Prediction method, device and equipment for wound healing process

Abstract

The embodiment of the invention discloses a prediction method, device and equipment for a wound healing process. The method comprises the following steps: acquiring a target image sequence; performing local feature extraction and global feature extraction processing on each target image in the target image sequence by using a first module of a pre-trained prediction model to obtain a local feature vector and a global feature vector; performing time sequence feature extraction processing on the local feature vector and the global feature vector based on different preset extraction periods by using a second module of the pre-trained prediction model to obtain a plurality of time sequence feature vectors; using the pre-trained prediction model to generate healing condition prediction information for the wound part based on the plurality of time sequence feature vectors; and determining a treatment scheme according to the healing condition prediction information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of life science, and in particular to a wound healing process prediction method, device and equipment. BACKGROUND

[0002] Currently, in primary hospitals, rehabilitation agencies and home care environments, wound healing condition prediction mainly relies on the experience and naked eye observation of medical staff, supplemented by ruler measurement of wound size. This wound healing condition prediction method is not only time-consuming and laborious, but also has subjective bias and cross-infection risk. Therefore, a technical solution is needed to use electronic digital data processing technology to objectively process wound images, thereby improving the prediction accuracy of wound healing conditions and improving the effect of wound care. Existing patents such as CN115426939A "Machine learning system and method for assessment, healing prediction and treatment of wounds" and CN116798024A "Burn and scald image rapid grading identification method and system based on artificial intelligence" both attempt to process and predict wound images through machine learning or artificial intelligence, reflecting the application trend of electronic digital data processing technology in this field. SUMMARY

[0003] The purpose of the embodiments of the present application is to provide a technical solution for improving the effect of wound care by improving the prediction accuracy of wound healing conditions.

[0004] To solve the above technical problems, the embodiments of the present application are implemented as follows: In a first aspect, the embodiments of the present application provide a wound healing process prediction method, which comprises: obtaining a target image sequence composed of target images containing a user's wound site and a standard reference object at different time points; using a first module of a pre-trained prediction model to perform local feature extraction and global feature extraction processing on each target image in the target image sequence to obtain a local feature vector and a global feature vector; using a second module of the pre-trained prediction model to perform time series feature extraction processing on the local feature vector and the global feature vector based on different preset extraction periods to obtain a plurality of time series feature vectors; using the pre-trained prediction model to generate healing condition prediction information for the wound site based on the plurality of time series feature vectors; determining a treatment plan for the user's wound site according to the healing condition prediction information.

[0005] In a second aspect, the embodiments of the present application provide a wound healing process prediction device, which comprises: a sequence acquisition module, configured to acquire a target image sequence, the target image sequence being composed of target images containing a wound site of a user and a standard reference object at different time points; a first extraction module, configured to perform local feature extraction and global feature extraction on each target image in the target image sequence respectively by using a first module of a pre-trained prediction model, to obtain a local feature vector and a global feature vector; a second extraction module, configured to perform time sequence feature extraction on the local feature vector and the global feature vector respectively based on different preset extraction periods by using a second module of the pre-trained prediction model, to obtain a plurality of time sequence feature vectors; an information prediction module, configured to generate healing condition prediction information for the wound site based on the plurality of time sequence feature vectors by using the pre-trained prediction model; a scheme determination module, configured to determine a treatment scheme for the wound site of the user according to the healing condition prediction information.

[0006] In a third aspect, an embodiment of the present application provides a wound healing process prediction device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, and when the computer program is executed by the processor, the steps of the wound healing process prediction method provided in the above embodiment are implemented.

[0007] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the wound healing process prediction method provided in the above embodiment are implemented.

[0008] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the wound healing process prediction method provided in the above embodiment are implemented. BRIEF DESCRIPTION OF DRAWINGS

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0010] Figure 1 It is a flowchart of the wound healing process prediction method of the present application; Figure 2 It is a flowchart of another wound healing process prediction method of the present application; Figure 3 A structural schematic diagram of a prediction device for a wound healing process according to the present application; Figure 4 A structural schematic diagram of a prediction device for a wound healing process according to the present application. DETAILED DESCRIPTION

[0011] Embodiments of the present application provide a prediction method, device and equipment for a wound healing process.

[0012] In order to make the personnel in the technical field better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by the person of ordinary skill in the art without creative labor should belong to the protection scope of the present application.

[0013] The embodiment of the present specification provides a wound healing process prediction method, device and equipment. At present, in primary hospitals, rehabilitation agencies and home care environments, wound healing prediction mainly depends on the experience and naked eye observation of medical staff, supplemented by ruler measurement of wound size. This healing prediction method is not only time-consuming and laborious, but also has subjective bias and cross-infection risk. Therefore, a technical solution is needed to improve the prediction accuracy of wound healing and improve the effect of wound care. In the solution, a target image sequence is obtained, wherein the target image sequence is composed of target images containing a user's wound site and a standard reference object at different time points. A first module of a pre-trained prediction model is used to perform local feature extraction and global feature extraction processing on each target image in the target image sequence to obtain local feature vectors and global feature vectors. A second module of the pre-trained prediction model is used to perform time series feature extraction processing on the local feature vectors and the global feature vectors based on different preset extraction periods to obtain a plurality of time series feature vectors. The pre-trained prediction model is used to generate healing condition prediction information for the wound site based on the plurality of time series feature vectors. The treatment scheme for the user's wound site is determined according to the healing condition prediction information. In this way, on the one hand, since the target image contains a standard reference object, the recognition accuracy of the wound site can be improved by comparing the wound site with the standard reference object. On the other hand, through the prediction model, time series feature extraction can be performed through multiple extraction periods based on local features and global features to obtain time series feature vectors that can represent the change rule of the wound site over time. Then, through the time series feature vectors, the healing condition of the wound site can be accurately predicted to generate a corresponding treatment scheme according to the healing condition prediction information, thereby improving the effect of wound care. Specific processing can be referred to the specific content in the following embodiments.

[0014] As Figure 1 shown, the embodiment of the present application provides a wound healing process prediction method. The execution subject of the method can be a terminal device or a server. The terminal device can be a mobile terminal device such as a mobile phone, a tablet computer, a smart watch, etc. It can also be a terminal device such as a computer. The server can be a standalone server or a server cluster composed of multiple servers. The method can specifically include the following steps: In step S102, a target image sequence is obtained.

[0015] The target image sequence can be composed of target images of the trauma site of the user and a standard reference object at different time points. The trauma site can be any part of the body, such as the arm, palm, face, or the like. The standard reference object can be any object of a size that can be obtained, such as a one-yuan coin, a band-aid, or a medical ruler.

[0016] In implementation, the server can obtain target images at multiple time points, and construct a target image sequence according to the time points of each target image. For example, medical staff or the user can take a photo of the trauma site by using a mobile phone or the like. The server can obtain photos taken of the trauma site and a standard reference object of the user every day in the past week, and determine the obtained photos as target images.

[0017] In addition, when taking the photo, the terminal device can guide the user to place the standard reference object when taking the photo.

[0018] In step S104, the first module of the pre-trained prediction model is used to perform local feature extraction and global feature extraction on each target image in the target image sequence, respectively, to obtain a local feature vector and a global feature vector.

[0019] In implementation, the pre-trained prediction model can automatically detect the contour of the standard reference object, and realize mapping conversion from image pixels to actual length by combining the known physical size, so as to accurately calculate the actual area and edge contour of the trauma site, and solve the size error caused by the lack of scale information in photo shooting.

[0020] Subsequently, the pre-trained prediction model can perform image preprocessing and wound state recognition, and output evaluation information including wound type, area, color feature, and the like.

[0021] Then, the first module of the pre-trained prediction model can perform local feature extraction and global feature extraction according to the evaluation information obtained by the image preprocessing and wound state recognition.

[0022] In step S106, the second module of the pre-trained prediction model is used to perform time sequence feature extraction on the local feature vector and the global feature vector based on different preset extraction periods, to obtain multiple time sequence feature vectors.

[0023] In implementation, the server can determine a plurality of preset extraction periods according to time points corresponding to target images included in the target image sequence. For example, assuming that the target image sequence includes target images in the past 15 days, the server can divide the 15 days into periods to obtain a plurality of different preset extraction periods. Specifically, the obtained preset extraction periods can include four extraction periods of the past 3 days, the past 5 days, the past 10 days, and 15 days.

[0024] Since wound healing is a dynamic and time-series process, it is difficult to grasp the overall situation by evaluating the wound healing state at a single time point. Therefore, the server can perform time-series feature extraction processing through a plurality of different preset extraction periods to obtain a plurality of time-series feature vectors of different time scales that can represent the change rule of the wound.

[0025] In step S108, the pre-trained prediction model is used to generate healing condition prediction information for the wound site based on the plurality of time-series feature vectors.

[0026] In implementation, the pre-trained prediction model can further include a fourth module. The server can use the fourth module to perform healing condition prediction processing based on the plurality of time-series feature vectors to obtain healing condition prediction information for the wound site.

[0027] The healing condition prediction information can include healing period prediction (e.g., X days are needed for complete wound healing), area reduction rate prediction (e.g., Y% reduction in area is predicted next week), and risk prompt for infection or healing stagnation.

[0028] Wound healing is a complex and dynamic process. Accurate prediction of its trend is crucial for timely intervention. Through time series analysis and deep learning technology, the healing state and potential risks of the wound can be accurately predicted based on continuously collected wound image data.

[0029] In step S110, a treatment plan for the wound site of the user is determined according to the healing condition prediction information.

[0030] In implementation, the server can use a large language model to determine a treatment plan for the wound site of the user according to the healing condition prediction information. Alternatively, the server can determine a healing condition score for the wound site according to the healing condition prediction information, and determine a treatment plan corresponding to the healing condition score according to a preset correspondence between scores and plans. Alternatively, the server can send the healing condition prediction information to a preset medical staff and receive a treatment plan fed back by the preset medical staff. In addition, there can be various methods for determining a treatment plan, which can be selected according to different actual application scenarios. The embodiments of the present specification do not limit the determination method.

[0031] In addition, in primary hospitals, many doctors are not wound care specialists. A system for generating healing condition prediction information and corresponding treatment plans (i.e., an intelligent monitoring and decision support system for wound healing management throughout the whole process) can be provided as an "AI assistant" to provide a second opinion. When a patient visits a doctor, the doctor can take a picture of the wound using a tablet or a mobile phone, and the system can quickly give a judgment on the category and severity of the wound and prompt whether there are signs of infection or the need for referral to a specialist hospital. Through AI standardized scoring, primary doctors can more confidently develop treatment plans or decide on referral, which can improve the accuracy and safety of primary wound care. In rehabilitation institutions (such as chronic wound clinics, rehabilitation care centers, etc.), nursing staff can use the above system to regularly monitor long-term chronic wound patients. The healing condition prediction information generated by the system can help determine the effectiveness of the current treatment plan: if the healing rate is found to be lower than expected, the doctor can adjust the treatment plan (such as changing the drug or debridement method); if the prediction shows that the wound is still difficult to heal in two weeks, more aggressive intervention measures can be considered in advance.

[0032] This data-driven decision support helps to manage chronic wound patients individually and avoid the risk of complications caused by long-term treatment of wounds. In the home care scenario, the above system can also provide services to users through a mobile App. For example, patients or family members can take pictures of the wound every day and upload them, and the prediction model can automatically analyze the color changes of the wound to determine the healing trend. If the system detects an anomaly (for example, the wound suddenly becomes red and swollen, which may be a sign of infection), it will immediately issue a warning through the App and recommend medical review, greatly reducing the time for medical staff to manually interpret. The system can introduce more wound types during model training, and through the healing prediction function, the system can become a general wound remote management tool. Through the mobile App, patients can also easily view their healing scores and progress charts, which not only improves their enthusiasm for participating in self-care, but also allows professional medical staff to keep track of patient conditions at all times, enabling true remote medical care and continuous care.

[0033] Based on the above process, an intelligent monitoring and decision support system for wound healing management throughout the whole process can be constructed. In terms of deployment and user experience of the system, HealScope-AI focuses on the integration of "high precision and low threshold". Medical staff or patients can upload wound images by taking pictures with a mobile phone, and the system automatically completes image processing, recognition analysis, trend prediction, and visualization display. Users can view healing indices, tissue structure maps, red and swollen area heat maps, report exports, and can also switch to the patient end to view graphical scores, expression prompts, and operation suggestions, truly realizing "professional-patient" two-way adaptation.

[0034] In addition, the system can adopt an ONNX export mechanism in model deployment, is compatible with multiple edge devices and platforms, can be embedded in basic medical devices, clinic terminals, or deployed on a hospital dedicated server for remote management, greatly enhancing its engineering capability and clinical promotion potential. The system can also support online model updating and automatic learning, continuously optimizing recognition accuracy and prediction ability as data accumulates, and has good self-evolution ability.

[0035] In summary, HealScope-AI, with its multi-dimensional innovations in image recognition, time series modeling, intelligent intervention suggestion generation, and interactive visualization, has built a complete and intelligent wound assessment-prediction-intervention integrated platform. It not only fills the gap between traditional system static assessment and suggestion output, but also greatly expands the practical application boundary of AI in skin wound management through general models and high usability design.

[0036] The embodiment of the present application provides a wound healing process prediction method, acquires a target image sequence, wherein the target image sequence is composed of target images containing a wound site of a user and a standard reference object at different time points, utilizes a first module of a pre-trained prediction model to perform local feature extraction and global feature extraction processing on each target image in the target image sequence respectively to obtain a local feature vector and a global feature vector, utilizes a second module of the pre-trained prediction model to perform time series feature extraction processing on the local feature vector and the global feature vector respectively based on different preset extraction periods to obtain a plurality of time series feature vectors, utilizes the pre-trained prediction model to generate healing condition prediction information for the wound site based on the plurality of time series feature vectors, and determines a treatment scheme for the wound site of the user according to the healing condition prediction information. In this way, on the one hand, since the target image contains the standard reference object, the recognition accuracy of the wound site can be improved through comparison between the wound site and the standard reference object, and on the other hand, through the prediction model, time series feature extraction can be performed through a plurality of extraction periods on the basis of taking into account local features and global features to obtain time series feature vectors capable of representing the change rule of the wound site over time, and then through the time series feature vectors, the healing condition of the wound site can be accurately predicted to generate a corresponding treatment scheme according to the healing condition prediction information, thereby improving the wound care effect.

[0037] In actual application, the first module can include a first submodule constructed by a residual neural network and a second submodule constructed by a visual Transformer, the first submodule is used for performing local feature extraction on each target image in the target image sequence to obtain the local feature vector, and the second submodule is used for performing local feature extraction on each target image in the target image sequence to obtain the global feature vector.

[0038] Convolutional Neural Networks (CNNs) have become the standard tool in medical image analysis due to their exceptional performance in image feature extraction. The system uses a pre-trained Residual Neural Network (ResNet) to extract local detail features from the target image through multiple convolution and pooling operations, resulting in a local feature vector.

[0039] While CNNs excel in local feature extraction, their ability to capture overall image information is relatively limited. To compensate for this deficiency, the system introduces a visual transformer (ViT) to construct the second module. ViT can analyze the global context information of the wound image through self-attention mechanisms, which can be used to assess the condition of the skin around the wound.

[0040] By combining the local feature extraction capabilities of CNNs with the global information analysis capabilities of ViT, a comprehensive portrayal of wound features can be achieved, enhancing the system's adaptability to complex wound scenarios.

[0041] In practical applications, the local feature vector is used to represent the clarity of the wound edge, the texture changes of the tissue, and the distribution of the color of the wound site, while the global feature vector is used to represent the condition of the skin around the wound site.

[0042] For example, by analyzing the blurring degree or color abnormalities (such as redness or blackening) of the wound edge, the system can identify early signs of infection; by observing texture changes, the necrosis degree of wound tissue or the regeneration of granulation tissue can be determined. The extraction of these local features plays a crucial role in distinguishing different types of wounds (such as surgical wounds, diabetic foot ulcers) and assessing the severity of the wound.

[0043] For example, when there are signs of inflammation spreading around the wound (such as redness or swelling of the skin), ViT can capture these subtle but important changes.

[0044] In practical applications, the second module includes a third sub-module constructed by a convolutional neural network and a fourth sub-module constructed by a long short-term memory network. The multiple time series feature vectors include time series feature vectors obtained by the third sub-module performing time series feature extraction processing on the local feature vector and the global feature vector respectively according to a preset first extraction period, and time series feature vectors obtained by the fourth sub-module performing time series feature extraction processing on the local feature vector and the global feature vector respectively according to a preset second extraction period. The first extraction period is less than the second extraction period.

[0045] Since the wound healing process involves changes at different time scales, such as short-term changes in color or exudate on a daily basis, and long-term trends in area reduction on a weekly basis, etc. To fully capture these dynamic features, a second module containing a multi-scale convolutional layer structure can be designed. Among them, the short-scale convolutional layer can focus on extracting short-term change features, such as subtle improvements in the wound surface (such as lighter tissue color or reduced exudate); long-scale convolutional layers can be responsible for analyzing long-term trends, such as the overall progress of wound closure and boundary changes. This multi-scale feature extraction method can ensure that the system has a comprehensive understanding of the wound healing process and can identify subtle changes in the healing process, such as early signs of infection or potential risks of healing stagnation, etc.

[0046] In practical applications, the time series feature vector extracted by the third sub-module can be used to represent the spatial structure features of the wound site, and the time series feature vector extracted by the fourth sub-module can be used to represent the changes in wound features of the wound site.

[0047] Since the ConvLSTM combines the advantages of CNN and LSTM, the fourth sub-module can be constructed based on ConvLSTM. Among them, CNN can be responsible for processing the spatial structure features of the wound image (i.e. the target image), such as the shape of the wound area and the distribution of the tissue; and LSTM can be used to capture the evolution law of these features in the time dimension, such as the daily reduction of wound area or the gradual process of tissue regeneration. Through ConvLSTM, the system can learn the wound closure speed, tissue regeneration pattern, and potential development trend of complications, and output specific healing condition prediction information accordingly, such as the number of days required for complete healing, the amplitude of area reduction in the future week, and the risk of infection or healing stagnation, etc.

[0048] In terms of wound healing trend prediction, the classic ConvLSTM structure is fully referenced and optimized, and a multi-scale convolution feature extraction strategy is combined to form a time series modeling module specifically for dynamic changes in the wound surface. ConvLSTM has been widely used in video analysis and spatial sequence modeling, and it can capture both spatial features and temporal evolution patterns of images. On this basis, combined with the phased feature changes in the wound healing process, a "multi-scale time series feature extraction module" is innovatively designed to extract short-term changes (such as redness and swelling, exudate increase and decrease within 12 days, etc.) and medium and long-term trends (such as closure rate, edge tissue growth within 714 days, etc.), and through fusion modeling, the accuracy and robustness of healing state prediction are improved.

[0049] In the experiment, the module showed good generalization ability and stability in the multi-class wound trend prediction task, with a healing time prediction accuracy of over 85%, especially in identifying high-risk healing abnormalities such as infection stagnation, which has important clinical reference value. By combining ConvLSTM with a time sequence structure optimized for wound data characteristics, HealScope-AI not only breaks through the limitations of traditional wound assessment systems, which are static in nature, but also provides more forward-looking assistance for intelligent wound monitoring.

[0050] In actual application, the second module of the pre-trained prediction model in step S106 performs time sequence feature extraction processing on the local feature vector and the global feature vector based on different preset extraction periods. The processing method for obtaining multiple time sequence feature vectors can be various, and an optional processing method is provided as follows. Figure 2 As shown in the figure, the processing can specifically include the following steps S1062-S1066.

[0051] In step S1062, the third module of the pre-trained prediction model determines the weight values corresponding to the local feature vector and the global feature vector according to the image features of each target image in the target image sequence.

[0052] In implementation, in order to further optimize the performance of image recognition, an adaptive local-global attention module, i.e., the third module of the prediction model, can be constructed. The third module can dynamically adjust the attention degree of the model to local features and global features according to the image features of the wound image.

[0053] For example, for a small wound, the system will pay more attention to local details, such as subtle changes in the wound edge; while for a complex wound with obvious inflammation spread, the system will preferentially analyze global features, such as the color distribution of the surrounding skin and the overall inflammation range.

[0054] In addition, the third module of the adaptive mechanism can be implemented through a lightweight network structure, which can enable the system to automatically select to focus on local details or global context information according to the complexity of the specific wound type (such as abrasions, burns, diabetic foot, etc.), thereby not only improving the accuracy of recognition, but also ensuring the robustness of the system, especially in scenes with large differences in light conditions or patient skin color.

[0055] In step S1064, the target feature vector is determined according to the local feature vector and the corresponding weight value, and the global feature vector and the corresponding weight value.

[0056] In implementation, the server can perform weighted processing on the local feature vector and the corresponding weight value, and the global feature vector and the corresponding weight value to obtain the target feature vector.

[0057] The target feature vector can be used to represent the wound type (e.g., surgical incision, abrasion, diabetic foot ulcer, pressure sore, etc.), severity classification (mild, moderate, severe), precise wound area segmentation result, and size measurement data of the wound site. Through the first module, the second module, and the third module constructed based on the deep learning algorithm described above, the wound area of the wound site and its changes can be automatically calculated, which can provide high-quality, quantitative basic data support for subsequent trend prediction and intervention decision-making.

[0058] In step S1066, the second module of the pre-trained prediction model is used to perform time series feature extraction processing on the target feature vector based on different preset extraction periods, to obtain a plurality of time series feature vectors.

[0059] In implementation, since wound healing is a dynamic time series process, the second module can be constructed by a deep learning framework combining multi-scale time series feature extraction and ConvLSTM. By the second module, subtle changes in the wound healing process can be captured more accurately, and scale features corresponding to multiple extraction periods (such as short-term (e.g., daily changes in tissue color, exudate increase / decrease, etc.) and long-term (e.g., weekly changes in wound area and boundary) can be extracted from the time series wound images (i.e., target images contained in the target image sequence) by the third submodule constructed by a multi-scale convolution feature extraction layer. These multi-scale features can then be input into the fourth submodule constructed by ConvLSTM to achieve accurate prediction of wound healing trends by the fourth submodule. The second model integrates the advantages of CNN and LSTM, and can effectively capture the relationship between spatial structure and time series features. Through the second module, the system can automatically learn and predict time series feature vectors that can be used to represent the change pattern of the wound over time, such as wound area reduction speed, edge tissue regeneration, and possible signs of infection, etc.

[0060] Innovatively, two mainstream model architectures in the field of computer vision, Vision Transformer (ViT) and Residual Neural Network (ResNet), are deeply integrated to build an image recognition module with local-global perception capability. In the case where most existing researches still mainly use single CNN architecture, this system first introduces ViT to process skin wound images to capture global feature information such as overall distribution, symmetry of edge redness area, and exudation and diffusion of surrounding skin, etc. ResNet, on the other hand, precisely focuses on local details in the image such as tissue texture, tiny bleeding points, and color and distribution state of granulation tissue. The two are integrated through a "local-global adaptive attention mechanism" to dynamically adjust attention allocation according to the complexity of the wound, effectively avoiding redundant stacking or missing of feature information, and significantly improving the adaptability and accuracy of the model in identifying different types of wounds and tissue classification. Experimental results show that the average accuracy of the module in image classification tasks reaches 98.2%, and the error in area measurement is controlled within ±4.5%, both of which are better than existing open-source models.

[0061] In actual application, the prediction model can also be trained, and the processing mode of training the prediction model can be various. The following provides an optional processing mode, as shown in Figure 2 The specific processing can include the following steps S202-S208.

[0062] In step S202, a sample image sequence is obtained.

[0063] In step S204, the sample images contained in the sample image sequence are subjected to data enhancement processing according to a preset data enhancement rule, to obtain an enhanced sample image sequence.

[0064] The preset data enhancement rule at least includes image rotation, color jitter, brightness adjustment, and contrast adjustment.

[0065] In implementation, since the quality and features of wound images differ greatly due to factors such as lighting conditions, patient skin color, and shooting angle, this poses a challenge to the training and performance of the model. Therefore, the diversity of training data can be increased through data enhancement technology, i.e., the server can perform data enhancement processing on the sample image sequence through image rotation, color jitter, brightness adjustment, and contrast adjustment, etc. data enhancement rules to obtain an enhanced sample image sequence.

[0066] In step S206, a first prediction model is obtained.

[0067] The first prediction model can be obtained according to first sample data, and the data amount of the first sample data is greater than that of the sample image sequence. The first sample data can be a general data set (such as ImageNet, etc.).

[0068] In step S208, the first prediction model is trained according to the enhanced sample image sequence to obtain the pre-trained prediction model.

[0069] In implementation, the server can use the first prediction model pre-trained on a large general dataset (such as ImageNet) through transfer learning, and then fine-tune it for specific features of wound images to obtain the pre-trained prediction model, thereby overcoming the problem of data scarcity in the medical field. The generalization ability of the model can ensure its stable performance in different scenarios.

[0070] In this way, by using data augmentation, transfer learning and other technical means, the model can be pre-trained using public databases and then fine-tuned on data in specific application scenarios, ensuring that the model can still work effectively when the data volume is insufficient. Through this multi-scale time series analysis method, the prediction accuracy and sensitivity can be effectively improved, providing more reliable decision support for clinical wound management.

[0071] In addition, to further enhance the generalization ability and prediction accuracy of the model, the system can also incorporate a variety of wound cases, including normal healing and delayed healing samples, during the training process, allowing the model to accurately distinguish between different healing trajectories and provide early warning of abnormal conditions.

[0072] HealScope-AI focuses on "generalization" and "scenario adaptability" in system construction. Unlike many models that only focus on diabetic foot, pressure sores and other single chronic wounds, the system's architecture is designed from the outset to address postoperative wounds, trauma, burns, animal trauma and other types of wounds. Through data augmentation and cross-wound knowledge transfer techniques, the model's ability to recognize rare types is improved, and a general model framework with wide applicability is constructed. At the same time, the system also supports task environment switching interfaces and interactive modes, adapting to various application scenarios such as primary care, rehabilitation institutions, home care and disaster relief.

[0073] In actual application, the processing mode of determining the treatment scheme for the user's wound site according to the healing condition prediction information in step S110 can be various, and the following provides an optional processing mode, as shown in Figure 2 The specific processing can include the following steps S1102-S1106.

[0074] In step S1102, the wound care rules corresponding to the user's wound site are determined according to the healing condition prediction information.

[0075] In implementation, the server can build a rule engine that can generate intervention recommendations based on internationally recognized wound care clinical guidelines. For example, when the system detects signs of infection (such as redness or abnormal exudation), the rule engine may recommend wound care rules using antimicrobial dressings or topical antibiotics; if healing is slow, it may suggest increasing the frequency of dressing changes or using a wound healing-promoting dressing (such as a hydrocolloid dressing). This rule-based approach can ensure the medical rationality and clinical operability of the recommendations, in line with current best practice standards.

[0076] In step S1104, the multi-modal user data corresponding to the user is obtained.

[0077] Among them, the multi-modal data can include individual characteristic data, medical history information, laboratory examination results and lifestyle data of the user.

[0078] In implementation, in addition to image data, the server can also obtain additional data of the user, including individual characteristics (such as age, comorbidity, etc.), medical history information (such as history of diabetes, history of hypertension, etc.), laboratory examination results (such as blood glucose level, inflammation index, etc.) and lifestyle data (such as whether to smoke, activity level, etc.). By comprehensively analyzing these multi-modal user data, the system can identify key factors affecting wound healing, such as malnutrition may cause delayed healing, and smoking may increase the risk of infection, so as to generate more accurate recommendations. For example, if the patient's blood glucose control is poor, the system will preferentially recommend blood glucose management strategies; if the risk of infection is high, it is recommended to use antimicrobial drugs. This multi-modal approach fully embodies the scientific principle that wound healing is affected by multiple systemic factors.

[0079] In step S1106, a pre-trained scheme determination model is used to determine a treatment scheme for the wound site of the user according to the wound care rules and the multi-modal user data.

[0080] Among them, the scheme determination model is a model trained by reinforcement learning.

[0081] In implementation, in order to further improve the flexibility and individualization level of decision-making, reinforcement learning technology can be introduced. Reinforcement learning can simulate the trial-and-error process, continuously optimize the treatment path, adjust the wound care rules according to the multi-modal data of the user, and obtain the treatment scheme. For example, for diabetic patients, the system may preferentially recommend blood glucose control aids (such as dietary adjustments) to promote healing; for patients with malnutrition, it may suggest supplementing protein or vitamins. In this process, the reinforcement learning agent learns from historical treatment effect data, evaluates the efficacy of different interventions, and dynamically adjusts the strategy to maximize healing speed and minimize complication risk.

[0082] Based on the healing condition prediction information obtained through image recognition and healing trend prediction, the introduction of personalized intervention suggestion generation mechanism can realize the function upgrade from simple monitoring to active intervention. By combining rule engine and reinforcement learning technology, the specific wound state of the patient, the predicted healing trend, and the patient's background data (such as age, chronic disease history, living habits, etc.) can be fully utilized to dynamically generate precise personalized treatment plans. The wound management provides a scientific and dynamic decision-making framework, which can ensure the personalization, precision and clinical effectiveness of the intervention plan.

[0083] In specific implementation, first, a preliminary judgment is made by the built-in rule engine, which is based on the latest clinical guidelines and professional treatment consensus to ensure that the generated wound care rules have clinical effectiveness and medical rationality. The rule engine can automatically identify specific wound conditions (such as high risk of infection and obvious delay in healing), and preliminarily recommend appropriate intervention measures (i.e. wound care rules), such as increasing wound debridement frequency, using specific antibacterial or wound healing promoting dressings, etc. Subsequently, the reinforcement learning algorithm further optimizes these suggestions by learning historical treatment effect data to continuously evaluate and optimize the effectiveness of intervention strategies. Reinforcement learning can identify and recommend specific measures that can best improve healing conditions, achieving gradual precision and personalization of the plan.

[0084] In addition, through real-time collection of user feedback (such as treatment effect data and patient satisfaction), the model continuously learns and adjusts the suggestion strategy in real time to maximize treatment effectiveness. Ultimately, specific and actionable intervention suggestions (i.e. treatment plans) can be output, including recommended debridement frequency, dressing type, drug treatment plan, and other auxiliary nursing measures. By achieving personalized and precise treatment guidance, wound care quality can be effectively improved, healing time can be shortened, complications can be reduced, and patient rehabilitation experience can be improved.

[0085] In terms of function design, the system breaks through the limitations of traditional AI evaluation tools that only recognize but do not make decisions, and innovatively introduces a "personalized intervention suggestion generation module". Based on reinforcement learning strategy and clinical rule engine, combined with individualized features of patients (such as medical history, previous medication, immune status) and wound characteristics (such as infection risk level, tissue structure proportion, exudate category, etc.), the module outputs actionable care suggestions, such as "recommend changing dressing 2 times a day, recommend using silver ion antibacterial dressing" and "need specialist consultation", achieving intelligent upgrade from "passive evaluation" to "active suggestion". This function not only improves the clinical guidance value of the system, but also greatly reduces the treatment differences caused by lack of experience of primary medical staff, promoting the standardization and scientization of wound management decisions.

[0086] In practical applications, the healing condition prediction information can include a healing trend curve. The processing manner of generating the healing condition prediction information for the wound site based on the plurality of time sequence feature vectors in step S108 can be various. An optional processing manner is provided as follows. Figure 3 As shown, the specific processing can include the following steps S1082-S1084.

[0087] In step S1082, the pre-trained prediction model is used to determine the healing condition corresponding to each prediction time point based on the time point corresponding to each target image in the target image sequence and the plurality of time sequence feature vectors.

[0088] In step S1084, the healing trend curve is generated according to the healing condition corresponding to each prediction time point.

[0089] In implementation, since the collection time interval of the target image can not be fixed, this poses an additional challenge to time series prediction. For example, the patient can not be able to upload images on some days, resulting in data gaps. To solve this problem, a time-aware mechanism can be introduced to consider the difference in collection time interval through Time Embedding technology. This method can enable the system to adapt to changes in observation gaps, ensuring the accuracy and reliability of the prediction results. For example, the system can infer potential changes in the missing time period based on historical data, thereby generating a continuous healing trend curve.

[0090] Through the above dynamic modeling technology, personalized treatment reference data can be provided for medical staff, and abnormal conditions such as infection risk or delayed healing can be warned in advance.

[0091] Since the intelligent monitoring and decision support system for wound healing whole-process management combines the medical scientific principles of wound healing, the clinical relevance of the system can be ensured. This is embodied in the following aspects: Identification of wound healing stages: wound healing is divided into four key stages: hemostasis, inflammation, proliferation, and remodeling. The system identifies visual markers of each stage through image analysis by the first module, such as exudate characteristics in the inflammation stage, granulation tissue formation in the proliferation stage, and scar maturation in the remodeling stage, ensuring accurate monitoring of the healing process.

[0092] Visual signs of complications: the system incorporates visual signs of wound complications, such as redness caused by infection, tissue necrosis or irregular edges caused by delayed closure. These signs are automatically identified by the AI model, providing a basis for early intervention.

[0093] Reference to Standard Care Practices: The system's recommendations reference standard wound care practices, including debridement techniques (removal of necrotic or infected tissue), principles of dressing selection (choosing appropriate dressings such as alginate or hydrocolloid based on wound moisture and exudate levels), and infection control strategies (monitoring for signs of infection and recommending antimicrobial therapy as necessary). These practices are based on clinical guidelines, ensuring the scientific rigor and operational feasibility of the recommended protocols.

[0094] Through deep integration with medical science, the clinical applicability of the technology output can be ensured, providing reliable decision support for healthcare professionals and patients.

[0095] In addition, to verify the scientificity and practicality of the system, the following measures can be taken to ensure its credibility and effectiveness in reality: Clinical Trials: By comparing with expert assessment results, test the recognition accuracy, prediction reliability and clinical effectiveness of intervention recommendations of the system. Indicators can include the accuracy of wound type recognition, the sensitivity and specificity of healing trend prediction, and the improvement effect of intervention measures on healing time and infection rate.

[0096] Outcome Assessment: Verify the clinical benefits of the system through randomized controlled trials (RCT), such as shortening healing time, reducing infection incidence, and reducing medical costs. These outcome assessments will provide scientific basis for the popularization and application of the system.

[0097] Interdisciplinary Collaboration: Work closely with medical experts, technical teams and clinical institutions to ensure seamless integration of technology implementation and clinical workflow. For example, work with primary hospitals to verify the practicality of the system, and work with rehabilitation institutions to optimize chronic wound management solutions.

[0098] Through the above verification process, the scientific rigor and clinical feasibility of the system can be ensured, setting a model for the application of medical AI.

[0099] To ensure the quality of system training and the wide applicability of actual application, a multi-source heterogeneous wound image database with complete structure, diverse types and standard annotation can be constructed. This database can cover different wound types, healing stages and use scenarios, with good representativeness, generalization ability and clinical transferability. The database mainly includes the following three types of data sources: 1. International public wound image database The system makes full use of existing standardized public database resources, including: Medetec Wound Database, Foot Ulcer Challenge Dataset (FUCD), CURA Wound Dataset. These datasets cover common chronic and acute wounds such as diabetic foot ulcers, pressure sores, postoperative incisions, and provide clear image quality and corresponding professional annotation information, suitable for image classification, region segmentation, healing trend modeling, etc. As an important basis for training models, the use of public datasets enhances the initial recognition ability and cross-domain versatility of the model.

[0100] 2. Real case images collected in cooperation with clinical institutions A real wound image collection cooperation mechanism can be established. The images obtained are desensitized and collected and used under the premise of compliance with ethical review. Data content can cover different wound types such as postoperative wounds, burns, lacerations, etc. Some also contain patient basic information, medical history, treatment process records and wound development stage labels.

[0101] Clinical image data are annotated by multiple wound repair specialists to ensure the accuracy and professionalism of the label information, mainly for: local adaptation and optimization of the model; generation and training of personalized intervention recommendations; modeling and prediction of multi-stage wound healing dynamic changes.

[0102] 3. Model-assisted generation and simulation of extended data For tasks such as small area burns, complex wounds or extreme infection wounds, which are rare in clinical practice or have a severe shortage of sample quantities, various enhancement mechanisms can be introduced for data expansion, including: image enhancement (rotation, scale, noise, crop, etc.); medical image style transfer (StyleGAN, CycleGAN); multi-modal joint construction and AI-assisted simulation generation.

[0103] Combined with the clinical expert review mechanism, the extended images are subjected to multi-stage quality screening to ensure that they meet the modeling requirements in terms of image authenticity, structural logic and medical interpretation consistency, especially for trend prediction and small sample learning tasks.

[0104] Based on artificial intelligence (AI) image recognition and generation technology, the above-mentioned intelligent system integrating wound image recognition, healing trend prediction and intervention decision support is constructed, which can include four core functional modules: (1) wound image intelligent recognition module, which uses deep learning technology to automatically analyze wound images, identify wound types, measure wound area and tissue condition; (2) healing trend intelligent prediction module, based on multi-scale time series feature extraction technology and ConvLSTM algorithm, accurately predicts the future healing process of the wound; (3) personalized intervention suggestion generation module, combined with rule engine and reinforcement learning, generates targeted treatment plan according to wound state, healing trend and patient background data; (4) scoring and visual feedback module, which converts the above analysis results into intuitive and easy-to-read scoring index and visual interface to support clinical decision-making and patient self-management.

[0105] Among them, the scoring and visual feedback module can display all analysis results in graphical form on the user interface. Including: wound area-time change curve (unit: cm 2 ), healing speed trend, risk warning mark, system generated intervention suggestions, etc., to help doctors and patients timely grasp the wound rehabilitation dynamics, enhance compliance and remote management ability. Through the closed-loop process of "image analysis + trend prediction + intelligent intervention + visual feedback", a precise, efficient and practical full-process one-stop AI wound management platform is constructed, which is suitable for primary medical care, rehabilitation nursing and home scenarios, and provides a new technical solution for wound care.

[0106] With the objective assessment of the current wound status and the prediction of future trends, the scoring and visual feedback module is responsible for translating the results into a form that is easy for the user to understand and operate. On the one hand, the system will generate a wound healing score for each wound, taking into account the size and depth of the wound, the health of the tissue, and the healing stage. The score is designed as a grade classification, for example, A indicates that it is expected to heal completely, and the lower the score, the more serious the wound problem or the slower the healing. The calculation of the score will refer to the commonly used wound assessment scale in clinical practice, and will be automatically generated in combination with the quantitative indicators identified by AI, to ensure that the score has medical significance and comparability. On the other hand, the system provides rich visual feedback. For example, the analysis and annotation map of the wound photo will be displayed on the application interface: the wound area will be highlighted and outlined, and the system will mark the areas of granulation tissue, new skin, and possible necrotic tissue with different colors, so that the user can intuitively understand the composition of the wound. The system will also draw a healing progress curve, such as automatically drawing a graph of the change of wound area over time according to the previous follow-up, and marking the predicted future trend on the curve with different colors. For abnormal situations that may occur (such as predicting a risk of infection in the future), the interface will provide warning signs or information prompts. All of these feedbacks are helpful for the user to understand the evaluation given by AI: not just an abstract judgment, but also data and graphics that can be traced. This intuitive presentation form meets the different needs of medical staff and patients - professionals can record cases and judge the effectiveness accordingly, and patients can also see more clearly whether their wounds are improving or worsening, thereby improving compliance. It is worth mentioning that the scoring and visual feedback module will also support report export function for medical institutions to archive or for experts to use for remote consultation.

[0107] Real-time monitoring of wound status, risk warning and personalized guidance can be achieved, and a quantitative evaluation report and treatment recommendations can be output. On the one hand, through analysis of wound photos, the system can achieve automatic identification and quantitative evaluation of wound type, area size, healing stage, etc., reducing dependence on human experience; on the other hand, using deep learning algorithms to extract features and time series analysis of continuously collected wound images, the current healing degree can be accurately judged, the future development trend can be predicted, and the risk of wound deterioration or slow healing can be predicted in advance, and reinforcement learning can be used to integrate multiple modalities of patient data to intelligently generate personalized intervention recommendations. In addition, the system will output unified, quantitative and interpretable intelligent auxiliary evaluation results, such as wound healing index, wound area change curve, etc., to help doctors and patients intuitively understand the current status and future trend of the wound. By integrating the above functions into an integrated platform, the present application is expected to improve the efficiency and quality of wound care: in primary hospitals, reduce the assessment pressure when there is a lack of professional wound department; in rehabilitation institutions, provide scientific basis to adjust the rehabilitation plan; in home care scenarios, provide technical support for patient self-management and telemedicine. Ultimately, the system is expected to reduce the incidence of wound complications, speed up the healing process, reduce medical costs, and improve patient quality of life.

[0108] The present application not only makes up for the limitations of traditional reliance on naked eye and experience diagnosis, but also can be effectively applied in various scenarios, producing tangible social and economic value.

[0109] First, at the level of hospital clinical nursing, the present application can be used for postoperative follow-up and chronic wound management in multiple departments such as surgery, burn department, dermatology department and rehabilitation department. At present, doctors or nurses usually rely on visual inspection and simple measurement (such as ruler measurement of wound size) for wound monitoring, which is highly subjective and time-consuming. However, the present application can automatically evaluate multiple indicators such as wound area, exudation degree, and tissue necrosis range within seconds through AI recognition and time series prediction models, and intuitively predict the future healing progress. For example, the system can predict whether the wound will heal smoothly in a few days or weeks, or whether there is a potential risk of infection or scar hyperplasia. For inpatient wards that require large-scale management, such intelligent evaluation can significantly improve clinical nursing efficiency, reduce repetitive labor for medical staff, and allow them to devote more time to high-value nursing interventions or communication.

[0110] Secondly, the value of the present application is particularly prominent in primary medical institutions. As we all know, community health service centers, township health centers and other primary medical units generally have the problems of shortage of professional wound care personnel, insufficient medical equipment and technical support. When patients have wounds or need to change dressings and follow-up after surgery, primary medical staff often cannot judge the risk of the wound in time due to lack of experience, thus delaying treatment. The AI evaluation system of the present application is deployed in a light-weight manner on a mobile phone or tablet. Only by taking a photo of the wound and inputting basic information, the system can quickly give a wound classification, healing trend and intervention suggestion, helping primary doctors to quickly make a judgment and decide whether to need referral. This not only relieves the nursing pressure of primary care, but also effectively reduces the patient's running between hospitals and saves medical resources.

[0111] In rehabilitation and nursing institutions such as chronic wound clinics, nursing homes or rehabilitation hospitals, the present application can also provide great help. Chronic wounds often require repeated care and observation for several weeks or even months. Under the traditional way, each detection requires manpower and time investment. The present application provides a quantitative healing curve and risk warning mechanism for the nursing team through regular photography and automatic analysis. Once the system detects that the healing rate is stagnant or the infection rate is rising, it will immediately alert medical staff and provide intervention suggestions, such as changing dressings, increasing dressing frequency or introducing new drug treatment programs. This can significantly reduce the deterioration of wounds caused by judgment errors or observation delays, and achieve better rehabilitation management and rational allocation of resources.

[0112] Home care is another important application scenario of the present application. Under the existing medical mode, many postoperative patients or chronic disease patients need to return to the hospital regularly for wound review after discharge. However, for patients with limited mobility or poor transportation, frequent visits to the hospital not only increases the economic burden, but also wastes medical resources. The present application provides a mobile application. Patients or family members can take photos of the wound at home every day and record the basic situation. The system will automatically evaluate the current situation, predict future changes and give daily care suggestions, such as whether to suggest compression bandaging, when to see a doctor, etc. Once acute deterioration is detected, the system can also send an early warning to the patient to seek professional help immediately. By achieving self-management and intelligent wound management in the home environment, the entire society can reduce the occupation of a large number of bed and outpatient resources, and significantly improve the comfort and compliance of patients.

[0113] Disaster emergency and special environment scenarios are also promising. In natural disasters such as earthquakes and floods, or in war field rescue, there are often a large number of wounded in a short period of time, and the medical resources are severely insufficient. With the aid of the intelligent recognition and prediction function carried on the portable device in the present application, rescue personnel can quickly evaluate the severity of the wound and the risk of infection of the wounded in the field environment, assist in grading diagnosis and emergency treatment. Doctors can also remotely view wound pictures and combine the automatic evaluation given by the system to decide whether to air transport or transfer the patient, shorten the valuable medical decision-making time, and improve the overall efficiency and success rate of disaster emergency rescue.

[0114] Animal wound management is also a promising extension direction. Pet hospitals or large-scale breeding farms also need to identify wound categories, predict healing progress, and intervene in time in animal trauma or postoperative care. For example, if the ulcer on the hoof of a dairy cow is not discovered and properly cared for in a timely manner, it will cause serious economic losses. The present application can adapt to the characteristics of animal skin and tissue structure after training or fine-tuning through cross-wound knowledge transfer and field adaptation, and realize universal wound management for multiple species.

[0115] Finally, at a deeper level, the present application also has scientific research and teaching value. The large number of wound images and healing process data collected and labeled have important reference significance for further research on tissue repair mechanisms, improvement of innovative dressings, and exploration of personalized treatment plans. Colleges and universities, research institutions and enterprises can jointly develop more medical AI products on this basis, further catalyze the development of the "Internet + smart medical" industry, and inject sustained impetus into public health undertakings.

[0116] The practical application value of the present application lies in long-term benefits in multiple levels, multiple scenarios, and multiple groups: it makes wound care work in hospitals more standardized and efficient, enables primary medical institutions and rehabilitation centers to obtain professional evaluation capabilities, makes home care and disaster emergency scenarios more timely and accurate, and brings new development space in the fields of agriculture and pet medical care. With the continuous deepening of the digitalization of the medical industry, the present application will play an indispensable role in reducing medical costs, narrowing the medical gap between urban and rural areas, and improving the quality of life of patients, and will show its true social public welfare and industrial potential.

[0117] The embodiment of the present specification provides a prediction method of a wound healing process, acquires a target image sequence, wherein the target image sequence is composed of target images containing a trauma site of a user and a standard reference object at different time points, uses a first module of a pre-trained prediction model to perform local feature extraction and global feature extraction processing on each target image in the target image sequence respectively to obtain a local feature vector and a global feature vector, uses a second module of the pre-trained prediction model to perform time sequence feature extraction processing on the local feature vector and the global feature vector respectively based on different preset extraction periods to obtain a plurality of time sequence feature vectors, uses the pre-trained prediction model to generate healing condition prediction information for the trauma site based on the plurality of time sequence feature vectors, and determines a treatment scheme for the trauma site of the user according to the healing condition prediction information. In this way, on the one hand, since the target image contains the standard reference object, the recognition accuracy of the trauma site can be improved through comparison between the trauma site and the standard reference object, and on the other hand, through the prediction model, time sequence feature extraction can be performed through a plurality of extraction periods on the basis of taking into account local features and global features to obtain time sequence feature vectors capable of representing the change rule of the trauma site over time, and then through the time sequence feature vectors, the healing condition of the trauma site can be accurately predicted to generate a corresponding treatment scheme according to the healing condition prediction information, thereby improving the wound care effect.

[0118] The above is a prediction method of a wound healing process provided by the embodiment of the present specification, based on the same idea, the embodiment of the present specification also provides a prediction device of a wound healing process, as shown in Figure 4 .

[0119] The prediction device of the wound healing process comprises a sequence acquisition module 301, a first extraction module 302, a second extraction module 303, an information prediction module 304 and a scheme determination module 305, wherein: The sequence acquisition module 301 is used to acquire a target image sequence, wherein the target image sequence is composed of target images containing a trauma site of a user and a standard reference object at different time points; The first extraction module 302 is used to use a first module of a pre-trained prediction model to perform local feature extraction and global feature extraction processing on each target image in the target image sequence respectively to obtain a local feature vector and a global feature vector; The second extraction module 303 is used to use a second module of the pre-trained prediction model to perform time sequence feature extraction processing on the local feature vector and the global feature vector respectively based on different preset extraction periods to obtain a plurality of time sequence feature vectors; The information prediction module 304 is configured to generate healing condition prediction information of the wound site based on the plurality of time sequence feature vectors by using the pre-trained prediction model. The scheme determination module 305 is configured to determine a treatment scheme for the wound site of the user according to the healing condition prediction information.

[0120] In the embodiments of the present specification, the first module includes a first sub-module constructed by a residual neural network, and a second sub-module constructed by a visual Transformer, the first sub-module is configured to perform local feature extraction on each target image in the target image sequence to obtain the local feature vector, and the second sub-module is configured to perform local feature extraction on each target image in the target image sequence to obtain the global feature vector.

[0121] In the embodiments of the present specification, the second module includes a third sub-module constructed by a convolutional neural network, and a fourth sub-module constructed by a long short-term memory network, the plurality of time sequence feature vectors include time sequence feature vectors obtained by performing time sequence feature extraction processing on the local feature vector and the global feature vector respectively according to a pre-set first extraction period by the third sub-module, and time sequence feature vectors obtained by performing time sequence feature extraction processing on the local feature vector and the global feature vector respectively according to a pre-set second extraction period by the fourth sub-module, the first extraction period being less than the second extraction period.

[0122] In the embodiments of the present specification, the second extraction module is configured to: determine, by a third module of the pre-trained prediction model, weight values corresponding to the local feature vector and the global feature vector respectively according to image features of each target image in the target image sequence; determine a target feature vector according to the local feature vector and the corresponding weight value, and the global feature vector and the corresponding weight value; perform, by a second module of the pre-trained prediction model, time sequence feature extraction processing on the target feature vector based on the different pre-set extraction periods to obtain the plurality of time sequence feature vectors.

[0123] In the embodiments of the present specification, the device further includes: a sample acquisition module configured to acquire a sample image sequence; a data enhancement module configured to perform data enhancement processing on a sample image included in the sample image sequence according to a pre-set data enhancement rule to obtain an enhanced sample image sequence, the pre-set data enhancement rule at least including image rotation, color jitter, brightness adjustment, and contrast adjustment; The model obtaining module is configured to obtain a first prediction model, wherein the first prediction model is trained according to first sample data, and a data quantity of the first sample data is greater than a data quantity of the sample image sequence. The model training module is configured to train the first prediction model according to the enhanced sample image sequence, to obtain the pre-trained prediction model.

[0124] In the embodiments of the present specification, the scheme determination module is configured to: determine a wound care rule corresponding to the wound site of the user according to the healing condition prediction information; obtain multi-modal user data corresponding to the user; wherein the multi-modal data includes individual feature data, medical history information, laboratory examination results, and living habit data of the user; determine a treatment scheme for the wound site of the user according to the wound care rule and the multi-modal user data by using a pre-trained scheme determination model, wherein the scheme determination model is a model trained by reinforcement learning.

[0125] In the embodiments of the present specification, the healing condition prediction information includes a healing trend curve, and the information prediction module is configured to: determine a healing condition corresponding to each prediction time point by using the pre-trained prediction model based on a time point corresponding to each target image in the target image sequence and the plurality of time sequence feature vectors; generate the healing trend curve according to the healing condition corresponding to each prediction time point.

[0126] In the embodiments of the present specification, the local feature vector is used to represent the definition of the wound edge of the wound site, the texture change of the tissue, and the distribution of the color, and the global feature vector is used to represent the skin condition around the wound site.

[0127] In the embodiments of the present specification, the time sequence feature vector extracted by the third sub-module is used to represent the spatial structure feature of the wound site, and the time sequence feature vector extracted by the fourth sub-module is used to represent the change of the wound feature of the wound site.

[0128] The embodiment of the present specification provides a wound healing process prediction device, acquires a target image sequence, wherein the target image sequence is composed of target images containing a wound site of a user and a standard reference object at different time points, uses a first module of a pre-trained prediction model to perform local feature extraction and global feature extraction processing on each target image in the target image sequence respectively to obtain local feature vectors and global feature vectors, uses a second module of the pre-trained prediction model to perform time sequence feature extraction processing on the local feature vectors and the global feature vectors respectively based on different preset extraction periods to obtain a plurality of time sequence feature vectors, uses the pre-trained prediction model to generate healing condition prediction information for the wound site based on the plurality of time sequence feature vectors, and determines a treatment scheme for the wound site of the user according to the healing condition prediction information. In this way, on the one hand, since the target image contains the standard reference object, the recognition accuracy of the wound site can be improved through comparison between the wound site and the standard reference object, and on the other hand, through the prediction model, time sequence feature extraction can be performed through a plurality of extraction periods based on taking into account local features and global features to obtain time sequence feature vectors that can represent the change rule of the wound site over time, and then through the time sequence feature vectors, the healing condition of the wound site can be accurately predicted to generate a corresponding treatment scheme according to the healing condition prediction information, thereby improving the wound care effect.

[0129] The above is the wound healing process prediction device provided by the embodiment of the present specification. Based on the same idea, the embodiment of the present specification also provides a wound healing process prediction device, as shown in Figure 1 to Figure 2 .

[0130] The wound healing process prediction device can be a terminal device or a server provided by the above embodiment.

[0131] The wound healing process prediction device can have relatively large differences due to different configurations or performances, and can include one or more processors 401 and memories 402. The memories 402 can store one or more storage applications or data. The memories 402 can be temporary storage or persistent storage. The applications stored in the memories 402 can include one or more modules (not shown in the figure), and each module can include a series of computer executable instructions in the wound healing process prediction device. Further, the processor 401 can be configured to communicate with the memory 402 and execute a series of computer executable instructions in the memory 402 on the wound healing process prediction device. The wound healing process prediction device can also include one or more power supplies 403, one or more wired or wireless network interfaces 404, one or more input / output interfaces 405, and one or more keyboards 406.

[0132] In particular, in the present embodiment, the prediction device of wound healing process comprises a memory, and one or more programs, wherein one or more programs are stored in the memory, and one or more programs can comprise one or more modules, and each module can comprise a series of computer executable instructions in the prediction device of wound healing process, and configured to be executed by one or more processors, the one or more programs comprise computer executable instructions for: obtaining a target image sequence, the target image sequence is composed of target images containing a wound site of a user and a standard reference object at different time points; using a first module of a pre-trained prediction model, respectively performing local feature extraction and global feature extraction processing on each target image in the target image sequence to obtain a local feature vector and a global feature vector; using a second module of the pre-trained prediction model, respectively performing time series feature extraction processing on the local feature vector and the global feature vector based on different preset extraction periods to obtain a plurality of time series feature vectors; using the pre-trained prediction model, generating healing condition prediction information for the wound site based on the plurality of time series feature vectors; determining a treatment plan for the wound site of the user according to the healing condition prediction information.

[0133] Each embodiment in the present specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the prediction device of wound healing process embodiment, it is basically similar to the method embodiment, so the description is relatively simple, and the relevant part can be referred to the part of the method embodiment.

[0134] The embodiment of the specification provides a prediction device for a wound healing process, acquires a target image sequence, wherein the target image sequence is composed of target images containing a trauma site of a user and a standard reference object at different time points, uses a first module of a pre-trained prediction model to perform local feature extraction and global feature extraction processing on each target image in the target image sequence respectively to obtain a local feature vector and a global feature vector, uses a second module of the pre-trained prediction model to perform time sequence feature extraction processing on the local feature vector and the global feature vector respectively based on different preset extraction periods to obtain a plurality of time sequence feature vectors, uses the pre-trained prediction model to generate healing condition prediction information for the trauma site based on the plurality of time sequence feature vectors, and determines a treatment scheme for the trauma site of the user according to the healing condition prediction information. In this way, on the one hand, since the standard reference object is contained in the target image, the recognition accuracy of the trauma site can be improved through comparison between the trauma site and the standard reference object, and on the other hand, through the prediction model, time sequence feature extraction can be performed through a plurality of extraction periods on the basis of taking into account local features and global features to obtain time sequence feature vectors capable of representing the change rule of the trauma site over time, and then through the time sequence feature vectors, the healing condition of the trauma site can be accurately predicted to generate a corresponding treatment scheme according to the healing condition prediction information, thereby improving the wound care effect.

[0135] Further, based on the method shown above Figure 1 to Figure 2 One or more embodiments of the specification also provide a storage medium for storing computer executable instruction information, in a specific embodiment, the storage medium can be a U disk, an optical disk, a hard disk, etc., and the computer executable instruction information stored in the storage medium can realize the following processes when executed by a processor: acquire a target image sequence, the target image sequence being composed of target images containing a trauma site of a user and a standard reference object at different time points; use a first module of a pre-trained prediction model to perform local feature extraction and global feature extraction processing on each target image in the target image sequence respectively to obtain a local feature vector and a global feature vector; use a second module of the pre-trained prediction model to perform time sequence feature extraction processing on the local feature vector and the global feature vector respectively based on different preset extraction periods to obtain a plurality of time sequence feature vectors; use the pre-trained prediction model to generate healing condition prediction information for the trauma site based on the plurality of time sequence feature vectors; determine a treatment scheme for the trauma site of the user according to the healing condition prediction information.

[0136] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the above-described storage medium embodiment is basically similar to the method embodiment, so the description is relatively simple; relevant parts can be referred to the description of the method embodiment.

[0137] This specification provides a storage medium for acquiring a target image sequence, wherein the target image sequence consists of target images containing the user's wound site and a standard reference object at different time points. A first module of a pre-trained prediction model is used to perform local feature extraction and global feature extraction processing on each target image in the target image sequence, obtaining local feature vectors and global feature vectors. A second module of the pre-trained prediction model is used to perform temporal feature extraction processing on the local feature vectors and global feature vectors based on different preset extraction periods, obtaining multiple temporal feature vectors. Based on the multiple temporal feature vectors, the pre-trained prediction model generates prediction information on the healing status of the wound site. Based on the healing status prediction information, a treatment plan for the user's wound site is determined. In this way, on the one hand, since the target image contains a standard reference object, the accuracy of wound identification can be improved by comparing the wound site with the standard reference object. On the other hand, through the prediction model, based on both local and global features, temporal features can be extracted through multiple extraction cycles to obtain a temporal feature vector that can characterize the change pattern of the wound site over time. Then, through the temporal feature vector, the healing status of the wound site can be accurately predicted, and corresponding treatment plans can be generated in a targeted manner based on the healing status prediction information, thereby improving the wound care effect.

[0138] Furthermore, based on the above Figure 1 The method shown in this specification, along with one or more embodiments, also provides a computer program product including a computer program that, when executed by a processor, performs the following process: A target image sequence is acquired, which consists of target images at different time points containing the user's trauma site and a standard reference object; Using the first module of the pre-trained prediction model, local feature extraction and global feature extraction are performed on each target image in the target image sequence to obtain local feature vectors and global feature vectors. Using the second module of the pre-trained prediction model, based on different preset extraction periods, temporal feature extraction processing is performed on the local feature vector and the global feature vector respectively to obtain multiple temporal feature vectors; generate healing condition prediction information for the trauma site based on the plurality of time sequence feature vectors by using the pre-trained prediction model; determine a treatment scheme for the trauma site of the user according to the healing condition prediction information.

[0139] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the above-mentioned computer program product embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment.

[0140] The embodiment of the specification provides a computer program product, acquires a target image sequence, wherein the target image sequence is composed of target images containing a trauma site of a user and a standard reference object at different time points, uses a first module of a pre-trained prediction model to perform local feature extraction and global feature extraction processing on each target image in the target image sequence respectively to obtain a local feature vector and a global feature vector, uses a second module of the pre-trained prediction model to perform time sequence feature extraction processing on the local feature vector and the global feature vector respectively based on different preset extraction periods to obtain a plurality of time sequence feature vectors, generates healing condition prediction information for the trauma site based on the plurality of time sequence feature vectors by using the pre-trained prediction model, and determines a treatment scheme for the trauma site of the user according to the healing condition prediction information. In this way, on the one hand, since the target image contains the standard reference object, the recognition accuracy of the trauma site can be improved through comparison between the trauma site and the standard reference object, and on the other hand, through the prediction model, time sequence feature extraction can be performed through multiple extraction periods on the basis of taking into account local features and global features to obtain time sequence feature vectors that can represent the change rule of the trauma site over time, and then through the time sequence feature vectors, the healing condition of the trauma site can be accurately predicted to generate a corresponding treatment scheme according to the healing condition prediction information, thereby improving the wound care effect.

[0141] The above describes specific embodiments of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different than the order in the embodiments and still achieve the desired result. In addition, the processes depicted in the figures do not necessarily require the particular order shown, or sequential order, to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous or possible.

[0142] In the 1990s, it was possible to distinguish whether an improvement in a technology was a hardware improvement (e.g., an improvement in the circuit structure of a diode, transistor, switch, etc.) or a software improvement (an improvement in a method flow). However, as technology has advanced, many improvements in method flows today can be considered as direct improvements in hardware circuit structures. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into a hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (e.g., a field programmable gate array (FPGA)) is an integrated circuit whose logic function is determined by user programming of the device. A designer programs a digital system "integrated" on a PLD by himself / herself, without having to ask a chip manufacturer to design and manufacture a special integrated circuit chip. Moreover, instead of manually manufacturing an integrated circuit chip, this programming is now mostly implemented using "logic compiler" software, which is similar to a software compiler used when developing a program, and the original code before compilation must also be written in a specific programming language, which is called a hardware description language (HDL), and there are many types of HDL, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., and the most commonly used are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also be aware that it is only necessary to logically program a method flow using the above-mentioned hardware description languages and program it into an integrated circuit to easily obtain a hardware circuit that implements the logical method flow.

[0143] The controller can be implemented in any suitable way, for example, the controller can take the form of a microprocessor or processor and a computer readable medium storing computer readable program code, such as software or firmware, executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller and an embedded microcontroller, examples of which include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320, the memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that, in addition to being implemented in pure computer readable program code, the controller can equally well be implemented to perform the same functions using logic gates, switches, application specific integrated circuits, programmable logic controllers and embedded microcontrollers, by logically programming the method steps. The controller can thus be considered a hardware component, and the means included therein for performing the various functions can be considered structures within the hardware component. Alternatively, or even in addition, the means for performing the various functions can be considered both software modules that implement the methods and structures within the hardware component.

[0144] The systems, apparatuses, modules or units illustrated by the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0145] Those skilled in the art will understand that the embodiments of the present specification can be provided as methods, systems, or computer program products. Therefore, one or more embodiments of the present specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) containing computer usable program code.

[0146] The embodiments of the present specification are described with reference to flowcharts and / or block diagrams of the method, device (system) and computer program product according to the embodiments of the present specification. It should be understood that each flow and / or block in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable electronic devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable electronic devices generate a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions of one or more flows and / or blocks in the flowcharts and / or block diagrams can be implemented by an apparatus. ​ The functions of one or more flows and / or blocks in the flowcharts and / or block diagrams can be implemented by an apparatus.

[0147] In one typical arrangement, the computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0148] The memory can include non-persistent memory, random access memory (RAM), and / or non-volatile memory, such as read only memory (ROM) or flash memory, in a computer readable medium. The memory is an example of computer readable media.

[0149] It should also be noted that the terms "comprising," "including," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements in the list, but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without limitation, an element preceded by "comprises a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the recited element.

[0150] Those skilled in the art should understand that the embodiments of the present specification can be provided as a method, system or computer program product. Therefore, one or more embodiments of the present specification can take the form of a completely hardware embodiment, a completely software embodiment or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0151] One or more embodiments of the specification can be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, and the like, that perform particular tasks or implement particular abstract data types. One or more embodiments of the specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules can be located in local and remote computer storage media including memory storage devices.

[0152] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiments.

[0153] The above only describes the embodiments of the specification and does not limit the file. For those skilled in the art, the specification can have various changes and variations. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the specification shall be included in the scope of claims of the specification.

Claims

1. A device for predicting the wound healing process, characterized in that, The device comprises: a sequence acquisition module configured to acquire a target image sequence, the target image sequence comprising target images of a wound site of a user and a standard reference object at different time points; a first extraction module configured to perform local feature extraction and global feature extraction on each target image in the target image sequence by using a first module of a pre-trained prediction model, to obtain a local feature vector and a global feature vector; a second extraction module configured to perform time sequence feature extraction on the local feature vector and the global feature vector based on different preset extraction periods by using a second module of the pre-trained prediction model, to obtain a plurality of time sequence feature vectors; an information prediction module configured to generate healing condition prediction information for the wound site based on the plurality of time sequence feature vectors by using the pre-trained prediction model; a scheme determination module configured to determine a treatment scheme for the wound site of the user according to the healing condition prediction information.

2. The apparatus of claim 1, wherein, The first module comprises a first submodule constructed by a residual neural network and a second submodule constructed by a visual Transformer, the first submodule being configured to perform local feature extraction on each target image in the target image sequence to obtain the local feature vector, and the second submodule being configured to perform local feature extraction on each target image in the target image sequence to obtain the global feature vector.

3. The apparatus of claim 1, wherein, The second module comprises a third submodule constructed by a convolutional neural network and a fourth submodule constructed by a long short-term memory network, the plurality of time sequence feature vectors comprising time sequence feature vectors obtained by performing time sequence feature extraction on the local feature vector and the global feature vector based on a preset first extraction period by using the third submodule, and time sequence feature vectors obtained by performing time sequence feature extraction on the local feature vector and the global feature vector based on a preset second extraction period by using the fourth submodule, the first extraction period being smaller than the second extraction period.

4. The apparatus of claim 1, wherein, The second extraction module is configured to: determine weight values corresponding to the local feature vector and the global feature vector based on image features of each target image in the target image sequence by using a third module of the pre-trained prediction model; determine a target feature vector based on the local feature vector and the corresponding weight values and the global feature vector and the corresponding weight values; perform time sequence feature extraction on the target feature vector based on the different preset extraction periods by using the second module of the pre-trained prediction model, to obtain the plurality of time sequence feature vectors.

5. The apparatus of claim 1, wherein, The device further comprises: a sample acquisition module configured to acquire a sample image sequence; a data enhancement module configured to perform data enhancement on sample images in the sample image sequence according to a preset data enhancement rule, to obtain an enhanced sample image sequence, the preset data enhancement rule at least comprising image rotation, color jitter, brightness adjustment, and contrast adjustment; The model obtaining module is configured to obtain a first prediction model, wherein the first prediction model is trained according to first sample data, and a data quantity of the first sample data is greater than a data quantity of the sample image sequence; The model training module is configured to train the first prediction model according to the enhanced sample image sequence, to obtain the pre-trained prediction model.

6. The apparatus of claim 3, wherein, The time sequence feature vector extracted by the third sub-module is used to represent the spatial structure feature of the trauma site, and the time sequence feature vector extracted by the fourth sub-module is used to represent the change of the wound feature of the trauma site.

7. A method of predicting a wound healing process, characterized by, The method comprises: obtaining a target image sequence, wherein the target image sequence is composed of target images containing a trauma site of a user and a standard reference object at different time points; using a first module of a pre-trained prediction model to perform local feature extraction and global feature extraction on each target image in the target image sequence, to obtain a local feature vector and a global feature vector; using a second module of the pre-trained prediction model to perform time sequence feature extraction on the local feature vector and the global feature vector based on different preset extraction periods, to obtain a plurality of time sequence feature vectors; using the pre-trained prediction model to generate healing condition prediction information for the trauma site based on the plurality of time sequence feature vectors; determining a treatment scheme for the trauma site of the user according to the healing condition prediction information.

8. The method of claim 7, wherein, The determination of the treatment scheme for the trauma site of the user according to the healing condition prediction information comprises: determining a wound care rule corresponding to the trauma site of the user according to the healing condition prediction information; obtaining multi-modal user data corresponding to the user, wherein the multi-modal data comprises individual feature data, medical history information, laboratory examination results, and living habit data of the user; using a pre-trained scheme determination model to determine the treatment scheme for the trauma site of the user according to the wound care rule and the multi-modal user data, wherein the scheme determination model is a model trained by reinforcement learning.

9. The method of claim 7, wherein, The healing condition prediction information comprises a healing trend curve, and the generation of the healing condition prediction information for the trauma site by the pre-trained prediction model based on the plurality of time sequence feature vectors comprises: using the pre-trained prediction model to determine a healing condition corresponding to each prediction time point based on the time point corresponding to each target image in the target image sequence and the plurality of time sequence feature vectors; generating the healing trend curve according to the healing condition corresponding to each prediction time point.

10. The method of claim 7, wherein, The local feature vector is used to represent the clarity of the wound edge of the trauma site, the texture change of the tissue, and the distribution of the color, and the global feature vector is used to represent the skin condition around the trauma site.

Images