Intelligent multi-spectrum vision-based patient stoma intelligent nursing method

Abstract

The application provides a patient stoma intelligent nursing method based on intelligent multispectral vision and belongs to the technical field of medical nursing. The method solves the problems of strong subjectivity of evaluation, lack of quantitative standard, poor standardization of nursing operation, fragmented data recording and inability to form a standardized and traceable intelligent nursing closed loop caused by excessive dependence on manual experience in the existing stoma nursing. The method comprises the following steps: S1, acquiring an overall posture image of a patient; S2, acquiring multispectral image data of a stoma part of the patient; S3, preprocessing the multispectral image data; S4, constructing a multimodal feature vector for each pixel position; S5, extracting a stoma part mask and calculating standardized physiological indexes; S6, encoding the standardized physiological indexes into a state vector and generating a preliminary nursing scheme based on a random forest algorithm; and S7, generating a final nursing scheme. The method is mainly used for intelligent nursing of intestinal stoma or urologic stoma of a patient.

Description

Technical Field

[0001] This invention belongs to the field of medical and nursing technology, and in particular relates to an intelligent nursing method for patient stoma based on intelligent multispectral vision. Background Technology

[0002] In current medical care, postoperative care for enterostomies or urostomies primarily relies on manual intervention. Stoma care is a high-frequency, high-precision, and extremely challenging task. In the early postoperative period, the stoma is usually in a edematous phase, accompanied by the continuous discharge of intestinal fluid and feces, which can easily lead to complications such as peristomal skin inflammation, infection, bleeding, and irritant dermatitis.

[0003] However, traditional nursing methods have many inherent defects: First, the nursing effect is highly dependent on the nurse's personal experience. The assessment of stoma complications (such as color judgment, swelling degree assessment, and infection sign identification) is highly subjective and lacks unified, quantifiable objective standards, which can easily lead to inappropriate nursing plans due to subjective misjudgment. Second, the nursing process is time-consuming and laborious. Frequent cleaning and dressing changes not only increase the workload of medical staff but also bring physical pain and psychological burden to patients. Third, manual operation makes it difficult to maintain a sterile environment for a long time, increasing the risk of cross-infection. Fourth, existing nursing records are mostly fragmented textual descriptions, lacking continuous imaging data and quantitative data support, which is not conducive to retrospective analysis and accurate prediction of the evolution of the patient's condition.

[0004] With the aging population and changes in the incidence of related diseases, the number of ostomy patients is on the rise, and the clinical demand for nursing solutions with high-precision perception, intelligent decision-making, and standardized execution capabilities is becoming increasingly urgent. Although some intelligent auxiliary solutions involving ostomy care assessment have been proposed, most of them only reach the suggestion generation level and fail to achieve a complete logical link from multimodal data perception and quantitative assessment of physiological status to automatic generation and closed-loop execution of nursing plans. Furthermore, there are significant shortcomings in the deep integration of multispectral image processing and clinical nursing decision-making. Summary of the Invention

[0005] In view of this, the present invention aims to propose an intelligent stoma care method based on intelligent multispectral vision, in order to solve the problems in existing stoma care that rely too much on human experience, resulting in strong subjectivity in assessment, lack of quantitative standards, poor standardization of nursing operations, fragmented data recording, and inability to form a standardized and traceable intelligent nursing closed loop.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: According to a first aspect of the present invention, a patient stoma intelligent care method based on intelligent multispectral vision includes the following steps: Step S1: Obtain the patient's overall posture image, compare the detected patient posture with the preset standard nursing position, and output voice guidance prompts if the posture does not meet the requirements until the detected posture meets the requirements. Step S2: Acquire multispectral image data of the patient's stoma site, wherein the multispectral image data includes at least synchronously acquired visible light images, infrared images, and ultraviolet images; Step S3: Preprocess the multispectral image data. The preprocessing includes at least multimodal data correction, wavelet transform-based image denoising, and image spatial coordinate registration. Each modal image is mapped to a unified analytical coordinate system to obtain spatially aligned multispectral image data. Step S4: Based on the spatially aligned multispectral image data, construct a multimodal feature vector for each pixel location. The multimodal feature vector includes at least visible light color features, thermal radiation energy index, and fluorescence metabolic intensity operator. Step S5: Extract the stoma site mask using a semantic segmentation network, and calculate standardized physiological indicators based on the multimodal feature vector and the stoma site mask using a weighted mapping model. The standardized physiological indicators include at least color parameters, quantified values ​​of swelling degree, classification of exudate properties, and results of infection sign determination. Step S6: Encode the standardized physiological indicators into state vectors, and generate a preliminary care plan based on the random forest algorithm. The preliminary care plan includes at least the type of cleaning solution, cleaning operation parameters, ointment application plan, and dressing type. Step S7: Output the preliminary nursing plan to the human-computer interaction interface, receive adjustment instructions and confirmation instructions from external input, and generate the final nursing plan; Step S8: According to the final nursing plan, perform cleaning and dressing changes for the stoma site in sequence; Step S9: During the cleaning and dressing change operations, real-time multispectral image data of the stoma site is continuously acquired, and closed-loop correction control is performed on the current operation process based on real-time visual feedback. Step S10: Record the multispectral imaging data, physiological indicator assessment data, and final nursing plan data throughout the entire nursing process, and update the data to the patient case database; Step S11: Retrieve historical nursing data from the patient case database, combine it with the current nursing data to perform trend analysis and time-series prediction, generate the optimal time window prediction information for the next nursing care and complication risk warning information, and update the prediction results to the patient case database.

[0007] Furthermore, the image denoising process based on wavelet transform in step S3 specifically includes: performing a three-level discrete wavelet transform on the image, decomposing the image signal into Symlets wavelet basis, applying a soft threshold function to shrink the high-frequency detail components, leaving the low-frequency approximation components unprocessed, and reconstructing the image by using the processed high-frequency and low-frequency information through inverse discrete wavelet transform to obtain the denoised image.

[0008] Furthermore, the image spatial coordinate registration in step S3 specifically includes: using the visible light image as a reference, the infrared image and the ultraviolet image are mapped to a unified analytical coordinate system with the visible light image using a feature point matching method, so as to ensure that the three images achieve pixel-level physical information alignment at spatial coordinate points.

[0009] Furthermore, the multimodal feature vector constructed for each pixel location in step S4 is represented as follows: in, For each pixel position The five-dimensional eigenvectors, , , These are color channel values ​​taken from a visible light image. This is the normalized thermal radiation energy index. This is the normalized fluorescence metabolic intensity operator.

[0010] Furthermore, in step S5, the quantification of swelling degree is calculated as follows: Each component in the five-dimensional feature vector is normalized to ensure its value falls within the [0,1] interval. Based on the physical characteristics of each spectral mode, a swelling height prediction model is constructed. The formula for calculating the swelling degree quantification index is: in, Contribute weight to thermal radiation For spatial geometric gradient weights, The rate of change of image morphological edges. This is the calibration constant.

[0011] Furthermore, in step S6, the random forest algorithm is... The system consists of several decision trees, each of which independently performs classification or regression prediction on the input state vector. A subset is randomly extracted from the feature vector space using a bootstrap sampling method to ensure that each decision tree learns nursing logic in different dimensions. Each non-leaf node of each decision tree uses information gain as the splitting criterion to automatically select the feature index that has the greatest impact on the current nursing decision for branching.

[0012] Furthermore, in step S6, the final option is determined by majority voting for discrete decision items such as cleaning fluid type, ointment application scheme, and dressing type, and the final output is the arithmetic mean of all decision tree predictions for continuous numerical items in the cleaning operation parameters.

[0013] Furthermore, in steps S8 and S9, the cleaning and dressing change operations are repeatedly performed multiple times based on the real-time visual feedback evaluation results of the stoma site until the real-time evaluation reaches the preset cleaning or dressing change standard.

[0014] According to a second aspect of the present invention, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the intelligent stoma care method for patients based on intelligent multispectral vision as described above.

[0015] According to a third aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, the computer program being configured to cause the computer to perform the intelligent patient stoma care method based on intelligent multispectral vision as described above.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention overcomes the limitations of human vision by introducing multispectral imaging technology that includes visible light, infrared and ultraviolet bands. It can comprehensively capture the structural, thermodynamic and metabolic information of stoma tissue. By constructing multimodal feature vectors and weighted mapping models, it achieves objective quantitative assessment of key physiological indicators such as stoma color, swelling degree, exudation nature and infection status, and completely overcomes the defects of traditional manual assessment, which is highly subjective and lacks unified standards. 2. This invention encodes quantitative physiological indicators into state vectors and uses a random forest algorithm to automatically generate a complete nursing plan that includes cleaning, medication administration, and dressing changes. This achieves the standardization and automation of nursing decisions, effectively reduces reliance on individual operational experience, and improves the consistency and reliability of nursing plans in different scenarios. 3. This invention constructs a closed-loop control logic of "real-time visual perception - dynamic deviation correction - automatic effect evaluation - on-demand repetitive execution" to ensure that nursing operations can be adaptively adjusted according to the real-time status of the stoma, which greatly improves the accuracy and thoroughness of cleaning and dressing changes, and effectively reduces the risk of secondary damage or infection caused by improper operation. 4. This invention systematically records multispectral images, quantitative indicators, and operational data throughout the entire process, and combines this with historical data for trend analysis and prediction. This enables a shift from a single "passive nursing" model to a continuous "active prediction" model, providing a solid data foundation for individualized precision nursing in clinical practice and scientific research data analysis. Attached Figure Description

[0017] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is an overall flowchart of a patient stoma intelligent care method based on intelligent multispectral vision as described in this invention; Figure 2 The nursing device described in this embodiment is an application of the intelligent stoma care method based on intelligent multispectral vision of the present invention.

[0018] 1-Device frame, 2-Base, 3-Nursing treatment bed, 4-Treatment tool library, 5-Medication dressing container, 6-Multi-spectral vision system, 7-Multi-functional multi-degree-of-freedom robotic arm, 8-Intelligent computing center. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other, and the described embodiments are only some embodiments of the present invention, not all embodiments.

[0020] Figure 1 This is an overall flowchart of a patient stoma intelligent care method based on intelligent multispectral vision as described in this invention.

[0021] like Figure 1 As shown, this intelligent stoma care method based on intelligent multispectral vision includes the following steps: In step S1, the overall posture image of the patient is acquired, and the detected patient posture is compared with the preset standard nursing position. If the posture does not meet the requirements, voice guidance prompts are output until the detected posture meets the requirements.

[0022] In step S2, multispectral image data of the patient's stoma site is acquired, wherein the multispectral image data includes at least synchronously acquired visible light images, infrared images, and ultraviolet images.

[0023] In some embodiments, to eliminate the influence of sensor wavelength dependence, grating efficiency, and lens transmittance during image acquisition, it is typically necessary to use a material with a flat spectral response to acquire a reference image. Subsequently, exposure is simultaneously driven by three sensors—visible light (RGB), infrared (IR), and ultraviolet (UV)—to obtain three raw images: .

[0024] In step S3, the multispectral image data is preprocessed. The preprocessing includes at least multimodal data correction, wavelet transform-based image denoising, and image spatial coordinate registration. Each modal image is mapped to a unified analytical coordinate system to obtain spatially aligned multispectral image data.

[0025] In some implementations, the image denoising process based on wavelet transform in step S3 specifically includes: performing a three-level discrete wavelet transform on the image, decomposing the image signal into a Symlets wavelet basis, applying a soft threshold function to shrink the high-frequency detail components, leaving the low-frequency approximation components unprocessed, and reconstructing the image by using the processed high-frequency and low-frequency information through inverse discrete wavelet transform to obtain the denoised image.

[0026] Specifically, multispectral imaging systems often suffer from high image noise, strong spatial correlation, and low contrast. Therefore, this invention employs wavelet denoising based on the transform domain to enhance the image, performing a three-level discrete wavelet transform. The image signal is decomposed into a Symlets wavelet basis, and high-frequency detail components are shrunk using a soft thresholding function, while low-frequency components are left unprocessed to preserve key structural information. The processed high and low frequency information is then reconstructed using inverse discrete wavelet transform to obtain the denoised image. .

[0027] In some implementations, the image spatial coordinate registration in step S3 specifically includes: using the visible light image as a reference, the infrared image and the ultraviolet image are mapped to a unified analytical coordinate system with the visible light image using a feature point matching method, so as to ensure that the three images achieve pixel-level physical information alignment at spatial coordinate points.

[0028] Specifically, due to physical spatial discrepancies among the sensors, it is necessary to map the images of each modality to a unified analytical coordinate system to obtain visible light images. Based on this, a feature point matching method is used to ensure that the three images achieve pixel-level physical information alignment at spatial coordinate points.

[0029] In step S4, based on the spatially aligned multispectral image data, a multimodal feature vector is constructed for each pixel location. The multimodal feature vector includes at least visible light color features, thermal radiation energy index, and fluorescence metabolic intensity operator.

[0030] In some embodiments, the multimodal feature vector constructed for each pixel location in step S4 is represented as follows: in, For each pixel position The five-dimensional eigenvectors, , , These are color channel values ​​taken from a visible light image. This is the normalized thermal radiation energy index. This is the normalized fluorescence metabolic intensity operator.

[0031] Specifically, in Extracting stoma contour and morphological edge features, in Extracting the extreme points of the thermal gradient distribution characterizing deep inflammation, in Extracting and characterizing bacterial metabolites for each pixel location. Construct a five-dimensional feature vector: in, For each pixel position The five-dimensional eigenvectors, , , These are color channel values ​​taken from visible light images, used to identify the general shape and color changes of the stoma. The normalized thermal radiation energy index is used to assess inflammatory responses or blood circulation in deep tissues. This is a normalized fluorescence metabolic intensity operator, which can be used to assist in the detection of specific types of bacterial metabolites or secretion boundaries.

[0032] In step S5, a semantic segmentation network is used to extract the stoma site mask, and based on the multimodal feature vector and the stoma site mask, a weighted mapping model is used to calculate standardized physiological indicators. The standardized physiological indicators include at least color parameters, quantification of swelling degree, classification of exudate properties, and determination of infection signs.

[0033] In some embodiments, in step S5, the quantification of swelling degree is calculated as follows: Each component in the five-dimensional feature vector is normalized to ensure its value falls within the [0,1] interval; based on the physical characteristics of each spectral mode, a swelling height prediction model is constructed; and the formula for calculating the swelling degree quantification index is: in, Contribute weight to thermal radiation For spatial geometric gradient weights, The rate of change of image morphological edges. This is the calibration constant.

[0034] Specifically, the U-Net semantic segmentation network is used to extract the stoma site mask. A linear or nonlinear weighted mapping model is established to transform the fused five-dimensional pixel feature vector into standardized physiological indicators. Taking tissue swelling degree as an example, each component in the five-dimensional vector is first normalized to ensure its value falls within the [0,1] range. Then, based on the physical characteristics of each spectral mode, a core operator for indicator calculation is constructed, and a swelling height prediction model is established to determine the swelling degree. The formula for calculating quantitative indicators is as follows: in, Contribute weight to thermal radiation For spatial geometric gradient weights, The rate of change of image morphological edges. This is a calibration constant. The formula outputs a quantified swelling height value by fitting the relationship between local temperature rise and spatial deformation.

[0035] In step S6, the standardized physiological indicators are encoded into state vectors, and a preliminary care plan is generated based on the random forest algorithm. The preliminary care plan includes at least the type of cleaning solution, cleaning operation parameters, ointment application plan, and dressing type.

[0036] In some embodiments, the random forest algorithm in step S6 is derived from... The system consists of several decision trees, each of which independently performs classification or regression prediction on the input state vector. A subset is randomly extracted from the feature vector space using a bootstrap sampling method to ensure that each decision tree learns nursing logic in different dimensions. Each non-leaf node of each decision tree uses information gain as the splitting criterion to automatically select the feature index that has the greatest impact on the current nursing decision for branching.

[0037] Specifically, based on the quantitative assessment results of the above steps, four physiological indicators were obtained: color parameter C, degree of swelling, etc. Properties of exudate (Serous, bloody, purulent) Signs of infection (Yes / No) Intelligent decision-making generates a preliminary care plan based on the following algorithms: Four key physiological indicators were encoded as state vectors: ; The random forest algorithm is used as the basic decision engine. The system consists of several decision trees, each independently paired with the input state vector. Classification and prediction are performed. The key feature is the use of a bootstrap sampling method to randomly extract a subset from the feature vector space, ensuring that each decision tree learns nursing logic across different dimensions, thereby improving the system's robustness.

[0038] For each non-leaf node of each decision tree, information gain is used as the splitting criterion to automatically select the feature indicator that has the greatest impact on the current nursing decision for branching. Feature vector The random forest rapidly descends along the branches of each tree to the leaf nodes. Each tree is trained using a random subset, so different trees may emphasize different physiological indicators. This mechanism enables the random forest to evaluate the stoma status from multiple dimensions and avoid the bias of a single perspective.

[0039] In some embodiments, in step S6, the final option is determined by majority voting for discrete decision items such as cleaning fluid type, ointment application scheme and dressing type, and the final output is the arithmetic mean of all decision tree predictions for continuous numerical items in the cleaning operation parameters.

[0040] Specifically, regarding the classification objective, such as which cleaning solution to choose... Is it necessary to apply a special ointment? Which type of dressing to choose? The majority voting method is used.

[0041] For regression targets, such as the cleaning method used, continuous pressure needs to be applied. The instruction is to use the average of all decision trees.

[0042] The final structured care vector set output by the random forest is as follows: The random forest model can automatically handle the nonlinear coupling between quantitative indicators, achieving more accurate automated decision-making than traditional logical judgment.

[0043] In step S7, the preliminary nursing plan is output to the human-computer interaction interface, and adjustment instructions and confirmation instructions are received from external input to generate the final nursing plan.

[0044] Specifically, to prevent algorithmic misjudgments, the generated preliminary nursing plan, key assessment data, and suggestions are displayed on the human-computer interaction interface for review by medical staff. Doctors or nurses can fine-tune the plan based on experience or directly confirm it, achieving a deep integration of artificial intelligence and professional experience.

[0045] In step S8, cleaning and dressing changes are performed sequentially for the stoma site according to the final care plan.

[0046] In step S9, during the cleaning and dressing change operations, real-time multispectral image data of the stoma site is continuously acquired, and closed-loop correction control is performed on the current operation process based on real-time visual feedback.

[0047] In some embodiments, in steps S8 and S9, the cleaning operation and dressing change operation are repeatedly performed multiple times based on the real-time visual feedback evaluation results of the stoma site until the real-time evaluation reaches the preset cleaning standard or dressing change standard.

[0048] In step S10, multispectral imaging data, physiological indicator assessment data, and final nursing plan data of the entire nursing process are recorded, and the data are updated to the patient case database.

[0049] In step S11, historical nursing data from the patient case database is retrieved, and trend analysis and time-series prediction are performed in conjunction with the current nursing data to generate the optimal time window prediction information for the next nursing care and complication risk warning information, and the prediction results are updated to the patient case database.

[0050] The following is based on Figure 2 The nursing device in the example demonstrates the specific application process of the present invention.

[0051] Figure 2 This embodiment describes a nursing system that utilizes the intelligent multispectral vision-based intelligent stoma care method of the present invention.

[0052] This nursing device includes: Device frame 1, used for integrating and installing various components and modules; Base 2 ensures the device is reliably and securely fixed to the ground; Nursing treatment bed 3 allows the patient to be in a position that facilitates treatment; Treatment tool library 4 provides various cleaning, dressing change and other operation tools that can be autonomously replaced by the robotic arm; Medication dressing bin 5 provides sterile disinfection and equipment cleaning materials that can be accessed by robotic arms; The multispectral vision system 6 is responsible for real-time acquisition of multispectral image data such as visible light, infrared, and ultraviolet at the patient's stoma location. Visible light imaging is used to identify the general shape of the stoma, color changes (such as red, pink, purple, and black), and the condition of the surrounding skin rash. Infrared imaging is used to capture differences in thermal radiation and to determine the inflammatory response or blood circulation status of deep tissues through temperature distribution. Ultraviolet imaging is used to assist in the detection of specific types of bacterial metabolites or secretion boundaries. The multi-functional, multi-degree-of-freedom robotic arm 7 is responsible for autonomously loading medications, dressings, etc., and changing operating tools to perform nursing operations. The Intelligent Computing Center 8 is primarily responsible for processing multispectral image data in real time using intelligent image processing methods, identifying patient posture and stoma site, issuing prompts through a voice interaction module, automatically assessing the stoma condition and generating a preliminary nursing plan, allowing for adjustments and confirmation of the nursing plan through human interaction, visually locating the stoma, controlling robotic arms to perform stoma cleaning, dressing changes, and other operations, cleaning and resetting equipment, recording nursing process images, operation plans, and other data, predicting future nursing plans, and storing all data in a case database for future review and analysis.

[0053] The specific process for applying this method is as follows: Step S1: After the device is started, a comprehensive hardware self-test is first performed to confirm that the communication of each module, such as the multi-functional multi-degree-of-freedom robotic arm 7, the multispectral vision system 6, the medicine dressing tank 5, and the treatment tool library 4, is normal and in good condition, in preparation for entering the working state. When the patient enters the nursing treatment bed 3, the multispectral vision system 6 starts a preliminary scan to detect the patient's current posture image. The intelligent computing center 8 compares the detected patient posture with the preset standard nursing posture. If the posture does not meet the requirements, a prompt is issued through the voice interaction module to guide the patient to adjust the posture until the detected posture meets the requirements, ensuring the safety and accessibility of the nursing process. Step S2: The device activates the multispectral vision system 6 to acquire multi-band images of the located stoma site. The intelligent computing center 8 uses intelligent image processing methods to fuse and analyze the acquired images, quantitatively assessing key physiological indicators such as stoma color parameters, tissue swelling, exudate properties, and the presence of signs of infection. To eliminate the influence of sensor wavelength dependence, grating efficiency, and lens transmittance during image acquisition, reference images are typically obtained using materials with a flat spectral response. Subsequently, the visible light (RGB), infrared (IR), and ultraviolet (UV) sensors are simultaneously driven for exposure to obtain three original images: ; Step S3: Multispectral imaging systems often exhibit high image noise, strong spatial correlation, and low contrast. Therefore, this invention selects wavelet denoising based on the transform domain for image enhancement, performing a three-level discrete wavelet transform on the image. The image signal is decomposed into a Symlets wavelet basis. High-frequency detail components are shrunk using a soft thresholding function, while low-frequency components are left unprocessed to preserve key structural information. The processed high and low frequency information is then reconstructed using inverse discrete wavelet transform to obtain the denoised image. Because of the physical spatial deviations between the sensors, it is necessary to map the images of each modality to a unified analytical coordinate system to obtain visible light images. Based on this, a feature point matching method is used to ensure that the three images achieve pixel-level physical information alignment at spatial coordinate points; Step S4: In Extracting stoma contour and morphological edge features, in Extracting the extreme points of the thermal gradient distribution characterizing deep inflammation, in Extracting and characterizing bacterial metabolites for each pixel location. Construct a five-dimensional feature vector: in, For each pixel position The five-dimensional eigenvectors, , , These are color channel values ​​taken from visible light images, used to identify the general shape and color changes of the stoma. The normalized thermal radiation energy index is used to assess inflammatory responses or blood circulation in deep tissues. This is a normalized fluorescence metabolic intensity operator, which can be used to assist in the detection of specific types of bacterial metabolites or secretion boundaries.

[0054] Step S5: Using the U-Net semantic segmentation network and extracting the stoma site mask, the system establishes a linear or nonlinear weighted mapping model to transform the fused five-dimensional pixel feature vector into standardized physiological indicators. Taking tissue swelling degree as an example, firstly, each component in the five-dimensional vector is normalized so that its value range is within the [0,1] interval. Then, based on the physical characteristics of each spectral mode, a core operator for indicator calculation is constructed, and the system establishes a swelling height prediction model to determine the swelling degree. The formula for calculating quantitative indicators is as follows: in, Contribute weight to thermal radiation For spatial geometric gradient weights, The rate of change of image morphological edges. This is a calibration constant. The formula outputs a quantified swelling height value by fitting the relationship between local temperature rise and spatial deformation.

[0055] Step S6: Based on the quantitative evaluation results of the above steps, four physiological indicators are obtained: color parameter C, degree of swelling, etc. Properties of exudate (Serous, bloody, purulent) Signs of infection (Yes / No) Intelligent decision-making generates a preliminary care plan based on the following algorithms: Four key physiological indicators were encoded as state vectors: ; The random forest algorithm is used as the basic decision engine. The system consists of several decision trees, each independently paired with the input state vector. Classification and prediction are performed. The key feature is that a subset is randomly selected from the feature vector space using a bootstrap sampling method, ensuring that each decision tree learns nursing logic in different dimensions, thereby improving the system's robustness. For each non-leaf node of each decision tree, information gain is used as the splitting criterion to automatically select the feature indicator that has the greatest impact on the current nursing decision for branching. Feature vector The random forest quickly descends along the branches of each tree to the leaf node. Each tree is trained using a random subset, so different trees may focus on different physiological indicators. This mechanism enables the random forest to evaluate the stoma status from multiple dimensions and avoid the bias of a single perspective. For classification objectives, such as which cleaning solution to choose. Is it necessary to apply a special ointment? Which type of dressing to choose? The majority voting method is adopted. For regression targets, such as the cleaning method used, continuous pressure needs to be applied. The instruction is to use the average value of all decision trees; The final structured care vector set output by the random forest is as follows: The random forest model can automatically handle the nonlinear coupling between quantitative indicators, achieving more accurate automated decision-making than traditional logical judgment. Step S7: The intelligent computing center 8 outputs the generated preliminary nursing plan to the human-computer interaction interface, receives adjustment and confirmation instructions from external input, and generates the final nursing plan. To prevent algorithm misjudgment, the generated preliminary nursing plan, key assessment data, and suggestions are displayed on the human-computer interaction interface for review by medical staff. Doctors or nurses can fine-tune the plan based on experience or directly confirm it, achieving a deep integration of artificial intelligence and professional experience. Step S8: According to the final care plan, the multi-functional multi-degree-of-freedom robotic arm 7 changes the cleaning tool in the treatment tool library 4 according to the instructions, takes the cleaning solution from the medicine dressing chamber 5, and performs the stoma cleaning operation and dressing change operation. Step S9: During the stoma cleaning and dressing change operations performed by the multi-functional multi-degree-of-freedom robotic arm 7, the multi-spectral vision system 6 continuously performs real-time visual tracking, providing the multi-functional multi-degree-of-freedom robotic arm 7 with high-precision stoma position coordinates. Based on real-time visual feedback, it performs closed-loop correction control on the current operation process, guiding it to perform the stoma cleaning operation. After cleaning, the multi-functional multi-degree-of-freedom robotic arm 7 returns to the treatment tool library 4 to change the dressing tools and retrieve the corresponding dressings and medications. Similarly, under the real-time feedback of the multi-spectral vision system 6, the multi-functional multi-degree-of-freedom robotic arm 7 accurately applies medication to the affected area and applies dressings around the stoma. Steps S8 and S9 can be repeated multiple times according to the actual situation of the stoma (such as excessive exudation or severe contamination) until the vision system assesses that the cleaning standard has been met. Step S10: The system automatically saves the multispectral images, robotic arm operation trajectory, various assessment parameters, and the final confirmed nursing plan for the entire nursing process; Step S11: The intelligent computing center 8 combines the current data with the patient's historical data in the database to perform trend analysis and deep learning, predict the optimal time window for the next nursing care, the risk of possible complications, and propose targeted preventive nursing care suggestions. All data from this nursing care and future prediction suggestions are updated to the patient's exclusive case database to form a continuous, complete, and traceable electronic medical record for subsequent clinical review, analysis, or research use. Finally, the system prompts that the nursing care is complete, and the nursing care process ends.

[0056] The intelligent stoma care method for patients based on intelligent multispectral vision proposed in this invention has the following beneficial effects: 1. This invention overcomes the limitations of human vision by introducing multispectral imaging technology that includes visible light, infrared and ultraviolet bands. It can comprehensively capture the structural, thermodynamic and metabolic information of stoma tissue. By constructing multimodal feature vectors and weighted mapping models, it achieves objective quantitative assessment of key physiological indicators such as stoma color, swelling degree, exudation nature and infection status, and completely overcomes the defects of traditional manual assessment, which is highly subjective and lacks unified standards. 2. This invention encodes quantitative physiological indicators into state vectors and uses a random forest algorithm to automatically generate a complete nursing plan that includes cleaning, medication administration, and dressing changes. This achieves the standardization and automation of nursing decisions, effectively reduces reliance on individual operational experience, and improves the consistency and reliability of nursing plans in different scenarios. 3. This invention constructs a closed-loop control logic of "real-time visual perception - dynamic deviation correction - automatic effect evaluation - on-demand repetitive execution" to ensure that nursing operations can be adaptively adjusted according to the real-time status of the stoma, which greatly improves the accuracy and thoroughness of cleaning and dressing changes, and effectively reduces the risk of secondary damage or infection caused by improper operation. 4. This invention systematically records multispectral images, quantitative indicators, and operational data throughout the entire process, and combines this with historical data for trend analysis and prediction. This enables a shift from a single "passive nursing" model to a continuous "active prediction" model, providing a solid data foundation for individualized precision nursing in clinical practice and scientific research data analysis.

[0057] This invention proposes an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the intelligent patient stoma care method based on intelligent multispectral vision.

[0058] This invention proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the intelligent patient stoma care method based on intelligent multispectral vision.

[0059] The memory in this application embodiment can be volatile memory or non-volatile memory, or it can include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DRRAM). It should be noted that the memory used in the methods described in this invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0060] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state drives (SSDs)).

[0061] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0062] It should be noted that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiments can be completed by the integrated logic circuitry in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0063] The above provides a detailed description of the intelligent stoma care for patients based on intelligent multispectral vision proposed in this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for intelligent stoma care of a patient based on intelligent multispectral vision, characterized in that: Includes the following steps: Step S1: Obtain the patient's overall posture image, compare the detected patient posture with the preset standard nursing position, and output voice guidance prompts if the posture does not meet the requirements until the detected posture meets the requirements. Step S2: Acquire multispectral image data of the patient's stoma site, wherein the multispectral image data includes at least synchronously acquired visible light images, infrared images, and ultraviolet images; Step S3: Preprocess the multispectral image data. The preprocessing includes at least multimodal data correction, wavelet transform-based image denoising, and image spatial coordinate registration. Each modal image is mapped to a unified analytical coordinate system to obtain spatially aligned multispectral image data. Step S4: Based on the spatially aligned multispectral image data, construct a multimodal feature vector for each pixel location. The multimodal feature vector includes at least visible light color features, thermal radiation energy index, and fluorescence metabolic intensity operator. Step S5: Extract the stoma site mask using a semantic segmentation network, and calculate standardized physiological indicators based on the multimodal feature vector and the stoma site mask using a weighted mapping model. The standardized physiological indicators include at least color parameters, quantified values ​​of swelling degree, classification of exudate properties, and results of infection sign determination. Step S6: Encode the standardized physiological indicators into state vectors, and generate a preliminary care plan based on the random forest algorithm. The preliminary care plan includes at least the type of cleaning solution, cleaning operation parameters, ointment application plan, and dressing type. Step S7: Output the preliminary nursing plan to the human-computer interaction interface, receive adjustment instructions and confirmation instructions from external input, and generate the final nursing plan; Step S8: According to the final nursing plan, perform cleaning and dressing changes for the stoma site in sequence; Step S9: During the cleaning and dressing change operations, real-time multispectral image data of the stoma site is continuously acquired, and closed-loop correction control is performed on the current operation process based on real-time visual feedback. Step S10: Record the multispectral imaging data, physiological indicator assessment data, and final nursing plan data throughout the entire nursing process, and update the data to the patient case database; Step S11: Retrieve historical nursing data from the patient case database, combine it with the current nursing data to perform trend analysis and time-series prediction, generate the optimal time window prediction information for the next nursing care and complication risk warning information, and update the prediction results to the patient case database.

2. A smart multi-spectral vision based intelligent stoma care method for patients as claimed in claim 1, wherein: The image denoising process based on wavelet transform in step S3 specifically includes: performing a three-level discrete wavelet transform on the image, decomposing the image signal into a Symlets wavelet basis, applying a soft threshold function to shrink the high-frequency detail components, leaving the low-frequency approximation components unprocessed, and reconstructing the image by using the processed high-frequency and low-frequency information through inverse discrete wavelet transform to obtain the denoised image.

3. A smart multi-spectral vision based intelligent stoma care method for patients as claimed in claim 1, wherein: The image spatial coordinate registration in step S3 specifically includes: using the visible light image as a reference, the infrared image and the ultraviolet image are mapped to the same analytical coordinate system as the visible light image using a feature point matching method, so as to ensure that the three images achieve pixel-level physical information alignment at spatial coordinate points.

4. The intelligent stoma care method based on intelligent multispectral vision according to claim 1, characterized in that: The multimodal feature vector constructed for each pixel location in step S4 is represented as follows: in, For each pixel position The five-dimensional eigenvectors, , , These are color channel values ​​taken from a visible light image. This is the normalized thermal radiation energy index. This is the normalized fluorescence metabolic intensity operator.

5. The intelligent stoma care method based on intelligent multispectral vision according to claim 1, characterized in that: In step S5, the degree of swelling is quantified as follows: Each component in the five-dimensional feature vector is normalized to ensure its value falls within the range of [0,1]. Based on the physical characteristics of each spectral mode, a swelling height prediction model is constructed. The formula for calculating the swelling degree quantification index is as follows: in, Contribute weight to thermal radiation For spatial geometric gradient weights, The rate of change of image morphological edges. This is the calibration constant.

6. The intelligent stoma care method based on intelligent multispectral vision according to claim 1, characterized in that: In step S6, the random forest algorithm is... The system consists of several decision trees, each of which independently performs classification or regression prediction on the input state vector. A subset is randomly extracted from the feature vector space using a bootstrap sampling method to ensure that each decision tree learns nursing logic in different dimensions. Each non-leaf node of each decision tree uses information gain as the splitting criterion to automatically select the feature index that has the greatest impact on the current nursing decision for branching.

7. The intelligent stoma care method based on intelligent multispectral vision according to claim 6, characterized in that: In step S6, the final option is determined by majority voting for discrete decision items such as cleaning fluid type, ointment application scheme and dressing type, and the final output is the arithmetic mean of all decision tree predictions for continuous numerical items in the cleaning operation parameters.

8. The intelligent stoma care method based on intelligent multispectral vision according to claim 1, characterized in that: In steps S8 and S9, the cleaning and dressing change operations are repeatedly performed based on the real-time visual feedback evaluation results of the stoma site until the real-time evaluation reaches the preset cleaning or dressing change standard.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, the processor executing the program to implement a smart stoma care method based on intelligent multispectral vision as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, It stores a computer program that enables the computer to perform a patient stoma intelligent care method based on intelligent multispectral vision as described in any one of claims 1-8.

Images