Clinical decision support system based on AI

Abstract

The invention relates to the technical field of decision support, in particular to an AI-based clinical decision support system, which comprises an image analysis module, a patient state analysis module, a case matching module, a case analysis module, a decision analysis module and an error warning module. According to the method, the medical image is analyzed through artificial intelligence, abnormal features in the image can be recognized and classified more accurately, the abnormal region is marked, a doctor is assisted to quickly recognize a diseased region, intervention measures are taken in advance, and in the aspect of treatment suggestions, through comparison and analysis with historical cases, the treatment accuracy is improved. The system can provide targeted treatment suggestions, effectively improve the treatment effect, reduce errors and deviations in the treatment process, timely discover the treatment deviations by monitoring the treatment execution condition in real time, ensure the correct implementation of the treatment scheme, improve the scientificity and safety of decision making, and improve the treatment efficiency. And each link of the medical activity can be effectively managed and controlled.

Description

Technical Field

[0001] The present invention relates to the technical field of decision support, and more particularly to an AI-based clinical decision support system. Background Art

[0002] The technical field of decision support focuses on the development and application of various tools and methods to assist the decision-making process and enhance its efficiency and effectiveness. The technical field of decision support involves multiple sub-fields such as data analysis, optimization algorithms, model prediction, and artificial intelligence. In specific implementations, decision support systems typically include integrated software applications that utilize databases, models, and user interfaces to help users compile useful information and extract insights from it in order to make more informed decisions. Key components of this technical field also include real-time data processing, visualization techniques, and interactive user interface design, all aimed at providing intuitive and easy-to-understand decision support tools.

[0003] Among them, a clinical decision support system is a form of decision support technology applied in the medical field, aiming to assist medical professionals and patients in making higher-quality health decisions. The system provides disease risk assessment and health management assistance recommendations by analyzing patient data and integrating medical knowledge bases. Its main uses include assisting medical staff in implementing diagnoses, optimizing treatment plan selection, preventing medical errors, and enhancing the overall effectiveness of patient treatment. The core function of the system is to support the selection of clinical pathways and help doctors make scientific, evidence-based medical decisions through data-driven insights.

[0004] When dealing with complex clinical decisions, traditional support systems are limited by data integration and processing capabilities, resulting in untimely or inaccurate decision support. For example, in the case where historical health records cannot be updated and integrated in real time, doctors cannot obtain all relevant information in a timely manner, which limits the ability to comprehensively evaluate the patient's condition. The lack of efficient real-time data analysis capabilities makes the existing systems insufficient in response speed when dealing with emergencies and unable to provide immediate medical advice or warnings. This lag in data processing and analysis leads to misunderstandings in medical decisions, such as misjudging or missing the diagnosis of the condition, thus affecting the timeliness and effectiveness of disease treatment. Summary of the Invention

[0005] The purpose of the present invention is to solve the drawbacks existing in the prior art, and to propose an AI-based clinical decision support system.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: an AI-based clinical decision support system, the system includes:

[0007] The image analysis module collects the patient's medical image data based on the medical device interface, processes the images through a convolutional neural network, annotates the abnormal areas, and sends the medical images with the annotated abnormal areas to medical staff to obtain the abnormal image identification result;

[0008] The patient status analysis module evaluates the changing trends of the patient's health status and disease status based on the patient's medical test data by analyzing the physiological indicators and symptom changes over a period of time to obtain the trend analysis result;

[0009] The case matching module retrieves the similar case features in the disease database based on the abnormal image identification result and the trend analysis result, and sends the patient symptom information and treatment measure information of the similar cases to medical staff to obtain the case reference information;

[0010] The case analysis module receives the confirmed disease type from medical staff based on the case reference information, evaluates the effectiveness of the differential treatment measures, provides the effect information of each treatment measure for medical staff to obtain the treatment measure evaluation information;

[0011] The decision analysis module analyzes the treatment measures provided by the doctor based on the treatment measure evaluation information, compares with the normal treatment measures, identifies the abnormal treatment measures, and immediately reminds the medical staff to obtain the abnormal plan warning information;

[0012] The error warning module monitors the implementation process of the treatment measures based on the abnormal plan warning information, identifies the treatment operations that deviate from the confirmed treatment measures, and immediately reminds the medical staff to obtain the treatment deviation warning information.

[0013] As a further solution of the present invention, the abnormal image identification result includes the position, size and shape of the abnormal area, the trend analysis result includes the change trend of health indicators and the symptom development speed, the case reference information includes the diagnosis results, treatment records and treatment responses of similar cases, the treatment measure evaluation information includes the effect information of each treatment measure, the recommended treatment measures and the expected treatment results, the abnormal plan warning information includes the identified risky treatment measures, drug use abnormal information and equipment misuse information, and the treatment deviation warning information includes the differences between the actual and planned treatments, potential treatment risks and emergency adjustment suggestions.

[0014] As a further solution of the present invention, the image analysis module includes:

[0015] The image acquisition sub-module collects the patient's X-ray, MRI and CT scan data based on the medical device interface, performs format conversion and quality inspection on the images, and verifies the integrity and availability of the image data to obtain the image data set;

[0016] The image processing sub-module preprocesses the data based on the image dataset, including denoising, contrast enhancement, and edge sharpening, optimizes the image quality, and performs image normalization processing to obtain the cleaned image;

[0017] The feature recognition sub-module analyzes the image based on the cleaned image through a convolutional neural network model, identifies and labels abnormal features in the image, including tumors and fracture regions, to obtain the abnormal image identification result.

[0018] As a further solution of the present invention, the patient status analysis module includes:

[0019] The symptom collection sub-module extracts the subjective symptom descriptions of the patient based on the medical test data of the patient, performs semantic analysis and key information extraction on the descriptions, and organizes them into structured data to obtain the symptom information record;

[0020] The status evaluation sub-module extracts key physiological health indicators based on the medical test data of the patient, compares the deviation between the patient's current data and the standard health data through statistical analysis methods, evaluates the patient's immediate health status, and obtains the health evaluation information;

[0021] The trend analysis sub-module evaluates the change trends of physiological indicators and symptoms based on the symptom information record and the health evaluation information through time series analysis, evaluates the health risks in the future period, and obtains the trend analysis result.

[0022] As a further solution of the present invention, the case matching module includes:

[0023] The case search sub-module receives the analysis information of the medical image by the medical staff, including the symptom type and severity, based on the abnormal image identification result and the trend analysis result, combines the health trend data and the symptom description, and extracts the medical records with similar features to obtain the candidate case list;

[0024] The feature comparison sub-module performs feature comparison based on the candidate case list, evaluates the matching degree of the candidate cases using the disease symptoms, blood indicators, and the severity of the symptoms, and performs matching degree ranking to identify the matching cases and generate the matching case information;

[0025] The information feedback sub-module organizes the symptom and treatment measure data of the matching cases based on the matching case information, summarizes the information and sends it to the medical staff to obtain the case reference information.

[0026] As a further solution of the present invention, the formula for evaluating the matching degree of the candidate cases is:

[0027]

[0028] where xi represents the i-th index value of the patient, h i represents the i-th index value of the candidate case, w i represents the weight coefficient of the i-th index, n represents the total number of indices, and M represents the matching degree score of the candidate case.

[0029] As a further solution of the present invention, the case analysis module includes:

[0030] The disease confirmation sub-module, based on the case reference information, receives the disease type confirmed by medical staff, combines medical images and symptom characteristics to verify the accuracy of the disease diagnosis. If there is any inconsistency, it immediately notifies the medical staff for confirmation to obtain the disease type confirmation information;

[0031] The data extraction sub-module, based on the disease type confirmation information, extracts the treatment records of patients with the same type of disease from the database, including drug usage records, treatment responses, and treatment outcomes, to obtain the treatment data of the same type of cases;

[0032] The effect evaluation sub-module, based on the treatment data of the same type of cases, evaluates the effectiveness of each treatment measure in the same type of cases through statistical analysis, provides the effect evaluation of each treatment measure for medical staff, and obtains the treatment measure evaluation information.

[0033] As a further solution of the present invention, the formula for evaluating the effectiveness of each treatment measure in the same type of cases is:

[0034]

[0035] where E is the effectiveness score of the treatment measure, S t represents the number of samples with successful treatment, R represents the overall treatment success rate, P represents the average satisfaction of patients, C represents the average cost of treatment, α and β are adjustment coefficients, and T is the total number of treatments.

[0036] As a further solution of the present invention, the decision analysis module includes:

[0037] The measure analysis sub-module, based on the treatment measure evaluation information, receives the treatment plan provided by medical staff, extracts the types of drugs used, dosages, and treatment equipment, compares them with the standard treatment guidelines, and identifies the deviations from the conventional treatment measures to obtain the measure deviation information;

[0038] The risk identification sub-module, based on the measure deviation information, analyzes the use of high-risk drugs and non-standard treatment equipment, evaluates the potential health risks, and obtains the risk identification information;

[0039] Based on the risk identification information, the abnormal warning sub-module sends a warning about high-risk treatment measures to medical staff, notifies the medical staff to confirm the adverse consequences of the treatment measures, and immediately reminds the medical staff to obtain abnormal plan warning information.

[0040] As a further solution of the present invention, the error warning module includes:

[0041] Based on the abnormal plan warning information, the treatment monitoring sub-module receives the treatment measures confirmed by medical staff, tracks the implementation of the confirmed treatment measures, and collects treatment operation data in real time to obtain treatment implementation monitoring data;

[0042] Based on the treatment implementation monitoring data, the deviation identification sub-module identifies the deviations in the implementation process by comparing the planned and actual treatment measures, including inconsistencies in treatment dosage, time, and method, evaluates the severity of the deviations and the potential impact on the patient's health, and obtains treatment deviation identification results;

[0043] Based on the treatment deviation identification results, the early warning notification sub-module immediately implements an early warning, sends a treatment deviation alarm to medical staff, and sends the operations with deviations to obtain treatment deviation early warning information.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0045] In the present invention, by using artificial intelligence to analyze medical images, it is possible to more accurately identify and classify abnormal features in the images, and by marking the abnormal areas, it helps doctors quickly identify the lesion sites. Through the evaluation of the patient's long-term health data, medical staff can observe the subtle changes in the disease progression, so as to take intervention measures in advance. In terms of treatment suggestions, through the comparative analysis with historical cases, targeted treatment suggestions can be provided, effectively improving the treatment effect while reducing errors and deviations in the treatment process. By real-time monitoring the treatment implementation situation, treatment deviations can be immediately discovered, ensuring the correct implementation of the treatment plan, enhancing the scientificity and safety of decision-making, and ensuring that all aspects of medical activities can be effectively managed and controlled. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0047] Figure 1 It is the system flowchart of the present invention;

[0048] Figure 2Schematic diagram of the system framework of the present invention;

[0049] Figure 3 Flowchart of the image analysis module of the present invention;

[0050] Figure 4 Flowchart of the patient status analysis module of the present invention;

[0051] Figure 5 Flowchart of the case matching module of the present invention;

[0052] Figure 6 Flowchart of the case analysis module of the present invention;

[0053] Figure 7 Flowchart of the decision analysis module of the present invention;

[0054] Figure 8 Flowchart of the error warning module of the present invention. Detailed implementation manners

[0055] The following describes the technical solutions in the present invention with reference to the accompanying drawings.

[0056] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either of the two can be selected.

[0057] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0058] In the embodiments of the present invention, sometimes subscripts such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0059] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0060] Please refer to Figure 1 , the AI-based clinical decision support system, the system includes:

[0061] The image analysis module collects the patient's medical image data based on the medical device interface, including X-rays, MRIs, and CT scans, cleans the image data, processes the images through a convolutional neural network, identifies and classifies the abnormal features in the images, labels the abnormal areas, and sends the medical images with labeled abnormal areas to medical staff to obtain the abnormal image identification result;

[0062] The patient status analysis module extracts the patient's physiological health indicators based on the patient's medical test data, including respiration, heart rate, and blood indices, and collects the patient's descriptions of symptoms. By analyzing the changes in physiological indicators and symptoms over a period of time, it evaluates the changing trends of the patient's health status and disease status to obtain the trend analysis result;

[0063] The case matching module obtains the medical staff's analysis information of the medical images based on the abnormal image identification result and the trend analysis result. According to the patient's symptom characteristics and the changing trends of the health status and disease status, it retrieves the similar case characteristics in the disease database, identifies similar cases, and sends the patient symptom information and treatment measure information of the similar cases to medical staff to obtain the case reference information;

[0064] The case analysis module receives the confirmed disease type from medical staff based on the case reference information, extracts the treatment record data of patients with the same type of disease, evaluates the effectiveness of different treatment measures by analyzing the deviations between the current patient's condition and that of patients with the same type of case and combining the effects of treatment measures, and provides the medical staff with the effect information of each treatment measure to obtain the treatment measure evaluation information;

[0065] The decision analysis module receives the treatment plan provided by medical staff based on the treatment measure evaluation information, analyzes the treatment measures provided by the doctor, compares them with normal treatment measures, identifies the abnormal treatment measures, including the use of dangerous and controlled drugs, abnormal drug use frequencies, and abnormal treatment device use, and immediately alerts the medical staff to obtain the abnormal plan warning information;

[0066] The error warning module receives the treatment measures confirmed by medical staff based on the abnormal plan warning information, monitors the implementation process of the treatment measures, identifies the treatment operations that deviate from the confirmed treatment measures, and immediately alerts the medical staff to obtain the treatment deviation warning information.

[0067] The abnormal image identification results include the location, size, and shape of the abnormal area. The trend analysis results include the change trend of health indicators and the development speed of symptoms. The case reference information includes the diagnosis results, treatment records, and treatment responses of similar cases. The treatment measure evaluation information includes the effectiveness information of each treatment measure, the recommended treatment measures, and the expected treatment results. The abnormal plan warning information includes the identified risky treatment measures, abnormal drug use information, and improper equipment use information. The treatment deviation warning information includes the differences between the actual and planned treatments, potential treatment risks, and emergency adjustment suggestions.

[0068] Please refer to Figure 2 and Figure 3 ,The image analysis module includes an image acquisition sub-module, an image processing sub-module, and a feature recognition sub-module;

[0069] Based on the medical device interface, the image acquisition sub-module collects the X-ray, MRI, and CT scan data of the patient, performs format conversion and quality inspection on the images, verifies the integrity and availability of the image data, and obtains an image data set;

[0070] Based on the medical device interface, extract the original image data from medical devices (such as X-ray machines, MRI scanners, and CT scanners), transfer the data through the dedicated communication protocol or interface (such as the DICOM protocol) of the connected device. Each frame of image data needs to perform a data packet integrity check during the transmission process. Verify whether there are packet losses or errors in the transmission through a checksum algorithm (such as CRC cyclic redundancy check). Subsequently, perform a unified conversion on the received image file format, convert the original data from different source devices into a standard format (such as PNG or TIFF format). During the format conversion process, parameter matching needs to be performed according to the pixel resolution, color depth, and file header information of each device's image to ensure that the data after format conversion retains the key information and resolution of the original image. Then, perform a quality inspection on each piece of image data. Screen out images with a lower degree of blurriness through an image sharpness measurement function (such as a gradient sharpness evaluation function), and analyze whether the contrast range and brightness distribution of the image are within the set range (such as the gray value in the range of [50, 200]) in combination with the gray histogram. If it is found that the data quality does not meet the standard, re-collect the image data through secondary scanning to obtain an image data set with verified integrity and quality.

[0071] Based on the image data set, the image processing sub-module preprocesses the data, including denoising, contrast enhancement, and edge sharpening, optimizes the image quality, and performs image normalization processing to obtain the cleaned image;

[0072] Based on the image dataset, data preprocessing is performed, including denoising, contrast enhancement, and edge sharpening. In the denoising step, the type of denoising algorithm required is determined by analyzing the noise distribution characteristics of the image data (such as Gaussian noise, salt-and-pepper noise, etc.). For image data with Gaussian noise distribution, the bilateral filtering algorithm is used for calculation. Specifically, the filtering weight matrix is obtained by calculating the spatial weight and gray value difference weight of each pixel point, and then the filtered pixel value is obtained through weighted averaging. For image data with salt-and-pepper noise distribution, the median filter is used to calculate the median of each pixel value and its neighborhood and replace the original pixel value. In the contrast enhancement stage, the histogram equalization technique is adopted. By analyzing the distribution of the image gray histogram, the gray values are redistributed, and the gray range of the low-contrast image is adjusted to the full range (such as [0, 255]) to optimize the visual quality of the image. Edge sharpening enhances the edge information of the image through the Laplacian operator convolution operation. Specifically, the kernel matrix of the Laplacian operator is applied to each pixel point, and the calculation results are superimposed on the original image to improve the clarity of the edge features. The processed image is standardized, including normalizing the size to the specified resolution (such as 1024x1024 pixels) and normalizing the gray values to the specified range (such as 0 to 1). The standardized data can adapt to the input requirements of subsequent processing steps, and finally, the cleaned image data is obtained.

[0073] Based on the cleaned image, the feature recognition sub-module analyzes the image through a convolutional neural network model, identifies and labels the abnormal features in the image, including tumor and fracture regions, and obtains the abnormal image identification result.

[0074] Based on the cleaned image, the image data is analyzed and processed through a convolutional neural network model. The cleaned image data is divided into a training set and a test set, and the division ratio is set according to the data distribution. The convolutional neural network model is trained using the training set. The model structure includes multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layer extracts local feature information through a sliding convolutional kernel, and the pooling layer performs feature dimensionality reduction and spatial information compression on the convolutional results. During the training process, the model parameters are updated through a loss function (such as cross-entropy loss) and the backpropagation algorithm. After training, the test set is used to verify the model performance, and the generalization ability of the model is judged according to evaluation metrics (such as accuracy, recall, etc.). For the input of the cleaned image into the trained convolutional neural network model, the features of each image are extracted and classified layer by layer. The model labels the detected feature regions at the output layer, including tumor regions and fracture regions. By post-processing the position coordinates of the labeling results (such as dilation operation), the shape of the labeled region is optimized into a regular polygon, and the structured record of the labeled image data and feature information is saved, and the abnormal image identification result is output.

[0075] Please refer to Figure 2 andFigure 4 , the patient status analysis module includes a symptom collection sub-module, a status assessment sub-module, and a trend analysis sub-module;

[0076] The symptom collection sub-module extracts the patient's subjective symptom descriptions based on the patient's medical test data, performs semantic analysis and key information extraction on the descriptions, and organizes them into structured data to obtain symptom information records;

[0077] Based on the patient's medical test data, the self-reported information of the patient is captured through a speech recognition interface. The information often includes subjective feelings such as pain, discomfort, and fatigue. After the data is obtained, preliminary text analysis is performed using natural language processing techniques. The process includes word segmentation, stop word removal, and part-of-speech tagging. For example, the description of the patient's "feeling very fatigued and often having headaches" is converted into keywords "fatigue" and "headache". The text mining technique is used to analyze the correlation between the keywords and the disease model. The key information in the text is extracted through a preset semantic analysis model (such as the bag-of-words model or the TF-IDF model), and the matching degree between the symptoms and the known diseases is evaluated. The analysis results are converted into structured data, and labels and corresponding weight values are established for each symptom and stored in the database to generate symptom information records. Each record details the symptom type, description intensity, and correlation score, providing data support for subsequent clinical diagnosis and treatment decisions.

[0078] The status assessment sub-module extracts key physiological health indicators based on the patient's medical test data, and through statistical analysis methods, compares the patient's current data with the standard health data to evaluate the patient's immediate health status and obtain health assessment information;

[0079] Based on the patient's medical test data, key physiological indicators such as heart rate, blood pressure, and blood oxygen saturation are extracted from the patient's test data. The data is usually directly digitized by medical test equipment and stored in the medical information database. These physiological indicators are retrieved through a data interface, and statistical analysis software is used to perform statistical descriptions on the patient's current physiological indicator data, including the mean, standard deviation, and data distribution. The actual measured values are compared with the standard health data stored in the medical database, and the deviation degree of each indicator is calculated using difference analysis methods (such as the Z-score or T-test). The evaluation criterion is to judge the health status through a set health threshold. For example, the normal range of heart rate is 60-100 beats per minute. If the patient's heart rate continuously exceeds 100 beats per minute, it is considered that there is a health risk. The evaluation results are summarized and converted into health assessment information. The record includes the measured value, standard value, deviation value of each indicator, and its health impact score, providing a decision-making basis for doctors.

[0080] The trend analysis sub-module evaluates the change trends of physiological indicators and symptoms through time series analysis based on the symptom information records and health assessment information, and evaluates the health risks in future periods to obtain trend analysis results.

[0081] Based on symptom information records and health assessment information, input the data into a time series analysis model. For example, use an ARIMA model or exponential smoothing method to predict the future trends of symptoms and physiological indicators. By determining the parameters of the model, such as seasonal cycles, trends, and random fluctuation factors, calculate the model parameters, such as autoregressive coefficients and moving average coefficients, using historical data. Apply the parameters to the prediction model to predict the patient's health status over a certain period in the future. The output of the evaluation model is a series of prediction points, indicating future health risks or stability. For example, if the model predicts that the patient's heart rate trend will gradually increase within the next three months, it may indicate an increased risk of cardiovascular disease. Organize the trend analysis results into a data form, describing the future trends of each indicator, health risk assessment, and its potential clinical significance, providing more comprehensive suggestions for the medical team on condition monitoring and preventive measures.

[0082] Please refer to Figure 2 and Figure 5 , the case matching module includes a case search sub-module, a feature comparison sub-module, and an information feedback sub-module;

[0083] Based on the abnormal image identification results and trend analysis results, the case search sub-module receives the analysis information of medical images from medical staff, including symptom types and severity. Combining health trend data and symptom descriptions, extract medical records with similar features to obtain a list of candidate cases;

[0084] Based on the abnormal image identification results and trend analysis results, receive the analysis information of medical images from medical staff, including symptom types such as tumors or fractures, and severity such as mild, moderate, or severe. Subsequently, combine health trend data, such as the change trends of time series data of body temperature, blood pressure, etc., perform semantic analysis on the symptom descriptions, use natural language processing technology to extract keywords and match cases with similar symptom descriptions, and after synthesizing the information, evaluate the matching degree of the candidate case list through a sorting algorithm, and screen out candidate cases that match the current patient's condition to obtain a list of cases.

[0085] Based on the list of candidate cases, the feature comparison sub-module conducts feature comparisons, uses disease symptoms, blood indicators, and the severity of symptoms to evaluate the matching degree of the candidate cases, and performs matching degree sorting to identify matching cases and generate matching case information;

[0086] The formula for evaluating the matching degree of candidate cases is:

[0087]

[0088] where, x i represents the i-th indicator value of the patient, h i represents the i-th indicator value of the candidate case, wi represents the weight coefficient of the i-th indicator, n represents the total number of indicators, and M represents the matching degree score of candidate cases.

[0089] Formula:

[0090]

[0091] Detailed Explanation of Parameters and Acquisition Methods:

[0092] x i : represents the i-th indicator value of the patient. For example, blood indicators. These data are usually directly obtained from the patient's medical records. In clinical practice, blood indicators include but are not limited to white blood cell count, erythrocyte sedimentation rate, etc., and these data are obtained through blood tests.

[0093] h i : the i-th indicator value of the candidate case, obtained by collecting the corresponding indicators of cases with the same symptoms.

[0094] w i : is the weight coefficient for each indicator. The coefficient is set according to the importance of the indicator in disease diagnosis. The determination of the weight coefficient usually requires domain experts to evaluate based on experience or previous clinical data. For example, if the disease is mainly manifested by immune response, the weight of white blood cell count may be set relatively high.

[0095] n: represents the number of indicators involved in the calculation. This is a fixed value, determined according to the actual number of indicator items participating in the matching.

[0096] Calculation Example:

[0097] Set the following parameters: only consider the white blood cell count indicator, x 1 = 8000 / mm3, h 1 = 10000 / mm3. Assume that the importance evaluation of white blood cell count in the diagnosis of this disease is 0.5, and the number of indicators n = 1.

[0098] Calculate M:

[0099]

[0100] The calculation result M≈0.316 indicates that according to the current weights and data set, the matching degree of this case is relatively low.

[0101] Based on the information of the matching cases, the information feedback sub-module sorts out the symptom and treatment measure data of the matching cases, summarizes the information and sends it to medical staff to obtain case reference information.

[0102] Based on the matching case types, organize the symptom and treatment measure data of the matching cases. Extract the detailed information of relevant cases through database queries, including symptom types, onset times, treatment courses, and their effects, etc. Use data summarization tools to classify and summarize the information, and send the information to medical staff for their reference and decision-making, so as to provide the most accurate and practical case reference information to medical staff, facilitating them to make more effective diagnostic and treatment decisions according to the specific condition of the current patient, and generating case reference information.

[0103] Please refer to Figure 2 and Figure 6 , the case analysis module includes a disease confirmation sub-module, a data extraction sub-module, and an effect evaluation sub-module;

[0104] Based on the case reference information, the disease confirmation sub-module receives the disease types confirmed by medical staff, combines medical images and symptom characteristics to verify the accuracy of the disease diagnosis. If there are inconsistencies, immediately notify the medical staff for confirmation to obtain the disease type confirmation information.

[0105] Based on the case reference information, receive the confirmed information from medical staff, and extract abnormal markers from medical images, such as the size, shape, and density characteristics of tumor or inflammation areas. Use image processing algorithms such as edge detection and region growing algorithms to quantify these characteristics. Then, combine symptom characteristics, such as pain level and development speed. The data is transformed into actionable numerical indicators through natural language processing technology. Use pattern recognition and machine learning models, such as decision trees or neural networks, to verify the disease type to ensure the accuracy of the diagnosis. If the model output is inconsistent with the preliminary diagnosis provided by medical staff, a prompt will be issued through an automated system, requiring medical staff to re-evaluate to ensure the accuracy and timeliness of the diagnosis, and ensure the reliability and scientific nature of the processing results.

[0106] Based on the disease type confirmation information, the data extraction sub-module extracts the treatment records of patients with the same type of disease from the database, including drug usage records, treatment responses, and treatment outcomes, to obtain the treatment data of similar cases.

[0107] Based on the disease type confirmation information, extract the relevant treatment records of specific disease types from the database. The process involves data retrieval technology and big data analysis. Use SQL query statements to retrieve all historical cases in the medical database that are the same as the diagnosed disease, including information such as patients' drug usage records, treatment responses, and treatment outcomes. Use data cleaning technology to remove incomplete or incorrect records. Through data mining technologies such as clustering analysis and association rule analysis, identify common treatment paths and drug reaction patterns from a large number of cases to ensure that the treatment data extracted from the database has a high degree of accuracy and reference value, providing a scientific basis for subsequent treatment.

[0108] The effect evaluation sub-module evaluates the effectiveness of each treatment measure in similar cases through statistical analysis based on the treatment data of similar cases, provides the effect evaluation of each treatment measure for medical staff, and obtains the treatment measure evaluation information.

[0109] The formula for evaluating the effectiveness of each treatment measure in similar cases is:

[0110]

[0111] Among them, E is the effectiveness score of the treatment measure, S t represents the number of samples with successful treatment, R represents the overall treatment success rate, P represents the average satisfaction of patients, C represents the average cost of treatment, α and β are adjustment coefficients, and T is the total number of treatments.

[0112] Detailed explanation of parameters and acquisition methods:

[0113] S t : The number of samples with successful treatment, representing the number of treatment events recorded as successful within the target time period. It is usually obtained by reviewing the medical records of patients, especially the cases marked as having good treatment effects in the hospital database.

[0114] R: Represents the overall treatment success rate, calculated as the number of successful treatments divided by the total number of treatments. This can be obtained from the data in the medical record system. For example, by calculating the ratio of successful treatment cases to total cases within a certain period of time.

[0115] P: Represents the average satisfaction of patients, and the data is obtained through patient satisfaction surveys. The survey can be conducted after treatment, usually using a scale scoring system such as 1 to 5.

[0116] C: Represents the average cost of treatment, obtained from the hospital's financial system or cost accounting department. The cost includes all direct and indirect expenses related to the treatment.

[0117] α and β: These are the weight coefficients of satisfaction and cost, determined by clinical research and cost-benefit analysis to reflect the relative importance of different indicators in the total score.

[0118] T: The total number of treatments, representing the number of treatment implementations included in the analysis, which can be directly obtained from the hospital database.

[0119] Calculation example:

[0120] Set the parameters as follows: The number of successful treatments S t = 200, the total number of treatments T = 1000, the success rate The average patient satisfaction P = 4.5 (out of 5), the average treatment cost C = 500 yuan, the weight coefficients α = 0.3, β = 0.2.

[0121] Logarithmic calculation: log(501) ≈ 2.7;

[0122] Apply the β weight: 0.2 + 0.3×4.5 - 0.2×2.7;

[0123] Calculate the result of the intermediate step: 0.2 + 1.35 - 0.54 = 1.01;

[0124] Calculate the final score:

[0125] The calculated result E = 40.4, representing the average cost of the treatment measure. The higher the score, the more effective the measure.

[0126] Please refer to Figure 2 and Figure 7 , the decision - making analysis module includes a measure analysis sub - module, a risk identification sub - module, and an anomaly warning sub - module;

[0127] Based on the treatment measure evaluation information, the measure analysis sub - module receives the treatment plan provided by medical staff, extracts the types of drugs used, dosage, and treatment equipment, compares them with the standard treatment guidelines, identifies the deviations from the conventional treatment measures, and obtains the measure deviation information;

[0128] Based on the treatment measure evaluation information, extract key information from the treatment plan received from medical staff, including the types of drugs used, dosage, and treatment equipment. Confirm the standard usage guidelines for each drug and equipment through database queries, and use data comparison algorithms to detect the deviations between the actual treatment plan and the standard treatment guidelines. For example, by calculating the percentage difference between the drug dosage used and the recommended dosage to evaluate the degree of deviation, and at the same time evaluate the usage of treatment equipment to check whether non - standard or outdated equipment is used, and identify the deviations from the conventional treatment measures, such as excessive drug dosage or equipment not meeting the current medical standards. The deviation information is then compiled into a report for medical staff to review and adjust to ensure the rationality and safety of the treatment plan.

[0129] Based on the measure deviation information, the risk identification sub - module analyzes the usage of high - risk drugs and non - standard treatment equipment, evaluates the potential health risks, and obtains the risk identification information;

[0130] Based on the measure deviation information, evaluate the potential risks in treatment, especially analyze the use of high-risk drugs and non-standard treatment equipment. Through medical databases and drug side effect data, collect and analyze the historical safety data of relevant drugs and equipment, such as the adverse reaction records of drugs and the failure rates of equipment. Use statistical models such as logistic regression to analyze the correlation between drug and equipment use and adverse events, and evaluate potential health risks, such as allergic reactions caused by drugs or complications caused by equipment operation errors. The risk identification results will be used to remind medical staff of possible safety issues during the treatment process to ensure patient safety.

[0131] Based on the risk identification information, the abnormal warning sub-module sends warnings about high-risk treatment measures to medical staff, notifies medical staff to confirm the adverse consequences of treatment measures, and immediately reminds medical staff to obtain abnormal plan warning information.

[0132] Based on the risk identification information, use an automated communication system, such as email and SMS services, to send warning messages to relevant medical staff in a timely manner. The information content includes warnings about the use of high-risk drugs, tips on the use of non-standard equipment, and their possible adverse consequences, so that medical staff can respond quickly, adjust the treatment plan or take necessary preventive measures, and update and re-warn based on the feedback from medical staff to ensure that all relevant personnel can immediately understand and handle potential treatment risks, ensuring the timeliness and accuracy of warning information, thereby effectively improving the efficiency and effectiveness of treatment safety management.

[0133] Please refer to Figure 2 and Figure 8 , the error warning module includes a treatment monitoring sub-module, a deviation identification sub-module, and a warning notification sub-module;

[0134] Based on the abnormal plan warning information, the treatment monitoring sub-module receives the treatment measures confirmed by medical staff, tracks the implementation of the confirmed treatment measures, and collects treatment operation data in real time to obtain treatment implementation monitoring data;

[0135] Based on the abnormal plan warning information, monitor the treatment measures confirmed by medical staff in real time. By receiving real-time data from medical equipment, including the infusion rate and set time of the drug pump, and the operation status of other relevant treatment equipment, use data collection software to ensure that all operation data is accurately recorded and stored. At the same time, through time series analysis of the implemented treatment measures, compare the implementation time, duration, and drug dose of each treatment operation to ensure consistency with the original plan of medical staff. Any deviation from the plan will be immediately recorded and generate implementation monitoring data, helping the medical team to promptly understand possible execution deviations or equipment failures during the treatment process, ensuring the accuracy of treatment and patient safety.

[0136] Based on the treatment implementation monitoring data, the deviation identification sub-module identifies the deviations in the implementation process by comparing the planned and actual treatment measures, including inconsistencies in treatment dosage, time, and method, evaluates the severity of the deviations and their potential impact on the patient's health, and obtains the treatment deviation identification result;

[0137] Based on the treatment implementation monitoring data, using data comparison techniques such as anomaly detection algorithms in machine learning, automatically compare the actual treatment data with the treatment plan, including the accuracy of drug dosage, the time accuracy of treatment application, and the consistency of methods. If the dosage or time deviation exceeds the preset critical value, the algorithm will mark it as a potential risk event. At the same time, evaluate the possible impact of the deviation on the patient's health, and use a statistical model to calculate the severity of the deviation for the clinical team to evaluate and decide whether to adjust the treatment plan or conduct subsequent diagnoses.

[0138] Based on the treatment deviation identification result, the early warning notification sub-module immediately implements an early warning, sends a treatment deviation alert to medical staff, and sends the operations with deviations to obtain the treatment deviation early warning information.

[0139] Based on the treatment deviation identification result, through the hospital's internal communication network, such as email and instant messaging, directly send the treatment deviation alert to the relevant medical staff. The notification content includes the specific deviation details, potential health risks, and urgency, so that the medical staff can respond quickly, take necessary measures to adjust the treatment plan or intervene immediately, and help the medical staff accurately understand and handle the abnormalities in the treatment process, thereby ensuring the continuity and safety of the patient's treatment.

[0140] It should be understood that the term "and / or" in this article is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context.

[0141] In the present invention, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0142] It should be understood that in various embodiments of the present invention, the sequence numbers of the above processes do not indicate the order of execution, and the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0143] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0144] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0145] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be in an electrical, mechanical, or other form.

[0146] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0147] In addition, the functional units in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0148] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0149] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. AI-based clinical decision support system, characterized by: The system comprises: The image analysis module collects the patient's medical image data based on the medical equipment interface, processes the image through a convolutional neural network, marks the abnormal area, and sends the medical image with the abnormal area marked to the medical staff to obtain the abnormal image identification result; The patient status analysis module is based on the patient's medical test data. By analyzing the changes in physiological indicators and symptoms over a period of time, it evaluates the changing trends of the patient's health status and disease status and obtains trend analysis results; The case matching module retrieves similar case features from the disease database based on the abnormal image identification results and trend analysis results, and sends patient symptom information and treatment measure information of similar cases to medical personnel to obtain case reference information; The case analysis module receives the disease type confirmed by the medical staff based on the case reference information, evaluates the effectiveness of the differentiated treatment measures, provides the medical staff with the effect information of each treatment measure, and obtains the treatment measure evaluation information; The decision analysis module analyzes the treatment measures provided by the doctor based on the treatment measure evaluation information, compares the normal treatment measures, identifies abnormal treatment measures, and immediately reminds the medical staff to obtain abnormal plan warning information; The error warning module monitors the implementation process of the treatment measures based on the abnormal plan warning information, identifies and confirms the treatment operations that deviate from the treatment measures, and immediately reminds the medical staff to obtain treatment deviation warning information.

2. The AI-based clinical decision support system according to claim 1, characterized in that: The abnormal image identification results include the location, size and shape of the abnormal area, the trend analysis results include the changing trend of health indicators and the speed of symptom development, the case reference information includes the diagnosis results, treatment records and treatment responses of similar cases, the treatment measure evaluation information includes the effect information of each treatment measure, recommended treatment measures and expected treatment results, the abnormal plan warning information includes identified risky treatment measures, abnormal drug use information and improper equipment use information, and the treatment deviation warning information includes the difference between actual and planned treatment, potential treatment risks and emergency adjustment suggestions.

3. The AI-based clinical decision support system according to claim 1, characterized in that: The image analysis module comprises: The image acquisition submodule collects the patient's X-ray, MRI and CT scan data based on the medical device interface, performs format conversion and quality inspection on the images, verifies the integrity and availability of the image data, and obtains the image data set; The image processing submodule pre-processes the data based on the image data set, including denoising, contrast enhancement and edge sharpening, optimizes the image quality, and performs image standardization to obtain a cleaned image; The feature recognition submodule analyzes the image based on the cleaned image through a convolutional neural network model, identifies and marks abnormal features in the image, including tumors and fracture areas, and obtains abnormal image identification results.

4. The AI-based clinical decision support system according to claim 1, characterized in that: The patient status analysis module comprises: The symptom collection submodule extracts the patient's subjective symptom description based on the patient's medical test data, performs semantic analysis and key information extraction on the description, and organizes it into structured data to obtain symptom information records; The status assessment submodule extracts key physiological health indicators based on the patient's medical test data, compares the deviation between the patient's current data and the standard health data through statistical analysis methods, assesses the patient's immediate health status, and obtains health assessment information; The trend analysis submodule evaluates the changing trends of physiological indicators and symptoms and the health risks in the future period through time series analysis based on the symptom information records and health assessment information to obtain trend analysis results.

5. The AI-based clinical decision support system according to claim 1, characterized in that: The case matching module includes: The case search submodule receives the analysis information of the medical image from the medical staff based on the abnormal image identification results and trend analysis results, including the symptom type and severity, and extracts medical records with similar characteristics in combination with the health trend data and symptom description to obtain a list of candidate cases; The feature comparison submodule performs feature comparison based on the candidate case list, evaluates the matching degree of the candidate cases by using disease symptoms, blood indicators and the severity of the symptoms, sorts the matching degrees, identifies matching cases, and generates matching case information; The information feedback submodule organizes the symptoms and treatment measures data of the matching cases based on the matching case information, summarizes the information and sends it to medical personnel to obtain case reference information.

6. The AI-based clinical decision support system according to claim 5, characterized in that: The formula for evaluating the matching degree of candidate cases is: Among them, x i represents the patient's ith index value, h i represents the i-th index value of the candidate case, w i represents the weight coefficient of the i-th indicator, n represents the total number of indicators, and M represents the matching score of the candidate case.

7. The AI-based clinical decision support system according to claim 1, characterized in that: The case analysis module includes: The disease confirmation submodule receives the disease type confirmed by the medical staff based on the case reference information, verifies the accuracy of the disease diagnosis by combining the medical images and symptom characteristics, and immediately notifies the medical staff for confirmation if there is any discrepancy, and obtains the disease type confirmation information; The data extraction submodule extracts the treatment records of patients with similar diseases from the database based on the disease type confirmation information, including drug use records, treatment responses and treatment results, to obtain treatment data of similar cases; The effect evaluation submodule evaluates the effectiveness of each treatment measure in similar cases based on the treatment data of similar cases through statistical analysis, provides medical personnel with an effect evaluation of each treatment measure, and obtains treatment measure evaluation information.

8. The AI-based clinical decision support system according to claim 7, characterized in that: The formula for evaluating the effectiveness of each treatment measure in similar cases is: Among them, E is the effectiveness score of the treatment measure, S t represents the number of samples with successful treatment, R represents the overall treatment success rate, P represents the average patient satisfaction, C represents the average cost of treatment, α and β are adjustment coefficients, and T is the total number of treatments.

9. The AI-based clinical decision support system according to claim 1, characterized in that: The decision analysis module includes: The measure analysis submodule receives the treatment plan provided by the medical personnel based on the treatment measure evaluation information, extracts the types of drugs used, dosages of drugs used, and treatment equipment, compares them with the standard treatment guidelines, identifies deviations from conventional treatment measures, and obtains measure deviation information; The risk identification submodule analyzes the use of high-risk drugs and non-standard treatment equipment based on the measure deviation information, evaluates potential health risks, and obtains risk identification information; Based on the risk identification information, the abnormal warning submodule sends warnings of high-risk treatment measures to medical personnel, notifies medical personnel to confirm the adverse consequences of the treatment measures, and immediately reminds medical personnel to obtain abnormal plan warning information.

10. The AI-based clinical decision support system according to claim 1, characterized in that: The error warning module comprises: The treatment monitoring submodule receives the treatment measures confirmed by the medical staff based on the abnormal plan warning information, tracks the execution of the confirmed treatment measures, collects treatment operation data in real time, and obtains treatment implementation monitoring data; The deviation identification submodule is based on the treatment implementation monitoring data, and by comparing the planned treatment measures with the actually implemented treatment measures, identifies the deviations in the implementation process, including the inconsistency of treatment dosage, time and method, evaluates the severity of the deviation and the potential impact on the patient's health, and obtains the treatment deviation identification result; The early warning notification submodule implements early warning based on the treatment deviation identification result, sends a treatment deviation alarm to medical personnel, and sends the operation with deviation to obtain treatment deviation early warning information.

Images