Intelligent tuberculosis screening and data analysis system

Abstract

The invention discloses an intelligent tuberculosis screening and data analysis system, relates to the technical field of medical data analysis systems, and aims to improve the diagnosis efficiency and accuracy of tuberculosis. The technical key points are that the system comprises a data acquisition module, a preprocessing module, an AI diagnosis module, a multi-modal data fusion module, a data analysis module, an intelligent follow-up visit module and a database management module. By collecting image data and clinical information of a patient, the system adopts a deep learning algorithm to automatically analyze an image, identifies a potential tuberculosis focus, and provides a personalized diagnosis result in combination with a multi-modal data fusion technology. Meanwhile, the system has intelligent follow-up visit and remote monitoring functions, physiological indexes of the patient are obtained in real time through the wearable device, and dynamic illness state tracking and early warning are carried out. According to the method, the accuracy and efficiency of tuberculosis screening can be remarkably improved, precise medical services are provided for patients, and epidemiological prediction and prevention and control in the public health field are supported.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data analysis systems, and in particular to an intelligent tuberculosis screening and data analysis system. Background Art

[0002] Tuberculosis (TB) is a chronic infectious disease caused by Mycobacterium tuberculosis and is one of the major public health issues worldwide. According to statistics from the World Health Organization (WHO), there are millions of new cases of TB worldwide each year, especially in economically underdeveloped areas, where the incidence and mortality of TB are more serious. Although medical technology has made significant progress in recent years, the detection and diagnosis of TB still face many challenges.

[0003] Traditional TB screening methods mainly rely on imaging examinations (such as X-rays, CT scans) and microbiological examinations (such as sputum culture). These methods have the following disadvantages:

[0004] High diagnostic dependency: Traditional imaging diagnosis requires experienced professional doctors to read and judge the images, which is easily affected by personal experience and subjective factors, resulting in high rates of missed diagnosis and misdiagnosis.

[0005] Long detection time: The detection cycle of microbiological examinations (such as sputum culture) is long, usually taking several weeks, which is difficult to meet the clinical needs of rapid diagnosis.

[0006] Resource limitations: In resource-limited areas, the lack of sufficient professional equipment and personnel makes large-scale screening and early diagnosis difficult.

[0007] In recent years, artificial intelligence (AI) technology, especially deep learning algorithms (such as convolutional neural networks (CNNs), has made significant progress in the field of medical image analysis. The introduction of AI technology has made automated and precise image diagnosis possible. Through training on a large amount of medical image data, AI systems can identify and annotate subtle lesion features, reduce the workload of manual film reading, and improve the accuracy of diagnosis.

[0008] In the field of tuberculosis screening, AI technology has shown great potential, including the following aspects:

[0009] Efficient data processing: AI can efficiently analyze and process large amounts of medical imaging data, significantly shortening the time for screening and diagnosis.

[0010] Automated lesion detection: The deep learning-based AI model can automatically identify tuberculosis features in lung images, such as nodules, fibrosis, cavitation, etc., reducing errors in manual judgment.

[0011] Reduced reliance on human resources: AI systems can provide auxiliary diagnosis when physician resources are scarce, improving access to medical services, especially in remote areas and developing countries.

[0012] In the medical field, the application of big data technology has gradually become mainstream. By mining and analyzing massive medical data, more accurate disease diagnosis and prediction models can be obtained. Especially in tuberculosis screening, the multimodal fusion of imaging data and clinical data can provide a more comprehensive disease analysis. The advantages of multimodal data fusion are reflected in the following aspects:

[0013] Multi-dimensional analysis: By integrating the patient's imaging data, blood test data, genetic information, etc., we can conduct in-depth analysis of tuberculosis from multiple dimensions to improve the accuracy of diagnosis.

[0014] Personalized diagnosis: Multimodal data fusion technology can take into account individual differences among patients and provide more personalized diagnostic results and treatment recommendations, which is in line with the development direction of modern precision medicine.

[0015] Disease prediction and prevention: Analysis based on multimodal big data can help public health agencies understand the spread of diseases and provide data support for disease prevention and control.

[0016] Traditional tuberculosis diagnosis and treatment methods usually lack long-term tracking and follow-up of patients, especially after the initial diagnosis and treatment. Many patients relapse or have poor treatment effects during treatment due to lack of adequate follow-up and monitoring. The introduction of intelligent follow-up systems and remote monitoring technology can make up for this shortcoming:

[0017] Real-time monitoring: Smart wearable devices can obtain patients’ physiological data in real time, such as body temperature, heart rate, respiratory rate, etc., to provide doctors with real-time health monitoring information.

[0018] Long-term tracking: Regularly obtaining symptom feedback and lifestyle information from patients through a follow-up system to monitor changes in the condition will help improve the continuity and effectiveness of treatment.

[0019] Early warning: Through AI and big data analysis, the intelligent follow-up system can predict the patient's condition and provide early warning, helping doctors intervene in time.

[0020] With the large-scale digitization and networking of medical data, data security and privacy protection issues have become increasingly prominent. Especially when using AI and big data technologies for analysis, how to ensure the security and privacy of patient data has become a key challenge. Existing technologies mainly face the following problems in data security:

[0021] Data leakage risk: Medical data is highly sensitive and may have serious impact on patients once leaked, so the data transmission and storage process must be encrypted and protected.

[0022] Data sharing problem: When sharing data between different medical institutions, it is necessary to ensure the security and privacy of the data, which requires the introduction of encryption and access control technology in the sharing mechanism.

[0023] Privacy protection: Traditional privacy protection methods often reduce risks through data de-identification or data masking, but these methods may have limited effectiveness when faced with complex data analysis needs. Therefore, new privacy protection technologies (such as differential privacy and federated learning) are being widely used.

[0024] In summary, the intelligent tuberculosis screening and data analysis system based on artificial intelligence, deep learning, big data analysis and multimodal fusion can significantly improve the efficiency and accuracy of tuberculosis diagnosis, reduce misdiagnosis and missed diagnosis, and improve the long-term management of patients. At the same time, it has important application value in public health prevention and control. However, how to ensure data security and privacy while improving diagnostic performance is an issue that needs to be considered in system design. This patent proposes an intelligent tuberculosis screening and data analysis system, which aims to achieve efficient, accurate and safe tuberculosis diagnosis and management through the comprehensive use of the above technologies.

[0025] To this end, we continue to design an intelligent tuberculosis screening and data analysis system to solve the above problems. Summary of the invention

[0026] The purpose of the present invention is to provide an intelligent tuberculosis screening and data analysis system to address the deficiencies of the prior art and to solve the problems raised in the background technology.

[0027] To achieve the above object, the present invention provides the following technical solution: an intelligent tuberculosis screening and data analysis system, comprising the following modules:

[0028] Data acquisition module: used to collect multimodal imaging data and basic information of patients, including but not limited to X-rays, CT scan imaging data, and synchronize the data from multiple sources;

[0029] Preprocessing module: standardize, normalize, denoise, enhance, segment and convert the collected image data to ensure the consistency of image data input and the accuracy of analysis.

[0030] AI diagnosis module: Analyzes preprocessed image data through the convolutional neural network (CNN) model and deep learning algorithm to identify tuberculosis-related image features and locate the lesion area.

[0031] Data analysis module: Based on the output results of the AI ​​diagnosis module, statistical analysis is performed on the screening data, providing functions such as grouping, classification, clustering, regression analysis and prediction.

[0032] Report generation module: Automatically generate standardized screening reports based on diagnosis and data analysis results, and display the analysis results in a variety of data visualization forms.

[0033] Database management module: used to store patient information, imaging data, analysis results and diagnostic reports, and supports data retrieval, management and multi-level access permission control.

[0034] As a preferred technical solution of the present invention, the Al diagnosis module adopts an improved deep convolutional neural network, whose architecture includes multi-layer convolutional layers, pooling layers, fully connected layers and Softmax classifiers. The feature extraction of the convolution process is implemented by the following formula:

[0035]

[0036] Where: x is the input image data; W j is the convolution kernel weight matrix of the jth layer; * indicates the convolution operation; b i is the bias; σ is the ReLU activation function.

[0037] As a preferred technical solution of the present invention, the AI ​​diagnosis module introduces a multi-level attention mechanism to improve the accuracy of the model in lesion detection. The attention mechanism uses a self-attention algorithm to represent the weighted average of key area features, and the formula is as follows:

[0038]

[0039] Where: Q is the query matrix; K is the key matrix; V is the value matrix; d k is the dimension of the key matrix.

[0040] As a preferred technical solution of the present invention, the image segmentation function in the preprocessing module uses a fully convolutional neural network (FCN) based on the U-Net architecture, and its loss function uses an improved Dice coefficient, and the formula is as follows:

[0041]

[0042] Where: p i is the pixel value predicted by the model; g i is the actual marked pixel value; N is the total number of pixels.

[0043] As a preferred technical solution of the present invention, the data analysis module uses a Bayesian classifier to classify the image feature data and uses multivariate regression analysis to assess the risk of disease. The multivariate regression analysis formula is as follows:

[0044]

[0045] Among them: y is the probability of the patient suffering from the disease; x i is the patient's characteristic variable; β i is the regression coefficient; ∈ is the error term.

[0046] As the preferred technical solution of the present invention, the data acquisition module has the function of real-time acquisition and transmission of image data, supports multiple medical image formats such as DICOM, JPEG, PNG, and introduces a data redundancy check mechanism during the acquisition process to ensure the integrity and accuracy of data transmission.

[0047] As a preferred technical solution of the present invention, the Al diagnosis module includes an image enhancement submodule, which performs contrast enhancement on the image by an algorithm based on contrast-limited adaptive histogram equalization (CLAHE), and the formula is as follows:

[0048] I output =CLAHE(I input )

[0049] Where: I input is the original input image; I output is the enhanced image; CLAHE is the contrast limited adaptive histogram equalization operation.

[0050] As a preferred technical solution of the present invention, when analyzing the image data, the Al diagnosis module adjusts and trains the multi-dimensional weights of the disease characteristics through a supervised learning method, and the training loss function used is the cross entropy loss, and the formula is as follows:

[0051]

[0052] Where: L CE is the cross entropy loss; y i is the actual label; p i is the predicted probability value.

[0053] As a preferred technical solution of the present invention, the report generation module can generate a complete report including screening results, image annotations, statistical analysis data and predictive analysis, supports image embedding and high-resolution chart generation, and supports automatic translation of multi-language reports.

[0054] As a preferred technical solution of the present invention, the database management module uses the Advanced Encryption Standard (AES) and hash algorithm (SHA-256) for data encryption and secure storage. The encryption algorithm formula is as follows:

[0055]

[0056] H(M)=SHA-256(M)

[0057] Where: C is the encrypted ciphertext; E k is the encryption function; M is the plaintext information; K is the encryption key; H(M) is the hash value obtained by SHA-256 hashing.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] This technical solution significantly improves the efficiency and accuracy of tuberculosis screening by introducing deep learning algorithms and artificial intelligence technologies. Automated image analysis can quickly identify lung lesions, reduce dependence on professional doctors, and reduce misdiagnosis and missed diagnosis rates. Multimodal data fusion technology combines imaging and clinical data to provide each patient with a comprehensive personalized diagnosis, improve the accuracy of diagnosis, and contribute to the development of precision medicine.

[0060] The system integrates intelligent follow-up and remote monitoring modules, which can monitor the patient's health status in real time and support long-term disease tracking. Through regular health follow-up and physiological data collection, the system can predict disease changes at an early stage, provide timely warnings, help doctors intervene effectively, improve patient treatment outcomes, and reduce the risk of disease recurrence.

[0061] To address the risk of medical data privacy leakage, this technical solution adopts a multi-level data security strategy, including data encryption, differential privacy and federated learning technologies, to ensure the security of patient privacy during data transmission, storage and analysis. At the same time, cross-institutional data sharing is achieved through the federated learning framework, avoiding the privacy risks brought by centralized data storage and providing secure technical support for collaboration between medical institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a system block diagram of an intelligent tuberculosis screening and data analysis system proposed in the present invention. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] The following is combined with Figure 1 With reference to the accompanying drawings and a number of embodiments, specific embodiments of the present invention are described in detail.

[0065] Example 1: Data collection and standardization preprocessing

[0066] Data acquisition: The system deploys multiple image acquisition devices in the hospital environment, including X-ray and CT imaging devices. Each device is connected to the system through the Internet of Medical Things (IoMT) interface. After acquisition, the image data is transmitted to the system's central server in real time. The data acquisition module uses the standard DICOM protocol to ensure the consistency of the image data format.

[0067] The data acquisition module supports both JPEG and PNG formats to meet the acquisition requirements of different medical devices. The system also integrates the hospital information management system (HIS) to automatically obtain the patient's basic information (such as age, gender, medical history, etc.) and physical data (such as body temperature, blood pressure, etc.).

[0068] The system identifies the data of each patient and uses a unique patient ID (such as UUID) to track and manage the data throughout the entire process, ensuring the accuracy and completeness of the data.

[0069] Data preprocessing: After data acquisition, the system automatically preprocesses the image data. First, a denoising algorithm based on Gaussian filtering is used to smooth the random noise in the image data. The filtering formula is as follows:

[0070]

[0071] Where: G(x,y) is the pixel value after filtering; σ is the standard deviation, which controls the degree of smoothing; x,y are the pixel coordinates.

[0072] CLAHE (Contrast Limited Adaptive Histogram Equalization) is used to enhance the contrast of the image and highlight the detailed features of the lungs. On the preprocessed image data, the system uses the U-Net network structure to automatically segment the lung ROI (region of interest). The goal of segmentation is to obtain the precise outline of the lungs as input for subsequent analysis.

[0073] The Dice coefficient is used as an evaluation criterion for segmentation accuracy:

[0074]

[0075] Where: p i is the predicted pixel value; g i is the actual pixel value; N is the total number of pixels.

[0076] Example 2: Automated tuberculosis image analysis based on deep learning

[0077] Model construction: The system adopts an improved convolutional neural network (CNN) architecture, which includes multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layer is used to extract low-level and high-level features of the image, the pooling layer is used to reduce the feature dimension, and the fully connected layer is used for feature fusion and classification.

[0078] To improve the accuracy of lesion detection, the model introduces a residual network (ResNet) structure to reduce the gradient vanishing problem in deep networks. The forward propagation formula of the model is as follows:

[0079] y=σ(W·x+b)

[0080] Where: x is the input image data; W is the convolution kernel weight matrix; b is the bias; σ is the ReLU activation function.

[0081] Lesion identification: The system extracts features from the preprocessed images, and the model outputs a classification probability map for each pixel to determine whether the area is a tuberculosis lesion.

[0082] The final result of lesion identification is determined by the Softmax classifier:

[0083]

[0084] Where: P(y=k|x) is the probability that the input image belongs to category k; z j is the activation value of the jth class.

[0085] The attention mechanism is introduced to weight the key areas of the image through the following formula:

[0086]

[0087] Where: Q, K, V are query, key and value matrices; d k The dimension of the key.

[0088] Example 3: Multimodal data fusion and personalized diagnosis

[0089] Data fusion: The system supports the fusion of imaging data with patients’ multimodal data (such as blood tests, genomic analysis, etc.).

[0090] The image features and clinical data features are fused through the feature weighted average strategy. The formula is as follows:

[0091] F fused =α·F image +(1-α)·F clinical

[0092] Among them: F fused is the fused feature vector; F image is the image feature; F clinical is the clinical feature; α is the fusion weight parameter.

[0093] Personalized diagnosis: The system is based on a multimodal deep neural network (MDNN) to generate personalized diagnosis results based on the individual characteristics of the patient.

[0094] The model supports the key annotation of high-risk patients and provides personalized health scores:

[0095]

[0096] Where: S is the health score; w i is the feature weight; f i is the patient's characteristic value.

[0097] Example 4: Intelligent follow-up and real-time monitoring

[0098] Patient follow-up module: The system keeps in touch with patients regularly through the intelligent follow-up module. Patients can fill out health status questionnaires through mobile devices, including information such as symptom changes and living habits.

[0099] The system uploads these follow-up data to the central server, and the data analysis module monitors the condition based on historical records and current data. The dynamic monitoring system supports connection with wearable devices to obtain real-time physiological data of patients, such as heart rate, blood pressure and respiratory rate. These data are transmitted to the system through the IoT gateway, and LSTM (Long Short-Term Memory Network) is used to perform time series analysis on the data to predict the development trend of the disease:

[0100] P t+1 =f(P t ,Δ t )

[0101] Where: P t+1 is the predicted value at the next time point; P t is the state at the current time point; Δ t For the time difference.

[0102] Example 5: Large-scale data analysis and epidemiological prediction

[0103] Data collection: The system collaborates with public health departments to conduct large-scale screening in specific areas and collect screening results and epidemiological data in the area.

[0104] The data is transmitted to the central database through a secure network. The system uses big data processing frameworks such as Hadoop and Spark to achieve distributed storage and analysis of large-scale data.

[0105] Epidemiological model construction: The system uses the SEIR model to simulate the tuberculosis transmission trend in a specific area, and the initial parameters of the model are dynamically adjusted based on actual data in the region.

[0106] The SEIR model formula is as follows:

[0107]

[0108] The prediction results are displayed through a visualization platform, and the reports generated by the system include key indicators such as incidence rate, transmission rate and potential risk areas.

[0109] Example 6: Data security and privacy protection

[0110] Data encryption and privacy protection: The system uses the AES encryption algorithm to encrypt and store patient data, and the data is encrypted using the TLS protocol during transmission. The AES algorithm formula is as follows:

[0111]

[0112] Where: C is the encrypted ciphertext; E k is the encryption function; M is the plaintext information; K is the encryption key.

[0113] Federated learning: The system introduces a federated learning framework, where each medical institution trains data locally and optimizes the global model through encrypted gradient and parameter sharing.

[0114] The parameter update formula for federated learning is as follows:

[0115]

[0116] Where: θ t is the current model parameter; η is the learning rate; L i (θ) is the loss function of the i-th institution; m is the number of institutions participating in federated learning.

[0117] Differential privacy protection: In order to further ensure patient privacy, the system introduces differential privacy technology in the process of data analysis and sharing.

[0118] At each data request, the system protects privacy by adding random noise to the data set. The noise generation adopts normal distribution, and the formula is as follows:

[0119]

[0120] Where: M(x) is the data after adding noise; f(x) is the original data function; The mean is 0 and the variance is σ 2 Normally distributed noise.

[0121] Multi-level access control: The system adopts a role-based access control (RBAC) mechanism to assign different permissions to users at different levels. Doctors, researchers, and system administrators will obtain corresponding permissions based on their roles, limiting access to sensitive data.

[0122] Data access logs within the database will be recorded in detail for auditing and security inspection.

[0123] Overall advantages of the system: Efficient tuberculosis screening: The system combines deep learning, data fusion and large-scale analysis technologies to significantly improve the efficiency of tuberculosis screening and reduce the probability of manual misdiagnosis.

[0124] Personalized diagnosis and treatment recommendations: The system uses multimodal data fusion technology to provide patients with personalized diagnosis reports and treatment recommendations, improving the accuracy and effectiveness of medical services.

[0125] Real-time dynamic monitoring: Through wearable devices and intelligent follow-up modules, the system can monitor the patient's health status in real time, detect potential changes in the condition early, and provide early warning information.

[0126] Strong data security and privacy protection: The system adopts a multi-level privacy protection strategy, including data encryption, differential privacy, and federated learning, to ensure the security of patient data during use and sharing.

[0127] Assisting public health decision-making: The system's large-scale data analysis capabilities can provide public health departments with accurate epidemiological forecasting and screening data support, helping to formulate effective disease prevention and control strategies.

[0128] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in the field. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent tuberculosis screening and data analysis system, characterized in that: Includes the following modules: Data acquisition module: used to collect multimodal imaging data and basic information of patients, including but not limited to X-rays, CT scan imaging data, and synchronize the data from multiple sources; Preprocessing module: standardize, normalize, denoise, enhance, segment and convert the collected image data; AI diagnosis module: Analyzes pre-processed image data through convolutional neural network models and deep learning algorithms to identify tuberculosis-related image features and locate lesion areas; Data analysis module: Based on the output results of the AI ​​diagnosis module, statistical analysis is performed on the screening data, providing grouping, classification, clustering, regression analysis and prediction functions; Report generation module: automatically generates standardized screening reports based on diagnosis and data analysis results, and displays analysis results in a variety of data visualization forms; Database management module: used to store patient information, imaging data, analysis results and diagnostic reports, and supports data retrieval, management and multi-level access permission control.

2. An intelligent tuberculosis screening and data analysis system according to claim 1, characterized in that: The Al diagnosis module adopts an improved deep convolutional neural network, whose architecture includes multi-layer convolutional layers, pooling layers, fully connected layers and Softmax classifiers. The feature extraction of the convolution process is implemented by the following formula: Where: x is the input image data; W j is the convolution kernel weight matrix of the jth layer; * represents the convolution operation; b i is the bias; σ is the ReLU activation function.

3. The intelligent tuberculosis screening and data analysis system according to claim 2, characterized in that: The AI ​​diagnosis module introduces a multi-level attention mechanism. The attention mechanism uses a self-attention algorithm to represent the weighted average of key area features. The formula is as follows: Where: Q is the query matrix; K is the key matrix; V is the value matrix; d k is the dimension of the key matrix.

4. The intelligent tuberculosis screening and data analysis system according to claim 1, characterized in that: The image segmentation function in the preprocessing module uses a fully convolutional neural network based on the U-Net architecture, and its loss function uses an improved Dice coefficient, and the formula is as follows: Where: p i is the pixel value predicted by the model; g i is the actual marked pixel value; N is the total number of pixels.

5. The intelligent tuberculosis screening and data analysis system according to claim 1, characterized in that: The data analysis module uses a Bayesian classifier to classify the image feature data and uses multiple regression analysis to assess the risk of disease. The multiple regression analysis formula is as follows: Among them: y is the probability of the patient suffering from the disease; x i is the patient's characteristic variable; β i is the regression coefficient; ∈ is the error term.

6. The intelligent tuberculosis screening and data analysis system according to claim 1, characterized in that: The data acquisition module has the function of real-time acquisition and transmission of image data. The data acquisition module supports multiple medical image formats, such as DICOM, JPEG and PNG, and introduces a data redundancy check mechanism during the acquisition process.

7. The intelligent tuberculosis screening and data analysis system according to claim 3, characterized in that: The A1 diagnosis module includes an image enhancement submodule, which performs contrast enhancement on the image by using an algorithm based on contrast-limited adaptive histogram equalization, and the formula is as follows: I output =CLAHE(I input ) Where: I input is the original input image; I output is the enhanced image; CLAHE is the contrast limited adaptive histogram equalization operation.

8. The intelligent tuberculosis screening and data analysis system according to claim 7, characterized in that: When analyzing the image data, the Al diagnosis module adjusts and trains the multi-dimensional weights of the disease characteristics through a supervised learning method, and the training loss function used is the cross entropy loss, and the formula is as follows: Where: L CE is the cross entropy loss; y i is the actual label; p i is the predicted probability value.

9. The intelligent tuberculosis screening and data analysis system according to claim 1, characterized in that: The report generation module can generate a complete report including screening results, image annotations, statistical analysis data and predictive analysis, supports image embedding and high-resolution chart generation, and supports automatic translation of multi-language reports.

10. The intelligent tuberculosis screening and data analysis system according to claim 1, characterized in that: The database management module uses the Advanced Encryption Standard and hash algorithm to encrypt and securely store data. The encryption algorithm formula is as follows: H(M)=SHA-256(M) Where: C is the encrypted ciphertext; E k is the encryption function; M is the plaintext information; K is the encryption key; H(M) is the hash value obtained by SHA-256 hashing.

Images