Gall bladder cholesterol polyp prediction system

Abstract

The invention discloses a gall bladder cholesterol polyp prediction system, and the system comprises a collection device which is used for obtaining historical patient pathological data; the analysis device is used for obtaining a single-factor analysis result of the non-cholesterol polyp related factors by adopting single-factor Logistic regression analysis according to the historical patient pathological data; according to the single-factor analysis result of the non-cholesterol polyp related factors, a multi-factor Logistic regression analysis result of the non-cholesterol polyp related factors is obtained through multi-factor Logistic regression analysis; the construction device is used for constructing a Bayesian network prediction model according to the multi-factor Logistic regression analysis result of the non-cholesterol polyp related factors; and the prediction device is used for inputting the to-be-processed patient pathological data into the Bayesian network prediction model to predict the cholecystic cholesterol polyp. By adopting the technical scheme provided by the invention, the identification capability of gallbladder polyp properties is effectively improved.

Description

Technical Field

[0001] The invention belongs to the field of biomedicine technology, and in particular relates to a gallbladder cholesterol polyp prediction system. Background Art

[0002] Gallbladder polyps are a general term for a type of lesion in which the gallbladder wall bulges into the gallbladder cavity. Pathologically, gallbladder polyps include cholesterol polyps, gallbladder adenomas, gallbladder adenomyomas, inflammatory polyps, adenomatous hyperplasia, fibromas, hemangiomas, and polyp-type early gallbladder cancer. Clinically, they are generally divided into three categories: cholesterol polyps, benign non-cholesterol polyps, and polyp-type early gallbladder cancer. This classification makes the diagnosis, differential diagnosis, and treatment of gallbladder polyps simpler. Since some gallbladder polyps are early gallbladder cancers, and some gallbladder polyps are benign but may become cancerous in the future, in order to avoid missed diagnosis of gallbladder cancer or prevent gallbladder polyps from becoming cancerous, domestic and foreign guidelines recommend cholecystectomy for polyps larger than 10 mm. However, the benefit-risk ratio of this strategy is extremely low, because postoperative pathology shows that cholesterol polyps account for the majority, with a proportion as high as 80.9-90.3%, while cholesterol polyps have no potential for canceration and do not require active surgical resection, while gallbladder cancer accounts for only a very small part, 0-0.7%. This shows that even if the gallbladder is removed strictly according to the surgical indications recommended by the guidelines, only a very small number of gallbladder cancers are found, and the vast majority are cholesterol polyps. Therefore, the positive predictive value of using 10 mm as the threshold for intervention on gallbladder polyps is extremely low, which not only fails to achieve the purpose of preventing gallbladder cancer, but also leads to the expansion of cholecystectomy, resulting in excessive medical treatment, and it is inevitable that excessive harm will be caused to patients. However, if doctors do not follow the guidelines and give up surgery for polyps larger than 1 cm and adopt a simple observation strategy, they will face the risk of missing gallbladder cancer. Due to the high malignancy of gallbladder cancer, once missed, the consequences are serious, especially in the current medical environment, which often leads to medical disputes. Doctors will have great concerns about this. Therefore, it is necessary to re-examine the rationality of the standard of using polyps larger than 10 mm in diameter as an indication for surgery to remove the gallbladder.

[0003] A large number of studies have shown that among gallbladder polyps larger than 1 cm, the most common pathological types are cholesterol polyps, gallbladder adenomas, and early gallbladder cancer, and gallbladder polyps of other pathological types account for a relatively small proportion. Gallbladder adenomas and early gallbladder cancer have clear surgical indications, while cholesterol polyps are benign lesions. Since they have no potential for malignant transformation, they will not produce symptoms in most cases, so they can continue to be observed even if the diameter is larger than 1 cm. Therefore, for gallbladder polyps found by ultrasound examination, if it can be determined that the possibility of cholesterol polyps is relatively high, an observation strategy can be adopted to avoid surgery. The currently developed gallbladder polyp prediction focuses on predicting the possibility of malignant lesions, and does not predict the possibility of cholesterol polyps. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a gallbladder cholesterol polyp prediction system that can predict the possibility of cholesterol polyps.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] A gallbladder cholesterol polyp prediction system, comprising:

[0007] A collection device for obtaining historical patient pathology data;

[0008] An analysis device is used to obtain univariate analysis results of non-cholesterol polyp-related factors by univariate Logistic regression analysis based on historical patient pathological data; and to obtain multivariate Logistic regression analysis results of non-cholesterol polyp-related factors by multivariate Logistic regression analysis based on the univariate analysis results of non-cholesterol polyp-related factors;

[0009] Constructing a device for constructing a Bayesian network prediction model based on the results of multivariate logistic regression analysis of non-cholesterol polyp-related factors;

[0010] The prediction device is used to input the patient's pathological data to be processed into the Bayesian network prediction model to predict gallbladder cholesterol polyps.

[0011] Preferably, the cutoff values ​​of age and polyp diameter for distinguishing cholesterol polyps from non-cholesterol polyps are determined according to the maximum value of the Youden Index.

[0012] Preferably, the historical patient pathological data include: gender, age, polyp diameter, polyp growth time, number of polyps, whether the gallbladder wall is thickened, whether stones are present, and polyp nature; wherein the polyp nature includes: cholesterol polyps and non-cholesterol polyps.

[0013] Preferably, the analysis device comprises:

[0014] The first analysis unit is used to obtain the univariate analysis results of non-cholesterol polyp-related factors through univariate logistic regression analysis based on gender, age, polyp diameter, polyp growth time, number of polyps, whether the gallbladder wall is thickened, whether there is concomitant stone, and polyp nature;

[0015] The second analysis unit was used to establish a binary multivariate Logistic regression analysis using gender, age, polyp diameter, number of polyps, whether the gallbladder wall was thickened, and whether stones were present in the univariate analysis results of non-cholesterol polyp-related factors as independent variables and non-cholesterol gallbladder polyps as dependent variables. Gender, number of polyps, whether the gallbladder wall was thickened, and whether stones were present were defined as categorical covariates to screen out independent risk factors for non-cholesterol polyps and obtain the multivariate Logistic regression analysis results of non-cholesterol polyp-related factors.

[0016] Preferably, the input nodes of the Bayesian network prediction model include: gender, age, polyp diameter, polyp growth time, number of polyps, whether the gallbladder wall is thickened, and whether stones are present. The target nodes of the Bayesian network prediction model include: polyp properties.

[0017] The present invention predicts gallbladder cholesterol polyps by using a Bayesian network prediction model, thereby effectively improving the ability to identify the properties of gallbladder polyps. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0019] Figure 1 This is a schematic diagram of the structure of a gallbladder cholesterol polyp prediction system according to an embodiment of the present invention;

[0020] Figure 2 ROC curve to determine the age cutoff value;

[0021] Figure 3 ROC curve to determine the cutoff value of polyp diameter;

[0022] Figure 4 This is the structure diagram of the Bayesian network prediction model;

[0023] Figure 5 Schematic diagram of the Bayesian network model inference of GCP for people with single gallbladder polyps and polyp diameter ≥ 1 cm;

[0024] Figure 6 Schematic diagram of the Bayesian network model inference of GCP in the population with multiple gallbladder polyps and polyp diameter ≥ 1 cm;

[0025] Figure 7 The Bayesian network model inference diagram of GCP for people with multiple gallbladder polyps, diameter ≥ 1 cm, female, and age ≤ 55 years old;

[0026] Figure 8 This is a schematic diagram of the Bayesian network model reasoning for the GCP of people with multiple gallbladder polyps, diameter ≥ 1 cm, female, age ≤ 55 years old, no gallbladder stones, and no thickening of the gallbladder wall;

[0027] Fig. 9 This is a schematic diagram of the Bayesian network model inference of GCP for people with multiple gallbladder polyps, diameter ≥ 1 cm, female, age ≤ 55 years, no gallbladder stones, no gallbladder wall thickening, and polyp growth time ≥ 2 years. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] Embodiment 1:

[0031] like Figure 1 As shown, an embodiment of the present invention provides a gallbladder cholesterol polyp prediction system, comprising:

[0032] A collection device for obtaining historical patient pathological data; wherein the historical patient pathological data includes: gender, age, polyp diameter, polyp growth time, number of polyps, whether the gallbladder wall is thickened, whether there are stones, and polyp nature; wherein the polyp nature includes: cholesterol polyps and non-cholesterol polyps; the age and polyp diameter cutoff values ​​for distinguishing cholesterol polyps and non-cholesterol polyps are determined according to the maximum value of the Youden index;

[0033] The analysis device is used to obtain the univariate analysis results of non-cholesterol polyp related factors by univariate Logistic regression analysis based on historical patient pathological data; and obtain the multivariate Logistic regression analysis results of non-cholesterol polyp related factors by multivariate Logistic regression analysis based on the univariate analysis results of non-cholesterol polyp related factors; wherein the analysis device includes: a first analysis unit, which is used to obtain the multivariate Logistic regression analysis results of non-cholesterol polyp related factors by univariate Logistic regression analysis based on gender, age, polyp diameter, polyp growth time, number of polyps, whether the gallbladder wall is thickened, whether there is concomitant stones, and polyp nature. The regression analysis obtained the univariate analysis results of non-cholesterol polyp-related factors; the second analysis unit was used to take the gender, age, polyp diameter, number of polyps, whether the gallbladder wall was thickened, and whether there were stones in the univariate analysis results of non-cholesterol polyp-related factors as independent variables, and non-cholesterol gallbladder polyps as dependent variables, to establish a binary multivariate Logistic regression analysis, define gender, number of polyps, gallbladder wall thickness, and whether there were stones as categorical covariates, screen out independent risk factors for non-cholesterol polyps, and obtain the multivariate Logistic regression analysis results of non-cholesterol polyp-related factors;

[0034] A device is constructed for constructing a Bayesian network prediction model based on the results of multi-factor Logistic regression analysis of non-cholesterol polyp-related factors; wherein the input nodes of the Bayesian network prediction model include: gender, age, polyp diameter, polyp growth time, number of polyps, whether the gallbladder wall is thickened, whether there are stones, and the target nodes of the Bayesian network prediction model include: polyp properties;

[0035] The prediction device is used to input the patient's pathological data to be processed into the Bayesian network prediction model to predict gallbladder cholesterol polyps.

[0036] As an implementation of the present invention, a data set of 696 patients who underwent cholecystectomy is used as historical patient pathology data, as shown in Tables 1 and 2. Among them, gender: 306 males (44.5%), 390 females (55.5%); age: 423 cases (60.8%) <55 years old, 273 cases (39.2%) ≥55 years old; polyp growth time: 414 cases (59.5%) <2 years, 282 cases (40.5%) ≥2 years; polyp diameter: 411 cases (59.1%) <1 cm, 285 cases (40.9%) ≥1 cm; number of polyps: single in 200 cases (28.7%), multiple in 496 cases (71.3%); gallbladder wall thickness: thickened in 198 cases (28.4%), normal in 498 cases (71.6%); 248 cases (35.6%) combined with stones, 448 cases (64.4%) without stones; 556 cases (79.9%) cholesterol polyps; 140 cases (20.1%) non-cholesterol polyps.

[0037] Table 1 Clinical and pathological data of patients

[0038]

[0039] Table 2 Clinical data of cholesterol polyp group and non-cholesterol polyp group

[0040]

[0041] As an implementation of an embodiment of the present invention, the cutoff value of age and polyp diameter for distinguishing cholesterol polyps from non-cholesterol polyps is determined according to the maximum value of the Youden index. Figure 2 , 3 As shown, the ROC curve of the age of patients with gallbladder polyps shows that the area under the curve (AUC) is 0.615 (95% CI: 0.560-0.670), the maximum Youden index is 0.187, and the corresponding age cutoff is 54.5 years old. At this time, the sensitivity of predicting non-cholesterol polyps is 49.3% and the specificity is 69.4%. The ROC curve of the polyp diameter of patients with gallbladder polyps shows that the area under the curve (AUC) is 0.745 (95% CI: 0.694-0.795), the maximum Youden index is 0.39, and the corresponding polyp diameter cutoff is 0.985cm. At this time, the sensitivity of predicting non-cholesterol polyps is 72.1% and the specificity is 66.9%. For the convenience of application, the embodiment of the present invention sets the cutoff value of polyp diameter to 1cm and the cutoff value of age to 55 years old.

[0042] As an implementation method of the present invention, gender, age, polyp diameter, number of polyps, whether there is concomitant stone, whether there is gallbladder wall thickening, polyp growth time and whether there is concomitant fatty liver were subjected to univariate regression analysis. The results showed that gender (P=0.002, OR=1.868, 95%CI: 1.262-2.764), age (P<0.001, OR=4.853, 95%CI: 3.218-7.320), polyp diameter (P<0.001, OR=5.236, 95%CI: 3.476-7.886), and number of polyps (P<0.001, OR=4.6 56, 95%CI: 3.152-6.877), whether the gallbladder wall was thickened (P<0.001, OR=2.756, 95%CI: 1.875-4.051), and whether there was concomitant stones (P<0.001, OR=2.554, 95%CI: 1.751-3.726) were related factors for non-cholesterol polyps (P<0.05); while the growth time of polyps (P=0.12, OR=1.107, 95%CI: 0.757-1.619) and whether there was concomitant fatty liver (P=0.420, OR=0.840, 95%CI: 0.549-1.284) had no statistically significant difference (P>0.05). The results of univariate regression analysis are shown in Table 3.

[0043] Table 3 Results of univariate analysis of factors associated with non-cholesterol polyps

[0044]

[0045] Six statistically significant (P<0.05) research variables in the univariate logistic regression analysis of factors related to non-cholesterol polyps were selected as independent variables: gender, age, polyp diameter, number of polyps, whether the gallbladder wall was thickened, and whether there was stone. The pathological result of non-cholesterol gallbladder polyps was used as the dependent variable. A binary multivariate logistic regression analysis was established, and gender, number of polyps, gallbladder wall thickness, and whether there was stone were defined as categorical covariates to screen out independent risk factors for non-cholesterol polyps. The results showed that the following factors were the clinical diagnosis and treatment: female (P<0.001, OR=2.331, 95% CI: 1.407-3.860), age ≥55 years (P<0.001, OR=4.621, 95% CI: 2.826-7.555), polyp diameter ≥1 cm (P<0.001, OR=10.947, 95% CI: 6.323-18.954), combined with stones (P<0.001, OR=3.56, 95% CI: 2.096-6.047), single polyp (P<0.001, OR=4.628, 95% CI: 2.864-7.478), and gallbladder wall thickening (P=0.002, OR=2.393, 95% CI: CI: 1.436-3.986) was an independent risk factor for non-cholesterol polyps (P<0.05). The results of multivariate regression analysis are shown in Table 4.

[0046] Table 4 Results of multivariate logistic regression analysis of factors associated with non-cholesterol polyps

[0047]

[0048] As an implementation method of an embodiment of the present invention, in constructing a Bayesian network prediction model, the dependency relationship between the input node and the target node is determined, and the dependency relationship is represented by directed edges in the Bayesian network to form a directed acyclic graph (DAG) structure. Through the K2 search algorithm and ten-fold cross validation, the structure of the Bayesian network is first determined, that is, an optimal set of parent nodes is selected so that the Bayesian network can effectively fit the data and improve the prediction accuracy; then, the EM algorithm is used for parameter learning, the conditional probability of each node is calculated, and finally the parameterized model of the Bayesian network is obtained. By combining the K2 search algorithm, ten-fold cross validation and the EM algorithm, the optimization learning of the structure and parameters of the Bayesian network can be realized, thereby obtaining an optimal Bayesian network model for the prediction of gallbladder cholesterol polyps.

[0049] like Figure 4 As shown, Figure 4 Each node box in has two different states and corresponding prior probabilities. For example, the value corresponding to the cholesterol polyp group is 75%, which means that the proportion of cholesterol polyp patients in the total gallbladder polyp population is 75%, which is the prior probability of the node. Figure 4 The directed line segments in represent the dependency relationship between nodes. For example, age points to polyp properties, indicating that age is the parent node of polyp properties, that is, age is directly related to polyp properties. Figure 4 The Bayesian network showing the prediction model of gallbladder polyp properties consists of 8 nodes and 9 directed edges. The directed edges indicate the dependency between nodes. If the arrow points from node A to node B (A→B), then A is the parent node of B and B is the child node of A. The uncertainty between the two nodes is quantified in the conditional probability table through parameter learning. The probability of occurrence of the target node can be calculated through the directed acyclic graph and the conditional probability table of each node. The percentage of each node represents the prior probability of the influencing factor. For example, the prior probability that the gallbladder polyp is cholesterol polyp in this model is 0.75, then P (cholesterol polyp) = 0.75. The Bayesian network can be used to predict the occurrence of the target node based on the known node X. i Estimate the probability P (YǀX) of the unknown node Y occurring i ), and predict the risk of disease occurrence.

[0050] Prediction and reasoning of gallbladder polyp Bayesian network model: The Bayesian network prediction model is constructed through GeNIe Academic software, and the prediction model is used to predict the properties of gallbladder polyps. The target node of the model is the properties of gallbladder polyps. Select the corresponding node to add evidence information to get the prediction percentage of the properties of gallbladder polyps. For example, if a person's gallbladder polyp is larger than 1cm in diameter and the polyp is single, then the probability that the polyp is GCP is P (GCP | single polyp, diameter ≥ 1cm) = 0.43, such as Figure 5 If the patient has multiple polyps and the maximum diameter of the polyps is greater than 1 cm, then P (GCP | multiple polyps, diameter ≥ 1 cm) = 0.70, such as Figure 6 If the patient is female and younger than 55 years old, then P(GCP|female, age ≤55 years old, multiple polyps, diameter ≥1 cm) = 0.75, as shown in Figure 7 If we continue to add evidence information, the patient does not have gallstones and the gallbladder wall is not thickened, then P (GCP | female, age ≤ 55 years old, multiple polyps, diameter ≥ 1 cm, no gallstones, no gallbladder wall thickening) = 0.81, such as Figure 8 If the patient's polyp growth time is greater than 2 years, then P (GCP | female, age ≤ 55 years, multiple polyps, diameter ≥ 1 cm, no gallstones, no thickening of the gallbladder wall, polyp growth time ≥ 2 years) = 0.86, such as Fig. 9 shown.

[0051] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A gallbladder cholesterol polyp prediction system, characterized in that: include: A collection device for obtaining historical patient pathology data; An analysis device is used to obtain univariate analysis results of non-cholesterol polyp-related factors by univariate Logistic regression analysis based on historical patient pathological data; and to obtain multivariate Logistic regression analysis results of non-cholesterol polyp-related factors by multivariate Logistic regression analysis based on the univariate analysis results of non-cholesterol polyp-related factors; Constructing a device for constructing a Bayesian network prediction model based on the results of multivariate logistic regression analysis of non-cholesterol polyp-related factors; The prediction device is used to input the patient's pathological data to be processed into the Bayesian network prediction model to predict gallbladder cholesterol polyps.

2. The gallbladder cholesterol polyp prediction system according to claim 1, characterized in that: The cutoff values ​​of age and polyp diameter for distinguishing cholesterol polyps from non-cholesterol polyps were determined according to the maximum value of Youden index.

3. The gallbladder cholesterol polyp prediction system according to claim 2, characterized in that: Historical patient pathological data included: gender, age, polyp diameter, polyp growth time, number of polyps, whether the gallbladder wall was thickened, whether there were stones, and the nature of the polyps; among them, the nature of the polyps included: cholesterol polyps and non-cholesterol polyps.

4. The gallbladder cholesterol polyp prediction system according to claim 3, characterized in that: The analytical device includes: The first analysis unit is used to obtain the univariate analysis results of non-cholesterol polyp-related factors through univariate logistic regression analysis based on gender, age, polyp diameter, polyp growth time, number of polyps, whether the gallbladder wall is thickened, whether there is concomitant stone, and polyp nature; The second analysis unit was used to establish a binary multivariate Logistic regression analysis using gender, age, polyp diameter, number of polyps, whether the gallbladder wall was thickened, and whether stones were present in the univariate analysis results of non-cholesterol polyp-related factors as independent variables and non-cholesterol gallbladder polyps as dependent variables. Gender, number of polyps, whether the gallbladder wall was thickened, and whether stones were present were defined as categorical covariates to screen out independent risk factors for non-cholesterol polyps and obtain the multivariate Logistic regression analysis results of non-cholesterol polyp-related factors.

5. The gallbladder cholesterol polyp prediction system according to claim 4, characterized in that: The input nodes of the Bayesian network prediction model include: gender, age, polyp diameter, polyp growth time, number of polyps, whether the gallbladder wall is thickened, and whether there are stones. The target nodes of the Bayesian network prediction model include: polyp properties.

Images