An out-of-hospital hierarchical diagnosis method based on ibs conditions
By acquiring multi-dimensional information outside the hospital and combining it with machine learning models for IBS classification and diagnosis, the problem of low accuracy in outpatient diagnosis has been solved, enabling accurate classification and early intervention of IBS and reducing the burden on the healthcare system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 侯晓华
- Filing Date
- 2026-03-31
- Publication Date
- 2026-06-05
Smart Images
Figure CN122158089A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of medical diagnostic technology, specifically to an outpatient tiered diagnostic method for IBS. Background Technology
[0002] Irritable bowel syndrome (IBS) is a common functional gastrointestinal disorder with a complex pathogenesis closely related to intestinal motility abnormalities, visceral hypersensitivity, intestinal microecological imbalance, and psychological factors. Clinical manifestations include recurrent abdominal pain and bloating, accompanied by changes in bowel habits (diarrhea, constipation, or alternating between the two), severely impacting patients' quality of life. Currently, the diagnosis of IBS mainly relies on in-hospital clinicians' history taking, physical examination, and exclusionary tests (such as stool analysis and colonoscopy). The diagnostic process is highly dependent on medical resources and the physician's clinical experience.
[0003] However, IBS patients often experience chronic, recurrent episodes, and in most cases, frequent hospitalizations are unnecessary. Convenient and accurate tiered diagnostic methods are lacking in out-of-hospital settings (such as home and community). Current IBS diagnostic methods are largely limited to in-hospital settings, and out-of-hospital diagnosis suffers from the following shortcomings: first, low diagnostic accuracy, relying heavily on single symptoms and failing to integrate multi-dimensional information available outside the hospital; second, unclear tiering, unable to distinguish between mild, moderate, and severe IBS, making it difficult to provide targeted intervention recommendations.
[0004] Therefore, the present invention provides an outpatient triage diagnostic method based on IBS symptoms to solve the above-mentioned problems. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a method for outpatient triage diagnosis of IBS symptoms, which solves the problems mentioned in the background section.
[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for outpatient triage diagnosis of IBS, comprising the following steps:
[0007] S1: Obtain multi-dimensional information related to IBS symptoms provided by patients in out-of-hospital settings;
[0008] S2: Standardize and preprocess the multi-dimensional information from outside the hospital to obtain feature data that can be input into the algorithm model;
[0009] S3: Input the standardized feature data into the preset IBS grading diagnosis algorithm model and output the patient's IBS disease grading diagnosis result;
[0010] The IBS tiered diagnostic algorithm model is trained using multiple sets of relevant outpatient sample data and is used to achieve outpatient tiered diagnosis of IBS.
[0011] Preferably, the multi-dimensional information outside the hospital includes at least one of symptom information, basic health information, lifestyle information, and past medical history information, which can be filled in by the patient or obtained through simple inquiry.
[0012] Preferably, the standardization preprocessing in S2 includes data cleaning, outlier handling, missing value imputation, feature encoding, and feature normalization, which are used to eliminate the dimensional differences and data errors of the original outpatient information and transform it into feature data in a unified format.
[0013] Preferably, the training process of the IBS hierarchical diagnostic algorithm model includes:
[0014] A1: Collect multi-dimensional information from multiple groups of samples outside the hospital, as well as the IBS grading diagnostic labels corresponding to each group of samples, and construct a sample dataset;
[0015] A2: Perform the standardization preprocessing described in claim 3 on the sample dataset to obtain standardized sample feature data;
[0016] A3: Based on the standardized sample feature data and the corresponding IBS grading diagnostic labels, an initial algorithm model is trained, and after performance verification and optimization, the preset IBS grading diagnostic algorithm model is obtained.
[0017] Preferably, the initial algorithm model is a machine learning model, selected from at least one of random forest model, support vector machine model, neural network model, and logistic regression model. A single model or a combination of models can be selected according to the diagnostic accuracy requirements.
[0018] Preferably, the IBS disease grading diagnosis results include four levels: no IBS, mild IBS, moderate IBS, and severe IBS, and the confidence level corresponding to each level is output. The confidence level is used to characterize the reliability of the diagnosis results.
[0019] Preferably, the method further includes model iterative updates: periodically collecting new multi-dimensional information about patients outside the hospital and corresponding clinical diagnosis results, constructing a new sample dataset, and fine-tuning the IBS grading diagnosis algorithm model based on the new sample dataset to achieve iterative optimization of model performance.
[0020] Preferably, the samples in the sample dataset cover people of different ages, genders, regions and symptom presentations, and the IBS grading diagnostic labels are determined based on the Rome IV diagnostic criteria and clinical diagnosis results to ensure sample diversity and label accuracy.
[0021] Preferably, the confidence level is calculated using the probability output by the model. When the confidence level is lower than a preset threshold, the patient is prompted to supplement relevant information outside the hospital for re-diagnosis, or it is recommended to go to the hospital for further examination.
[0022] Beneficial effects
[0023] This invention provides a method for outpatient triage diagnosis of IBS. Compared with existing technologies, it has the following advantages:
[0024] (1) This outpatient triage diagnosis method based on IBS integrates multi-dimensional outpatient information, combines standardized preprocessing to eliminate data errors, adopts multi-feature screening and model optimization, trains the model with reference to the Rome IV diagnostic criteria, and judges the reliability of diagnosis through confidence level to reduce the rate of misdiagnosis and missed diagnosis; the model iteration and update mechanism can continuously improve the diagnostic accuracy and adapt to changes in clinical needs.
[0025] (2) This outpatient triage diagnosis method based on IBS symptoms does not rely on professional medical equipment and doctor guidance in the hospital. Patients can provide outpatient information on their own. The process is simple and convenient. It is suitable for outpatient screening and triage of a wide range of people, which helps to detect IBS disease early and provide targeted intervention, and reduces the burden on the medical system. Attached Figure Description
[0026] Figure 1 The flowchart of an outpatient triage diagnostic method for IBS is provided by the present invention. Detailed Implementation
[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0028] This invention provides a technical solution: an outpatient triage diagnostic method for IBS, specifically including the following:
[0029] S1: Obtain multi-dimensional information related to IBS symptoms provided by patients in out-of-hospital settings;
[0030] Multi-dimensional information outside the hospital does not require specialized medical equipment for testing within the hospital. It can be obtained through methods such as patient self-reporting, online inquiries, and paper questionnaires, including but not limited to:
[0031] (1) Symptom information: frequency of abdominal pain (times / week), duration of abdominal pain (minutes / time), degree of abdominal distension (1-5 points, 1 point for no abdominal distension, 5 points for severe abdominal distension), bowel habits (diarrhea / constipation / alternating between the two), frequency of bowel movements (times / day), stool characteristics (formed / loose / hard / mucous stool), and the relationship between abdominal pain and bowel movements (relief after defecation / no relief / worsening).
[0032] (2) Basic health information: age, gender, height, weight, BMI, blood pressure, blood sugar, allergy history;
[0033] (3) Lifestyle information: Dietary structure (spicy / greasy / light / balanced), frequency of drinking (times / week), smoking status (yes / no), regularity of work and rest (regular / irregular), frequency of exercise (times / week), mental stress score (1-10 points);
[0034] (4) Past medical history information: whether there is a history of gastrointestinal diseases, whether there is a history of mental illness (anxiety, depression, etc.), whether the patient has taken intestinal-related drugs (antidiarrheal drugs, laxatives, probiotics, etc.).
[0035] S2: Standardize and preprocess multi-dimensional information from outside the hospital to obtain feature data that can be input into the algorithm model;
[0036] The preprocessing workflow includes data cleaning, outlier handling, missing value imputation, feature encoding, and feature normalization. This is used to eliminate dimensional differences and data errors in the original outpatient information, transforming it into feature data in a unified format, as detailed below:
[0037] (1) Data cleaning: Remove obviously erroneous data (such as negative abdominal pain frequency, BMI index greater than 50 or less than 15, mental stress score exceeding the range of 1-10, etc.), and correct data format errors (such as date and numerical format consistency).
[0038] (2) Outlier handling: Outliers are identified and handled using the interquartile range (IQR). For outliers that exceed the range of [Q1-1.5IQR, Q3+1.5IQR], Q1 (first quartile) or Q3 (third quartile) is used to replace them to avoid outliers affecting the model's diagnostic accuracy.
[0039] (3) Missing value imputation: For numerical features (such as abdominal pain frequency, BMI index, and mental stress score), the mean imputation method is used, and the calculation formula is as follows:
[0040]
[0041] in, Let n be the mean of the feature, and n be the effective data volume for that feature. This is the i-th valid eigenvalue;
[0042] For categorical features (such as bowel habits and smoking status), the mode imputation method is used to select the category with the highest frequency as the imputation value for missing values; for missing values of key symptom features, they are marked as "unknown" and participate in subsequent coding as independent categories;
[0043] (4) Feature encoding: One-hot encoding is used for unordered categorical features (such as bowel habits and dietary structure). For example, the bowel habits are encoded as: diarrhea = 100, constipation = 010, alternation between the two = 001. Ordinal encoding is used for ordered categorical features (such as abdominal distension and mental stress score), and the level is directly converted into the corresponding value (1-5 points, 1-10 points). 0-1 encoding is used for binary features (such as smoking status) (yes = 1, no = 0).
[0044] (5) Feature normalization: The min-max normalization method is used to normalize all numerical features to the [0,1] interval to eliminate differences in units and ensure that the influence of different features on the model is balanced. The normalization formula is as follows:
[0045]
[0046] in, These are the normalized eigenvalues. These are the original eigenvalues. This is the minimum value of the feature. This is the maximum value of this feature.
[0047] Through the above preprocessing, the non-standardized multi-dimensional information from outside the hospital is transformed into standardized feature data in a unified format, denoted as X=[x1,x2,...,x...]. n ], where n is the feature dimension (n=32 in this embodiment).
[0048] S3: Input standardized feature data into the preset IBS grading and diagnosis algorithm model, and output the patient's IBS disease grading and diagnosis results;
[0049] The training process of the IBS hierarchical diagnostic algorithm model includes:
[0050] A1: Collect multi-dimensional information from multiple groups of samples outside the hospital, as well as the IBS grading diagnostic labels corresponding to each group of samples, and construct a sample dataset;
[0051] A2: Perform standardization preprocessing on the sample dataset to obtain standardized sample feature data;
[0052] A3: Based on standardized sample feature data and corresponding IBS grading diagnostic labels, an initial algorithm model is trained, and after performance verification and optimization, a preset IBS grading diagnostic algorithm model is obtained.
[0053] The initial algorithm model is a machine learning model, selected from at least one of the following: random forest model, support vector machine model, neural network model, and logistic regression model. A single model or a combination of models can be selected according to the diagnostic accuracy requirements.
[0054] The sample dataset includes individuals of different ages, genders, regions, and symptom presentations. IBS grading diagnostic labels are determined based on the Rome IV diagnostic criteria and clinical diagnosis results, ensuring sample diversity and label accuracy; details are as follows:
[0055] A1: Constructing the Sample Dataset: 12,000 sets of sample data were collected, including 9,000 training sets and 3,000 test sets. Each set of samples contains the multi-dimensional outpatient information from S1, as well as the corresponding IBS grading diagnostic labels. The samples cover populations of different ages (18-70 years), genders, regions (north and south), and symptom presentations to ensure sample diversity. The IBS grading diagnostic labels are determined based on the Rome IV diagnostic criteria and clinical diagnosis results, and are divided into 4 levels: 0 = no IBS, 1 = mild IBS, 2 = moderate IBS, and 3 = severe IBS. The label determination criteria are as follows:
[0056] No IBS: No abdominal pain or bloating, normal bowel habits, and no related past medical history or triggering factors;
[0057] Mild IBS: Abdominal pain ≤ 2 times per week, bloating ≤ 2 points, mild bowel abnormalities, no significant impact on daily life;
[0058] Moderate IBS: Abdominal pain occurs 3-5 times per week, abdominal bloating is rated 3, bowel habits are significantly abnormal, and it has a certain impact on life;
[0059] Severe IBS: Abdominal pain ≥6 times per week, abdominal bloating ≥4 points, severely abnormal bowel habits, seriously affecting life, and may be accompanied by mental and psychological symptoms.
[0060] A2: Sample Data Preprocessing: Perform the S2 standardization preprocessing procedure on 12,000 sets of sample data to obtain standardized sample feature data and the corresponding label matrix Y=[y1,y2,...,y 12000 ](y i ∈{0,1,2,3});
[0061] A3: Model Training and Optimization: Random forest model is selected as the initial algorithm model (it can be replaced by support vector machine model or neural network model, all of which are within the scope of protection of this invention). The model parameters are set as follows: the number of decision trees is 120, the maximum depth of each decision tree is 12, the minimum number of sample splits is 6, and the minimum number of sample leaf nodes is 3.
[0062] During model training, the Gini coefficient was used to calculate the importance of each feature, and the top 20 features with the greatest impact on IBS grading diagnosis (such as frequency of abdominal pain, degree of abdominal distension, bowel habits, mental stress score, and history of gastrointestinal diseases) were selected. The formula for calculating feature importance is as follows:
[0063]
[0064] in, Number of IBS grades (in this embodiment) =4), For the first The percentage of samples of a certain class in the current decision tree node;
[0065] The model outputs the graded diagnosis results of the sample through a voting mechanism of multiple decision trees. Each decision tree outputs the probability that the sample belongs to a certain grade. The final model output is the average of the probabilities of all decision trees, and the grade with the highest probability is taken as the final diagnosis result.
[0066] A4: Model Performance Validation: Input the standardized feature data of 3000 test samples into the trained model, calculate the model's diagnostic accuracy, recall, and F1 score to validate the model's performance. In this embodiment, the model's test accuracy is 93.1%, the recall rate for mild IBS is 90.2%, the recall rate for moderate IBS is 94.3%, the recall rate for severe IBS is 95.7%, and the recall rate for no IBS is 92.5%, which meets the requirements for outpatient tiered diagnosis. If the model performance does not meet the standards, adjust the model parameters (such as the number of decision trees and the maximum depth) or supplement the sample data, and retrain until the standards are met to obtain the preset IBS tiered diagnosis algorithm model.
[0067] In S3, the output IBS disease grading diagnosis results include four levels: no IBS, mild IBS, moderate IBS, and severe IBS. It also outputs the confidence score for each level. The confidence score characterizes the reliability of the diagnostic results and is calculated using the probability output by the model. When the confidence score is lower than a preset threshold, the patient is prompted to obtain relevant information from outside the hospital for re-diagnosis, or is advised to undergo further in-hospital examination. The confidence score calculation formula is as follows:
[0068]
[0069] in, For a certain level of confidence, The number of decision trees that vote in favor of this level. The total number of decision trees (in this embodiment) =120);
[0070] The preset reliability threshold is 60%. When the confidence level is ≥80%, the diagnosis result is considered highly reliable, and a diagnostic suggestion is directly output. When the confidence level is ≤60% and <80%, the result is considered moderately reliable, and the patient is prompted to supplement relevant information from outside the hospital (such as symptom details and lifestyle habits) before a re-diagnosis is made. When the confidence level is <60%, the result is considered low reliable, and the patient is advised to go to the hospital for further examination (such as colonoscopy and fecal calprotectin testing) to avoid misdiagnosis or missed diagnosis.
[0071] S4: Model Iteration and Update: Regularly collect new multi-dimensional information about patients outside the hospital and corresponding clinical diagnosis results, construct a new sample dataset, and fine-tune the IBS hierarchical diagnosis algorithm model based on the new sample dataset to achieve iterative optimization of model performance;
[0072] Every 3 months, collect new outpatient multidimensional information and corresponding clinical diagnosis results (at least 600 sets) to construct a new sample dataset; after performing S2 standardization preprocessing on the new sample dataset, merge it with the original training sample dataset, and use incremental learning method to fine-tune the existing IBS hierarchical diagnosis algorithm model. Only the new samples are trained, without retraining the entire model, thus reducing computational costs.
[0073] After the model is updated, it is validated using the A4 performance validation standard to ensure that the diagnostic accuracy of the updated model is not less than 90%. The generalization ability and diagnostic accuracy of the model are continuously optimized to adapt to the IBS diagnostic needs of different populations and at different times.
[0074] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.
[0075] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0076] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for outpatient triage diagnosis of IBS symptoms, characterized in that, Includes the following steps: S1: Obtain multi-dimensional information related to IBS symptoms provided by patients in out-of-hospital settings; S2: Standardize and preprocess the multi-dimensional information from outside the hospital to obtain feature data that can be input into the algorithm model; S3: Input the standardized feature data into the preset IBS grading diagnosis algorithm model and output the patient's IBS disease grading diagnosis result; The IBS grading diagnosis algorithm model is trained using multiple sets of relevant outpatient sample data and is used to achieve outpatient grading diagnosis of IBS.
2. The method for outpatient triage diagnosis of IBS based on claim 1, characterized in that: The multi-dimensional information outside the hospital includes at least one of the following: symptom information, basic health information, lifestyle information, and past medical history information. This information can be filled out by the patient or obtained through a simple inquiry.
3. The method for outpatient triage diagnosis of IBS based on claim 1, characterized in that: The standardization preprocessing in S2 includes data cleaning, outlier handling, missing value imputation, feature encoding, and feature normalization, which are used to eliminate the dimensional differences and data errors of the original outpatient information and transform it into feature data in a unified format.
4. The method for outpatient triage diagnosis of IBS based on claim 1, characterized in that: The training process of the IBS hierarchical diagnostic algorithm model includes: A1: Collect multi-dimensional information from multiple groups of samples outside the hospital, as well as the IBS grading diagnostic labels corresponding to each group of samples, and construct a sample dataset; A2: Perform the standardization preprocessing described in claim 3 on the sample dataset to obtain standardized sample feature data; A3: Based on the standardized sample feature data and the corresponding IBS grading diagnostic labels, an initial algorithm model is trained, and after performance verification and optimization, the preset IBS grading diagnostic algorithm model is obtained.
5. The method for outpatient triage diagnosis of IBS based on claim 4, characterized in that: The initial algorithm model is a machine learning model, selected from at least one of the following: random forest model, support vector machine model, neural network model, and logistic regression model. A single model or a combination of models can be selected according to the diagnostic accuracy requirements.
6. The method for outpatient triage diagnosis of IBS based on claim 1, characterized in that: The IBS disease grading diagnosis results include four levels: no IBS, mild IBS, moderate IBS, and severe IBS. At the same time, the confidence level corresponding to each level is output, and the confidence level is used to characterize the reliability of the diagnosis results.
7. The method for outpatient triage diagnosis of IBS based on claim 1, characterized in that: It also includes model iteration and updates: regularly collecting new multi-dimensional information about patients outside the hospital and corresponding clinical diagnosis results, constructing a new sample dataset, and fine-tuning the IBS grading diagnosis algorithm model based on the new sample dataset to achieve iterative optimization of model performance.
8. A method for outpatient triage diagnosis of IBS based on claim 4, characterized in that: The samples in the dataset cover people of different ages, genders, regions, and symptom presentations. The IBS grading diagnostic labels are determined based on the Rome IV diagnostic criteria and clinical diagnosis results, ensuring the diversity of the samples and the accuracy of the labels.
9. A method for outpatient triage diagnosis of IBS based on claim 6, characterized in that: The confidence level is calculated using the probability output by the model. When the confidence level is lower than a preset threshold, the patient is prompted to supplement relevant information outside the hospital for re-diagnosis, or it is recommended to go to the hospital for further examination.