Pressure injury wound detection method and computer equipment

Abstract

The present application relates to the field of pressure injury wound detection, and specifically to a pressure injury wound detection method and computer equipment. A specific implementation of the method includes: controlling an image acquisition device to take a picture of a pressure injury wound area; controlling a temperature sensor to detect the temperature of a pressure injury wound area; controlling a wound humidity detector to detect the humidity of a pressure injury wound area; performing image quality detection on an image of a pressure injury wound area; in response to determining that the image quality detection result meets a preset image quality condition, performing pressure injury detection on the image of the pressure injury wound area, the temperature data of the wound area, and the humidity data of the wound area; and generating pressure injury wound treatment information based on each pressure injury detection result. This implementation can calculate the trauma level of a pressure injury wound by integrating humidity data, humidity data, and wound image features.

Description

Technical Field

[0001] The present application relates to the field of pressure injury wound detection, and specifically to a pressure injury wound detection method and computer equipment. Background Art

[0002] Due to the limited physical activity, long-term bedridden patients need to perform regular nursing operations such as wiping, turning over, and changing clothes to prevent the occurrence of pressure sores. This is because the patient maintains the same posture for a long time, and the local body is subjected to continuous pressure, which will affect blood circulation, cause damage to the skin and subcutaneous tissue, and then cause pressure sores. Different individuals have different physical constitutions and types of diseases, which makes the time and severity of pressure sores different. For example, patients with weak constitutions and chronic diseases such as diabetes may be more likely to develop pressure sores, and pressure sores may develop more rapidly and seriously. The manifestation of pressure sores itself has many clinical stages, the common stages 1 to 4, and the stages of unstageable and deep tissue injury. This stage is very professional, and nurses need to master professional knowledge to accurately judge the stage of pressure sores and then provide further treatment. At present, the nursing method mainly relies on the visual judgment of clinical nurses. When the nurse cannot confirm, a specialist nurse for pressure sores is required to help judge. This method has many inconveniences, such as possible delays in time and impact on timely treatment; it is also not rigorous enough, because visual inspection is easily affected by subjective factors, and the judgments of different nurses may differ, leading to misjudgment or inaccurate assessment of the pressure ulcer condition. Summary of the invention

[0003] The content of this application is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this application is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0004] Some embodiments of the present application propose a pressure injury wound detection method, a computer device, and a computer-readable storage medium to solve the technical problems mentioned in the above background technology section.

[0005] In a first aspect, some embodiments of the present application provide a pressure injury wound detection method, which is applied to a pressure injury wound detection system. The pressure injury wound detection system includes: an image acquisition device, a temperature sensor and a wound humidity detector. The method includes: in response to receiving a pressure injury wound detection instruction for a target user, for each pressure injury wound area of ​​the target user, performing the following detection steps: controlling the image acquisition device to take a photo of the pressure injury wound area to obtain an image of the pressure injury wound area; controlling the temperature sensor to perform temperature detection on the pressure injury wound area to obtain wound area temperature data; controlling the wound humidity detector .... The method comprises the steps of: performing humidity detection on the pressure injury wound area to obtain humidity data of the wound area; performing image quality detection on the pressure injury wound area image to obtain image quality detection results; in response to determining that the image quality detection results meet the preset image quality conditions, performing pressure injury detection on the pressure injury wound area image, the wound area temperature data and the wound area humidity data to obtain pressure injury detection results, wherein the historical pressure injury detection result sequence corresponds to the pressure injury wound area; generating pressure injury wound treatment information according to each pressure injury detection result, and sending the pressure injury wound treatment information to an associated doctor terminal.

[0006] In a second aspect, the present application also provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, it implements the method described in any implementation of the first aspect.

[0007] In a third aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored, wherein when the computer program is executed by a processor, the method described in any implementation of the first aspect is implemented.

[0008] The above-mentioned embodiments of the present application have the following beneficial effects: The pressure injury wound detection method provided by the present application can calculate the trauma level of the wound image by integrating humidity data, humidity data and wound image features (such as pixel channel information and edge contrast). At the same time, the trauma level is further detected by combining the area of ​​the wound surface, the changes in tissue damage of the wound, the temperature and humidity of the wound, and other information. For example, the size of the wound surface area can reflect the degree of development of the pressure sore; the changes in tissue damage can help understand the depth of the pressure sore and the degree of tissue damage; information such as temperature and humidity may be related to the infection risk of the wound, etc. The integration of this information can more comprehensively and accurately evaluate the pressure sore level (pressure injury detection results). BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The above and other features, advantages and aspects of the embodiments of the present application will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar figure labels represent the same or similar elements. It should be understood that the drawings are schematic and that the components and elements are not necessarily drawn to scale.

[0010] Figure 1 is a flow chart of some embodiments of the pressure injury wound detection method according to the present application;

[0011] Figure 2 is a schematic diagram of a computer device suitable for implementing some embodiments of the present application;

[0012] Figure 3 is a schematic screenshot illustrating various stages of a pressure injury. DETAILED DESCRIPTION

[0013] Embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not intended to limit the scope of protection of the present application.

[0014] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0015] It should be noted that the concepts such as "first" and "second" mentioned in this application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0016] It should be noted that the modifications of "one" and "plurality" mentioned in the present application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0017] The names of the messages or information exchanged between multiple devices in the embodiments of the present application are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0018] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0019] Figure 1The present invention is a flowchart of some embodiments of the pressure injury wound detection method according to the present invention. The flowchart 100 of some embodiments of the pressure injury wound detection method according to the present invention is shown. The pressure injury wound detection method is applied to a pressure injury wound detection system, which includes: an image acquisition device, a temperature sensor and a wound humidity detector, and includes the following steps:

[0020] Step 101, in response to receiving a pressure injury wound detection instruction for a target user, for each pressure injury wound area of ​​the target user, the following detection steps are performed:

[0021] Step 1011, controlling the image acquisition device to take a picture of the pressure injury wound area to obtain an image of the pressure injury wound area.

[0022] In some embodiments, the execution subject (e.g., a computer device) of the pressure injury wound detection method may control an image acquisition device to take a picture of the pressure injury wound area to obtain an image of the pressure injury wound area. The image acquisition device may refer to a medical camera that is communicatively connected to the execution subject, and may be used to take an image of the pressure injury wound of the target user. The target user may be a patient with a pressure injury (pressure sore). The pressure injury wound detection instruction may refer to an instruction to perform injury detection on each pressure injury wound of the target user.

[0023] Step 1012: Control the temperature sensor to detect the temperature of the pressure injury wound area to obtain temperature data of the wound area.

[0024] In some embodiments, the execution subject may control a temperature sensor to detect the temperature of the pressure injury wound area to obtain temperature data of the wound area. The temperature sensor may be a medical temperature sensor for detecting the temperature of the pressure injury wound area (pressure sore wound).

[0025] Step 1013, controlling the wound humidity detector to perform humidity detection on the pressure injury wound area to obtain humidity data of the wound area.

[0026] In some embodiments, the execution subject may control a wound humidity detector to detect humidity in the pressure injury wound area to obtain humidity data of the wound area. The wound humidity detector may be a medical humidity detector for detecting humidity in the pressure injury wound area.

[0027] Step 1014: Perform image quality detection on the pressure injury wound area image to obtain an image quality detection result.

[0028] In some embodiments, the above-mentioned execution subject may perform image quality detection on the above-mentioned pressure injury wound area image to obtain an image quality detection result. The image quality detection may represent clarity detection, image distortion detection, and wound area image integrity detection of the pressure injury wound area image. Among them, the clarity detection may be to detect the clarity of the pressure injury wound area image to determine whether the pressure injury wound area image can be used. The image distortion detection may detect whether the pressure injury wound area image is distorted. The wound area image integrity detection may detect whether the pressure injury wound area image completely covers the pressure injury wound area. The image quality detection result may indicate whether the clarity of the pressure injury wound area image is greater than or equal to the preset clarity, and whether the image is distorted.

[0029] For example, the methods for image clarity detection mainly include the following:

[0030] ‌1. Use an image clarity meter: An image clarity meter is an instrument specifically used to measure the clarity of a sample. By measuring the maximum luminous flux M and the minimum luminous flux m of the sample, the clarity is calculated using the formula C%=(Mm) / (M+m)*100%. The larger the value, the better the clarity‌1.

[0031] 2. Use image processing technology: In the field of artificial intelligence, you can use a preset scene text detection model to locate the text area in the image, crop the text area, perform frequency domain transformation based on the reference-free discrete cosine transform domain, and then use the preset text clarity calculation formula to calculate the clarity of each text area, and finally get the comprehensive clarity of the image.

[0032] For example, the methods of image distortion detection mainly include the following:

[0033] ‌1. Blur and distortion detection‌: Blur and distortion are manifested as unclear image details, blurred edges, and the whole image looks like it is covered by a layer of mist. The detection method is to observe whether the details in the image are clear and whether the edges are sharp. If the whole image is blurred and there is no obvious change in geometry, it may be blur and distortion.

[0034] 2. Color distortion detection: Color distortion is manifested as the color in the image being inconsistent with the color of the original scene, with color cast or abnormal color saturation. The detection method is to compare the color in the image with the reference color or the actual scene to observe whether there is obvious color deviation.

[0035] ‌3. Geometric distortion detection‌: Geometric distortion is manifested as straight lines in the image becoming curved or the image shape becoming distorted. The detection method is to look for obvious straight line elements in the image (such as the edges of buildings, the boundaries of roads, etc.) and observe whether these lines remain straight or have abnormal curvature.

[0036] ‌4. Increased noise detection‌: Increased noise is manifested as an increase in the number of random granular interference points in the image, affecting the clarity and fineness of the image. The detection method is to observe whether there is obvious granular noise in the dark area of ​​the image or at a high ISO setting.

[0037] ‌5. Detail loss detection‌: Detail loss is manifested as blurred image edges, reduced texture details, or unnatural color transitions. The detection method is to compare with the original image or high-resolution image to observe whether there is any loss of detail information.

[0038] ‌6. Aliasing distortion detection‌: Aliasing distortion is manifested as the overlap of high-frequency components in the image, resulting in jagged or wavy edges. The detection method is to observe whether the image edge is smooth and whether there is obvious jagged or wavy distortion.

[0039] For example, the wound area image integrity detection can be performed by a target detection algorithm; the target detection algorithm may include: Cascade9 + HOG / DPM + Haar / SVM and other algorithms. It can also be performed by a pre-trained wound area image integrity detection neural network model. The wound area image integrity detection neural network model may be: R-CNN (Selective Search + CNN + SVM) model, Fast R-CNN (Selective Search + CNN + ROI) model.

[0040] In an actual application scenario, the execution entity may input the pressure injury wound area image into a pre-trained image detection model to obtain an image quality detection result.

[0041] Furthermore, the image detection model can be trained by the following steps:

[0042] The first step is to obtain a target pressure injury wound area image sample. The target pressure injury wound area image sample includes: a sample pressure injury wound area image. The sample pressure injury wound area image can be a sample image showing a pressure injury wound area.

[0043] In the second step, the sample pressure injury wound area image is input into the first encoder of the initial image detection model to obtain the first wound area image feature. The first encoder is used to extract the features related to the distorted image in the sample pressure injury wound area image. The first encoder can also adopt the encoder structure in the generator of the distortion compensation model based on the GAN (Generative Adversarial Network) framework. The first wound area image feature of the sample pressure injury wound area image can be extracted by the first encoder. For example, the first encoder can adopt a multi-layer convolutional neural network structure. For example, the first encoder can include multiple layers of dilated convolutional layers. The initial image detection model can be an untrained neural network model for detecting image quality (distortion / clarity). For example, the initial image detection model can be an initial CNN neural network model, an initial YOLOv1 or a ‌DiffusionEdge‌ model.

[0044] In the third step, the sample pressure injury wound area image is input into the second encoder of the initial image detection model to obtain the second wound area image feature. The second encoder is used to extract the clarity feature of the sample pressure injury wound area image. The second encoder can be used to extract the clarity feature of the sample pressure injury wound area image. The second encoder here can use the encoder in the pre-trained YOLOv1 or ‌DiffusionEdge‌ model.

[0045] The fourth step is to fuse the first wound area image feature with the second wound area image feature to generate a fused wound area image feature. The execution subject can fuse the first wound area image feature output by the middle layer of the first encoder with the second wound area image feature extracted by the second encoder in the Attention Fusion network to obtain the fused wound area image feature.

[0046] In the fifth step, the above-mentioned fused wound area image features are input into the image quality output layer included in the initial image detection model to obtain the initial image quality detection result. The image quality output layer can be a decoder, which can include a deconvolution layer corresponding to the first encoder and a deconvolution layer corresponding to the second encoder, and restore the feature representation encoded by the first encoder to image distortion data, and restore the feature representation encoded by the second encoder to clarity data.

[0047] The sixth step is to determine the loss value between the above-mentioned initial image quality detection result and the corresponding image quality sample label. For example, the loss value between the above-mentioned initial image quality detection result and the corresponding image quality sample label can be determined by a preset loss function. The preset loss function can be a hinge loss function, a cross-entropy loss function, or a Huber loss function.

[0048] In the seventh step, in response to determining that the above loss value is less than or equal to the preset loss value, the initial image detection model is determined as the trained image detection model.

[0049] Step 1015, in response to determining that the above image quality detection result meets the preset image quality condition, pressure injury detection is performed on the above pressure injury wound area image, the above wound area temperature data and the above wound area humidity data to obtain a pressure injury detection result.

[0050] In some embodiments, the execution subject may perform pressure injury detection on the pressure injury wound area image, the wound area temperature data and the wound area humidity data in response to determining that the image quality detection result meets the preset image quality condition, and obtain the pressure injury detection result. Among them, the historical pressure injury detection result sequence corresponds to the pressure injury wound area. The preset image quality condition may be: the image quality detection result characterizes that the image quality is normal. For example, the area, depth, edge contour, wound necrotic tissue type, wound necrotic tissue amount, wound exudate type, skin color around the wound, wound exudate amount, tissue edema around the wound, tissue sclerosis around the wound, wound granulation tissue, wound epithelialization and the like of the pressure injury wound area image may be identified. For another example, the inflammation of the current wound may be determined by the wound area temperature data. For another example, the amount of exudate of the current wound may be determined by the wound area humidity data. The pressure injury detection result may indicate the pressure injury level (stage of pressure sore) of the target user.

[0051] It should be noted that pressure sores (pressure injuries) are also called bedsores, which can be divided into four stages and are more common in patients who are bedridden for a long time and have weak constitutions. They are caused by poor overall nutritional status, passive lying position, loss of skin sensation, prolonged pressure on the skin of the bony protuberances between the mattress and the bony protuberances, blood circulation disorders, neurotrophic changes, and skin necrosis. Figure 3 The example shows a schematic screenshot of a pressure injury at various stages.

[0052] Stage 1 (I): This is the early stage of pressure sores. The local skin of the patient is intact, but there will be red spots that do not turn white when pressed. For patients with darker skin, the whitening phenomenon may not be seen when pressing with fingers, but the color will be different from the surrounding skin. At the same time, this area will also have symptoms such as pain, hardness, coolness or fever.

[0053] Stage II: Part of the patient's skin will be lost, showing as shallow open ulcers, the wound surface is pink, and there is no slough. However, it can also show as complete or open serous blisters, with a translucent or dry ulcer appearance and no slough;

[0054] Stage III: Ulcers form on the patient's skin surface, which may be shallow and reach the subcutaneous fat layer, or deep and reach the bone tissue. If secondary infection occurs, there will be more pus and a foul odor.

[0055] Stage IV: The patient has full-thickness skin tissue loss with exposed bones, tendons, or muscles that are visible and palpable. There may be slough and eschar under the wound, and there are usually sinus tracts. Since pressure sores can extend to muscles and supporting structures such as fascia, tendons, or joint capsules, they may also cause osteomyelitis.

[0056] Deep tissue injury stage: During the deep tissue injury stage, the skin may appear purple or brown-purple locally, and the epidermis may appear congested. The evolution of the injury may start with a small blister on a dark wound. Even if proper treatment is given to the pressure sores at this stage, the injured area may rapidly transform into exposed subcutaneous tissue.

[0057] Unstageable: Unstageable pressure ulcers involve full-thickness tissue loss, with the wound surface completely covered by necrotic tissue and eschar. Therefore, to determine the depth of the ulcer, it is necessary to first remove the surface necrotic tissue and eschar.

[0058] As an example, the above-mentioned execution entity can detect pressure injuries through the following 15 indicators: 1. Area of ​​pressure injury wound; 2. Depth of pressure injury wound; 3. Contour of edge of pressure injury wound; 4. Type of necrotic tissue of pressure injury wound; 5. Amount of necrotic tissue of pressure injury wound; 6. Type of exudate of pressure injury wound; 7. Amount of exudate of pressure injury wound; 8. Color of skin around pressure injury wound; 9. Edema of tissue around pressure injury wound; 10. Hardening of tissue around pressure injury wound; 11. Granulation tissue of pressure injury wound; 12. Epithelialization of pressure injury wound.

[0059] Pressure injury wound depth conditions may include: 1. Intact skin with non-blanchable erythema; 2. Partial-thickness loss of the epidermis and / or dermis; 3. Full-thickness loss, including damage or necrosis of subcutaneous tissue; may extend to but not through the deep fascia; and / or a mixture of partial and full-thickness loss and / or tissue layers covered by granulation tissue; 4. Covered by necrotic tissue; 5. Full-thickness loss with extensive destruction, damage or necrosis of muscle, bone, or supporting structures.

[0060] The contours of pressure injury wound edges can include: 1. fuzzy, diffuse, not clearly visible; 2. clear, clearly visible, attached, even attached to the wound base; 3. well-defined, not attached to the wound base; 4. well-defined, not attached to the wound base, involuted, thickened; 5. soft to hard, elastic to the touch. Hyperkeratosis = well-defined, fibrotic, scarred or hyperkeratotic.

[0061] The types of necrotic tissue in pressure injury wounds can include: 1. None; 2. White / gray non-viable tissue and / or non-adherent yellow slough; 3. Loose yellow slough; 4. Adherent, soft, black scab; 5. Firmly adherent, hard, black scab.

[0062] The amount of necrotic tissue in pressure injury wounds can include: 1. None; 2. Covering the wound bed <25%; 3. Covering the wound bed 25%-50%; 4. Covering the wound bed 50%-75%; 5. Covering the wound bed 75%-100%.

[0063] The types of exudate from pressure injury wounds may include: 1. None; 2. Bloody; 3. Serous bloody fluid: thin, watery, light pink to pink; 4. Serous: thin, watery, clear; 5. Purulent: thin or thick, opaque brown to yellow; with or without odor.

[0064] The amount of exudate in pressure injury wounds can be as follows: 1 = None: dry wound tissue; 2. Very little: moist wound tissue with no measurable exudate; 3. Small amount: moist wound tissue; exudate evenly distributed on the wound surface; the proportion of exudate wetting the dressing is ≤ 25%; 4. Moderate amount: saturated wound tissue; exudate evenly or unevenly distributed on the wound; the proportion of exudate wetting the dressing is between 25% and 75%; 5. Large amount: wound tissue is soaked in exudate; a large amount of exudate can be seen; evenly or unevenly distributed on the wound surface; the proportion of exudate wetting the dressing is ≥ 75%.

[0065] The skin color around pressure injury wounds can include: 1. Pink or ethnically normal skin color; 2. Bright red and / or blanch to touch; 3. White or grayish white or hypopigmented; 4. Dark red or purple and / or does not blanch with finger pressure; 5. Black or hyperpigmented.

[0066] Edema of tissues around pressure injury wounds may include: 1. No swelling or edema; 2. Non-pitting edema extends <4cm around the wound; 3. Non-pitting edema extends >4cm around the wound; 4. Pitting edema extends <4cm around the wound; 5. Crepitus and / or pitting edema extends >4cm around the wound.

[0067] Hardening of tissues around pressure injury wounds may include: 1. None; 2. Hardening range <2cm; 3. Hardening range 2-4cm, accounting for <50% of the wound area; 4. Hardening range 2-4cm, accounting for ≥50% of the wound area; 5. Hardening of any part around the wound >4cm.

[0068] Granulation tissue in pressure injury wounds may include: 1. Full or partial thickness wounds; 2. Bright, beef red: fills 75%-100% of the wound and / or tissue proliferation; 3. Bright, beef red: fills 25%-75% of the wound; 4. Pink, and / or dark red and / or fills <25% of the wound; 5. No granulation tissue.

[0069] Epithelialization of pressure injury wounds can include: 1. 100% wound coverage with intact surface; 2. 75%-100% wound coverage and / or epithelial tissue extending into the wound bed >0.5cm; 3. 50%-75% wound coverage and / or epithelial tissue extending into the wound bed <0.5cm; 4. 25%-50% wound coverage; 5. <25% wound coverage.

[0070] In an actual application scenario, the execution subject may, in response to determining that there is a historical pressure injury detection result sequence in the pressure injury wound area, input the historical pressure injury detection result sequence, the pressure injury wound area image, the wound area temperature data, and the wound area humidity data into a pre-trained pressure injury detection model to obtain a pressure injury detection result. The historical pressure injury detection result sequence may refer to each pressure injury detection result of the target user's history. The pre-trained pressure injury detection model may be a pre-trained neural network model that takes the historical pressure injury detection result sequence, the pressure injury wound area image, the wound area temperature data, and the wound area humidity data as input and takes the pressure injury detection result as output. For example, the pressure injury detection model may be an image recognition model based on a deep learning algorithm (CNN), an R-CNN model, a Faster R-CNN model, a YOLO series model, and the like.

[0071] Furthermore, the pre-trained pressure injury detection model can be trained by the following steps:

[0072] The first step is to obtain a pressure injury sample set, wherein the pressure injury samples include: historical pressure injury detection result sequence samples, pressure injury collection information samples, and the pressure injury collection information samples include: pressure injury wound area image samples, wound area temperature data samples, and wound area humidity data samples.

[0073] The second step is to determine the initial pressure injury detection model. The initial pressure injury detection model includes: an initial historical pressure injury prediction network, an initial pressure injury recognition network, and an initial fusion network. The initial historical pressure injury prediction network includes: an initial detection result sequence feature extraction layer, an initial multi-level pressure injury feature encoding layer, and an initial pressure injury prediction output layer. The initial pressure injury recognition network includes: an initial image feature extraction layer, an initial image contour extraction layer, an initial image area feature extraction layer, an initial multi-level image feature encoding layer, an initial temperature feature extraction layer, an initial humidity feature extraction layer, an initial feature fusion layer, and an initial pressure injury recognition output layer.

[0074] The initial detection result sequence feature extraction layer may be a feature extraction network layer for extracting each historical pressure injury detection result in the historical pressure injury detection result sequence sample. For example, the initial detection result sequence feature extraction layer may be a convolutional neural network layer (CNN), a recurrent neural network (RNN), or a long short-term memory network (LSTM). The initial multi-level pressure injury feature encoding layer may be a convolutional neural network (CNN) or a long short-term memory network (LSTM). The initial pressure injury prediction output layer may be an output layer (Output Layer) for outputting the predicted pressure injury level based on the historical pressure injury detection result sequence.

[0075] The initial image feature extraction layer may be a convolutional neural network (CNN), a recurrent neural network (RNN), a long short-term memory network (LSTM), etc. The initial image contour extraction layer may be an extraction network layer for extracting the contour of the pressure injury wound area in the pressure injury wound area image. For example, the initial image contour extraction layer may be an edge detection layer. The initial image area feature extraction layer may be an extraction layer that takes the output of the initial image contour extraction layer as input and takes the image area feature as output. The initial image area feature extraction layer may be a convolutional neural network layer (CNN). The initial multi-level image feature encoding layer may include: an initial image feature encoding layer (encoder) and an initial image feature decoding layer (decoder), the initial image feature encoding layer includes: an initial image feature decoupling layer (including an autoencoder network, an LSTM network) and an initial image feature fusion layer (VAE (Variational AutoEncoder, variational encoder), the initial image feature fusion layer includes: an initial fusion layer, an initial re-parameterization layer. The initial fusion layer may be a Feature Pyramid Network (FPN) or a Self-Attention Mechanism network. The initial re-parameterization layer may include: DyRep‌ (Dynamic Re-parameterization) and ACNet‌ (Asymmetric Convolution Net).

[0076] The initial temperature feature extraction layer may be an extraction layer for extracting temperature data features, such as a convolutional neural network (CNN). The initial humidity feature extraction layer may be an extraction layer for extracting humidity data features, such as a convolutional neural network (CNN).

[0077] The initial pressure injury identification output layer may be an output layer (Output Layer) for outputting pressure injury detection results.

[0078] The third step is to train the initial pressure injury detection model based on the pressure injury sample set to obtain a trained pressure injury detection model.

[0079] The third step may include the following sub-steps:

[0080] In the first sub-step, a pressure injury sample is selected from the pressure injury sample set as a pressure injury training sample. A pressure injury sample can be randomly selected from the pressure injury sample set as a pressure injury training sample.

[0081] In the second sub-step, the historical pressure injury detection result sequence samples included in the pressure injury training samples are input into the initial historical pressure injury prediction network to obtain the initial historical pressure injury prediction results.

[0082] The second sub-step may include:

[0083] First, the above historical pressure injury detection result sequence samples are input into the initial detection result sequence feature extraction layer to obtain the initial detection result sequence features.

[0084] Then, the above-mentioned initial detection result sequence features are input into the initial multi-level pressure injury feature encoding layer to obtain the initial multi-level pressure injury features.

[0085] Finally, the above initial multi-level pressure injury features are input into the initial pressure injury prediction output layer to obtain the initial historical pressure injury prediction results.

[0086] In a third sub-step, the pressure injury collection information sample included in the pressure injury training sample is input into an initial pressure injury recognition network to obtain an initial pressure injury recognition result.

[0087] The third sub-step may include:

[0088] 1. Input the above-mentioned pressure injury wound area image samples into the initial image feature extraction layer to obtain the initial pressure injury wound area image features.

[0089] 2. Input the above-mentioned initial pressure injury wound area image features into the initial multi-level image feature encoding layer to obtain the initial multi-level wound image features.

[0090] 3. Input the above-mentioned pressure injury wound area image sample into the initial image contour extraction layer to obtain the initial image contour feature.

[0091] 4. Input the above initial image contour features into the initial image area feature extraction layer to obtain the initial image area features.

[0092] 5. Input the above wound area temperature data samples into the initial temperature feature extraction layer to obtain the initial temperature feature.

[0093] 6. Input the above wound area humidity data samples into the initial humidity feature extraction layer to obtain the initial humidity features.

[0094] 7. Input the above-mentioned initial multi-level wound image features, the above-mentioned initial image area features, the above-mentioned initial temperature features and the above-mentioned initial humidity features into the initial feature fusion layer to obtain fusion features.

[0095] 8. Input the above fusion features into the initial pressure injury recognition output layer to obtain the initial pressure injury recognition results.

[0096] In the fourth sub-step, the initial historical pressure injury prediction result and the initial pressure injury recognition result are input into the initial fusion network to obtain the initial pressure injury fusion detection result.

[0097] The fifth sub-step is to determine the detection loss value between the initial pressure injury fusion detection result and the corresponding pressure injury detection label. The detection loss value between the initial pressure injury fusion detection result and the corresponding pressure injury detection label can be determined by a damage function. The loss function can be: mean square error loss (MSE Loss) function, cross-entropy loss (Cross-Entropy Loss) function, and mean absolute error (MAE) function.

[0098] A sixth sub-step, in response to determining that the detection loss value is less than or equal to a preset detection loss value, determines the initial pressure injury detection model as a trained pressure injury detection model.

[0099] Therefore, the pressure injury detection model constructed in this application can fully capture the multi-level information of the wound image area and enhance the efficiency and accuracy of the autoencoder in feature unwrapping. In addition, an initial pressure injury recognition output layer is constructed to help the model output the complex nonlinear relationship in the wound area image, which can improve the accuracy of wound grade detection.

[0100] Further, in response to determining that the above-mentioned image quality detection result does not meet the preset image quality condition, the above-mentioned image acquisition device is controlled to take a picture of the above-mentioned pressure injury wound area again to obtain a new pressure injury wound area image, and the image quality detection is performed on the new pressure injury wound area image again.

[0101] Step 102: Generate pressure injury wound treatment information according to each pressure injury detection result, and send the pressure injury wound treatment information to an associated doctor terminal.

[0102] In some embodiments, the execution subject may generate pressure injury wound treatment information based on each pressure injury detection result, and send the pressure injury wound treatment information to an associated doctor terminal. For example, each pressure injury detection result may correspond to a pressure injury level (stage of pressure sore). Thus, a corresponding treatment plan may be determined based on each pressure injury level. That is, a corresponding treatment plan may be preset for each pressure injury level. Thus, each treatment plan may be combined into pressure injury wound treatment information. The doctor terminal may refer to the computer terminal of the target user's attending physician.

[0103] In an actual application scenario, the above execution subject can generate pressure injury wound treatment information through the following steps:

[0104] In the first step, the injury treatment plan information set corresponding to the pressure injury is determined using the pre-constructed pressure injury level knowledge graph. The knowledge graph may be a pressure injury graph constructed based on a treatment guideline knowledge base. The treatment guideline knowledge base may store a knowledge database of guideline knowledge for treating each type of pressure injury. The pressure injury level knowledge graph may be a graph constructed based on multiple information bases. Among them, the multiple information bases include: a treatment guideline knowledge base, a disease information base, a patient information base, and a drug information base. The disease information base may be a knowledge base storing various disease information. The patient information base may be an information base storing patient information with the authorization of the patient. The drug information base may be an information base storing various drug-related information. Drug information may include: drug name, drug composition, and indication. The treatment guideline knowledge base includes: pressure injury disease name, gender, age, disease medical indicator name, disease medical indicator threshold, disease registration, drug treatment name under different registered diseases, diagnosis and treatment plan steps corresponding to the disease, and precautions (including but not limited to medication contraindications, implicit, work and rest and other behavioral habits). The disease information in the disease information database includes: symptom name, disease name, and classification. The patient information in the patient information database includes: name, age, gender, current disease name, historical disease name, family disease name, and name of the medication being taken. For the pressure injury level knowledge graph constructed for multiple knowledge bases, the corresponding nodes can be: injury level node, pressure injury type node, clinical guideline node, and medication node. The edges between nodes represent the corresponding relationship between the two nodes.

[0105] As an example, first, the execution subject can use the knowledge graph to find the nodes corresponding to the keyword set corresponding to pressure injury to obtain a node set. Then, the guideline knowledge corresponding to each node in the node set is obtained to obtain a guideline knowledge set. Finally, the injury treatment plan information is extracted from the guideline knowledge set to obtain an injury treatment plan information set.

[0106] In the second step, for each pressure injury wound area of ​​the target user, perform the following treatment steps:

[0107] 1. Determine the pressure injury detection result corresponding to the pressure injury wound area as the target pressure injury detection result.

[0108] 2. Use the pre-constructed pressure injury level knowledge graph to determine the injury treatment plan information set corresponding to the target pressure injury detection result. The above-mentioned execution subject can use the knowledge graph to find the nodes corresponding to the keyword set corresponding to the target pressure injury detection result to obtain a node set. After that, obtain the guideline knowledge corresponding to each node in the node set to obtain the guideline knowledge set. Finally, extract the injury treatment plan information from the guideline knowledge set to obtain the injury treatment plan information set.

[0109] 3. In response to determining that the pressure injury wound area has corresponding historical injury treatment plan information, extract user treatment adaptation information from the historical injury treatment plan information. The user treatment adaptation information may indicate precautions and taboos for the pressure injury wound area of ​​the target user.

[0110] 4. Selecting the injury treatment plan information corresponding to the user treatment adaptation information from the injury treatment plan information set as the wound treatment plan information corresponding to the pressure injury wound area. That is, the injury treatment plan information that meets the precautions and contraindications can be selected from each injury treatment plan information as the wound treatment plan information corresponding to the pressure injury wound area.

[0111] The third step is to merge the information of various wound treatment plans into pressure injury wound treatment information.

[0112] Therefore, based on the analysis of the above multi-dimensional data, the patient's current pressure sore level can be accurately judged. Compared with traditional manual judgment, this method avoids human errors and is more objective and accurate. Combined with the built-in pre-established pressure sore treatment plan library, treatment plan recommendations are given based on the determined pressure sore level. This enables nursing staff to obtain scientific and reasonable nursing guidance in a timely manner, provide patients with more accurate nursing services, improve the effectiveness and efficiency of pressure sore treatment, promote patient recovery, and improve the quality of life of patients.

[0113] The present application also provides a computer device 200. Figure 2 As shown, the computer device 200 includes: a bus 201, a processor 202, a memory 203 and a communication interface 204. The processor 202, the memory 203 and the communication interface 204 communicate with each other through the bus 201. The computer device 200 can be a server or a terminal device. It should be understood that the present application does not limit the number of processors and memories in the computer device 200.

[0114] The bus 201 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 The bus 201 is represented by only one line, but it does not mean that there is only one bus or one type of bus. The bus 201 may include a path for transmitting information between various components of the computer device 200 (for example, the processor 202, the memory 203, and the communication interface 204).

[0115] The processor 202 may include any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0116] The memory 203 may include a volatile memory, such as a random access memory (RAM). The memory 203 may also include a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0117] The memory 203 stores executable program codes, and the processor 202 executes the executable program codes to respectively implement the functions of the aforementioned acquisition module, sampling module, determination module, and mixing module, thereby implementing the aforementioned pressure injury wound detection method. That is, the memory 203 stores instructions for executing the aforementioned pressure injury wound detection method.

[0118] The communication interface 204 uses a transceiver module such as, but not limited to, a network interface card or a transceiver to implement communication between the computer device 200 and other devices or communication networks.

[0119] An embodiment of the present application also provides a chip, which includes a processor and a data interface. The processor reads instructions stored in a memory through the data interface to execute the above-mentioned method for pressure injury wound detection.

[0120] The embodiment of the present application also provides a computer-readable storage medium. The above-mentioned computer-readable storage medium can be any available medium that can be stored by a computing device or a data storage device such as a data center containing one or more available media. The above-mentioned available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid-state hard disk), etc. The computer-readable storage medium includes instructions, and the above-mentioned instructions instruct the computing device to execute the above-mentioned method for detecting pressure injury wounds.

[0121] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0122] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the protection scope of the technical solutions of the embodiments of the present application.

Claims

1. A pressure injury wound detection method, applied to a pressure injury wound detection system, the pressure injury wound detection system comprising: Image acquisition equipment, temperature sensor and wound moisture detector, the method comprises: In response to receiving a pressure injury wound detection instruction for a target user, the following detection steps are performed for each pressure injury wound area of ​​the target user: Controlling an image acquisition device to take a picture of the pressure injury wound area to obtain an image of the pressure injury wound area; Controlling a temperature sensor to detect the temperature of the pressure injury wound area to obtain temperature data of the wound area; Controlling a wound humidity detector to detect humidity in the pressure injury wound area to obtain humidity data of the wound area; Performing image quality detection on the pressure injury wound area image to obtain an image quality detection result; In response to determining that the image quality detection result meets a preset image quality condition, performing pressure injury detection on the pressure injury wound area image, the wound area temperature data, and the wound area humidity data to obtain a pressure injury detection result; Generate pressure injury wound treatment information according to each pressure injury detection result, and send the pressure injury wound treatment information to an associated doctor terminal; The step of performing pressure injury detection on the pressure injury wound area image, the wound area temperature data and the wound area humidity data to obtain a pressure injury detection result includes: In response to determining that there is a historical pressure injury detection result sequence in the pressure injury wound area, inputting the historical pressure injury detection result sequence, the pressure injury wound area image, the wound area temperature data, and the wound area humidity data into a pre-trained pressure injury detection model to obtain a pressure injury detection result; Wherein, before inputting the historical pressure injury detection result sequence, the pressure injury wound area image, the wound area temperature data and the wound area humidity data into a pre-trained pressure injury detection model to obtain the pressure injury detection result, the method further includes: Acquire a pressure injury sample set, wherein the pressure injury sample includes: a historical pressure injury detection result sequence sample, a pressure injury collection information sample, and the pressure injury collection information sample includes: a pressure injury wound area image sample, a wound area temperature data sample, and a wound area humidity data sample; Determine an initial pressure injury detection model, wherein the initial pressure injury detection model includes: an initial historical pressure injury prediction network, an initial pressure injury recognition network, and an initial fusion network; the initial historical pressure injury prediction network includes: an initial detection result sequence feature extraction layer, an initial multi-level pressure injury feature encoding layer, and an initial pressure injury prediction output layer; the initial pressure injury recognition network includes: an initial image feature extraction layer, an initial image contour extraction layer, an initial image area feature extraction layer, an initial multi-level image feature encoding layer, an initial temperature feature extraction layer, an initial humidity feature extraction layer, an initial feature fusion layer, and an initial pressure injury recognition output layer; Based on the pressure injury sample set, the initial pressure injury detection model is trained to obtain a trained pressure injury detection model; The step of training the initial pressure injury detection model based on the pressure injury sample set to obtain a trained pressure injury detection model includes: Selecting a pressure injury sample from the pressure injury sample set as a pressure injury training sample; Inputting the historical pressure injury detection result sequence samples included in the pressure injury training samples into the initial historical pressure injury prediction network to obtain the initial historical pressure injury prediction results; Inputting the pressure injury collection information sample included in the pressure injury training sample into the initial pressure injury recognition network to obtain an initial pressure injury recognition result; Inputting the initial historical pressure injury prediction result and the initial pressure injury recognition result into an initial fusion network to obtain an initial pressure injury fusion detection result; Determining a detection loss value between the initial pressure injury fusion detection result and a corresponding pressure injury detection tag; In response to determining that the detection loss value is less than or equal to the preset detection loss value, the initial pressure injury detection model is determined as the trained pressure injury detection model.

2. The pressure injury wound detection method according to claim 1, characterized in that: The detection step also includes: In response to determining that the image quality detection result does not meet the preset image quality conditions, the image acquisition device is controlled to take a photo of the pressure injury wound area again to obtain a new pressure injury wound area image, and image quality detection is performed on the new pressure injury wound area image again.

3. The pressure injury wound detection method according to claim 1, characterized in that: The performing image quality detection on the pressure injury wound area image to obtain an image quality detection result includes: The pressure injury wound area image is input into a pre-trained image detection model to obtain an image quality detection result.

4. A computer device, wherein: The computer device comprises a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 3 are implemented.

5. A computer-readable storage medium, wherein: The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images