tumor marker data of the cancer group, and determine peripheral blood-derived inflammatory marker data based on the routine blood data and blood biochemical data;

[0008] an abnormal factor determination module, configured to compare the tumor marker data and peripheral blood-derived inflammatory marker data with corresponding data of a corresponding normal group to determine key abnormal factors with statistical significance;

[0009] A primary influencing factor screening module is configured to construct a Lasso regression model based on all the key abnormal factors, and screen out the primary influencing factors from all the key abnormal factors according to the selected target parameters;

[0010] A secondary influencing factor screening module is configured to screen secondary influencing factors from all the primary influencing factors using a univariate logistic regression analysis and a parameter estimation method;

[0011] The significant influencing factor determination module is configured to determine the significant influencing factors from all the secondary influencing factors using multi-factor logistic regression analysis and parameter estimation methods.

[0012] Further, the abnormal factor determination module is specifically configured to determine the candidate abnormal factors using the Mann-Whitney U test method based on the tumor marker data and the peripheral blood-derived inflammatory marker data and the corresponding data of the corresponding normal group;

[0013] The P value of each of the candidate abnormal factors is determined, and the candidate abnormal factor whose P value is less than the first preset value is determined as the key abnormal factor.

[0014] Furthermore, the first-level influencing factor screening module is specifically configured to select the first regularization parameter λ1 corresponding to the minimum average error as the first target parameter;

[0015] According to the first target parameter, the key abnormal factor whose regression coefficient in the Lasso regression model is not compressed to zero is determined as a first-level influencing factor.

[0016] Furthermore, the first-level influencing factor screening module is specifically configured to select a second regularization parameter λ2 with an average error within a range of one times the standard error as the second target parameter;

[0017] According to the second target parameter, the key abnormal factor whose regression coefficient in the Lasso regression model is not compressed to zero is determined as a primary influencing factor.

[0018] Furthermore, the secondary impact factor screening module is specifically configured as follows:

[0019] Calculate the ROC curve for each of the first-level influencing factors respectively, and compare the area under the ROC curve (AUC) with a threshold value, and select the first-level influencing factor whose area AUC is greater than the threshold value as the candidate influencing factor;

[0020] For each cutoff value on each ROC curve, the corresponding sensitivity and specificity are calculated, and then the Youden index is calculated; the point with the largest Youden index is found on each ROC curve, and the cutoff value corresponding to the point is the optimal cutoff value;

[0021] According to the optimal cutoff value corresponding to each of the influencing factors to be selected, each of the influencing factors to be selected is associated with the nature of the disease;

[0022] Perform univariate logistic regression analysis on each of the candidate influencing factors after being associated with the nature of the disease, and determine the P value of each of the first-level influencing factors in combination with the parameter estimation method;

[0023] The candidate influencing factors whose P values ​​are smaller than the corresponding second preset values ​​are determined as secondary influencing factors.

[0024] Furthermore, the significant influencing factor determination module is specifically configured to perform a multifactor logistic regression analysis based on all the secondary influencing factors, combine the parameter estimation method, determine the P value of each of the primary influencing factors, and determine the secondary influencing factors whose P values ​​are less than the corresponding third preset values ​​as significant influencing factors.

[0025] Furthermore, the system also includes a risk warning module, which is configured to construct a logistic regression warning model based on each of the significant influencing factors, determine the probability value Logit(P) of the logistic regression warning model, and if the probability value Logit(P) is greater than a fourth preset value, it is predicted that there is an early cancer risk; if the probability value Logit(P) is less than or equal to the fourth preset value, it is predicted that there is no early cancer risk.

[0026] Further, the cancer group includes a male gastric cancer group and a female gastric cancer group; the healthy group includes a male healthy group and a female healthy group;

[0027] The data acquisition module is configured to collect tumor marker data of the male gastric cancer group including carcinoembryonic antigen (CEA) and carbohydrate antigen 199 (CA199), and collect tumor marker data of the female gastric cancer group including carcinoembryonic antigen (CEA), carbohydrate antigen 125 (CA125) and carbohydrate antigen 199 (CA199).

[0028] In a second aspect, the present application provides a method for determining early-stage cancer risk influencing factors, the method comprising:

[0029] Obtain routine blood data, blood biochemistry data, and tumor marker data of the cancer group;

[0030] Determine peripheral blood-derived inflammatory marker data based on the routine blood data and blood biochemical data; compare the tumor marker data and peripheral blood-derived inflammatory marker data with the corresponding data of the normal group to determine the key abnormal factors with statistical significance;

[0031] Building a Lasso regression model based on all the key abnormal factors, and screening out the first-level influencing factors from all the key abnormal factors according to the selected target parameters;

[0032] Using univariate logistic regression analysis and parameter estimation methods, secondary influencing factors were screened out from all the primary influencing factors;

[0033] Multivariate logistic regression analysis and parameter estimation methods were used to determine the significant influencing factors from all the secondary influencing factors.

[0034] Further, the method further comprises: for males, the early cancer risk influencing factors include carcinoembryonic antigen (CEA), lymphocyte count to monocyte count ratio (LMR), systemic inflammatory aggregation index (AISI) and prognostic nutritional index (PNI); for females, the early cancer risk influencing factors include carcinoembryonic antigen (CEA), lymphocyte count to monocyte count ratio (LMR) and prognostic nutritional index (PNI);

[0035] The peripheral blood derived inflammatory markers include neutrophil count to lymphocyte count ratio (NLR), platelet count to lymphocyte count ratio (PLR), lymphocyte count to monocyte count ratio (LMR), albumin to lymphocyte count ratio (ALR), (neutrophil + monocyte) to lymphocyte ratio (NMLR), systemic inflammatory aggregation index (AISI), prognostic nutritional index (PNI), systemic immune inflammatory index (SII) and systemic inflammatory response index (SIRI).

[0036] The system for determining early cancer risk influencing factors provided by the embodiment of the present application has beneficial technical effects compared with the prior art: comprehensive consideration of multiple biomarker data such as blood routine, blood biochemistry and tumor markers is helpful to discover effective risk influencing factors; by comparing with the corresponding data of the normal group, key abnormal factors with statistical significance can be screened out. This method constructs a multi-level screening model to gradually screen out primary influencing factors and secondary influencing factors from all key abnormal factors, and finally screens out significant influencing factors. It can more effectively determine factors that have a significant impact on early cancer, thereby improving the accuracy and reliability of cancer risk warning.

[0037] The method for determining early cancer risk influencing factors provided in the embodiment of the present application has the same beneficial technical effects as the system for determining early cancer risk influencing factors provided in the above embodiment of the present application, and will not be elaborated on. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings described herein are for explanation purposes only and are not intended to limit the scope of the present application in any way. In addition, the shapes and proportional dimensions of the components in the drawings are only for illustration purposes and are used to help understand the present application. They do not specifically limit the shapes and proportional dimensions of the components of the present application. Under the guidance of the present application, those skilled in the art can select various possible shapes and proportional dimensions to implement the present application according to specific circumstances. In the drawings:

[0039] Figure 1 A schematic diagram of the structure of a system for determining early cancer risk influencing factors provided in an embodiment of the present application;

[0040] Figure 2 A schematic flow chart of a method for determining early-stage cancer risk influencing factors provided in an embodiment of the present application;

[0041] Figure 3 11 are coefficient curves of the 11 key abnormal factors screened out in the male patient group in the embodiment of the present application in the Lasso regression model;

[0042] Figure 4 Based on Figure 3 The coefficient curve shown is a schematic diagram of the primary influencing factors of the male patient group screened out;

[0043] Figure 5 10 key abnormal factors screened out from the female patient group in the embodiment of the present application are coefficient curves in the Lasso regression model;

[0044] Figure 6 This is a schematic diagram of the primary influencing factors of the female patient group screened out by the coefficient curve in the Lasso regression model of the embodiment of the present application;

[0045] Figure 7 The ROC curve of the indicators of gastric cancer occurrence in the male group with the area under the curve (AUC) of CEA, NLR, PLR, ALR and AISI greater than the threshold value of 0.6 in the embodiment of the present application is shown in FIG.

[0046] Figure 8 The ROC curve of the indicators of gastric cancer occurrence in the male group with the area under the curve (AUC) of LMR and PNI greater than the threshold value of 0.6 in the embodiment of the present application is shown in FIG.

[0047] Fig. 9 The ROC curve diagram of gastric cancer occurrence in the female group for indicators with the area under the curve (AUC) of CEA, NLR, PLR and ALR greater than the threshold value of 0.6 in the embodiment of the present application;

[0048] Fig.10 This is a ROC curve diagram of the occurrence of gastric cancer in the female group when the area under the curve (AUC) of LMR and PNI is greater than the threshold value of 0.6 in the example of the present application.

[0049] Fig.11 The ROC curve is drawn based on the probability of gastric cancer in the male group in the embodiment of the present application;

[0050] Fig.12 The ROC curve is drawn based on the probability of gastric cancer in the female group in the embodiment of the present application;

[0051] Fig.13 This is a schematic diagram of the logistic multivariate regression analysis results of gastric cancer in the male group in the embodiment of the present application;

[0052] Fig.14 This is a schematic diagram of the results of the logistic multivariate regression analysis of gastric cancer in the female group in the examples of this application. DETAILED DESCRIPTION

[0053] In order to enable those skilled in the art to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0054] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more features.

[0055] like Figure 1 As shown, an embodiment of the present application provides a system for determining factors affecting early cancer risk, including a data acquisition module, an abnormal factor determination module, a primary influencing factor screening module, a secondary influencing factor screening module and a significant influencing factor determination module.

[0056] In an embodiment, the data collection module is configured to obtain routine blood test, blood biochemistry and tumor marker data of the cancer group.

[0057] The abnormal factor determination module is configured to determine the peripheral blood-derived inflammatory markers of the cancer group based on the routine blood data and blood biochemistry data of the cancer group; compare the tumor marker data and the peripheral blood-derived inflammatory markers with the corresponding data of the corresponding normal group to determine the key abnormal factors with statistical significance.

[0058] The first-level influencing factor screening module is configured to build a Lasso regression model based on all key abnormal factors, and screen out the first-level influencing factors from all key abnormal factors according to the selected target parameters.

[0059] The secondary impact factor screening module is configured to screen the secondary impact factors from all the primary impact factors using univariate logistic regression analysis and parameter estimation methods.

[0060] The significant influencing factor determination module is configured to determine the significant influencing factors from all secondary influencing factors using multi-factor logistic regression analysis and parameter estimation methods.

[0061] In one implementation, the diagnostic criteria used for the collected data is that the patient is diagnosed with primary gastric cancer based on postoperative pathology, and the pathological TNM staging is based on the 8th edition of the American Joint Committee on Cancer (AJCC) staging criteria.

[0062] As an example, the applicant's study collected a total of 1,196 gastric cancer patients eligible for inclusion from October 2020 to September 2024, and all cases came from Jiangsu Cancer Hospital. Among them, there were 865 male patients aged 21-75 years old, with a median age of 64 years old; 331 female patients aged 21-75 years old, with a median age of 63 years old. There were 687 healthy control groups, including 350 male patients aged 21-74 years old, with a median age of 44 years old; 337 female patients aged 20-75 years old, with a median age of 54 years old, all from Jiangsu Cancer Hospital. The research protocol complies with the ethical standards for human trials and has been approved by the Ethics Committee. The subjects were informed before the trial and obtained written consent. The results of blood routine, biochemical routine and serum tumor markers of patients were collected one week before surgery through the LIS system.

[0063] As an example, the group design includes a male gastric cancer group and a healthy control group; a female gastric cancer group and a healthy control group.

[0064] In the example, clinical pathological data include: gender, age, pathological type, TNM stage. Laboratory indicators one week before surgery, blood routine: neutrophil (NE), lymphocyte (LY), monocyte (MO), platelet count (Plt); biochemical routine: albumin (Alb).

[0065] Tumor markers: carcinoembryonic antigen (CEA), carbohydrate antigen 125 (CA125), carbohydrate antigen 199 (CA199).

[0066] Peripheral blood-derived inflammatory markers were calculated: neutrophil count to lymphocyte count ratio (NLR), platelet count to lymphocyte count ratio (PLR), lymphocyte count to monocyte count ratio (LMR), albumin to lymphocyte count ratio (ALR), (neutrophil + monocyte) to lymphocyte ratio (NMLR), systemic inflammatory aggregation index (AISI), prognostic nutritional index (PNI), systemic immune inflammatory index (SII), and systemic inflammatory response index (SIRI).

[0067] The data acquisition module calculates peripheral blood-derived inflammatory markers for the cancer group using the following formula:

[0068] NLR=NE / LY; PLR=Plt / LY; LMR=LY / MO; ALR=Alb(g / L) / LY; NMLR=(NE+MO) / LY; AISI=NE×MO×Plt / LY; PNI=Alb(g / L)+5×LY; SII=Plt×NE / LY; SIRI=(NE×MO) / LY.

[0069] In an embodiment, after the data acquisition module obtains the relevant data, it can transmit the blood routine, blood biochemistry and tumor marker data to the abnormal factor determination module through wired (such as USB, Ethernet and other interfaces) or wireless (such as Wi-Fi, Bluetooth and other communication protocols) electronic communication methods.

[0070] In some embodiments, the abnormal factor determination module is specifically configured as follows: based on the tumor marker data of the cancer group and the peripheral blood-derived inflammatory markers and the corresponding data of the corresponding normal group, the candidate abnormal factors are determined using the Mann-Whitney U test method; the P value of each candidate abnormal factor is determined, and the candidate abnormal factor with a P value less than a first preset value (the first preset value can be set to 0.05) is determined as a key abnormal factor. The Mann-Whitney U test, also known as the Wilcoxon rank sum test, is a non-parametric statistical test method, the core of which is to compare whether there is a significant difference in the medians of two independent samples. In a specific embodiment, data analysis can be performed using SPSS20.0 and GraphPad Prism v9.4.1 software. The P value is used to evaluate the statistical significance of the relationship between each candidate abnormal factor and the dependent variable. Generally, the smaller the P value, the more significant the relationship between the influencing factor and the dependent variable.

[0071] In some embodiments, the first-level influencing factor screening module is specifically configured to select the first regularization parameter λ1 corresponding to the minimum average error as the first target parameter; based on the first target parameter, the key abnormal factor whose regression coefficient in the Lasso regression model is not compressed to zero is determined as the first-level influencing factor.

[0072] In some other embodiments, the first-level influencing factor screening module is specifically configured to select a second regularization parameter λ2 whose average error is within one times the standard error as the second target parameter; based on the second target parameter, the key abnormal factor whose regression coefficient in the Lasso regression model is not compressed to zero is determined as the first-level influencing factor.

[0073] In some embodiments, the secondary influencing factor screening module is specifically configured as follows: the ROC curve is calculated for each primary influencing factor, and the area AUC under the ROC curve is compared with the threshold value, and the primary influencing factor with an area AUC greater than the threshold value is selected as the candidate influencing factor; for each cutoff value on each ROC curve, the corresponding sensitivity and specificity are calculated, and then the Youden index (sensitivity + specificity - 1) is calculated; the point with the largest Youden index is found on each ROC curve, and the cutoff value corresponding to the point is the optimal cutoff value; according to the optimal cutoff value corresponding to each candidate influencing factor, each candidate influencing factor is assigned a value of negative or positive; according to each candidate influencing factor after the assignment, a univariate logistic regression analysis is performed, and the P value of each primary influencing factor is determined in combination with the parameter estimation method; the candidate influencing factor with a P value less than each corresponding second preset value is determined as a secondary influencing factor. The embodiment compares the P value of each primary influencing factor with the set significance level. If the P value is less than the significance level, it is considered that there is a significant relationship between the influencing factor and the dependent variable.

[0074] The embodiment calculates the ROC curve and AUC value and compares them with the threshold value to preliminarily screen out the candidate influencing factors with statistical significance, objectively evaluates the predictive ability of each primary influencing factor on the risk of early cancer, and avoids the interference of subjective judgment; by calculating the sensitivity and specificity corresponding to each cutoff value, and then calculating the Youden index, the optimal cutoff value can be found, which helps to more accurately divide the negative and positive states of the candidate influencing factors, and provides a reliable data basis for the subsequent univariate logistic regression analysis. After determining the candidate influencing factors and their optimal cutoff values, a univariate logistic regression analysis is performed to screen out factors that have a greater impact on the risk of early cancer, and provide candidate variables for the subsequent multivariate logistic regression analysis.

[0075] In some embodiments, the significant influencing factor determination module is specifically configured to perform a multifactor logistic regression analysis based on all secondary influencing factors, and determine the P value of each primary influencing factor in combination with a parameter estimation method, and determine the secondary influencing factor whose P value is less than each corresponding third preset value as a significant influencing factor. Multifactor logistic regression analysis can simultaneously consider the impact of multiple secondary influencing factors on the risk of early cancer, avoiding the limitations of single factor analysis. This method can more comprehensively evaluate the role of each factor in cancer occurrence, thereby improving the accuracy of screening, and can provide some non-invasive, highly sensitive, highly specific and easily accessible biological indicators for the early diagnosis and treatment of cancer (such as gastric cancer).

[0076] Based on the above embodiments, the system for determining the influencing factors of early cancer risk provided in some other embodiments also includes a risk warning module (not shown in the figure), and the risk warning module is configured to construct a logistic regression warning model based on each significant influencing factor, and determine the probability value Logit (P) of the logistic regression warning model. If the probability value Logit (P) is greater than a fourth preset value, it is predicted that there is an early cancer risk; if the probability value Logit (P) is less than or equal to the fourth preset value, it is predicted that there is no early cancer risk.

[0077] In some embodiments, the risk warning module can be arranged in a backend server and connected to the significant influencing factor determination module through a communication line. The risk warning module can warn whether there is an early cancer risk through sound signals, image signals, text data, etc.

[0078] Based on the same inventive concept as the system of early cancer risk influencing factors provided in the above embodiments, correspondingly, the embodiments of the present application also provide a method of early cancer risk influencing factors, such as Figure 2 As shown, including:

[0079] Step S1: obtaining routine blood data, blood biochemical data and tumor marker data of the cancer group; determining peripheral blood-derived inflammatory marker data of the cancer group based on the routine blood data and blood biochemical data;

[0080] Step S2: comparing the tumor marker data and peripheral blood-derived inflammatory marker data with the corresponding data of the normal group to determine the key abnormal factors with statistical significance;

[0081] Step S3: construct a Lasso regression model based on all key abnormal factors, and screen out the first-level influencing factors from all key abnormal factors according to the selected target parameters;

[0082] Step S4: Using univariate logistic regression analysis and parameter estimation methods, the secondary influencing factors were screened out from all the primary influencing factors;

[0083] Step S5: Use multivariate logistic regression analysis and parameter estimation methods to determine the significant influencing factors from all secondary influencing factors.

[0084] The data collected in step S1 of the embodiment can refer to the above example. In step S2, the clinical data of two control groups, a gastric cancer patient group and a healthy group, are compared. The Kolmogorov-Smirnov test is used to determine whether the continuous variable does not obey the normal distribution. As an example, the Mann-Whitney U test method is used for inter-group comparison.

[0085] In the example, in the male patient group, 11 indicators including tumor markers CEA and CA199, and peripheral blood derived inflammatory markers NLR, PLR, LMR, ALR, NMLR, AISI, PNI, SII, and SIRI were compared. The results showed that compared with the control group, the levels of CEA, CA199, NLR, PLR, ALR, NMLR, AISI, SII, and SIRI were significantly increased, with statistically significant differences (P<0.01); the levels of LMR and PNI were significantly decreased, with statistically significant differences (P<0.01). Therefore, for men, CEA, CA199, NLR, PLR, ALR, NMLR, AISI, SII, SIRI, LMR, and PNI, these 11 indicators can be used as key abnormal factors.

[0086] In the female patient group: tumor markers CEA, CA199 and CA125 were compared; peripheral blood derived inflammatory markers NLR, PLR, LMR, ALR, NMLR, AISI, PNI, SII, SIRI, a total of 12 indicators. The results showed that compared with the control group, CEA, NLR, PLR, ALR, NMLR, AISI, SII, SIRI levels were significantly increased, and the difference was statistically significant (P<0.01); LMR and PNI levels were significantly decreased, and the difference was statistically significant (P<0.01). Therefore, for women, CEA, NLR, PLR, ALR, NMLR, AISI, SII, SIRI, LMR and PNI, these 10 indicators can be used as key abnormal factors.

[0087] In the example, step S3, the key abnormal factors with statistically significant differences between the control group and the gastric cancer group in the two cohorts are included in the Lasso regression analysis to construct a Lasso regression model, and the first-level influencing factors are screened by a ten-fold cross-validation method. Optionally, the regularization parameter λ corresponding to the minimum average error is selected as the first target parameter; according to the first target parameter, the key abnormal factors whose regression coefficients are not compressed to zero are determined as the first-level influencing factors.

[0088] In the male patient group, at λ = 0.0008798, Logλ = -7.036, 8 predictive variables with non-zero coefficients were screened out, including CEA, CA199, NLR, PLR, LMR, ALR, AISI, and PNI, such as Figure 3 and Figure 4 shown.

[0089] In the female group: at λ = 0.0014605, Logλ = -6.529, six predictor variables with non-zero coefficients were screened out, including CEA, NLR, PLR, LMR, ALR, and PNI, such as Figure 5 and Figure 6 shown.

[0090] In the example, step S4 specifically includes calculating the ROC curve for each first-level influencing factor, and comparing the area AUC under the ROC curve with the threshold, selecting the first-level influencing factor with an area AUC greater than the threshold as the candidate influencing factor; for each cutoff value on each ROC curve, calculating its corresponding sensitivity and specificity, and then calculating the Youden index; finding the point with the largest Youden index on each ROC curve, and the cutoff value corresponding to the point is the optimal cutoff value; according to the optimal cutoff value corresponding to each candidate influencing factor, assigning each candidate influencing factor a value of negative or positive; performing univariate logistic regression analysis on each candidate influencing factor after assignment, and determining the P value of each first-level influencing factor in combination with the parameter estimation method; determining the candidate influencing factor with a P value less than the corresponding second preset value as the second-level influencing factor.

[0091] In the example, step S2 uses healthy people as the control group and gastric cancer as the disease group, performs ROC curve analysis on the predictive variables screened in the two cohorts, and selects the first-level influencing factors with AUC values ​​greater than the threshold of 0.6 as candidate influencing factors.

[0092] For example, for the male group, the AUC value of CEA was 0.677, and the optimal cutoff value was 1.87, the AUC value of NLR was 0.727, and the optimal cutoff value was 1.91, the AUC value of PLR was 0.704, and the optimal cutoff value was 131.13, the AUC value of LMR was 0.803, and the optimal cutoff value was 4.84, the AUC value of ALR was 0.661, and the optimal cutoff value was 24.91, the AUC value of AISI was 0.661, and the optimal cutoff value was 200.13, and the AUC value of PNI was 0.908, and the optimal cutoff value was 55.33. Figure 7 and Figure 8 shown.

[0093] For the female group, the AUC value of CEA was 0.681, and the optimal cutoff value was 2.04, the AUC value of NLR was 0.688, and the optimal cutoff value was 1.75, the AUC value of PLR was 0.669, and the optimal cutoff value was 140.22, the AUC value of LMR was 0.77, and the optimal cutoff value was 5.47, the AUC value of ALR was 0.686, and the optimal cutoff value was 25.12, and the AUC value of PNI was 0.890, and the optimal cutoff value was 54.18. Fig. 9 and Fig.10 shown.

[0094] Taking whether the patient develops gastric cancer as the dependent variable (negative=0, positive=1), each candidate influencing factor is associated with the nature of the disease (i.e., whether the patient is ill, such as whether gastric cancer develops). For example, for the male group, CEA (CEA≤1.87ng / ml, then CEA=0; CEA>1.87ng / ml, then CEA=1), NLR (NLR≤1.91, then NLR=0; NLR>1.91, then NLR=1), PLR (PLR≤131.13, then PLR=0; PLR>131.13, then PLR=1), LMR (LMR≤4.84, then LMR=0, LMR>4.84, then LMR=1), ALR (ALR≤24.91, then ALR=0, ALR>24.91, then ALR=1), AISI (AISI≤200.13, then AISI=0, AISI>200.13, then AISI=1), PNI (PNI≤55.33, then PNI=0, PNI>55.33, then PNI=1).

[0095] A univariate logistic regression analysis was performed on each candidate influencing factor after being associated with the nature of the disease, and the P value of each first-level influencing factor was determined in combination with the parameter estimation method; the candidate influencing factors with a P value less than the corresponding second preset value (such as 0.01) were determined as second-level influencing factors. Each first-level influencing factor with a P value < 0.01 in the univariate analysis was included in the multivariate logistic regression analysis, and the likelihood ratio test (forward: LR) of the maximum partial likelihood estimate was used. Among the second-level influencing factors, each second-level influencing factor with a P value < the third set value (which can be 0.01) was regarded as a significant influencing factor. The results showed that the four indicators of CEA, LMR, AISI and PNI were independent risk factors for gastric cancer in male patients. Fig.13 shown.

[0096] For the female group, CEA (CEA≤2.04ng / ml, then CEA=0, CEA>2.04ng / ml, then CEA=1), NLR (NLR≤1.75, then NLR=0, NLR>1.75, then NLR=1), PLR (PLR≤140.22, then PLR=0, PLR>140.22, then PLR=1), LMR (LMR≤5.47, then LMR=0, LMR>5.47, Univariate logistic analysis was performed with CEA (ALR≤25.12, then ALR=0, ALR>25.12, then ALR=1), and PNI (PNI≤54.18, then PNI=0, PNI>54.18, then PNI=1) as independent variables. The indicators with P<0.01 in the univariate analysis were included in the multivariate logistic regression analysis, and the likelihood ratio test (forward: LR) of the maximum partial likelihood estimate was used. The secondary influencing factors with P value P<the third set value (which can be 0.01) were regarded as significant influencing factors. The results showed that CEA, LMR and PNI were independent risk factors for gastric cancer in female patients. Fig.14 shown.

[0097] The method for determining early cancer risk influencing factors provided in the embodiments of the present application can provide strong support for early warning and individualized treatment of cancer by determining early cancer risk influencing factors.

[0098] In some other embodiments, the method for determining factors affecting early cancer risk also includes: constructing a logistic regression early warning model based on each significant influencing factor, determining the probability value Logit(P) of the logistic regression early warning model, if the probability value Logit(P) is greater than a fourth preset value, predicting that there is an early cancer risk; if the probability value Logit(P) is less than or equal to the fourth preset value, predicting that there is no early cancer risk.

[0099] As an example, Logit(P)>0.5 predicts the occurrence of early cancer, and Logit(P≤0.5 predicts that early cancer will not occur. Based on the probability value Logit(P) of judging the occurrence of early cancer, the AUC area under the ROC curve (i.e., C statistic) is plotted. The larger the AUC area, the better the risk discrimination ability of the logistic regression early warning model. Generally, AUC<0.6 is considered to have poor discrimination, 0.6-0.75 is considered to have a certain discrimination ability of the risk logistic regression early warning model, and >0.75 is considered to have good discrimination ability. Finally, the Hosmer-Lemeshow goodness of fit test is used to evaluate the calibration ability of the risk early warning module. The P value in the Hosmer-Lemeshow goodness of fit test is>0.05, indicating that there is no statistically significant difference between the predicted value and the actual observed value of the logistic regression early warning model, and the risk early warning module has good calibration ability.

[0100] The results showed that the logistic regression early warning model of the male cohort was: Logit (P) = 2.137 + 1.026 * CEA - 0.776 * LMR + 0.732 * AISI - 3.021 * PNI. This model was statistically significant (χ 2 =592.216, P = 0.000), the prediction accuracy was 83.70%, and the AUC was 0.895 (95% CI: 0.876-0.915, P = 0.000). The Hosmer-Lemeshow test results suggested that the model had good prediction accuracy in this study population (χ 2 =11.460,P=0.120) Fig.11 .

[0101] The logistic regression early warning model of the female group was: Logit(P)=1.620+1.015*CEA-0.917*LMR-2.727*PNI. This model was statistically significant (χ 2 =336.892, P = 0.000), the prediction accuracy was 81.74%, and the area under the ROC curve was 0.868 (95% CI: 0.841-0.896, P = 0.000). The Hosmer-Lemeshow test results suggested that the model had good prediction accuracy in this study population (χ 2 =4.388, P=0.495) Fig.12 shown.

[0102] By monitoring changes in these influencing factors, doctors can promptly detect potential risks of cancer and take appropriate preventive measures or treatment plans, thereby improving patients' survival rate and quality of life.

[0103] The system for determining early cancer risk influencing factors provided in the embodiments of the present application can implement the method for determining early cancer risk influencing factors provided in the embodiments of the present application, but the implementation device of the method for determining early cancer risk influencing factors provided in the present application includes but is not limited to the system for determining early cancer risk influencing factors listed in the embodiments. All structural deformations and replacements of the prior art made according to the principles of the present application are included in the protection scope of the present application.

[0104] The system and method for determining the factors affecting early cancer risk provided by the present application are introduced in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the concept of the present application and should not be construed as limiting the scope of protection of the present application.

Claims

1. A system for determining factors affecting early cancer risk, characterized in that: The system comprises: A data acquisition module is configured to obtain routine blood data, blood biochemical data and tumor marker data of the cancer group, and determine peripheral blood-derived inflammatory marker data based on the routine blood data and blood biochemical data; an abnormal factor determination module, configured to compare the tumor marker data and the peripheral blood-derived inflammatory marker data with corresponding data of a corresponding normal group to determine key abnormal factors with statistical significance; A primary influencing factor screening module is configured to construct a Lasso regression model based on all the key abnormal factors, and screen out the primary influencing factors from all the key abnormal factors according to the selected target parameters; A secondary influencing factor screening module is configured to screen secondary influencing factors from all the primary influencing factors using a univariate logistic regression analysis and a parameter estimation method; The significant influencing factor determination module is configured to determine the significant influencing factors from all the secondary influencing factors using multi-factor logistic regression analysis and parameter estimation methods.

2. The system for determining early cancer risk influencing factors according to claim 1, characterized in that: The abnormal factor determination module is specifically configured to determine candidate abnormal factors using a Mann-Whitney U test method based on the tumor marker data and peripheral blood-derived inflammatory marker data of the cancer group and corresponding data of the corresponding normal group; The P value of each of the candidate abnormal factors is determined, and the candidate abnormal factor whose P value is less than the first preset value is determined as the key abnormal factor.

3. The system for determining early cancer risk influencing factors according to claim 1, characterized in that: The first-level influencing factor screening module is specifically configured to select the first regularization parameter λ1 corresponding to the minimum average error as the first target parameter; According to the first target parameter, the key abnormal factor whose regression coefficient in the Lasso regression model is not compressed to zero is determined as a first-level influencing factor.

4. The system for determining early cancer risk influencing factors according to claim 1, characterized in that: The first-level influencing factor screening module is specifically configured to select a second regularization parameter λ2 within a range of one times the standard error as the second target parameter; According to the second target parameter, the key abnormal factor whose regression coefficient in the Lasso regression model is not compressed to zero is determined as a primary influencing factor.

5. The system for determining early cancer risk influencing factors according to claim 1, characterized in that: The secondary impact factor screening module is specifically configured as follows: Calculate the ROC curve for each of the first-level influencing factors respectively, and compare the area under the ROC curve (AUC) with a threshold value, and select the first-level influencing factor whose area AUC is greater than the threshold value as the candidate influencing factor; For each cutoff value on each ROC curve, the corresponding sensitivity and specificity are calculated, and then the Youden index is calculated; the cutoff value corresponding to the point with the largest Youden index on each ROC curve is found to be the optimal cutoff value; According to the optimal cutoff value corresponding to each of the influencing factors to be selected, each of the influencing factors to be selected is associated with the nature of the disease; According to each of the candidate influencing factors associated with the nature of the disease, a univariate logistic regression analysis is performed, and the P value of each of the first-level influencing factors is determined in combination with a parameter estimation method; The candidate influencing factors whose P values ​​are smaller than the corresponding second preset values ​​are determined as secondary influencing factors.

6. The system for determining early cancer risk influencing factors according to claim 1, characterized in that: The significant influencing factor determination module is specifically configured to perform a multifactor logistic regression analysis based on all the secondary influencing factors, combine the parameter estimation method, determine the P value of each of the primary influencing factors, and determine the secondary influencing factors whose P values ​​are less than the corresponding third preset values ​​as significant influencing factors.

7. The system for determining early-stage cancer risk influencing factors according to claim 1, characterized in that: The system also includes a risk warning module; The risk warning module is configured to construct a logistic regression warning model based on each of the significant influencing factors, determine the probability value Logit(P) of the logistic regression warning model, and if the probability value Logit(P) is greater than a fourth preset value, it is predicted that there is an early cancer risk; if the probability value Logit(P) is less than or equal to the fourth preset value, it is predicted that there is no early cancer risk.

8. The system for determining early cancer risk influencing factors according to claim 1, characterized in that: The cancer group includes a male gastric cancer group and a female gastric cancer group; the healthy group includes a male healthy group and a female healthy group; The data acquisition module is configured to collect tumor marker data of the male gastric cancer group including carcinoembryonic antigen (CEA) and carbohydrate antigen 199 (CA199), and collect tumor marker data of the female gastric cancer group including carcinoembryonic antigen (CEA), carbohydrate antigen 125 (CA125) and carbohydrate antigen 199 (CA199).

9. A method for determining factors affecting early cancer risk, characterized in that: The method comprises: Obtaining routine blood data, blood biochemical data and tumor marker data of the cancer group; determining peripheral blood-derived inflammatory marker data based on the routine blood data and blood biochemical data; comparing the tumor marker data and peripheral blood-derived inflammatory marker data with the corresponding data of the corresponding normal group to determine the key abnormal factors with statistical significance; Building a Lasso regression model based on all the key abnormal factors, and screening out the first-level influencing factors from all the key abnormal factors according to the selected target parameters; Using univariate logistic regression analysis and parameter estimation methods, secondary influencing factors were screened out from all the primary influencing factors; Multivariate logistic regression analysis and parameter estimation methods were used to determine the significant influencing factors from all the secondary influencing factors.

10. The system for determining early-stage cancer risk influencing factors according to claim 9, characterized in that: The method further comprises: For males, the early cancer risk influencing factors include carcinoembryonic antigen (CEA), lymphocyte count to monocyte count ratio (LMR), systemic inflammatory aggregation index (AISI) and prognostic nutritional index (PNI); for females, the early cancer risk influencing factors include carcinoembryonic antigen (CEA), lymphocyte count to monocyte count ratio (LMR) and prognostic nutritional index (PNI); The peripheral blood derived inflammatory markers include neutrophil count to lymphocyte count ratio (NLR), platelet count to lymphocyte count ratio (PLR), lymphocyte count to monocyte count ratio (LMR), albumin to lymphocyte count ratio (ALR), (neutrophil + monocyte) to lymphocyte ratio (NMLR), systemic inflammatory aggregation index (AISI), prognostic nutritional index (PNI), systemic immune inflammatory index (SII) and systemic inflammatory response index (SIRI).

Images