Patient pressure sore identification and nursing auxiliary system based on PDA and deep learning

Abstract

The invention relates to the technical field of medical care, in particular to a patient pressure sore recognition and nursing auxiliary system based on PDA and deep learning, and the system comprises PDA equipment which has a double-photographing function and is used for photographing a human face, uploading the human face to the system, inputting human face information, photographing a skin image of a patient and uploading the skin image; the server side is provided with a deep learning algorithm model and a nursing scheme database; the server side is used for analyzing the skin image of the patient uploaded by the PDA so as to judge the pressure sore condition; and the nursing scheme database provides corresponding scientific nursing measures according to the judgment result of the server side. According to the system, accurate pressure sore recognition is achieved by means of a deep learning algorithm and a thermal imaging technology, personalized nursing measures are generated by means of a nursing scheme database, accurate patient identity recognition is guaranteed by means of face recognition, medical risks are comprehensively reduced, and nursing efficiency and quality are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical care, and particularly to a patient pressure ulcer recognition and nursing assistance system based on PDA and deep learning. Background Art

[0002] In clinical nursing work, ensuring medical safety and improving nursing quality are the core objectives. However, currently, two severe challenges are faced: the accuracy of the doctor's order execution link and the scientific nature of pressure ulcer prevention and nursing.

[0003] Whether the doctor's order is executed accurately directly relates to the patient's life and health. In the actual execution process, patient identification is the primary checkpoint to ensure the accurate execution of the doctor's order. Currently, patient identification mainly relies on manual inquiry, which is extremely vulnerable to various factors. Once misidentification occurs, it may lead to the incorrect execution of the doctor's order, and in severe cases, even cause irreparable medical accidents, posing a great threat to the patient's life safety.

[0004] The prevention and nursing of pressure ulcers are also the top priorities of nursing work. Traditional methods for pressure ulcer judgment overly rely on nurses' subjective experience, and there are significant differences in the experience levels and professional cognitions of different nurses, and their understandings and grasps of the pressure ulcer judgment criteria are also different. This leads to frequent errors in actual judgment. Especially in the initial stage of pressure ulcers, the symptoms are hidden, and it is difficult for inexperienced nurses to detect and accurately judge the occurrence and stage of pressure ulcers in a timely manner, thus delaying the best treatment opportunity and seriously affecting the patient's treatment effect and recovery process. At the same time, the existing methods have obvious shortcomings in quantitatively evaluating the development of pressure ulcers and cannot accurately record and deeply analyze the dynamic change process of pressure ulcers, which makes the formulation of personalized nursing plans lack strong data support and is difficult to meet the differentiated needs of patients.

[0005] In summary, it is extremely urgent to develop a system integrating functions such as accurate patient identification, automated pressure ulcer analysis, and precise nursing measures provision, which has extremely important practical significance for reducing medical risks, improving nursing quality, and improving the patient's prognosis. Summary of the Invention

[0006] In view of this, the purpose of the present invention is to provide a patient pressure ulcer recognition and nursing assistance system based on PDA and deep learning to solve the problems in the prior art that the prior art mainly relies on nurses' subjective experience to judge pressure ulcers, and there are also potential hazards in patient identification and doctor's order execution.

[0007] According to the first aspect of the embodiments of the present invention, a patient pressure ulcer recognition and nursing assistance system based on PDA and deep learning is provided, including:

[0008] A PDA device with dual camera functions, which is used to take pictures of faces and upload the face information to the system for entry, and is also used to take pictures of patients' skin images and upload them;

[0009] A server side, on which a deep learning algorithm model and a nursing plan database are deployed;

[0010] The server side is responsible for analyzing the patients' skin images uploaded by the PDA to judge the pressure ulcer situation; the nursing plan database provides corresponding scientific nursing measures according to the judgment results of the server side.

[0011] Further, the deep learning algorithm model includes: a convolutional neural network model.

[0012] Further, the deep learning algorithm model includes:

[0013] Multiple convolutional layers, which are used to slide the convolutional kernel on the image to extract local features of color and texture in the skin image;

[0014] A pooling layer, which is used to downsample the features extracted by the convolutional layer to reduce the data volume;

[0015] A fully connected layer, which is used to integrate the features extracted after convolution and pooling; a softmax classifier, which is used to classify and judge the integrated and extracted features to determine whether a pressure ulcer appears and the stage of the pressure ulcer.

[0016] Further, the system also includes a thermal imaging technology module, which is used to obtain skin temperature information, and fuse the thermal imaging image with the ordinary skin image and input it into the CNN model for training and analysis to identify local temperature changes caused by pressure ulcers.

[0017] Further, the nursing plan database includes: a MySQL database and an Oracle database, which are used to automatically generate corresponding pressure ulcer nursing measures according to the pressure ulcer level analyzed by the system.

[0018] Further, the nursing plan database also includes:

[0019] A knowledge extraction module: which is used to extract key information related to pressure ulcer nursing from a large number of clinical cases; the key information covers symptom manifestations, influencing factors, past nursing measures and their effects corresponding to different pressure ulcer levels;

[0020] An entity and relationship determination module: which is used to determine the entities in the knowledge graph by using the key information in the knowledge extraction module, including pressure ulcer levels, nursing measures, nursing supplies, patient characteristics, and clarify the relationships between entities;

[0021] Knowledge Fusion Module: It is used to integrate the entities in the knowledge graph determined by the Entity and Relationship Determination Module, eliminate data redundancy and conflicts, connect the entities and relationships in a graphical way, and construct a knowledge network;

[0022] Intelligent Matching Module: It is used to convert the pressure ulcer level determined by the system into a feature vector containing key information by using the constructed knowledge network, represent the nursing plans in the database as feature vectors covering nursing measure types, applicable pressure ulcer situations, and nursing frequency information, and match the nursing plans by calculating the similarity of the two feature vectors.

[0023] Furthermore, the PDA device is equipped with a high-definition camera, a high-performance processor, a large-capacity memory and storage to meet the requirements of image acquisition, processing and data storage.

[0024] Furthermore, the PDA device conducts data interaction with the server through the network to achieve image upload, receipt of algorithm analysis results, and acquisition of nursing measure suggestions.

[0025] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0026] Accurate Pressure Ulcer Identification: This system uses deep learning algorithms, especially convolutional neural network models, and combines multi-dimensional features such as the color, texture, and temperature of skin images to accurately judge whether a pressure ulcer appears and the stage it is in, reducing human judgment errors. Through the fusion analysis of thermal imaging technology and ordinary images, it can more sensitively capture the local temperature changes caused by pressure ulcers and improve the ability to identify early pressure ulcers.

[0027] Personalized Nursing Plan Formulation: The nursing plan database is constructed based on a large number of clinical cases and medical research results. The system automatically generates personalized nursing measures according to the pressure ulcer level and can be adjusted in combination with the individual characteristics of the patient, improving the pertinence and effectiveness of nursing.

[0028] Reducing Medical Risks: The face recognition function of the PDA device ensures the accurate identification of the patient's identity, effectively avoids the wrong execution of medical orders, and reduces the risk of medical accidents.

[0029] Improving Nursing Efficiency: The system automatically analyzes pressure ulcers and generates nursing measures, reducing the time for nurses' manual operations and subjective judgments, enabling nurses to devote more energy to actual nursing work and improving the overall nursing efficiency.

[0030] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present invention. Description of the Drawings

[0031] The accompanying drawings here are incorporated into and form a part of the specification, showing embodiments in line with the present invention and, together with the specification, are used to explain the principles of the present invention.

[0032] Figure 1 It is a schematic diagram of the composition of a patient pressure ulcer recognition and nursing assistance system based on a PDA and deep learning shown according to an exemplary embodiment. Detailed implementation manners

[0033] Here, the exemplary embodiments will be described in detail, and the examples are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0034] Embodiment 1

[0035] Please refer to Figure 1 , Figure 1 It is a schematic diagram of the composition of a patient pressure ulcer recognition and nursing assistance system based on a PDA and deep learning shown according to an exemplary embodiment. The system includes:

[0036] A PDA device 1 with a dual-photographing function, used for taking face photos and uploading the face information to the system, and for taking pictures of the patient's skin and uploading them;

[0037] A server-side 2, on which a deep learning algorithm model 21 and a nursing plan database 22 are deployed;

[0038] The server-side 2 is responsible for analyzing the patient's skin image uploaded by the PDA to judge the pressure ulcer situation; the nursing plan database 22 provides corresponding scientific nursing measures according to the judgment result of the server-side 2.

[0039] In specific implementation, the deep learning algorithm model 21 includes: a convolutional neural network (CNN) model, and this model includes:

[0040] Multiple convolutional layers: By carefully designing convolutional kernels of different sizes and weights and sliding them on the patient's skin image for convolutional operations, it can accurately and efficiently extract subtle local features such as skin color and texture. Especially for the specific features at each stage of pressure ulcers, such as the color change of erythema in the initial stage and the abnormal texture caused by skin damage in the development stage, targeted extraction is achieved, providing key data support for subsequent judgment.

[0041] Pooling layer: Immediately following the convolutional layer, it downsamples the rich but redundant data extracted by the convolutional layer. While retaining key features, it effectively reduces the data volume, lowers the computational complexity, significantly improves the model's running efficiency, and ensures fast response even with limited hardware resources.

[0042] Fully connected layer: It comprehensively integrates the features after convolutional and pooling processing, breaks the local isolation between features, and constructs a complete feature expression system, enabling the model to analyze pressure ulcer-related features from a global perspective.

[0043] Softmax classifier: Based on the features integrated by the fully connected layer, it uses the softmax algorithm for classification judgment. By calculating the probability distribution of each pressure ulcer stage and normal skin state, it accurately determines whether the patient's skin has a pressure ulcer and the specific stage it is in, providing a direct and accurate basis for formulating subsequent nursing measures.

[0044] In specific implementation, in the CNN model, the fully connected layer integrates the features extracted and processed by the previous convolutional layer and pooling layer to form a feature vector of a fixed length. Suppose the feature vector output by the fully connected layer is Z = {Z1, Z2, Z3,... Z n}, where n is the dimension of the feature vector, and each Z i represents the value of the feature vector in the i-th dimension. These values comprehensively reflect various feature information related to pressure ulcers in the input image.

[0045] The Softmax classifier uses the Softmax function to convert the feature vector Z output by the fully connected layer into a probability distribution. The formula of the Softmax function is as follows:

[0046]

[0047] Among them, P(y = j|z) represents the probability that the sample belongs to the j-th class under the condition of the input feature vector Z.

[0048] N is the total number of classification categories, corresponding to the number of categories of pressure ulcer stages and normal skin states. For example, there may be normal skin, stage I pressure ulcer, stage II pressure ulcer, stage III pressure ulcer, stage IV pressure ulcer, and unstageable pressure ulcer, with a total of N = 6 categories.

[0049] Zj is the j-th element of the feature vector Z.

[0050] Through this formula, a corresponding probability value will be calculated for each category, and the sum of the probability values of all categories is 1. For example, for a specific input image, after being calculated by the Softmax function, it may obtain: the probability of normal skin is 0.1, the probability of stage I pressure ulcer is 0.7, the probability of stage II pressure ulcer is 0.1, the probability of stage III pressure ulcer is 0.05, the probability of stage IV pressure ulcer is 0.03, and the probability of unstageable pressure ulcer is 0.02.

[0051] According to the probability distribution calculated by the Softmax function, select the category with the highest probability as the final classification result. In practical applications, in order to improve the accuracy and reliability of judgment, a probability threshold can also be set. When the maximum probability value is greater than this threshold, the classification result is considered valid; if the maximum probability value is less than the threshold, it is considered that the current judgment result is unreliable and further examination or re - acquisition of images for analysis may be required. For example, set the threshold to 0.6. In the above example, the probability of stage I pressure ulcer is 0.7 which is greater than 0.6, so the judgment is valid; if the maximum probability value is 0.5, re - evaluation may be needed.

[0052] Through the above steps, the Softmax classifier can accurately determine whether the patient's skin has a pressure ulcer and the specific stage based on the features integrated by the fully - connected layer, providing a direct and accurate basis for the formulation of subsequent nursing measures.

[0053] In specific implementation, the patient pressure ulcer recognition and nursing assistance system based on PDA and deep learning further includes a thermal imaging technology module, which is used to obtain skin temperature information, and fuse the thermal imaging image with the ordinary skin image and then input it into the CNN model for training and analysis to identify local temperature changes caused by pressure ulcers.

[0054] In specific implementation, in the patient pressure ulcer recognition and nursing assistance system based on PDA and deep learning, the thermal imaging technology module can use infrared induction to obtain the temperature information of the patient's skin and generate a thermal imaging image. Since pressure ulcers can cause local skin temperature abnormalities, such as inflammation reactions may lead to temperature increases. After fusing the thermal imaging image and the ordinary skin image and inputting them into the CNN model, the model can learn the associations between skin temperature and other features (such as color, texture) in normal and abnormal states. In this way, when analyzing new images, the CNN model can integrate these multi - modal information to more accurately identify local temperature changes caused by pressure ulcers, thereby improving the accuracy and reliability of pressure ulcer recognition, especially playing an important role in early detection of pressure ulcer signs.

[0055] In specific implementation, the following are the specific implementation steps:

[0056] Training process

[0057] 1. Data collection and pre - processing

[0058] Data collection: Collect a large number of thermal imaging images of pressure ulcers at different patients, different locations, and different stages, as well as corresponding normal skin images. At the same time, record the detailed information of each case, such as the stage of pressure ulcer, the basic situation of the patient, etc.

[0059] Data annotation: Professional medical staff annotate the images to clarify the pressure ulcer area, normal skin area, and the stage of pressure ulcer. For thermal imaging images, mark the areas with abnormal local temperature.

[0060] Image preprocessing: Perform preprocessing operations on the collected thermal imaging images and normal skin images, including resizing the images, normalizing, enhancing the contrast, etc., to improve the image quality and the training effect of the model. For example, uniformly resize all images to a size of 224×224 pixels and normalize the pixel values to the range of [0,1].

[0061] 2. Image fusion

[0062] Feature-level fusion: Fuse the features of thermal imaging images and normal skin images. The feature vectors of the images can be extracted and then spliced together to form a feature vector containing more information. For example, use a pre-trained convolutional neural network (such as ResNet) to extract the features of thermal imaging images and normal skin images respectively, and then splice these two feature vectors in the channel dimension.

[0063] Decision-level fusion: Independently analyze thermal imaging images and normal skin images respectively, and then fuse the analysis results. For example, use two different CNN models to classify thermal imaging images and normal skin images respectively, and then vote or perform weighted averaging according to the classification results of the two models to obtain the final classification result.

[0064] 3. Model training

[0065] Initialize the model: Select a suitable CNN model architecture, such as VGG, Inception, or a custom CNN architecture, and randomly initialize the parameters of the model.

[0066] Define the loss function: According to the nature of the task, select a suitable loss function, such as the cross-entropy loss function. The cross-entropy loss function can measure the difference between the model prediction result and the true label, and optimize the parameters of the model by minimizing the loss function.

[0067] Training the model: Input the fused image data into the CNN model for training. During the training process, use the backpropagation algorithm to update the model's parameters to minimize the loss function. At the same time, adopt techniques such as batch normalization and dropout to prevent the model from overfitting. The training process usually requires multiple epochs, and in each epoch, all the training data is input into the model for one training.

[0068] Analysis process

[0069] 1. Image input and preprocessing

[0070] When analyzing new patient skin images, use a PDA device to collect thermal imaging images and ordinary skin images simultaneously.

[0071] Perform the same preprocessing operations on the collected images as in the training stage to ensure the quality and format of the input images are consistent.

[0072] 2. Feature extraction and fusion

[0073] Input the preprocessed thermal imaging images and ordinary skin images into the trained CNN model, and extract the features of the images through convolutional layers and pooling layers.

[0074] According to the image fusion method in the training stage, fuse the features of the thermal imaging images and ordinary skin images to obtain a feature vector containing temperature information and visual information.

[0075] 3. Classification and judgment

[0076] Input the fused feature vector into the fully connected layer and the softmax classifier to obtain the probability distribution of each pressure ulcer stage.

[0077] According to the probability distribution, select the category with the highest probability as the final judgment result of the pressure ulcer stage. For example, if the probability of being judged as stage I pressure ulcer is the highest, it is considered that the patient is currently in stage I pressure ulcer.

[0078] 4. Result output and feedback

[0079] Output the pressure ulcer stage judgment result to the PDA device for nurses to view.

[0080] According to the judgment result, extract the corresponding nursing measure suggestions from the nursing plan database 22 and display them on the PDA device together.

[0081] Record the analysis results and relevant data to provide data support for subsequent research and model optimization.

[0082] In specific implementation, the nursing plan database 22 is formulated by clinical nursing experts based on a large number of clinical cases and medical research results, covering the nursing key points and detailed operation procedures for different stages of pressure ulcers. When the system analyzes the pressure ulcer level, corresponding pressure ulcer nursing measures are automatically generated based on this database.

[0083] It should be noted that during the operation of the system, the PDA accurately captures the patient's skin image using its high-resolution camera. At the same time, the integrated thermal imaging technology module synchronously obtains the skin temperature information, and these multi-modal data are uploaded to the server in real time. The deep learning algorithm model 21 - Convolutional Neural Network (CNN) deployed on the server side 2 plays a key role. Its multiple convolutional layers slide on the image through carefully designed convolutional kernels to efficiently extract local features such as skin color and texture; the pooling layer downsamples the extracted features to reduce the data volume and computational complexity; the fully connected layer integrates the features, and finally the softmax classifier classifies and judges the integrated features to accurately determine the specific level of the patient's pressure ulcer, covering different situations such as stage I to stage IV and unstageable.

[0084] Once the system determines the pressure ulcer level, the intelligent matching mechanism of the nursing plan database 22 is immediately activated. This database is constructed using advanced knowledge graph technology, which structurally organizes the nursing knowledge summarized by clinical nursing experts based on a large number of clinical cases and cutting-edge medical research results to form a clear knowledge network. The database management system uses efficient indexing algorithms and query optimization techniques to quickly locate the nursing key points and detailed operation procedures corresponding to the current pressure ulcer level from the vast nursing plan data in a very short time.

[0085] For example, when the system determines that the patient is in stage I of pressure ulcer, the database uses natural language processing technology to output the matched nursing measures in a clear and easy-to-understand text form. The content may include using an intelligent turning mattress to assist turning regularly, using a pressure sensor to monitor the pressure of the compressed part in real time, keeping the skin clean with the help of a skin cleaning device with automatic sensing function, and using a skin protectant containing an intelligent monitoring chip to provide real-time feedback on the skin condition.

[0086] For stage II pressure ulcers, the nursing measures generated by the database may involve an intelligent wound monitoring system, which can monitor indicators such as wound exudation and healing progress in real time and automatically adjust the nursing plan according to the monitoring results; at the same time, it is recommended to use a dressing with intelligent temperature control function to promote wound healing.

[0087] For stage III pressure ulcers, the database will provide debridement operation training guidance combining virtual reality (VR) and augmented reality (AR) technologies to help nurses perform debridement more accurately; and it is recommended to use a dressing with intelligent drug delivery function to automatically release an appropriate amount of drug according to the wound infection situation.

[0088] When the patient is in stage IV pressure ulcers, the database will call on big data analysis technology, combine the treatment experience and effects of previous similar cases, and provide comprehensive treatment suggestions for the multidisciplinary collaborative team. At the same time, using telemedicine technology, it realizes remote real-time guidance of wound care by experts.

[0089] For unstageable pressure ulcers, the database analyzes the necrotic tissue and eschar at the wound site with the help of image recognition technology, and formulates a precise removal plan. And using an intelligent early warning system, it monitors the patient's vital signs and wound changes in real time, and issues an alarm in a timely manner once abnormalities occur.

[0090] In addition, the nursing plan database 22 also combines machine learning algorithms, continuously optimizes and adjusts the nursing plan according to the actual nursing effects and feedback information of the patients, realizes the dynamic update and personalized customization of nursing measures, and provides more scientific, efficient and precise pressure ulcer nursing services for patients.

[0091] In specific implementation, key information related to pressure ulcer nursing is extracted from a large number of clinical cases, such as the symptom manifestations, influencing factors, previous nursing measures and their effects corresponding to different pressure ulcer grades. For the latest medical research results, the latest nursing concepts, treatment methods and evidence-based bases are mainly extracted. For example, if it is found that a specific nutritional supplement has a positive effect on pressure ulcer healing, this information will be extracted.

[0092] Furthermore, the entities in the knowledge graph are determined, including pressure ulcer grades (such as stage I, stage II, etc.), nursing measures (such as turning over, wound cleaning, etc.), nursing supplies (such as dressings, disinfectants, etc.), patient characteristics (such as age, underlying diseases), etc. At the same time, the relationships between these entities are clarified. For example, "stage I pressure ulcers" correspond to the nursing measure of "turning over regularly", and "disinfectants" are needed for "wound cleaning", etc.

[0093] Furthermore, the extracted knowledge is integrated to eliminate redundancy and conflicts in the data. For example, the descriptions of the same nursing measure in different cases may be slightly different and need to be unified and standardized. The entities and relationships are connected in a graphical way to form a clear knowledge network. In this network, each node represents an entity, and the edge represents the relationship between entities, thus intuitively showing the whole picture of pressure ulcer nursing knowledge. Using the structure and rules of the knowledge graph, inferences are made to discover new knowledge. For example, if it is known that a certain new type of dressing has a good effect on "shallow ulcers (common symptoms of stage II pressure ulcers)", and the "current patient" is in "stage II pressure ulcers" and has "shallow ulcers", it can be inferred that this patient may be suitable for using this new type of dressing.

[0094] Furthermore, the pressure ulcer level determined by the system is converted into a series of feature vectors, which contain key information about the pressure ulcer level, such as the degree of skin damage, signs of infection, etc.

[0095] Each nursing plan in the database is also represented as a feature vector, covering information such as the type of nursing measures, applicable pressure ulcer conditions, nursing frequency, etc. By calculating the similarity between the pressure ulcer level feature vector and the nursing plan feature vector, the most suitable nursing plan is found. Common similarity calculation methods include cosine similarity, Euclidean distance, etc. For example, when using cosine similarity calculation, the closer the similarity value is to 1, the more matching they are.

[0096] Through the above knowledge graph technology to build a database, intelligent matching mechanism, as well as efficient indexing algorithms and query optimization techniques, the nursing plan database 22 can quickly and accurately locate the corresponding nursing key points and detailed operation procedures after the system determines the pressure ulcer level, providing timely and effective nursing guidance for medical staff.

[0097] In specific implementation, in the patient pressure ulcer recognition and nursing assistance system based on PDA and deep learning, the high-definition camera equipped on the PDA device is a key component for image acquisition. The high-definition camera has a high pixel and good imaging quality, and can clearly capture the subtle features of the patient's skin. For pressure ulcer recognition, this is particularly important because pressure ulcers will show various subtle changes at different stages, such as early skin redness, color change, and late skin breakage, ulcer morphology, etc.

[0098] For example, in stage I of pressure ulcers, it may only show non-blanching erythema on the skin surface. The high-definition camera can accurately capture this slight color change, providing high-quality image data for subsequent deep learning algorithm analysis. If the camera pixel is too low or the imaging quality is poor, it may lead to the loss of key features, thus affecting the accuracy of pressure ulcer recognition.

[0099] Furthermore, the high-performance processor is the core of the PDA device for initial image processing and system operation. After collecting the patient's skin image, the processor needs to perform a series of preprocessing operations on the image, such as adjusting the brightness, contrast of the image, removing noise, etc., to improve the quality and clarity of the image. At the same time, during the data interaction with the server, the processor also needs to handle tasks such as network communication, data encryption and decryption.

[0100] For example, when a PDA device collects a large amount of image data, a high-performance processor can quickly process and compress this data, reducing the data transmission volume and improving the data transmission efficiency. In addition, when running various application programs of the system, the high-performance processor can ensure the smoothness and response speed of the system, avoid stuttering, and ensure that medical staff can operate the device in a timely and accurate manner.

[0101] A large-capacity memory and storage are crucial for a PDA device. The memory is used to temporarily store running programs and data. Sufficient memory can ensure that the system can run multiple application programs simultaneously and can quickly process and exchange data. For example, when performing image acquisition and processing, the memory can store image data and intermediate results during the processing, avoiding data loss and a decrease in processing speed.

[0102] The large-capacity storage is used to store data such as the collected patient skin images and nursing records for a long time. As the system is used, the collected images and data will continuously increase. If the storage capacity is insufficient, it may lead to data loss or the inability to collect new data. At the same time, these stored data can also be used for subsequent data analysis, research, and comparison, providing more reference bases for the treatment and nursing of pressure ulcers.

[0103] The PDA device conducts data interaction with the server through a network, and this network can be Wi-Fi, Bluetooth, mobile data network (such as 4G, 5G), etc. Different network methods are applicable to different scenarios. For example, in the hospital ward, the Wi-Fi network is usually relatively stable and fast, and the PDA device can transmit data to the server through Wi-Fi; while during the mobile process, the mobile data network can ensure the real-time connection of the device.

[0104] After medical staff use the PDA device to collect the patient's skin image, they upload the image to the server through the network. During the upload process, in order to ensure the security and integrity of the data, encryption technology may be used to encrypt the image data. After receiving the uploaded image, the server will store and manage the image and input it into the deep learning algorithm model 21 for analysis.

[0105] For example, when medical staff check the pressure ulcer condition of a patient in the ward, they use the PDA device to take a picture of the patient's skin, and then click the upload button. The PDA device will automatically send the image to the server through the network. The server side 2 will number and classify the storage of the image for subsequent query and analysis.

[0106] After the deep learning algorithm model 21 on the server side 2 analyzes the uploaded images, it will obtain the recognition results of pressure ulcers, such as whether there is a pressure ulcer, the stage of the pressure ulcer, etc. The server will send these analysis results back to the PDA device through the network. After receiving the results, the PDA device will display them on the screen for medical staff to view.

[0107] For example, if the server analyzes and determines that the patient's pressure ulcer is in stage II, it will send this result to the PDA device in the form of text or a chart. Medical staff can directly see the stage information of the pressure ulcer on the PDA device, thus providing a basis for subsequent nursing work.

[0108] This system realizes accurate pressure ulcer recognition by means of deep learning algorithms and thermal imaging technology, generates personalized nursing measures relying on the nursing plan database 22, and uses face recognition to ensure accurate patient identification, comprehensively reducing medical risks and significantly improving the efficiency and quality of nursing.

[0109] It can be understood that the same or similar parts in the above embodiments can be referred to each other, and the content not detailed in some embodiments can be seen in the same or similar content of other embodiments.

[0110] It should be noted that in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "a plurality" refers to at least two.

[0111] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of code including one or more executable instructions for implementing a specific logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in the reverse order according to the functions involved, rather than in the order shown or discussed, which should be understood by those skilled in the technical field to which the embodiments of the present invention belong.

[0112] It should be understood that each part of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following well-known technologies in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0113] Those of ordinary skill in the art can understand that all or part of the steps carried out in the methods of the above embodiments can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0114] In addition, each functional unit in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. When the above integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0115] The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disc, or the like.

[0116] In the description of this specification, the descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0117] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A patient pressure ulcer identification and nursing assistance system based on PDA and deep learning, characterized in that: include: PDA device with dual camera function, used to take pictures of faces and upload them to the system to input face information, and to take pictures of patients' skin and upload them; On the server side, deep learning algorithm models and nursing plan databases are deployed; The server is responsible for analyzing the patient's skin image uploaded by the PDA to determine the condition of the pressure sore; The nursing program database provides corresponding scientific nursing measures according to the judgment result of the server.

2. The patient pressure sore identification and nursing assistance system based on PDA and deep learning according to claim 1 is characterized in that: The deep learning algorithm model includes: a convolutional neural network model.

3. The patient pressure sore identification and nursing assistance system based on PDA and deep learning according to claim 2 is characterized in that: The deep learning algorithm model includes: A plurality of convolution layers, used for extracting color and texture local features in the skin image by sliding the convolution kernel on the image; The pooling layer is used to downsample the features extracted by the convolutional layer to reduce the amount of data; The fully connected layer is used to integrate and extract the features after convolution and pooling processing; the softmax classifier is used to classify and judge the integrated and extracted features to determine whether pressure sores occur and the stage of pressure sores.

4. The patient pressure sore identification and nursing assistance system based on PDA and deep learning according to claim 2 is characterized in that: The system also includes a thermal imaging technology module for acquiring skin temperature information, and fusing the thermal imaging image with the ordinary skin image and inputting it into the CNN model for training and analysis to identify local temperature changes caused by pressure sores.

5. The patient pressure sore identification and nursing assistance system based on PDA and deep learning according to claim 1 is characterized in that: The nursing program database includes: a MySQL database and an Oracle database, which are used to automatically generate corresponding pressure sore nursing measures according to the pressure sore level analyzed by the system.

6. The patient pressure sore identification and nursing assistance system based on PDA and deep learning according to claim 1 is characterized in that: The nursing plan database also includes: Knowledge extraction module: used to extract key information related to pressure ulcer care from massive clinical cases; the key information includes symptoms, influencing factors, past care measures and their effects corresponding to different pressure ulcer levels; Entity and relationship determination module: used to determine entities in the knowledge graph using key information in the knowledge extraction module, including pressure ulcer level, nursing measures, nursing supplies, patient characteristics, and clarify the relationship between entities; Knowledge fusion module: used to integrate the entities in the knowledge graph determined in the entity and relationship determination module, eliminate data redundancy and conflict, connect entities and relationships in a graphical way, and build a knowledge network; Intelligent matching module: used to use the constructed knowledge network to convert the pressure ulcer level determined by the system into a feature vector containing key information, and to represent the nursing plan in the database as a feature vector covering the type of nursing measures, applicable pressure ulcer conditions, and nursing frequency information, and to match the nursing plan by calculating the similarity of the feature vectors of the two.

7. The patient pressure sore identification and nursing assistance system based on PDA and deep learning according to claim 1 is characterized in that: The PDA device is equipped with a high-definition camera, a high-performance processor, and large-capacity memory and storage to meet image acquisition, processing and data storage requirements.

8. The patient pressure sore identification and nursing assistance system based on PDA and deep learning according to claim 1 is characterized in that: The PDA device exchanges data with the server via the network to upload images, receive algorithm analysis results, and obtain nursing measures recommendations.

Images