Method, device and equipment for early warning and detecting severe diseases of digestive system

Abstract

The invention relates to the technical field of disease prediction, in particular to an early warning detection method, device and equipment for severe diseases of a digestive system, and the method comprises the steps: obtaining multi-modal data information of a patient, the multi-modal data information of the patient comprises clinical symptoms, laboratory examination and iconography examination; constructing and training a clustering model to obtain a plurality of clustering sets, inputting the multi-modal data information of the patient into a target clustering set in the clustering model, judging whether the multi-modal data information of the patient is abnormal or not, if the multi-modal data information of the patient is abnormal, constructing a probability model, and if the multi-modal data information of the patient is abnormal, determining whether the multi-modal data information of the patient is abnormal; generating a severity label from the target cluster set, and inputting the severity label into a probability model to obtain a probability result of severity assessment of the severe disease; and performing graded early warning according to a probability result, and performing clinical decision making according to the graded early warning. Therefore, the problems that in the prior art, various digestive system diseases cannot be evaluated by using the same method, and results are different and not timely when the diseases are evaluated are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of disease prediction, in particular to a digestive system severe disease early warning detection method, device and equipment. BACKGROUND

[0002] The digestive system severe disease is the most common in clinic, and is a major class of diseases endangering people's life and health. The incidence of this kind of disease has increased significantly, and the incidence is gradually close to that in western developed countries, but the mortality rate in China is still higher than that in developed countries. Due to the continuous promotion of medical education at the government level and the increase of medical investment in primary medical institutions, the incidence and mortality rate of digestive system severe diseases in developed countries such as the United States and Japan have shown a significant downward trend in the past 20 years. The incidence of digestive system severe diseases in China, especially severe acute pancreatitis and acute and subacute liver failure, is still in a state of significant increase. Although the diagnosis and treatment ability of digestive system severe diseases in many large third-grade hospitals in China has approached the overall level of developed countries, the average level in China is far behind developed countries, and the mortality rate is still significantly higher than that in developed countries such as the United States. The mortality rate of severe pancreatitis is about 13-35%, the mortality rate of acute gastrointestinal hemorrhage is about 4-20%, the mortality rate of medical comprehensive treatment of acute-on-chronic liver failure is as high as 50-90%, and the mortality rate of enterogenic sepsis is about 50%, which seriously threatens the health of the people.

[0003] The management of critical patients, the proposal of scientific and standardized rescue process and the strengthening of basic nursing and psychological nursing can improve the success rate of rescue of acute and critical patients.

[0004] In related technologies, the evaluation of digestive system severe diseases mainly uses scoring systems such as CTSI score, SOFA score and MELD score. Each scoring system is suitable for different disease types, and none of the scoring systems can be applied to the four common digestive system severe diseases at the present stage. Therefore, before using the scoring system, doctors of digestive medicine need to first determine the exact severe disease type of the patient according to the symptoms and examination data of the patient, and then select the corresponding scoring system according to the disease type to evaluate the severity of the patient. In clinical judgment of disease type, on the one hand, isolated indicators are often used as the basis for judgment, such as serum amylase higher than 3 times the normal upper limit value to diagnose acute pancreatitis. This method has poor specificity, and about 20% of non-acute pancreatitis patients also have elevated serum amylase. On the other hand, doctors with different years of experience have different judgments on the same case, which also affects the selection of scoring systems, is not conducive to the evaluation and treatment of the patient, and also causes waste of medical resources.

[0005] Furthermore, for severe digestive system diseases, it is crucial to identify high-risk patients as early as possible and prevent mild or moderate cases from progressing to severe illness. The aforementioned manual assessment of the condition is time-consuming in practice, hindering early intervention. Summary of the Invention

[0006] This application provides a method for early warning detection of severe digestive system diseases, in order to solve the problems in the prior art, such as the inability to use the same method to evaluate multiple types of digestive system diseases, inconsistencies in the results of disease evaluation, and lack of timeliness.

[0007] The first aspect of this application provides a method for early warning detection of severe digestive system diseases, comprising the following steps: acquiring multimodal patient data information, wherein the multimodal patient data information includes clinical symptoms, laboratory tests, and imaging examinations; constructing and training a clustering model to obtain multiple cluster sets; inputting the multimodal patient data information into a target cluster set in the clustering model; determining whether the multimodal patient data is abnormal; wherein, if the multimodal patient data information is abnormal, constructing a probability model; generating severity labels for the target cluster sets and inputting them into the probability model to obtain a probability result for severe disease severity assessment; performing graded early warning based on the probability result; and making clinical decisions based on the graded early warning.

[0008] Optionally, the step of inputting patient multimodal data information into the target cluster set in the clustering model includes: constructing a similarity matrix based on the patient multimodal data information to obtain multimodal similarity matrix data; fusing the multimodal similarity matrix data to generate a degree matrix; constructing a Laplacian matrix based on the similarity matrix and the degree matrix; performing eigenvalue decomposition on the Laplacian matrix to obtain feature vectors; performing K-means clustering on the feature vectors to obtain the final category label; and outputting the target clustering result.

[0009] Optionally, the formula for the similarity matrix is: The similarity matrix for clinical symptom data is as follows:

[0010] in, It is the set of symptoms of patient i. Let j represent the set of symptoms of patient j, and let elements be... This indicates the similarity of clinical symptoms between the i-th patient and the j-th patient; The similarity matrix for laboratory test data is:

[0011] in, It's a bandwidth parameter. represents a set of laboratory examination data of the i-th patient, represents a set of laboratory examination data of the j-th patient, element represents the similarity of the laboratory examination data of the i-th patient and the j-th patient; The similarity matrix for the imaging examination data is:

[0012] wherein B is the number of histogram bins, is the frequency of the i-th image in the K-th bin, is the frequency of the j-th image in the K-th bin; element represents the similarity of the imaging examination data of the i-th patient and the j-th patient.

[0013] Optionally, the formula of the multi-modal similarity matrix is:

[0014]

[0015] wherein M is 3, indicating that there are 3 modalities of data, is the weight of the m-th modality data similarity matrix.

[0016] Optionally, the formula of the degree matrix is: The formula of the Laplacian matrix is: , The formula of the objective function of the clustering is:

[0017] wherein Tr represents the sum of the diagonals, F is a feature vector extracted from the Laplacian matrix and subjected to regularization processing, is the transpose of F, and I is a unit diagonal matrix.

[0018] Optionally, the target cluster set generates a severity label, comprising: performing secondary clustering on the target cluster set to obtain a severity label, wherein the severity label comprises a primary label, a secondary label, and a tertiary label.

[0019] Optionally, the probability model comprises:

[0020]

[0021]

[0022] wherein, M is the total number of decision trees, K is the number of classification species, and is set to 3, is the prediction contribution of the mth decision tree to the sample x belonging to the class K, is the accumulation of the prediction results of the M decision trees, is the probability of the sample x belonging to the class k.

[0023] Optionally, the grading early warning according to the probability result comprises: determining an initial disease degree category according to the probability result output by the probability model, wherein the initial disease degree category comprises a first disease, a second disease, and a third disease, and the priority is set as the first disease being greater than the second disease, and the second disease being greater than the third disease; determining a first target value, a second target value, and a third target value according to the probability result of the first disease, the second disease, and the third disease, and if the first target value is greater than or equal to a first preset value, determining a final disease degree category, otherwise, judging whether the sum of the first target value and the second target value is greater than or equal to a second preset value: wherein if the sum of the first target value and the second target value is greater than or equal to the second preset value, judging whether the difference between the first target value and the second target value is greater than or equal to a third preset value, if the difference between the first target value and the second target value is greater than or equal to the third preset value, determining the disease degree category of the first target value as the final disease degree category, otherwise, determining the final disease degree category according to the priority of the disease degree categories of the first target value and the second target value; otherwise, if the sum of the first target value and the second target value is less than the second preset value, determining the second disease as the final disease degree category; judging the early warning level according to the final disease degree category, if the final disease degree category is the first disease, starting a first early warning, if the final disease degree category is the second disease, starting a second early warning, and if the final disease degree category is the third disease, starting a third early warning.

[0024] The second aspect embodiment of the present application provides a severe disease early warning detection device of a digestive system, comprising: an acquisition module configured to acquire patient multi-modal data information, wherein the patient multi-modal data information comprises clinical symptoms, laboratory examination, and imaging examination; a construction module configured to construct and train a clustering model to obtain a plurality of clustering sets, input the patient multi-modal data information into a target clustering set in the clustering model, and judge whether the patient multi-modal data is abnormal, wherein if the patient multi-modal data information is abnormal, a probability model is constructed, the target clustering set is generated into a severity label and input into the probability model to obtain a probability result of severe disease severity evaluation; An electronic device is provided in a third aspect of the present application, comprising: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are configured to execute a severe disease early warning detection method for the digestive system as described in the above embodiments.

[0025] The beneficial effects achieved by the present application are as follows: the embodiments of the present application realize comprehensive evaluation of the health status of the patient by integrating the clinical symptoms, laboratory examination, imaging examination and other multi-modal data of the patient, overcome the limitations of using a single data source, significantly improve the accuracy and reliability of disease evaluation; automatic classification is performed by using spectral clustering to construct a clustering model, which reduces the workload of physicians in classifying diseases and shortens the examination time, thereby gaining more treatment time for the patient; the severity of the disease of the patient is evaluated by using a probability model, which provides objective and quantitative evaluation basis for clinicians and helps to distinguish patients with different severity and optimize the allocation of medical resources; the grading early warning system based on the probability result helps the gastroenterologists to quickly make clinical decisions, improves the treatment efficiency and improves the prognosis of the patient. Thus, the problems in the prior art that the same method cannot be used to evaluate multiple digestive system diseases, the evaluation results are different, and the evaluation is not timely are solved.

[0026] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter in the description of embodiments of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0027] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description of embodiments of the present application, taken in conjunction with the accompanying drawings: Figure 1 A flowchart of a severe disease early warning detection method for the digestive system according to an embodiment of the present application is provided; Figure 2 A flowchart of a method for obtaining a probability result of severe disease severity evaluation for the digestive system according to an embodiment of the present application is provided; Figure 3 A flowchart of a clustering model construction method according to an embodiment of the present application is provided; Figure 4 A flowchart of a method for determining a target value according to a probability model result according to an embodiment of the present application is provided; Figure 5 A flowchart of a method for determining a final disease degree category according to an embodiment of the present application is provided; Figure 6 A flowchart of a grading early warning method for a disease clustering result of acute pancreatitis according to an embodiment of the present application is provided; Figure 7 A flow chart of a disease clustering result is a grading early warning method for gastrointestinal massive hemorrhage according to an embodiment of the present application; Figure 8 A flow chart of a disease clustering result is a grading early warning method for intestinal sepsis according to an embodiment of the present application; Figure 9 A flow chart of a disease clustering result is a grading early warning method for acute-on-chronic liver failure according to an embodiment of the present application; Figure 10 A structural schematic diagram of a digestive system severe disease early warning detection device provided by the embodiment of the present application.

[0028] Figure 11 A structural schematic diagram of an electronic device provided by the embodiment of the present application.

[0029] Legend: P(1): probability of a patient suffering from a first-level disease; P(2): probability of a patient suffering from a second-level disease; P(3): probability of a patient suffering from a third-level disease. DETAILED DESCRIPTION

[0030] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0031] A digestive system severe disease early warning detection method and device are described below with reference to the accompanying drawings. In view of the long time for human assessment of the disease, the inability to identify the type of digestive system severe disease early, and the timely assessment of the severity of the patient's disease mentioned in the background art, the present application provides a digestive system severe disease early warning detection method. In this method, the patient's health status is comprehensively evaluated by integrating the patient's clinical symptoms, laboratory tests, imaging tests, and other multi-modal data, overcoming the limitations of using a single data source and significantly improving the accuracy and reliability of disease assessment. Automatic classification is performed using spectral clustering to construct a clustering model, reducing the workload of physicians in classifying diseases and shortening the examination time, which allows patients to gain more treatment time. The severity of the patient's disease is assessed using a probability model to provide objective and quantitative assessment criteria for clinicians, which helps to distinguish patients with different severity and optimize the allocation of medical resources. The grading early warning system based on the probability results can help gastroenterologists make quick clinical decisions, improve treatment efficiency, and improve patient outcomes. Thus, the problems of being unable to use the same method to assess multiple digestive system diseases, having different results when assessing diseases, and not being timely in the prior art are solved.

[0032] Specifically, Figure 1 A flowchart of a digestive system severe disease early warning detection method provided by an embodiment of the present application is shown.

[0033] As Figure 1 shown, the digestive system severe disease early warning detection method includes the following steps: In step S101, the patient's multi-modal information is obtained.

[0034] The patient's multi-modal data information includes clinical symptoms, laboratory tests, and imaging tests. The clinical symptoms, which can be referred to as clinical manifestations, refer to a series of abnormal changes in the body after the body has a disease, which can be used as an important basis for disease diagnosis. Laboratory tests refer to determining the characteristics of the submitted material, such as content, properties, concentration, and quantity, through physical or chemical tests in the laboratory. Imaging tests refer to using ultrasonic waves, X-rays, and other media to display the structure and density of human tissues and organs in the form of imaging, which is used to determine whether there is a local lesion.

[0035] Specifically, the clinical symptoms include heart rate, blood pressure, body temperature, stool condition, whether abdominal pain, whether fatigue, whether bleeding, whether hematemesis, whether dyspnea, whether syncope. Laboratory tests include blood routine (white blood cell count and classification, red blood cell count and hemoglobin, platelet count), C-reactive protein, blood amylase, lipase, blood glucose, urine amylase, liver (direct bilirubin, indirect bilirubin), renal function (creatinine, urine output) tests, imaging tests include standing abdominal plain film, abdominal CT, abdominal ultrasound.

[0036] It can be understood that the embodiments of the present application can obtain similar features between diseases by acquiring and analyzing common clinical symptoms, laboratory examination, and imaging examination data indicators of severe diseases of the digestive system, which facilitates subsequent clustering and evaluation.

[0037] In step S102, a clustering model is constructed and trained to obtain a plurality of clustering sets. The patient multi-modal data information is input into a target clustering set in the clustering model, and it is judged whether the patient multi-modal data is abnormal. If the patient multi-modal data information is abnormal, a probability model is constructed, the target clustering set is generated to generate a severity label and is input into the probability model to obtain a probability result of severe disease severity evaluation.

[0038] Specifically, as shown in Figure 2 , clustering analysis is performed by the features of the multi-modal data. The clustering model is constructed by a multi-modal spectral clustering algorithm. The target clustering set is the output of the clustering model and is divided into six kinds, one of which is a normal set and the other five are abnormal sets. The abnormal sets are four common severe diseases of the digestive system and other digestive system diseases. The four common severe diseases of the digestive system are acute pancreatitis, massive gastrointestinal bleeding, intestinal sepsis, and acute-on-chronic liver failure. The probability model is constructed using a LightGBM model. LightGBM is a high-efficiency gradient boosting tree framework developed by Microsoft. Compared with XGBoost and CatBoost algorithms, LightGBM is more suitable for large-scale data and high-dimensional features and is more suitable for medical scenarios. In this embodiment, LightGBM is used as a probability model, and the output is a three-class probability, which facilitates subsequent secondary evaluation of the disease.

[0039] It can be understood that the embodiments of the present application can describe the same object from different angles by fusing multi-modal data information for clustering analysis, make up for the noise and information loss of a single modality, and improve the comprehensive characterization of patient features. Through cross-validation of multi-modal data, the robustness is improved. The classification result of the disease is output as a probability, which can provide uncertainty measurement and support subsequent probability calibration, and provide more detailed decision support.

[0040] In this embodiment of the application, inputting patient multimodal data information into the target cluster set in the clustering model includes: constructing a similarity matrix based on the patient multimodal data information to obtain multimodal similarity matrix data; fusing the multimodal similarity matrix data to generate a degree matrix; constructing a Laplacian matrix based on the similarity matrix and the degree matrix; performing eigenvalue decomposition on the Laplacian matrix to obtain feature vectors; performing K-means clustering on the feature vectors to obtain the final category label; and outputting the target clustering result.

[0041] Specifically, such as Figure 3 As shown, the similarity matrix is ​​used to quantify the similarity between samples. When constructing the similarity matrix based on multimodal patient data, the Jaccard similarity coefficient is used for clinical symptom data. For laboratory test data, a Gaussian kernel function generated by Euclidean distance is used. For imaging test data, the histogram intersection similarity method is used. A weighted average method is used to fuse the three similarity matrices to obtain the multimodal similarity matrix. During the weighted linear fusion process, cross-validation is used to optimize the weights. The degree matrix is ​​a diagonal matrix derived from the similarity matrix and is used to describe the connection strength of each node in the graph. The degree matrix is ​​generated based on the multimodal similarity matrix. The Laplacian matrix is ​​used to describe the structural features of the graph. A normalized Laplacian matrix is ​​constructed based on the degree matrix, and the eigenvectors F corresponding to the first K smallest eigenvalues ​​of the Laplacian matrix are calculated. In this embodiment, K is set to 6. After regularization, the eigenvectors F are clustered using the K-means clustering algorithm, outputting K cluster sets.

[0042] It is understood that the embodiments of this application employ Jaccard similarity coefficient, Gaussian kernel function generated by Euclidean distance, and histogram intersection similarity method to construct similarity matrices for multimodal data, respectively. Jaccard similarity coefficient is suitable for binary features; for clinical symptom data, symptomatic data is represented by 1, and asymptomatic data by 0, with symptom severity expressed numerically, effectively measuring the similarity of sets. Laboratory test data is generally represented by specific numerical values; the Gaussian kernel function generated by Euclidean distance is sensitive to data and can smooth similarity through the kernel function. Imaging examination data outputs a set of images; the histogram intersection similarity method is suitable for matching the distribution of image data.

[0043] The embodiments of the present application improve the accuracy of the modality similarity matrix through customized processing of each modality data. The cross-validation optimization weight ensures the scientific and reasonable contribution of different modalities, avoids the deviation caused by subjective weight setting, and enhances the reliability of the fused multi-modal similarity matrix. The conversion from the similarity matrix to the degree matrix and then to the Laplacian matrix converts the data into a graph structure, captures the global relationship between samples by using the multi-modal spectral clustering method, and mines the inherent similarity. Selecting the first K smallest eigenvectors of the Laplacian matrix can effectively extract the low-dimensional embedding representation of the data, remove noise and retain the main structure. K-means clustering has high computational efficiency in low-dimensional space, and the clustering effect of K-means can be effectively improved by performing clustering after regularizing the feature vector.

[0044] In the embodiments of the present application, the formula of the similarity matrix is: The similarity matrix of the clinical symptom data is:

[0045] wherein, is the symptom set of the patient i, is the symptom set of the patient j, and the element represents the similarity of the clinical symptoms of the i th patient and the j th patient; The similarity matrix of the laboratory examination data is:

[0046] wherein, is a bandwidth parameter, is the laboratory examination data set of the i th patient, is the laboratory examination data set of the j th patient, and the element represents the similarity of the laboratory examination data of the i th patient and the j th patient; The similarity matrix of the imaging examination data is:

[0047] wherein, B is the number of histogram bins, is the frequency of the i th image in the K th bin, is the frequency of the j th image in the K th bin; and the element represents the similarity of the imaging examination data of the i th patient and the j th patient.

[0048] In the embodiments of the present application, the multi-modal similarity matrix data is obtained, wherein the formula of the multi-modal similarity matrix is:

[0049]

[0050] wherein M is 3, indicating that there are 3 modalities of data, is a weight of the mth modality data similarity matrix.

[0051] In the embodiment of the present application, the formula of the degree matrix is: The formula of the Laplacian matrix is: The formula of the objective function of clustering is:

[0052] wherein Tr represents summing along the diagonal, F is a feature vector extracted from the Laplacian matrix and subjected to regularization processing, is a transpose of F, and I is a unit diagonal matrix.

[0053] In the embodiment of the present application, the target cluster set is generated with a severity label, comprising: performing secondary clustering on the target cluster set to obtain the severity label, wherein the severity label comprises a first label, a second label and a third label.

[0054] wherein the K-means method is used to perform secondary clustering on the target cluster set respectively. In the embodiment of the present application, K of clustering is set to 3, obtaining 3 cluster sets, which are a first severity, a second severity and a third severity respectively. According to the cluster set to which the patient belongs, the first label, the second label and the third label are respectively assigned to the patient data.

[0055] It can be understood that the embodiment of the present application performs secondary clustering on the target cluster set. The first clustering realizes the classification of diseases, but does not directly correspond to the severity of the disease. The secondary clustering of the target cluster can further induce the severity label, so that the data is suitable for subsequent supervised learning, realizing the smooth transition from unsupervised learning to supervised learning. In addition, the severity label generated by the secondary clustering reduces the subjective bias of manual labeling, and intuitively reflects the real disease level based on data distribution.

[0056] In the embodiment of the present application, the probability model comprises:

[0057]

[0058]

[0059] wherein M is the total number of decision trees, K is the number of categories, and is set to 3, ​is the prediction contribution of the mth decision tree to the sample x belonging to the class K, is the accumulation of the prediction results of the M decision trees, is the probability of the sample x belonging to the class k. The probability model is constructed using a LightGBM model, which is a gradient boosting framework based on decision trees. Gradient boosting is an iterative decision tree ensemble method, and a new decision tree is trained in each iteration to fit the residual of the previous iteration, and finally a model that can accurately classify is obtained. The prediction value output by the model is the original prediction value, which is converted into probability using the Softmax normalization method, and the sum of the probabilities of each class is 1.

[0060] Specifically, the number of classification categories is 3, namely the first-level label, the second-level label and the third-level label. If The original prediction value of a certain sample is [1.2, 0.5, -0.8], and the probability after transformation using the Softmax normalization method is [0.62, 0.30, 0.08], indicating that the probability of the sample being a first-level label is 0.62, the probability of being a second-level label is 0.30, and the probability of being a third-level label is 0.08. The prediction probabilities of the three classifications sum to 1.

[0061] It can be understood that the embodiments of the present application output the probability values of the patient data as the first-level label, the second-level label and the third-level label through the LightGBM probability model, the first-level label corresponds to the first disease, the second-level label corresponds to the second disease, and the third-level label corresponds to the third disease, thereby realizing rapid assessment of the severity of the patient's condition and providing data support for subsequent grading and early warning. The probability result provides an objective reference for doctors to develop personalized rehabilitation plans for patients and improve the rehabilitation effect of patients.

[0062] In step S103, grading and early warning is performed according to the probability result of severe disease severity assessment, and clinical decision-making is performed according to the grading and early warning.

[0063] wherein the probability result of severe disease severity assessment is a three-class probability result, the three-class probability result is evaluated twice to determine the final disease degree category, each disease degree category corresponds to a warning level, and the warning scheme is started according to the final disease degree category.

[0064] Specifically, if the clustering result is abnormal and the abnormal category is acute pancreatitis, the three-class probabilities output in the probability model are: first-level disease probability 0.2, second-level disease probability 0.5, and third-level disease probability 0.3. According to the evaluation process, it is determined that the final disease degree category is the second-level disease, and the second-level warning scheme is started. For the digestive system severe disease acute pancreatitis, the second-level warning scheme is to be diagnosed to a general ward for related supportive treatment.

[0065] It can be understood that the embodiments of the present application increase the probability uncertainty management, on the basis of the probability model output probability result, the probability result is evaluated twice, the final disease degree category is determined, and the risk of misjudgment is reduced. A multi-level early warning mechanism is adopted to optimize resource allocation and improve the accuracy of clinical decision-making.

[0066] In the embodiments of the present application, the hierarchical early warning according to the probability result includes: setting the priority of the first disease greater than the second disease, and the second disease greater than the third disease; determining a first target value, a second target value and a third target value according to the probability results of the first disease, the second disease and the third disease, if the first target value is greater than or equal to a first preset value, determining the final disease degree category, otherwise, judging whether the sum of the first target value and the second target value is greater than or equal to a second preset value: wherein, if the sum of the first target value and the second target value is greater than or equal to the second preset value, judging whether the difference between the first target value and the second target value is greater than or equal to a third preset value, if the difference between the first target value and the second target value is greater than or equal to the third preset value, determining the disease degree category of the first target value as the final disease degree category, otherwise, determining the final disease degree category according to the priority of the disease degree category of the first target value and the second target value; otherwise, if the sum of the first target value and the second target value is less than the second preset value, determining the second disease as the final disease degree category; determining the early warning level according to the final disease degree category, if the final disease degree category is the first disease, starting the first early warning, if the final disease degree category is the second disease, starting the second early warning, if the final disease degree category is the third disease, starting the third early warning.

[0067] Specifically, as shown in Figure 4 , if the three classification probability output by the probability model is [0.62, 0.30, 0.08], the first target value is determined as 0.62, the second target value is 0.30, and the third target value is 0.08. As shown in Figure 5 , wherein the first preset value is 0.45, the second preset value is 0.7, and the third preset value is 0.1. Because the first target value is greater than the first preset value, the disease degree category of the first target value is determined as the final disease degree category, that is, the final disease degree category is determined as the first disease, and the first early warning scheme corresponding to the disease is started.

[0068] It can be understood that the embodiments of the present application evaluate the three classification probability result by multiple preset values, A priority system is also used, in the case of two probabilities close to each other, the one with higher priority is determined to avoid delaying the treatment opportunity, causing the patient to develop into a severe case. The embodiments of the present application are not only suitable for the case where the probability result is clear, but also suitable for the case where the probability results of two or three classifications tend to be balanced. According to the probability difference and the priority of the disease, the appropriate disease level is selected as the final disease level, so as to improve the treatment effect and survival rate of the patient.

[0069] According to the severe digestive system disease early warning detection method provided in the embodiments of the present application, in the method, by integrating the multi-modal data of the patient's clinical symptoms, laboratory examination, imaging examination and the like, comprehensive evaluation of the patient's health state is realized, the limitation of using a single data source is overcome, and the accuracy and reliability of disease evaluation are significantly improved; the clustering model is constructed by using spectral clustering for automatic classification, the workload of the physician for classifying the disease is reduced, and the examination time is shortened, so as to gain more treatment time for the patient; the severity of the patient's disease is evaluated by using the probability model, objective and quantitative evaluation basis is provided for the clinician, which is helpful to distinguish patients with different severity and optimize the allocation of medical resources; the grading early warning system based on the probability result can help the gastroenterologist to make clinical decisions quickly, improve the treatment efficiency and improve the prognosis of the patient. Thus, the problems in the prior art that the same method cannot be used to evaluate multiple digestive system diseases, the evaluation result is different, and the evaluation is not timely are solved.

[0070] In the following, a severe digestive system disease early warning detection method will be described in detail, and the content is as follows: S1, acquiring multi-modal information of a patient.

[0071] S2, inputting the multi-modal information of the patient into a clustering model to obtain a target clustering result.

[0072] The specific steps are as shown in Figure 3 If the clustering result is normal, it indicates that the patient's digestive system is normal, and the program is terminated. If the clustering result is abnormal other disease, the disease is not processed by the model, and the doctor performs disease evaluation, and the program is terminated.

[0073] S3, if the clustering result is one of acute pancreatitis, gastrointestinal hemorrhage, intestinal sepsis and acute-on-chronic liver failure, the patient data is input into a secondary clustering model to generate a severity label. According to the severity of the disease, the severity label is first, second and third respectively.

[0074] S4, inputting the patient data with the severity label into a probability model to output the probability of the patient suffering from a first disease, a second disease and a third disease.

[0075] S5, obtaining a target value from the probabilities of the primary disease, secondary disease and tertiary disease to determine the final disease degree of the patient.

[0076] S6, performing hierarchical early warning according to the one-time clustering result and the final disease degree.

[0077] Specifically, if the one-time clustering result of the patient is acute pancreatitis, the hierarchical early warning scheme is as shown in Figure 6 If the final disease degree is a tertiary disease, a tertiary early warning is started to perform routine drug treatment on the patient. If the final disease degree is a secondary disease, a secondary early warning is started to triage the patient to a general ward for relevant supportive treatment. If the final disease degree is a primary disease, a primary early warning is started to perform vital sign monitoring and maintenance in a digestive ICU room, early fluid resuscitation, inflammation suppression, intestinal protection, early contact with etiology treatment, etc.

[0078] If the one-time clustering result of the patient is gastrointestinal massive hemorrhage, the hierarchical early warning scheme is as shown in Figure 7 If the final disease degree is a tertiary disease, a tertiary early warning is started to perform standardized drug treatment on the patient and monitor the vital signs of the patient. If the final disease degree is a secondary disease, a secondary early warning is started to perform expansion, blood transfusion and necessary drug treatment on the patient. If the final disease degree is a primary disease, a primary early warning is started to perform expansion, blood transfusion and necessary drug treatment on the patient, start a gastrointestinal massive hemorrhage endoscopic treatment program, make an endoscopic diagnosis and treatment decision, and perform emergency surgical treatment.

[0079] If the one-time clustering result of the patient is intestinal sepsis, the hierarchical early warning scheme is as shown in Figure 8 If the final disease degree is a tertiary disease, a tertiary early warning is started to perform primary disease treatment and drug treatment. If the final disease degree is a secondary disease, a secondary early warning is started to perform hospitalization procedures and fluid resuscitation treatment on the patient. If the final disease degree is a primary disease, a primary early warning is started to let the patient enter an ICU monitoring ward to perform close vital sign monitoring and comprehensive organ function support on the patient. Fluid resuscitation, infection control, nutritional support, intestinal decontamination and other treatment measures are performed.

[0080] If the one-time clustering result of the patient is acute-on-chronic liver failure, the hierarchical early warning scheme is as shown in Figure 9As shown, if the final disease degree is a tertiary disease, a tertiary early warning is started, and the patient is given etiological analysis and symptomatic and supportive drug treatment. If the final disease degree is a secondary disease, a secondary early warning is started, and the patient is given hospitalization procedures, etiological analysis and symptomatic and supportive drug treatment. The changes in liver function, coagulation function, blood routine, blood ammonia and other indicators are closely monitored, and complications are prevented and treated. If the final disease degree is a primary disease, a primary early warning is started, and the patient is admitted to an ICU monitoring ward for close monitoring of vital signs and supportive treatment. For patients with progressive deterioration and ineffective internal medicine treatment, liver transplantation evaluation is performed as soon as possible. During the waiting period for liver sources, supportive treatment and complication management are strengthened.

[0081] In summary, the embodiments of the present application distinguish the patient's condition according to the patient's multi-modal data by constructing a clustering model and a probability model, analyze the patient's disease type and disease severity. Through this mode recognition, the critical condition recognition window period can be advanced by 1 to 3 hours. For the four common digestive system severe diseases, a multi-level early warning scheme is formulated, and the diagnosis and treatment path of digestive system severe diseases is standardized, which provides more intervention time for critical patients. In addition, through the multi-level early warning scheme, intelligent triage is realized, the allocation efficiency of ICU beds is improved, excessive examination is avoided, and the optimization of medical resources is realized.

[0082] It should be noted that the foregoing explanation and description of the embodiment of the method for early warning detection of a digestive system severe disease also applies to the embodiment of the device for early warning detection of a digestive system severe disease, which will not be described here.

[0083] Next, a device for early warning detection of a digestive system severe disease according to an embodiment of the present application is described with reference to the accompanying drawings.

[0084] Figure 10 is a structural schematic diagram of a device for early warning detection of a digestive system severe disease according to an embodiment of the present application.

[0085] As shown in Figure 10 , the device for early warning detection of a digestive system severe disease 10 comprises an acquisition module 100, a construction module 200 and an early warning module 300.

[0086] The acquisition module 100 is configured to acquire patient multi-modal data information, wherein the patient multi-modal data information includes clinical symptoms, laboratory examination, and imaging examination.

[0087] In summary, the method for detecting severe disease of the digestive system according to the embodiments of the present application is proposed, in which the method, by integrating the multi-modal data of the clinical symptoms, laboratory examination, and imaging examination of the patient, achieves comprehensive assessment of the health status of the patient, overcomes the limitation of using a single data source, significantly improves the accuracy and reliability of disease assessment, automatically classifies by using the spectral clustering to construct a clustering model, reduces the workload of the physician in classifying the disease, shortens the examination time, and helps the patient to gain more treatment time, assesses the severity of the disease of the patient by using the probability model, provides objective and quantitative assessment basis for the clinician, helps to distinguish patients with different severity, and optimizes the allocation of medical resources, the hierarchical warning system based on the probability result helps the physician of the department of gastroenterology to quickly make clinical decisions, improves the efficiency of treatment, and improves the prognosis of the patient. Thus, the problems in the prior art that the same method cannot be used to assess multiple types of digestive system diseases, the results are different when assessing the disease, and the assessment is not timely are solved.

[0088] Figure 11 The electronic device provided by the embodiments of the present application is shown in the structural schematic diagram. The electronic device can include: The memory 1101, the processor 1102, and the computer program stored in the memory 1101 and executable on the processor 1102.

[0089] The processor 1102 executes the program to implement the method for detecting severe disease of the digestive system provided in the above embodiments.

[0090] Further, the electronic device further includes: The communication interface 1103 is configured to communicate between the memory 1101 and the processor 1102.

[0091] The memory 1101 is configured to store the computer program executable on the processor 1102.

[0092] The memory 1101 can include a high-speed RAM memory and can also include a non-volatile memory, such as at least one disk memory.

[0093] If the memory 1101, the processor 1102 and the communication interface 1103 are implemented independently, the communication interface 1103, the memory 1101 and the processor 1102 can be connected to each other through a bus and complete communication between each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 11 In the figure, only one thick line is used to represent the bus, but it does not mean that there is only one bus or only one type of bus.

[0094] Optionally, in a specific implementation, if the memory 1101, the processor 1102 and the communication interface 1103 are integrated on a chip, the memory 1101, the processor 1102 and the communication interface 1103 can complete communication between each other through an internal interface.

[0095] The processor 1102 can be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement one or more embodiments of the present application.

[0096] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, the skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples without contradiction.

[0097] Furthermore, the terms "first", "second", etc. are used herein only to describe different steps or features and do not imply a relative importance or a specific order of steps or features. Thus, features defined with "first", "second" etc. can include at least one of the features implicitly or explicitly. In the description of the application, the meaning of "N" is at least two, such as two, three, etc., unless otherwise expressly specified.

[0098] Any process or method descriptions or blocks in flow charts described herein and elsewhere can be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps in the process. Alternate implementations are included within the scope of the preferred embodiments of the present application in which the functions performed by the various processes described herein and elsewhere are allocated differently among the various components, are performed by a different sequence of processes, are performed by different processes altogether, or are performed by a different combination of processes.

[0099] It should be understood that aspects of the present application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, the various steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. As such, if implemented in hardware and in another embodiment, the hardware can include any or a combination of the following technologies, which are all well known in the art: a discrete logic circuit having logic gates for implementing logic functions upon an application of data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0100] Those of skill in the art would understand that information and signals can be represented using any of a variety of technologies and techniques. Further, those of skill in the art would understand that the mechanisms of the preferred embodiments can be implemented using any of a variety of appropriate technologies and techniques. For example, the preferred embodiments can be implemented using digital circuitry, or using one or more processors programmed with appropriate software routines.

Claims

1. A method for early detection of severe disease of digestive system, characterized in that, The method comprises the following steps: Obtaining patient multi-modal data information, wherein the patient multi-modal data information comprises clinical symptoms, laboratory examination, and imaging examination; Constructing and training a clustering model to obtain a plurality of clustering sets, inputting the patient multi-modal data information into a target clustering set in the clustering model, and judging whether the patient multi-modal data is abnormal, wherein if the patient multi-modal data information is abnormal, a probability model is constructed, the target clustering set is generated into a severity label and input into the probability model to obtain a probability result of severe disease severity assessment; According to the probability result, a hierarchical early warning is performed, and a clinical decision is made according to the hierarchical early warning.

2. The method according to claim 1, wherein Inputting the patient multi-modal data information into a target clustering set in the clustering model comprises: Constructing a similarity matrix according to the patient multi-modal data information to obtain multi-modal similarity matrix data; Fusing the multi-modal similarity matrix data to generate a degree matrix; According to the similarity matrix and the degree matrix, a Laplacian matrix is constructed, the Laplacian matrix is subjected to feature decomposition to obtain a feature vector, the feature vector is subjected to K-means clustering to obtain a final class label, and a target clustering result is output.

3. The method according to claim 2, wherein The formula of the similarity matrix is: The similarity matrix for clinical symptom data is: ； wherein, is a set of symptoms of patient i, denotes a set of symptoms of patient j, elements denotes the similarity of the clinical symptoms of the i-th patient and the j-th patient; The similarity matrix for laboratory examination data is: ； wherein, is a bandwidth parameter, represents a set of laboratory examination data of the i-th patient, represents a set of laboratory examination data of the j-th patient, element represents a similarity of laboratory examination data of the i-th patient and the j-th patient; The similarity matrix for imaging examination data is: ； where B is the number of histogram bins, is the frequency of the i-th image in the K-th bin, is the frequency of the j-th image in the K-th bin; element represents the similarity of the i-th patient and the j-th patient imaging data.

4. The method according to claim 2, wherein The multi-modal similarity matrix data is obtained, wherein the formula of the multi-modal similarity matrix is: ； ； wherein M is 3, indicating that there are 3 kinds of modal data, is the weight of the mth modal data similarity matrix.

5. The method according to claim 2, wherein The formula of the degree matrix is: ； The formula of the Laplacian matrix is: ； ； The target function formula of the clustering is: ； wherein Tr denotes the diagonal sum, F is the feature vector extracted from the Laplacian matrix and regularized, is the transpose of F, and I is the identity diagonal matrix.

6. The method of claim 1, wherein the method is a method for predicting a severe disease of the digestive system. Generating a severity label for the target clustering set comprises: performing secondary clustering on the target clustering set to obtain a severity label, wherein the severity label comprises a first-level label, a second-level label, and a third-level label.

7. The method for early warning detection of severe digestive system diseases according to claim 1, characterized in that, The probability model comprises: ； ； ； where M is the total number of decision trees, K is the number of classification categories, set to 3, is the prediction contribution of the mth decision tree to the classification of sample x into class K, is the accumulation of the prediction results of the M decision trees, is the probability that sample x belongs to class k.

8. The method of claim 1, wherein the method is for early detection of a severe digestive disease. According to the probability result, a hierarchical early warning is performed, and a clinical decision is made according to the hierarchical early warning. According to the probability result output from the probability model, an initial disease degree category is determined, wherein the initial disease degree category comprises a first-level disease, a second-level disease, and a third-level disease, and the priority is set as first-level disease>second-level disease>third-level disease; According to the probability results of the first-level disease, the second-level disease, and the third-level disease, a first target value, a second target value, and a third target value are determined, if the first target value is greater than or equal to a first preset value, a final disease degree category is determined, otherwise, it is judged whether the sum of the first target value and the second target value is greater than or equal to a second preset value: if the sum of the first target value and the second target value is greater than or equal to the second preset value, it is judged whether the difference between the first target value and the second target value is greater than or equal to a third preset value, if the difference between the first target value and the second target value is greater than or equal to the third preset value, the disease degree category of the first target value is determined as the final disease degree category, otherwise, the final disease degree category is determined according to the priority of the disease degree categories of the first target value and the second target value; otherwise, if the sum of the first target value and the second target value is less than the second preset value, the second-level disease is determined as the final disease degree category. According to the final disease degree category, a warning level is determined. If the final disease degree category is a first-level disease, a first-level warning is initiated. If the final disease degree category is a second-level disease, a second-level warning is initiated. If the final disease degree category is a third-level disease, a third-level warning is initiated.

9. A digestive system severe disease early warning detection device, characterized by, The method comprises the following steps: An acquisition module is configured to acquire multi-modal data information of a patient, wherein the multi-modal data information of the patient comprises clinical symptoms, laboratory examination, and imaging examination; A construction module is configured to construct and train a clustering model to obtain a plurality of clustering sets, input the multi-modal data information of the patient into a target clustering set in the clustering model, and determine whether the multi-modal data of the patient is abnormal. If the multi-modal data information of the patient is abnormal, a probability model is constructed, the target clustering set is generated with a severity label and input into the probability model to obtain a probability result of severe disease severity assessment. A warning module is configured to perform hierarchical warning according to the probability result and clinical decision-making according to the hierarchical warning.

10. An electronic device, comprising: The method comprises the following steps: A memory, a processor, and a computer program stored in the memory and executable on the processor are provided. The processor executes the program to implement the severe digestive disease warning detection method according to any one of claims 1-8.

Images