Image-based PICC intravenous therapy information monitoring system and method

Abstract

The present application provides an image-based PICC intravenous therapy information monitoring system and method, which relates to the field of intelligent image analysis, and determines whether there are signs of infection at the patient's puncture site by performing time series analysis on high-definition images of the patient's puncture site at different time points within the target monitoring time period. In this way, the accuracy and timeliness of infection monitoring can be improved, which helps to detect signs of infection early, so that timely measures can be taken to reduce the incidence and severity of infection.

Description

Technical Field

[0001] The present application relates to the field of intelligent image analysis, and more specifically, to an image-based PICC intravenous treatment information monitoring system and method. Background Art

[0002] In today's medical field, intravenous therapy has become one of the most common technical operations in clinical practice. With the changes in people's lifestyles in modern society and the influence of many factors such as the environment, the spectrum of diseases has become increasingly complex, and various difficult diseases have emerged one after another. At the same time, medicine continues to develop and progress, the types of medication prescribed by doctors have become more and more numerous, and the forms of medication have also evolved in a diversified direction, which has led to significant trends in intravenous therapy such as multiple routes, rapid speed, and long retention.

[0003] For example, in the treatment of tumor diseases, many patients need long-term chemotherapy, and patients with chronic diseases often rely on continuous nutritional support, regular antibiotic treatment or long-term use of other specific drugs to maintain their physical condition. Such situations that require long-term intravenous infusion are everywhere. PICC (Peripherally Inserted Central Catheter) came into being under such practical needs. As an extremely important medical device for long-term intravenous treatment, it can reduce the pain caused by repeated punctures to patients, is relatively safe, and can be left in place for a long time. It is widely used in patients who need long-term infusion, chemotherapy, nutritional support, antibiotic treatment or other drugs.

[0004] However, during the process of PICC intravenous treatment, infection is always a serious complication that cannot be ignored. Traditional methods of monitoring infection mostly rely on regular observation by medical staff during busy work breaks, such as checking whether the patient's puncture site has symptoms such as redness, swelling, and exudation at the prescribed time nodes. But many times, the infection may only show some very subtle and easily overlooked signs in the early stages. Since medical staff cannot always keep an eye on every patient using PICC for treatment, they are likely to miss the early stages of these signs of infection. Once early detection is missed, the infection may further develop. Subsequent treatment must not only deal with the infection itself, but also consider a series of other complex problems caused by the infection, such as thrombosis and catheter dysfunction. This greatly increases the complexity of subsequent treatment and also brings considerable obstacles to the patient's recovery process.

[0005] Therefore, an image-based PICC intravenous therapy information monitoring solution is needed. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The present application provides an image-based PICC intravenous treatment information monitoring system and method.

[0007] According to one aspect of the present application, there is provided an image-based PICC intravenous therapy information monitoring system, comprising:

[0008] The patient puncture site data acquisition module is used to obtain high-definition image data of the patient's puncture site at different time points within the target monitoring time period;

[0009] A patient puncture site data analysis module is used to analyze the high-definition image data of the patient's puncture site at different time points to obtain a fusion feature vector of the temporal change of the patient's puncture site;

[0010] A module for generating analysis results of infection conditions at the puncture site of a patient, which is used to fuse information of feature vectors according to temporal changes in the conditions at the puncture site of the patient, to determine the infection conditions at the puncture site of the patient;

[0011] Wherein, the patient puncture site data analysis module includes:

[0012] A patient puncture site condition time series feature extraction unit, used for extracting image features from the high-definition image data of the patient's puncture site at different time points to obtain multiple patient puncture site condition time series feature matrices;

[0013] A patient puncture site condition time series change feature capture unit, used for dynamically capturing the multiple patient puncture site condition time series feature matrices to obtain a patient puncture site condition time series change feature graph;

[0014] A dimension reduction unit for the time series change feature of the patient's puncture site condition, used for pooling the time series change feature map of the patient's puncture site condition to obtain a time series change feature vector of the patient's puncture site condition;

[0015] A patient puncture site condition time series change feature emphasis unit, used for performing feature emphasis processing on the patient puncture site condition time series change feature graph to obtain a patient puncture site condition time series change emphasized feature graph;

[0016] A dimensionality reduction unit for emphasizing the time series change feature of the patient's puncture site condition, used for pooling the emphasizing feature map of the time series change feature of the patient's puncture site condition to obtain a emphasizing feature vector of the time series change feature of the patient's puncture site condition;

[0017] The patient's puncture site condition temporal feature fusion unit is used to fuse the patient's puncture site condition temporal change feature vector and the patient's puncture site condition temporal change emphasis feature vector to obtain the patient's puncture site condition temporal change fusion feature vector.

[0018] Furthermore, the patient puncture site condition temporal feature extraction unit is used to: respectively pass the high-definition image data of the patient's puncture site at different time points through the puncture site condition capture device to obtain the multiple patient puncture site condition temporal feature matrices.

[0019] Furthermore, the patient puncture site condition time series change feature capturing unit is used to: pass the multiple patient puncture site condition time series feature matrices through the puncture site condition change focus device to obtain the patient puncture site condition time series change feature graph.

[0020] Furthermore, the patient's puncture site condition time series change feature emphasis unit is used to: pass the patient's puncture site condition time series change feature diagram through a puncture site condition change emphasizer to obtain the patient's puncture site condition time series change emphasized feature diagram.

[0021] Furthermore, the puncture site condition capturer is a convolutional neural network model including a deep and shallow feature fusion module, the puncture site condition change attention device is a convolutional neural network model using a three-dimensional convolution kernel, and the puncture site condition change emphasizer is a convolutional neural network model using a channel attention mechanism.

[0022] Furthermore, the patient puncture site infection analysis result generation module includes:

[0023] A feature optimization unit, configured to perform feature aggregation optimization on the fused feature vector of the time series change of the patient's puncture site condition based on the selection of key decision parameters to obtain an optimized fused feature vector of the time series change of the patient's puncture site condition;

[0024] The feature analysis and classification unit is used to fuse the optimized time-series changes of the patient's puncture site with the feature vector through an infection monitoring classifier to obtain a classification result, and the classification result is used to indicate whether there are signs of infection at the patient's puncture site.

[0025] Furthermore, the feature optimization unit is used to:

[0026] Extracting a target puncture site infection classification parameter matrix of the infection monitoring classifier;

[0027] Performing node-wise decomposition of the target puncture site infection classification parameter matrix in units of row vectors to obtain a set of target puncture site infection classification parameter node encoding vectors;

[0028] Taking each target puncture site infection classification parameter node encoding vector in the set of target puncture site infection classification parameter node encoding vectors as the wandering topological space, topological space constraints are respectively performed on the patient puncture site condition time series change fusion feature vectors to obtain a set of constrained patient puncture site condition time series change fusion feature vectors;

[0029] The position mean vector of the set of the constrained patient puncture site condition time series change fusion feature vectors is calculated to obtain the optimized patient puncture site condition time series change fusion feature vector.

[0030] According to another aspect of the present application, a method for monitoring PICC intravenous therapy information based on an image is provided, comprising:

[0031] Obtain high-definition image data of the patient's puncture site at different time points within the target monitoring time period;

[0032] Analyzing the high-definition image data of the patient's puncture site at different time points to obtain a fusion feature vector of the temporal change of the patient's puncture site;

[0033] According to the time series changes of the puncture site of the patient, the information of the eigenvector is fused to determine the infection status of the puncture site of the patient;

[0034] The high-definition image data of the patient's puncture site at different time points are analyzed to obtain a fusion feature vector of the temporal change of the patient's puncture site, including:

[0035] Performing image feature extraction on the high-definition image data of the patient's puncture site at different time points to obtain a time series feature matrix of multiple patient puncture site conditions;

[0036] Capturing dynamic changes of the multiple patient puncture site condition time series feature matrices to obtain a patient puncture site condition time series change feature graph;

[0037] Pooling the patient's puncture site condition time series change feature map to obtain a patient's puncture site condition time series change feature vector;

[0038] Performing feature emphasis processing on the patient's puncture site condition time series change characteristic graph to obtain a patient's puncture site condition time series change emphasized characteristic graph;

[0039] Pooling the patient's puncture site condition time series change emphasis feature map to obtain the patient's puncture site condition time series change emphasis feature vector;

[0040] The patient's puncture site condition temporal change feature vector and the patient's puncture site condition temporal change emphasis feature vector are fused to obtain the patient's puncture site condition temporal change fused feature vector.

[0041] Compared with the prior art, the image-based PICC intravenous therapy information monitoring system and method provided by the present application can determine whether there are signs of infection at the patient's puncture site by performing time series analysis on high-definition images of the patient's puncture site at different time points within the target monitoring time period. In this way, the accuracy and timeliness of infection monitoring can be improved, which helps to detect signs of infection at an early stage, so as to take timely measures to reduce the incidence and severity of infection. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0043] Figure 1 It is a system block diagram of an image-based PICC intravenous treatment information monitoring system according to an embodiment of the present application.

[0044] Figure 2 It is a block diagram of a patient puncture site data analysis module in an image-based PICC intravenous treatment information monitoring system according to an embodiment of the present application.

[0045] Figure 3 This is a block diagram of a patient puncture site infection analysis result generation module in an image-based PICC intravenous treatment information monitoring system according to an embodiment of the present application.

[0046] Figure 4 The present invention is a flowchart of an image-based PICC intravenous treatment information monitoring method according to an embodiment of the present application. DETAILED DESCRIPTION

[0047] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0048] As mentioned in the above background technology, intravenous therapy is the most common technical operation in clinical practice. With the complexity of the disease spectrum and the diversification of the types and forms of medication prescribed by doctors, intravenous therapy has shown a trend of multi-path, fast speed and long retention. PICC is a medical device used for long-term intravenous treatment, often used for patients who need long-term infusion, chemotherapy, nutritional support, antibiotic treatment or other drugs. Infection is a serious complication during PICC intravenous treatment. Traditional monitoring methods may rely mainly on regular observation by medical staff, which may miss the early stages of signs of infection, thereby affecting the complexity of subsequent treatment. Therefore, an image-based PICC intravenous treatment information monitoring solution is expected.

[0049] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks provides new solutions and solutions for image-based PICC intravenous treatment information monitoring.

[0050] Figure 1 FIG. 1 is a system block diagram of an image-based PICC intravenous treatment information monitoring system according to an embodiment of the present application. Figure 1 As shown, in the image-based PICC intravenous treatment information monitoring system 100, it includes: a patient puncture site data acquisition module 110, which is used to obtain high-definition image data of the patient's puncture site at different time points within the target monitoring time period; a patient puncture site data analysis module 120, which is used to analyze the high-definition image data of the patient's puncture site at different time points to obtain a fused feature vector of the patient's puncture site condition time series change; a patient puncture site infection condition analysis result generation module 130, which is used to determine the infection condition of the patient's puncture site based on the information of the fused feature vector of the patient's puncture site condition time series change.

[0051] In the embodiment of the present application, the patient puncture site data acquisition module 110 is used to obtain high-definition image data of the patient's puncture site at different time points within the target monitoring time period. It should be understood that the high-definition image data of the patient's puncture site can reflect information about skin color and texture. In detail, the skin at a normal puncture site usually presents a healthy skin color, and when an infection occurs, local skin redness may occur, which is one of the common external manifestations of inflammatory response. For example, in the early stage of infection, there may be only a faint blush, and as the infection worsens, the redness will become more obvious and the range may gradually expand. High-definition images can clearly capture subtle changes in skin color, providing an intuitive basis for judging the progress of infection. Moreover, the texture of healthy skin is relatively flat and smooth. If an infection occurs, it may become swollen, raised, or even hardened. Through high-definition images, it is possible to carefully observe whether the skin surface has unevenness, bulges, and other texture changes, so as to infer whether there are problems such as tissue edema or abscess formation caused by inflammation. Considering that infection is a pathological process that develops over time, the various changes caused by it at the puncture site are not instantaneous and fixed, but have a gradual evolution process. By acquiring image data at different time points, the skin condition of the puncture site at different time points can be fully recorded, so as to accurately determine whether there is an infection at the puncture site and avoid misjudgment due to insufficient or inaccurate data.

[0052] In the embodiment of the present application, the patient puncture site data analysis module 120 is used to analyze the high-definition image data of the patient's puncture site at different time points to obtain a fusion feature vector of the temporal change of the patient's puncture site. Specifically, Figure 2 FIG. 1 is a block diagram of a patient puncture site data analysis module in an image-based PICC intravenous treatment information monitoring system according to an embodiment of the present application. Figure 2As shown, the patient puncture site data analysis module 120 includes: a patient puncture site situation time series feature extraction unit 121, which is used to extract image features from the high-definition image data of the patient puncture site at different time points to obtain multiple patient puncture site situation time series feature matrices; a patient puncture site situation time series change feature capture unit 122, which is used to capture dynamic changes of the multiple patient puncture site situation time series feature matrices to obtain a patient puncture site situation time series change feature map; a patient puncture site situation time series change feature dimension reduction unit 123, which is used to pool the patient puncture site situation time series change feature map to obtain a patient puncture site situation time series change feature vector ; A patient puncture site condition temporal change feature emphasis unit 124, used for performing feature emphasis processing on the patient puncture site condition temporal change feature graph to obtain a patient puncture site condition temporal change emphasis feature graph; a patient puncture site condition temporal change emphasis feature dimension reduction unit 125, used for pooling the patient puncture site condition temporal change emphasis feature graph to obtain a patient puncture site condition temporal change emphasis feature vector; a patient puncture site condition temporal feature fusion unit 126, used for fusing the patient puncture site condition temporal change feature vector and the patient puncture site condition temporal change emphasis feature vector to obtain the patient puncture site condition temporal change fusion feature vector.

[0053] In the embodiment of the present application, the patient puncture site condition time series feature extraction unit 121 is used to extract image features from the high-definition image data of the patient puncture site at different time points to obtain multiple patient puncture site condition time series feature matrices. Specifically, in the embodiment of the present application, the patient puncture site condition time series feature extraction unit 121 is used to: pass the high-definition image data of the patient puncture site at different time points through the puncture site condition capturer to obtain the multiple patient puncture site condition time series feature matrices. It should be understood that when observing the puncture site image, the human eye may only be able to intuitively see some surface and relatively obvious changes, such as obvious redness, swelling or exudation. In order to be able to dig out the information hidden in the image that is not easily directly perceived by the naked eye but is closely related to the infection, such as subtle changes in skin texture, image features corresponding to changes in local tissue density, etc., so as to more comprehensively capture signs of infection, it is necessary to perform image feature extraction processing on the high-definition image data of the patient puncture site at different time points. The specific method of image feature extraction is to input the high-definition image data of the patient puncture site at different time points into the puncture site condition capturer for processing. In particular, the puncture site condition capturer described in the present application is a convolutional neural network model including a deep and shallow feature fusion module. The shallow convolution layer in the convolutional neural network can capture some local, low-level detail features in the image, such as edge information in the image, color changes in small areas, etc. These detail features are very critical for reflecting the slight changes in the skin at the puncture site, and are an important basis for early detection of signs of infection. For example, the slight abnormality of the skin texture that has just appeared can be reflected through the shallow features; while the deep convolution layer in the convolutional neural network can extract more abstract and holistic semantic features, and grasp the information contained in the image from a macro level; by fusing the features of two different levels together, a comprehensive feature representation of the puncture site image can be formed, which neither misses the detail changes nor can accurately judge the situation as a whole, so that the extracted features are more comprehensive, accurate and reliable in reflecting the real puncture site status.

[0054] In the embodiment of the present application, the patient puncture site situation time series change feature capture unit 122 is used to capture the dynamic changes of the multiple patient puncture site situation time series feature matrices to obtain the patient puncture site situation time series change feature map. Specifically, in the embodiment of the present application, the patient puncture site situation time series change feature capture unit 122 is used to: pass the multiple patient puncture site situation time series feature matrices through the puncture site situation change attention device to obtain the patient puncture site situation time series change feature map. It should be understood that the patient puncture site situation time series feature matrix can capture the texture features of the puncture site, the local shape features of the puncture site, such as the shape of the red and swollen area of ​​the skin around the puncture point, and the color distribution feature information of the puncture site, such as the gradient of the red color depth change. Each patient puncture site situation time series feature matrix represents the patient puncture site situation information at a time point. In order to understand the dynamic performance of the patient's puncture site after the initial puncture, it is necessary to input the multiple patient puncture site situation time series feature matrices into the puncture site situation change attention device to capture the dynamic changes of the puncture site situation. In particular, the puncture site change observer described in the present application is a convolutional neural network model using a three-dimensional convolution kernel. When processing the time series feature matrix, the three-dimensional convolution kernel can take into account the information interaction of the spatial and temporal dimensions at the same time. It can capture the trend of continuous changes in features over time along the time axis, such as whether the redness and swelling feature value of a local skin area is slowly rising or tending to be stable, which can clearly show the dynamic changes of the puncture site, and thus help to determine whether the patient has an infection.

[0055] In the embodiment of the present application, the patient puncture site condition time series change feature dimension reduction unit 123 is used to pool the patient puncture site condition time series change feature map to obtain the patient puncture site condition time series change feature vector. It should be understood that although the patient puncture site condition time series change feature map can show the detailed information of the puncture site condition in the time series change process, it will contain some redundant and local details, which are not very useful for judging whether there are signs of infection at the puncture site, but will increase the complexity of subsequent calculations. In order to extract representative abstract information that can reflect the overall change trend and key features, it is helpful to focus on the feature performance that is truly valuable for infection judgment, and at the same time can reduce the amount of calculation and reduce the complexity of calculation. In the technical solution of the present application, it is necessary to pool the patient puncture site condition time series change feature map to obtain the patient puncture site condition time series change feature vector. Moreover, in the process of collecting high-definition images of the patient's puncture site and the previous series of processing, it may be affected by factors such as different lighting conditions, slightly different shooting angles, and slight changes in the patient's body position, resulting in some subtle and unnecessary fluctuations in the feature map. The pooling operation can eliminate these subtle interferences caused by external factors to a certain extent by integrating local areas, making the final feature vector more robust, that is, less sensitive to these non-essential and accidental changes, and more stably reflecting the actual state changes of the puncture site.

[0056] In the embodiment of the present application, the patient puncture site condition time series change feature emphasis unit 124 is used to perform feature emphasis processing on the patient puncture site condition time series change feature map to obtain the patient puncture site condition time series change emphasized feature map. Specifically, in the embodiment of the present application, the patient puncture site condition time series change feature emphasis unit 124 is used to: pass the patient puncture site condition time series change feature map through the puncture site condition change emphasizer to obtain the patient puncture site condition time series change emphasized feature map. It should be understood that in the puncture site condition time series change feature map, although it has included a lot of various feature information reflecting the change of the puncture site over time, the importance of this information varies. Among them, the features directly related to the signs of infection, such as increased skin redness and swelling, increased exudate, abnormal texture changes, etc., play a key role in judging whether the infection occurs and the severity, but they may be covered by some relatively minor or background features. By performing feature emphasis processing, these features that are crucial to judging the infection can be highlighted, so that subsequent analysis can focus more on these key information, improving the accuracy and efficiency of infection judgment. Specifically, the patient's puncture site condition temporal change feature map is input into the puncture site condition change emphasizer for processing. In the present application, the puncture site condition change emphasizer is a convolutional neural network model using a channel attention mechanism. The channel attention mechanism has automatic learning ability, which can automatically analyze and learn the importance weight corresponding to each channel based on the information carried by each channel in the input puncture site condition temporal change feature map (here the channel dimension represents the time dimension). In other words, the model can determine by itself which of the feature channels corresponding to different time points are more critical to reflecting signs of infection and which are relatively minor, without the need to manually set the weights, which not only reduces the workload of manual intervention, but also can more objectively and accurately highlight important features according to the characteristics of the data itself. Since the model assigns different importance weights to channels corresponding to different time dimensions, it can highlight the feature information at the most critical time points in infection judgment. For example, the feature channel corresponding to the time point when the signs of infection first begin to appear may be emphasized, and the features it carries, such as slight redness and swelling of the skin and subtle changes in texture, will be more prominent. This will help to accurately locate the starting moment of the infection and provide strong clues for the early detection of infection.

[0057] In an embodiment of the present application, the patient's puncture site condition time series change emphasis feature dimensionality reduction unit 125 is used to pool the patient's puncture site condition time series change emphasis feature map to obtain a patient's puncture site condition time series change emphasis feature vector. It should be understood that the patient's puncture site condition time series change emphasis feature map obtained after processing by the puncture site condition change emphasizer, although the key infection-related features have been emphasized, it still exists in the form of an image, and the data volume is often large and the dimension is relatively high. In subsequent analysis, continuing to process such large-scale data will increase the computational complexity, occupy more storage resources, and may lead to low computational efficiency. Through the pooling operation, the patient's puncture site condition time series change emphasis feature map can be processed by dimensionality reduction and converted into a feature vector with lower dimensionality, which can make the data structure more concise and facilitate the efficient implementation of subsequent further data analysis.

[0058] In the embodiment of the present application, the patient puncture site situation time series feature fusion unit 126 is used to fuse the patient puncture site situation time series change feature vector and the patient puncture site situation time series change emphasis feature vector to obtain the patient puncture site situation time series change fusion feature vector. It should be understood that the patient puncture site situation time series change feature vector extracts the comprehensive situation of various feature changes at different time points of the puncture site as a whole, covers the information after a certain degree of integration of various features such as skin appearance and surrounding tissue changes, and reflects the feature performance at a relatively macro level and its evolution trend over time; the patient puncture site situation time series change emphasis feature vector is obtained by performing feature emphasis processing on the time series change feature map (using the channel attention mechanism to highlight key infection-related features) and then pooling, and its focus is on highlighting the most critical and recognizable feature information for judging signs of infection. In order to integrate the respective advantages of these two vectors, it is necessary to fuse the patient puncture site situation time series change feature vector and the patient puncture site situation time series change emphasis feature vector. Through fusion, the completeness of the feature representation of the puncture site status can be greatly improved, which can lay the foundation for more accurate judgment of whether the patient has an infection.

[0059] In the embodiment of the present application, the patient puncture site infection analysis result generation module 130 is used to determine the infection status of the patient's puncture site by fusing the information of the feature vector according to the temporal changes of the patient's puncture site status. Specifically, Figure 3 FIG. 1 is a block diagram of a patient puncture site infection analysis result generation module in an image-based PICC intravenous treatment information monitoring system according to an embodiment of the present application. Figure 3As shown, the patient puncture site infection situation analysis result generation module 130 includes: a feature optimization unit 131, which is used to perform feature aggregation optimization on the patient puncture site situation time series change fusion feature vector based on key decision parameter selection to obtain an optimized patient puncture site situation time series change fusion feature vector; a feature analysis classification unit 132, which is used to pass the optimized patient puncture site situation time series change fusion feature vector through an infection monitoring classifier to obtain a classification result, and the classification result is used to indicate whether there is a sign of infection at the patient's puncture site.

[0060] In an embodiment of the present application, the feature optimization unit 131 is used to perform feature aggregation optimization based on the selection of key decision parameters on the fusion feature vector of the time series change of the patient's puncture site condition to obtain an optimized fusion feature vector of the time series change of the patient's puncture site condition. In particular, considering that the high-definition image data of the patient's puncture site contains a large amount of detailed information, such as color changes, redness and swelling, exudation, etc., the changes in these features in time series may be an important indicator of signs of infection. However, important information related to signs of infection in the original high-definition image data may be lost during the feature extraction and fusion process. The loss of such key information will directly lead to the classifier's inability to accurately determine whether there are signs of infection at the patient's puncture site, increasing the risk of misdiagnosis or missed diagnosis. In order to maintain the main characteristics of the original data, in the technical solution of the present application, it is necessary to perform feature aggregation optimization based on the selection of key decision parameters on the fusion feature vector of the time series change of the patient's puncture site condition to obtain an optimized fusion feature vector of the time series change of the patient's puncture site condition.

[0061] Specifically, in an embodiment of the present application, the feature optimization unit 131 is used to: extract the target puncture site infection classification parameter matrix of the infection monitoring classifier; perform node decomposition on the target puncture site infection classification parameter matrix in units of row vectors to obtain a set of target puncture site infection classification parameter node encoding vectors; use each target puncture site infection classification parameter node encoding vector in the set of target puncture site infection classification parameter node encoding vectors as the wandering topological space, and perform topological space constraints on the patient's puncture site condition time series change fusion feature vectors to obtain a set of constrained patient puncture site condition time series change fusion feature vectors; calculate the positional mean vector of the set of constrained patient puncture site condition time series change fusion feature vectors to obtain the optimized patient puncture site condition time series change fusion feature vector.

[0062] More specifically, in an embodiment of the present application, each target puncture site infection classification parameter node encoding vector in the set of the target puncture site infection classification parameter node encoding vector is used as a wandering topological space, and the patient's puncture site condition time series change fusion feature vector is respectively subjected to topological space constraints to obtain a set of constrained patient puncture site condition time series change fusion feature vectors, including: multiplying the patient's puncture site condition time series change fusion feature vector and the transposed vector of the target puncture site infection classification parameter node encoding vector, and calculating the natural exponential function value of the multiplication result to obtain the puncture site infection classification weighted exponential response weight; calculating the patient The Euclidean distance between the target puncture site infection classification parameter node coding vector and the puncture site infection classification node coding distance value is obtained; the puncture site infection classification node coding distance value is multiplied with the patient's puncture site condition time series change fusion feature vector, and the natural exponential function value is calculated for each eigenvalue of the vector after the dot product to obtain the puncture site infection classification distance guidance index feature vector; the puncture site infection classification weighted index response weight is multiplied with the puncture site infection classification distance guidance index feature vector to obtain the constrained patient puncture site condition time series change fusion feature vector.

[0063] In the embodiment of the present application, specifically, the feature aggregation optimization based on the selection of key decision parameters is performed on the patient's puncture site condition time series change fusion feature vector to obtain the optimized patient's puncture site condition time series change fusion feature vector, including: processing the patient's puncture site condition time series change fusion feature vector with the following optimization formula to obtain the optimized patient's puncture site condition time series change fusion feature vector; wherein the optimization formula is:

[0064] ;

[0065] ;

[0066] ;

[0067] in, represents the target puncture site infection classification parameter matrix, The first, second, and third nodes of the set of target puncture site infection classification parameter node encoding vectors are represented , target puncture site infection classification parameter node encoding vector, represents the transpose of a vector, represents the natural exponential function, The fusion feature vector representing the temporal changes of the patient's puncture site, represents matrix multiplication, It means point multiplication by position. Represents the calculation vector and vector The Euclidean distance between The first one represents the set of fused feature vectors of the temporal changes of the puncture site of the patient after the constraint After the constraints, the fusion feature vector of the time series change of the patient's puncture site is The total number of sets of fused feature vectors representing the temporal changes in the puncture site of the patient after the constraint, Represents the optimized fused feature vector of the temporal changes in the patient's puncture site.

[0068] That is, in response to the above technical problems, in the technical solution of the present application, the feature aggregation optimization based on the selection of key decision parameters is performed on the fusion feature vector of the time series change of the patient's puncture site. The process first extracts the key parameters for decision-making from the trained infection monitoring classifier. These parameters form a matrix in a high-dimensional space, and each row represents the weight or influencing factor on a different dimension. The target puncture site infection classification parameter matrix can provide insight into the location and shape of the model decision boundary, and then infer which input features are most critical to the prediction results.

[0069] Next, the target puncture site infection classification parameter matrix is ​​node-decomposed in units of row vectors to obtain a set of node encoding vectors of target puncture site infection classification parameters. Here, each row vector is a node in the graph theory, which means that each set of parameters is now regarded as an entity with potential connectivity. This transformation allows the application of methods from graph theory and network science to explore the interactions between features. The node encoding vector not only carries information about the original parameters, but also implies knowledge about the topological structure of the entire system. Node decomposition further reveals the intrinsic connection pattern or structure of the data, so that the optimized feature vector can better adapt to new task requirements.

[0070] Then, each target puncture site infection classification parameter node encoding vector in the set of the target puncture site infection classification parameter node encoding vector is used as the walking topological space, and the patient puncture site condition time series change fusion feature vector is respectively subjected to topological space constraints to obtain a set of constrained patient puncture site condition time series change fusion feature vectors. "Walking" in the topological space defined by the node encoding vector is actually simulating an exploratory process, the purpose of which is to find feature transformations that can best maintain the structural characteristics of the original data. Each step determines the position of the next step based on the probability distribution of the current state. The topological space constraints ensure that even in different contexts, the feature representation still retains certain invariances. At the same time, it can also promote cross-domain transfer learning because it emphasizes the universal relationship between features rather than the details of a specific field. In this way, the reconstruction of the feature space is achieved, making the optimized feature vector more compact and having better generalization ability.

[0071] Finally, the position mean vector of the set of fused feature vectors of the temporal changes in the patient's puncture site conditions after the constraints is calculated to obtain the optimized fused feature vector of the temporal changes in the patient's puncture site conditions. Calculating the mean vector is a statistical aggregation method used to integrate the optimal solution from multiple perspectives. The idea behind this step is to reduce the deviation caused by a single estimate by fusing the information provided by different sample points. The averaging process is equivalent to performing a soft voting, which enhances the expressiveness of common features and makes the optimized feature vector more stable and reliable.

[0072] In the embodiment of the present application, the feature parsing classification unit 132 is used to pass the optimized patient puncture site situation time series change fusion feature vector through the infection monitoring classifier to obtain a classification result, and the classification result is used to indicate whether there is a sign of infection at the patient's puncture site. It should be understood that the optimized patient puncture site situation time series change fusion feature vector integrates multiple dimensions and multiple levels to reflect the key information related to the puncture site state and infection. In order to determine whether the patient is infected based on these key information, it is necessary to input the optimized patient puncture site situation time series change fusion feature vector into the infection monitoring classifier. The classification result output by the infection monitoring classifier clearly indicates whether there is a sign of infection at the patient's puncture site, which is a clear binary classification ("infection exists" or "no infection"), which directly provides medical staff with an intuitive, easy-to-understand and applicable decision reference, which facilitates them to quickly take corresponding follow-up treatment, care or further observation measures, and avoids decision delays or improper handling due to fuzzy judgments. Since the infection monitoring classifier is a model built based on a large amount of data for learning and training, it can fully explore the potential relationship between each feature in the feature vector and the infection, and can comprehensively consider many factors and their mutual influence when making classification judgments. Compared with manual judgment based on experience, it can more accurately identify early subtle signs of infection and distinguish different degrees of infection. Moreover, the judgment process of the classifier is relatively stable and is not affected by subjective factors (such as fatigue and emotions of medical staff), which is conducive to improving the accuracy and reliability of the entire PICC intravenous treatment information monitoring, thereby ensuring the safety of patients during treatment and reasonable medical intervention.

[0073] In summary, the image-based PICC intravenous treatment information monitoring system 100 according to the embodiment of the present application is explained, which determines whether there are signs of infection at the patient's puncture site by performing time series analysis on high-definition images of the patient's puncture site at different time points within the target monitoring time period. In this way, the accuracy and timeliness of infection monitoring can be improved, which helps to detect signs of infection early, so as to take timely measures to reduce the incidence and severity of infection.

[0074] Figure 4 FIG. 1 is a flow chart of a method for monitoring PICC intravenous treatment information based on an image according to an embodiment of the present application. Figure 4 As shown, in the image-based PICC intravenous treatment information monitoring method, the method includes: S1, obtaining high-definition image data of the patient's puncture site at different time points within the target monitoring time period; S2, analyzing the high-definition image data of the patient's puncture site at different time points to obtain a fused feature vector of the patient's puncture site condition time series change; S3, judging the infection condition of the patient's puncture site based on the information of the fused feature vector of the patient's puncture site condition time series change.

[0075] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned image-based PICC intravenous treatment information monitoring method have been described in the above reference. Figures 1 to 3 The image-based PICC intravenous therapy information monitoring system has been introduced in detail, and therefore, its repeated description will be omitted.

[0076] In summary, the image-based PICC intravenous treatment information monitoring method according to the embodiment of the present application is explained, which determines whether there are signs of infection at the patient's puncture site by performing time series analysis on high-definition images of the patient's puncture site at different time points within the target monitoring time period. In this way, the accuracy and timeliness of infection monitoring can be improved, which helps to detect signs of infection early, so as to take timely measures to reduce the incidence and severity of infection.

Claims

1. An image-based PICC intravenous therapy information monitoring system, characterized in that: include: The patient puncture site data acquisition module is used to obtain high-definition image data of the patient's puncture site at different time points within the target monitoring time period; A patient puncture site data analysis module is used to analyze the high-definition image data of the patient's puncture site at different time points to obtain a fusion feature vector of the temporal change of the patient's puncture site; A module for generating analysis results of infection conditions at the puncture site of a patient, which is used to fuse information of feature vectors according to temporal changes in the conditions at the puncture site of the patient, to determine the infection conditions at the puncture site of the patient; Wherein, the patient puncture site data analysis module includes: A patient puncture site condition time series feature extraction unit, used for extracting image features from the high-definition image data of the patient's puncture site at different time points to obtain multiple patient puncture site condition time series feature matrices; A patient puncture site condition time series change feature capture unit, used for dynamically capturing the multiple patient puncture site condition time series feature matrices to obtain a patient puncture site condition time series change feature graph; A dimension reduction unit for the time series change feature of the patient's puncture site condition, used for pooling the time series change feature map of the patient's puncture site condition to obtain a time series change feature vector of the patient's puncture site condition; A patient puncture site condition time series change feature emphasis unit, used for performing feature emphasis processing on the patient puncture site condition time series change feature graph to obtain a patient puncture site condition time series change emphasized feature graph; A dimensionality reduction unit for emphasizing the time series change feature of the patient's puncture site condition, used for pooling the emphasizing feature map of the time series change feature of the patient's puncture site condition to obtain a emphasizing feature vector of the time series change feature of the patient's puncture site condition; The patient's puncture site condition temporal feature fusion unit is used to fuse the patient's puncture site condition temporal change feature vector and the patient's puncture site condition temporal change emphasis feature vector to obtain the patient's puncture site condition temporal change fusion feature vector.

2. The image-based PICC intravenous treatment information monitoring system according to claim 1, characterized in that: The patient puncture site condition time series feature extraction unit is used to: respectively pass the high-definition image data of the patient's puncture site at different time points through the puncture site condition capture device to obtain the multiple patient puncture site condition time series feature matrices.

3. The image-based PICC intravenous treatment information monitoring system according to claim 2, characterized in that: The patient puncture site condition time series change feature capture unit is used to: pass the multiple patient puncture site condition time series feature matrices through the puncture site condition change focus device to obtain the patient puncture site condition time series change feature graph.

4. The image-based PICC intravenous treatment information monitoring system according to claim 3, characterized in that: The patient's puncture site condition time series change feature emphasis unit is used to: pass the patient's puncture site condition time series change feature diagram through the puncture site condition change emphasizer to obtain the patient's puncture site condition time series change emphasized feature diagram.

5. The image-based PICC intravenous treatment information monitoring system according to claim 4, characterized in that: The puncture site condition capturer is a convolutional neural network model including a deep and shallow feature fusion module, the puncture site condition change attention device is a convolutional neural network model using a three-dimensional convolution kernel, and the puncture site condition change emphasizer is a convolutional neural network model using a channel attention mechanism.

6. The image-based PICC intravenous treatment information monitoring system according to claim 5, characterized in that: The patient puncture site infection analysis result generation module comprises: A feature optimization unit, configured to perform feature aggregation optimization on the fused feature vector of the temporal change of the patient's puncture site condition based on the selection of key decision parameters to obtain an optimized fused feature vector of the temporal change of the patient's puncture site condition; The feature analysis and classification unit is used to fuse the optimized time-series changes of the patient's puncture site with the feature vector through an infection monitoring classifier to obtain a classification result, and the classification result is used to indicate whether there are signs of infection at the patient's puncture site.

7. The image-based PICC intravenous treatment information monitoring system according to claim 6, characterized in that: The feature optimization unit is used to: Extracting a target puncture site infection classification parameter matrix of the infection monitoring classifier; Performing node-wise decomposition of the target puncture site infection classification parameter matrix in units of row vectors to obtain a set of target puncture site infection classification parameter node encoding vectors; Taking each target puncture site infection classification parameter node encoding vector in the set of target puncture site infection classification parameter node encoding vectors as the wandering topological space, topological space constraints are respectively performed on the patient puncture site condition time series change fusion feature vectors to obtain a set of constrained patient puncture site condition time series change fusion feature vectors; The position mean vector of the set of the constrained patient puncture site condition time series change fusion feature vectors is calculated to obtain the optimized patient puncture site condition time series change fusion feature vector.

8. An image-based PICC intravenous treatment information monitoring method, characterized in that: include: Obtain high-definition image data of the patient's puncture site at different time points within the target monitoring time period; Analyzing the high-definition image data of the patient's puncture site at different time points to obtain a fusion feature vector of the temporal change of the patient's puncture site; According to the time series changes of the puncture site of the patient, the information of the eigenvector is fused to determine the infection status of the puncture site of the patient; The high-definition image data of the patient's puncture site at different time points are analyzed to obtain a fusion feature vector of the temporal change of the patient's puncture site, including: Performing image feature extraction on the high-definition image data of the patient's puncture site at different time points to obtain a time series feature matrix of multiple patient puncture site conditions; Capturing dynamic changes of the multiple patient puncture site condition time series feature matrices to obtain a patient puncture site condition time series change feature graph; Pooling the patient's puncture site condition time series change feature map to obtain a patient's puncture site condition time series change feature vector; Performing feature emphasis processing on the patient's puncture site condition time series change characteristic graph to obtain a patient's puncture site condition time series change emphasized characteristic graph; Pooling the patient's puncture site condition time series change emphasis feature map to obtain the patient's puncture site condition time series change emphasis feature vector; The patient's puncture site condition temporal change feature vector and the patient's puncture site condition temporal change emphasis feature vector are fused to obtain the patient's puncture site condition temporal change fused feature vector.

9. The image-based PICC intravenous treatment information monitoring method according to claim 8, characterized in that: According to the information of the feature vector of the temporal change of the puncture site of the patient, the infection situation of the puncture site of the patient is judged, including: Performing feature aggregation optimization based on the selection of key decision parameters on the patient's puncture site condition time series change fused feature vector to obtain an optimized patient's puncture site condition time series change fused feature vector; The optimized fusion feature vector of the time series change of the patient's puncture site condition is passed through an infection monitoring classifier to obtain a classification result, and the classification result is used to indicate whether there is a sign of infection at the patient's puncture site.

Images