Method and system for assessing a patient's skin pressure ulcer risk image

Abstract

The application discloses a patient skin pressure sore risk image evaluation method and system, and relates to the field of medical image processing.The method comprises the following steps: acquiring a skin image by using a standardized acquisition device and performing color correction; converting the image to a Lab color space to extract a channel abnormal value, calculate a redness degree index and a red spot area proportion; calculating a roughness and uniformity index of skin texture based on a gray level co-occurrence matrix; registering and fusing a visible light image and an infrared thermal image to detect a temperature abnormal area; registering and comparing the current image with historical records to quantify a skin state change trend; and calculating a pressure sore risk score by weighted fusion of multi-dimensional features and outputting a nursing suggestion.The application realizes early warning and objective quantitative evaluation of pressure sore risk, and the accuracy rate is above 92%.

Description

Technical Field

[0001] This invention relates to the field of medical image processing and clinical nursing decision support technology, specifically to a method and system for image assessment of patient pressure ulcer risk. Background Technology

[0002] Pressure injuries, clinically known as pressure ulcers, are skin and subcutaneous tissue damage caused by prolonged pressure on local tissues, leading to impaired blood circulation and subsequent soft tissue ischemia, hypoxia, and malnutrition. Pressure ulcers are a major complication faced by long-term bedridden patients, elderly patients, patients in the postoperative recovery period, and patients with neurological injuries. They not only severely impact patients' quality of life and prolong hospital stays but also impose a heavy nursing burden and financial pressure on medical institutions. According to international pressure ulcer prevention guidelines, the incidence of pressure ulcers in hospitalized patients can reach 3% to 14%, while the incidence in intensive care unit patients is as high as 10% to 40%. Early identification of pressure ulcer risk and implementation of targeted preventive measures are crucial for reducing the incidence of pressure ulcers and improving patient prognosis.

[0003] Currently, clinical assessment of pressure ulcer risk primarily relies on standardized assessment scales, such as the Braden Scale, Norton Scale, and Waterlow Scale. These scales score patients based on dimensions such as sensory perception, mobility, nutritional status, friction and shear forces, and skin moisture to determine pressure ulcer risk levels. However, these assessment methods have significant limitations: First, the assessment results largely depend on the subjective judgment and clinical experience of nursing staff; different nurses may yield significantly different assessment results for the same patient, making it difficult to standardize assessment criteria. Second, traditional scales focus on assessing systemic risk factors, while neglecting direct observation and quantitative analysis of local skin conditions, making it difficult to capture early microscopic changes in the skin. Third, scale assessments typically require a significant amount of nursing staff's time; under conditions of limited nursing resources, high-risk patients may be missed due to untimely assessments.

[0004] With the development of computer vision and image processing technologies, image analysis-based pressure ulcer assessment methods have gradually attracted attention. Existing technologies include several image recognition-based pressure ulcer staging methods. For example, Chinese Patent Publication No. CN119515853A discloses an image recognition-based pressure ulcer staging method and device. This technical solution mainly includes: segmenting pressure ulcer images from acquired images of the affected area; identifying pressure ulcer features from the images, including ulcer color, tissue type and area, and skin integrity; and determining the pressure ulcer stage based on the identified features. Specifically, this method uses semantic segmentation technology to process the affected area image, analyzes the pressure ulcer color features using RGB color histograms, identifies tissue types using artificial neural networks, and uses a classifier or K-means clustering method to determine the pressure ulcer stage.

[0005] However, the aforementioned existing technical solutions still have the following technical problems to be solved: First, this method mainly targets the staging of existing pressure ulcers, which is a post-event diagnosis rather than a pre-event warning, and cannot meet the clinical needs for early risk identification of pressure ulcers; Second, the use of the RGB color space for color analysis is easily affected by lighting conditions and differences in patient skin color, and the accuracy and stability of erythema detection need to be improved; Third, it lacks a systematic analysis of skin texture characteristics and cannot identify early signs of pressure ulcers such as dry skin and desquamation; Fourth, it does not involve thermal imaging technology and cannot detect abnormal skin temperature caused by local blood circulation disorders; Fifth, it only performs a single static assessment and lacks the ability to track and analyze the dynamic changes in skin condition.

[0006] Therefore, there is an urgent need to develop an intelligent image assessment method and system that can provide early warning of pressure ulcer risk, has objective quantitative assessment capabilities, and can dynamically track changes in skin condition, so as to improve the pertinence and effectiveness of pressure ulcer prevention. Summary of the Invention

[0007] To address the aforementioned problems in existing technologies, this invention provides a method and system for image-based assessment of pressure ulcer risk in patients, which achieves intelligent early assessment of pressure ulcer risk through multi-feature analysis of skin images and risk quantification scoring.

[0008] The first aspect of this invention provides a method for image assessment of patient pressure ulcer risk, comprising the following steps: S1 Image acquisition and preprocessing step, using a standardized acquisition device equipped with a ring-shaped uniform light source and a color calibration plate, periodically photographing skin images of areas prone to pressure ulcers, and performing color correction processing on the skin images to ensure consistent imaging colors; S2 Color feature analysis step, converting the corrected skin image from RGB color space to Lab color space, extracting the pixel value distribution of the skin region in the a-channel, determining the outlier threshold based on the statistical characteristics of the a-channel of normal skin regions, identifying erythema regions and calculating the degree of redness index and erythema area ratio; S3 Skin texture feature extraction step, converting the skin image into a grayscale image, calculating the texture statistics of the skin surface based on the gray-level co-occurrence matrix, and calculating the coarse texture feature based on the texture statistics. Roughness and uniformity indices are used to identify early signs of pressure ulcers such as dry skin and desquamation. The S4 temperature thermal imaging fusion step acquires infrared thermal images simultaneously with visible light images, geometrically registers and overlays the visible light and infrared thermal images, detects abnormal temperature areas caused by local blood circulation disorders, and quantifies the degree of abnormality. The S5 dynamic tracking and comparison step registers and compares the currently acquired skin image with stored historical images, calculates the temporal changes in skin color and texture features, and quantifies the trend of skin condition changes. The S6 risk assessment output step performs multi-feature weighted fusion of redness index, erythema area ratio, roughness index, uniformity index, degree of temperature abnormality, and skin condition change trend to calculate a comprehensive pressure ulcer risk score, and outputs pressure ulcer stage prediction and nursing intervention recommendations based on the score.

[0009] The second aspect of this invention provides a patient pressure ulcer risk image assessment system, comprising: an image acquisition module equipped with a ring-shaped uniform light source, a color calibration plate, and an infrared thermal imaging sensor, used to periodically capture visible light skin images and infrared thermal images of common pressure ulcer sites, and to perform color correction processing on the visible light skin images; a color feature analysis module used to convert the corrected skin images from the RGB color space to the Lab color space, extract the outlier distribution of the a channel, and calculate the redness index and erythema area ratio; a texture feature extraction module used to convert the skin images into grayscale images and calculate texture statistics based on the gray-level co-occurrence matrix, and calculate roughness and uniformity indices based on the texture statistics; a thermal imaging fusion module used to geometrically register and overlay the visible light image and the infrared thermal image, and detect and quantify abnormal temperature areas; a dynamic tracking module used to register and compare the current skin image with historical images to quantify the trend of skin condition changes; and a risk assessment module used to perform multi-feature weighted fusion of various feature indices to calculate a comprehensive pressure ulcer risk score, and output pressure ulcer stage prediction and nursing intervention suggestions.

[0010] Compared with existing technologies, this invention has the following beneficial effects: First, it uses the Lab color space for erythema region analysis, and the a-channel can effectively separate red and green hue information, eliminating interference from changes in lighting and skin color differences, improving the accuracy of erythema detection by more than 35% compared with the RGB method; Second, it introduces gray-level co-occurrence matrix for skin texture analysis, which can identify early signs of pressure ulcers such as dry skin and desquamation, advancing the pressure ulcer risk identification time by 24 to 48 hours; Third, it integrates visible light and infrared thermal imaging technologies, which can detect temperature abnormalities caused by local blood circulation disorders, making up for the detection blind spots of single visible light imaging; Fourth, it establishes a dynamic tracking and comparison mechanism, realizing risk warning by quantifying the trend of skin condition changes, rather than relying solely on a single assessment result; Fifth, it adopts a multi-feature weighted fusion risk scoring model, achieving a comprehensive assessment accuracy of over 92%, providing objective and quantitative reference for clinical nursing decisions. Attached Figure Description

[0011] Figure 1 This is a flowchart of a patient skin pressure ulcer risk assessment method provided in an embodiment of the present invention.

[0012] Figure 2 This is an architecture diagram of the patient skin pressure ulcer risk image assessment system provided in an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0014] See Figure 1 As shown, this invention provides a method for image-based assessment of pressure ulcer risk in patients. This method achieves intelligent early assessment of pressure ulcer risk through multi-feature analysis of skin images and risk quantification scoring. The overall method includes six core steps, forming a deeply coupled closed-loop collaborative architecture: the output of the previous step serves as the key input for the next step, and the assessment results of the subsequent steps can inversely adjust the parameter thresholds of the previous steps, thereby achieving adaptive optimization.

[0015] Step S1: Image acquisition and preprocessing steps.

[0016] In one embodiment of the present invention, the core objective of the image acquisition and preprocessing steps is to obtain high-quality, color-consistent skin images, laying a reliable data foundation for subsequent feature analysis. Because lighting conditions vary considerably in the clinical environment, and different patients exhibit individual differences in skin color, standardized image acquisition and color correction are prerequisites for ensuring the accuracy and repeatability of the assessment results.

[0017] Specifically, this step involves using a standardized acquisition device equipped with a ring-shaped uniform light source and a color calibration plate to periodically photograph the patient's pressure ulcer-prone areas. Preferably, these areas include the sacrum, coccyx, heel, scapula, hip, and occipital region, which are clinically recognized as high-risk areas for pressure ulcers, especially important for monitoring patients who are bedridden for extended periods. The acquisition device is held at a distance of 15cm to 30cm from the skin surface, preferably 20cm, at which point clear images of skin details are obtained while covering a sufficient assessment area. The image resolution is set to 1920×1080 pixels or higher to ensure the capture of microscopic skin texture features.

[0018] Regarding the design of the annular uniform light source, in a preferred embodiment of this invention, the color temperature range of the annular light source is set to 5000K to 6500K. This color temperature range is close to natural sunlight, which can accurately reproduce skin color without introducing color shift. The color rendering index is not less than 95. A high color rendering index ensures that the red components of the skin can be accurately presented, which is crucial for the subsequent identification of erythema areas. The annular arrangement of the light source structure can eliminate shadows produced by unidirectional illumination, so that all areas of the skin surface receive uniform illumination, with the illuminance uniformity coefficient controlled above 0.90. The power range of the light source is 8W to 15W, ensuring sufficient illumination while avoiding excessive heat that may affect patient comfort.

[0019] The color calibration chart plays a crucial role in color correction in this invention. Preferably, the color calibration chart contains at least 24 standard color patches, covering a color range from neutral gray to saturated colors, including 6 grayscale color patches, 6 skin tone color patches, and 12 color patches. The Lab color space reference value of each color patch is precisely calibrated and serves as the benchmark for color correction. Before each image acquisition, a reference image containing the color calibration chart is first captured. By comparing the deviations between the measured values ​​of the color patches in the reference image and the standard values, a color correction matrix is ​​established.

[0020] The specific process for color correction is as follows: First, white balance correction is performed on the acquired original image, adjusting the white point of the image to the standard value using a neutral gray block as a reference. Then, a color correction matrix is ​​applied to transform the entire image, eliminating color shifts caused by differences in device response and lighting fluctuations. After correction, the coefficient of variation for color consistency in the skin images is controlled within 3%, meaning that for images of the same skin condition acquired at different time points, the color value difference is less than 3%, providing a reliable guarantee for subsequent time-series comparative analysis.

[0021] The data collection interval is differentiated based on the patient's risk level. For high-risk patients, it is recommended to collect data every 4 to 6 hours; for medium-risk patients, every 8 to 12 hours; and for low-risk patients, every 12 to 24 hours. This differentiated collection strategy ensures close monitoring of high-risk patients while also allowing for the rational allocation of nursing resources. Collected images are simultaneously stored in a database, indexed by patient identifier and collection time, facilitating subsequent historical comparison and analysis.

[0022] Preferably, environmental factors need to be controlled and recorded during image acquisition. The ambient temperature should be maintained between 20°C and 26°C. Too low a temperature may cause vasoconstriction in the skin, affecting erythema manifestation, while too high a temperature may cause false positives due to skin redness. The ambient humidity should be maintained between 40% and 60%. An excessively dry environment may exacerbate skin dryness, affecting the baseline for texture assessment. Environmental parameters should be recorded at each acquisition as a reference for interpreting the assessment results.

[0023] Furthermore, this invention also incorporates an automatic image quality detection mechanism. After acquisition, the system automatically evaluates the image's sharpness, exposure, and color integrity. Sharpness assessment uses the Laplacian operator to calculate the image gradient variance; when the variance value is below a preset threshold, it indicates the image is blurry and requires re-acquisition. Exposure assessment analyzes the histogram distribution to determine if the image is overexposed or underexposed; when the proportion of highlights or shadows exceeds 15%, it indicates an exposure anomaly. Color integrity assessment ensures the reliability of color correction by detecting the color block recognition rate in the color chart area; when the color block recognition rate is below 90%, it indicates the need to adjust the acquisition angle or light source position. Only images that pass the quality detection process proceed to subsequent analysis, ensuring the reliability of the evaluation data from the source.

[0024] Step S2: Color feature analysis step.

[0025] In one embodiment of the present invention, the core innovation of the color feature analysis step lies in using the Lab color space for the detection and quantitative analysis of erythema regions. Compared with the traditional RGB color space, the Lab color space has the advantage of perceptual uniformity, and its a channel specifically represents red-green hue information, which can effectively separate skin tone base and erythema variations, thereby improving the accuracy and stability of erythema detection.

[0026] First, the color-corrected skin image is converted from the RGB color space to the Lab color space. The RGB to Lab conversion is a non-linear process, requiring two steps: RGB to XYZ and then to Lab. Since this conversion follows the standard method of the International Commission on Illumination (ICI) and is a well-known technique in the field, the specific formula will not be detailed here. After the conversion, each pixel receives three channel values: the L channel represents luminance, ranging from 0 to 100; the a channel represents red-green hue, ranging from -128 to 127, with positive values ​​leaning towards red and negative values ​​towards green; and the b channel represents blue-yellow hue, ranging from -128 to 127, with positive values ​​leaning towards yellow and negative values ​​towards blue.

[0027] In the pressure ulcer risk assessment application of this invention, the α-channel is a key analytical object. Skin redness is one of the typical early manifestations of pressure ulcers, characterized by a significantly higher α-channel value in the local area compared to the surrounding normal skin area. To accurately identify erythematous areas, this invention designs an adaptive threshold detection algorithm based on the statistical characteristics of normal skin.

[0028] Specifically, a normal skin reference area is first selected in the skin image. The selection of the reference area follows these principles: it is located at the edge of the area to be evaluated, has no obvious redness or pigmentation abnormalities, and its area is no less than 10% of the total evaluation area. Within the selected normal skin reference area, the a-channel values ​​of all pixels are extracted, and their mean is calculated. and standard deviation .

[0029] Channel A outlier detection threshold The calculation formula is as follows: ,in, This represents the mean pixel value of the a-channel in normal skin areas, with a typical value range of 5 to 25, reflecting the basic red component of patients with different skin tones. This represents the standard deviation of the pixel values ​​of the a-channel in normal skin areas. Its typical range is 2 to 8, reflecting the natural fluctuation of the a-channel values ​​in normal skin areas. This is the threshold coefficient, ranging from 2.5 to 3.5, with a preferred value of 3.0. A larger value should be selected. Values ​​can reduce the false positive rate but may miss mild erythema; therefore, smaller values ​​should be selected. While higher threshold values ​​can improve detection sensitivity, they may increase false positives; a balance must be struck based on clinical needs in practical applications. The technical advantage of this threshold setting method lies in its adaptability to patients with different skin tones, eliminating interference from individual skin color differences in erythema detection.

[0030] Iterate through each pixel within the skin evaluation area, when the pixel's a-channel value Exceeding the outlier threshold When a pixel is identified as a erythema region, it is determined to belong to that region. After traversal processing, all pixels identified as erythema regions form an erythema region mask. A value of 1 represents a erythematous pixel, and a value of 0 represents a non-erythematous pixel.

[0031] Based on the erythema region mask, this invention calculates two key color feature indicators: the degree of redness. and the percentage of erythema area Redness level index The calculation formula is as follows: ,in, The total number of pixels in the erythema region, when hour, Set it directly to 0; is the normalization coefficient, with a value of 20, used to normalize the redness index to the range of 0 to 1; pixels within the erythema area The a-channel value; The threshold for outlier determination is calculated above. The physical meaning of this formula is to calculate the average offset of each pixel in the erythema region that exceeds the threshold, and then normalize it. A value closer to 1 indicates a more severe degree of redness. The redness severity index can distinguish between mild redness and severe congestion, providing a quantitative basis for risk level determination.

[0032] Percentage of erythema area The calculation formula is as follows: ,in, This represents the number of pixels in the erythema region. This represents the total number of pixels in the skin analysis area. This reflects the size of the erythema within the assessment area; a larger area generally indicates a higher risk of pressure ulcers. Clinical experience suggests that when... A value exceeding 15% requires close attention.

[0033] The color feature analysis method of this invention has significant technical advantages over existing technologies. In a test on 2000 clinical images, the erythema detection method based on the Lab color space a channel improved the detection accuracy from 68% to 92% and reduced the false positive rate from 22% to 8% compared to the RGB color histogram method. The detection stability was particularly improved in patients with darker skin tones.

[0034] In a further preferred embodiment of the present invention, the color feature analysis also includes morphological analysis of the erythema region. After identifying the erythema region, connected component analysis is performed on the erythema region to extract morphological feature parameters of the erythema. These parameters include the number of erythema regions, reflecting whether the redness is concentrated or dispersed; the maximum connected component area of ​​the erythema, reflecting the extent of the main erythema; the edge regularity of the erythema, characterized by a roundness coefficient, which is equal to the ratio of area to the square of perimeter multiplied by a constant 4π, with a value ranging from 0 to 1. The closer the value is to 1, the closer the shape is to a circle. Early pressure ulcer erythema is usually a relatively regular circle or ellipse, while irregular shapes may indicate other skin problems; the spatial distribution characteristics of the erythema, calculating the distance between the centroid of the erythema region and the center of the assessment region, as well as the relative positional relationship between multiple erythema regions. These morphological feature parameters provide richer information dimensions for clinical diagnosis and help distinguish early pressure ulcer erythema from skin redness caused by other reasons.

[0035] Furthermore, this invention also incorporates a time-based analysis of erythema changes. By comparing the changes in the location and extent of the erythema area between two consecutive assessments, it determines whether the erythema is persistent or transient. Persistent erythema refers to erythema that repeatedly appears in the same location in two or more consecutive assessments, indicating that the local tissue is under continuous pressure and has a high risk of pressure ulcers. Transient erythema refers to erythema that appears and subsides within a short period of time, which is usually a normal physiological response and does not require excessive attention. This time-dimensional analysis further enhances the clinical relevance of erythema detection.

[0036] Step S3: Skin texture feature extraction step.

[0037] In one embodiment of this invention, the skin texture feature extraction step aims to identify early warning signs of pressure ulcers, such as dryness, peeling, and roughness, by analyzing the microscopic texture structure of the skin surface. These warning signs often appear before visible erythema, providing an earlier intervention opportunity for pressure ulcer risk warning. This invention uses a gray-level co-occurrence matrix as the core tool for texture analysis. This method can capture the spatial correlation of pixel gray-level values ​​in an image, thereby characterizing the texture properties of the skin surface.

[0038] First, the skin image is converted to a grayscale image. The grayscale conversion uses a weighted average method, weighting the RGB channels based on their respective contributions to human eye brightness perception. The conversion formula is as follows: The converted grayscale image retains the light and dark details of the skin surface, providing basic data for texture analysis. Preferably, the grayscale level is quantized to 256 levels, that is, the grayscale value ranges from 0 to 255, ensuring sufficient grayscale resolution to capture subtle texture changes.

[0039] The construction of the gray-level co-occurrence matrix (GLCM) is the core step in this process. The GLCM describes the gray-level distribution of pixel pairs with specific spatial relationships in an image; its mathematical definition is: given a pixel spacing... and direction angle Under the given conditions, calculate the grayscale value. The pixel and grayscale values The probability of pixels appearing at the same time .

[0040] In a preferred embodiment of the present invention, the construction parameters of the gray-level co-occurrence matrix are set as follows: pixel spacing. The value is set to 1 to 5 pixels, with preferred values ​​of 1 and 3, to capture texture features at the micro and meso scales, respectively. The calculation directions include four directions: 0 degrees, 45 degrees, 90 degrees, and 135 degrees. Multi-directional calculations can yield isotropic texture descriptions. For each group... The parameters are combined to construct a 256×256 co-occurrence matrix, and the matrix elements are... Indicates direction ,distance Under these conditions, grayscale values Normalized frequency of occurrence.

[0041] Based on the constructed gray-level co-occurrence matrix, this invention extracts four texture statistics: contrast, entropy, energy, and homogeneity. The calculation methods for these four statistics belong to the classic texture descriptor proposed by Haralick and are well-known techniques in the field. Contrast reflects the magnitude of the gray-level difference between adjacent pixels in an image; a higher contrast value indicates more drastic local gray-level changes and a rougher skin surface. Entropy reflects the randomness and complexity of the image's gray-level distribution; a higher entropy value indicates a more irregular texture. Energy, also known as the second moment of the angle, reflects the uniformity of the image's gray-level distribution; a higher energy value indicates a more uniform texture. Homogeneity reflects the similarity of gray levels in local areas of the image; a higher homogeneity value indicates a smoother local texture.

[0042] The innovation of this invention lies in the design of a calculation method for the comprehensive index of skin roughness and the comprehensive index of skin uniformity, which integrates multi-directional and multi-scale texture statistics into clinically interpretable quantitative indicators.

[0043] Comprehensive index of skin roughness The calculation formula is as follows: ,in, In this embodiment, the total number of direction-scale combinations involved in the calculation is... This corresponds to 2 scales multiplied by 4 directions; For the first Contrast value under group parameters and These are the normalized lower and upper bounds of the contrast, respectively, with values ​​ranging from 0 to 10000, determined through statistical analysis of a large number of clinical images. For the first Entropy value under group parameters and These are the normalized lower and upper bounds of entropy, respectively, with values ​​ranging from 0 to 8; and These are the weighting coefficients for contrast and entropy, respectively, both defaulting to 0.5 and can be adjusted based on clinical validation results. (Normalized) The value ranges from 0 to 1, with higher values ​​indicating rougher skin surfaces and more severe dryness and flaking. This index quantifies the roughness of skin texture. A value exceeding 0.6 indicates a significant skin dryness problem.

[0044] Skin evenness composite index The calculation formula is as follows: ,in, The total number of direction-scale combinations involved in the calculation; For the first The energy values ​​under the group parameters already range from 0 to 1, so no additional normalization is needed; For the first The homogeneity value under the group parameter also ranges from 0 to 1; and These are the weighting coefficients for energy and homogeneity, respectively, both defaulting to 0.5. The value ranges from 0 to 1; a larger value indicates a more even and smooth skin surface, while a smaller value indicates a more irregular skin texture. A value below 0.35 indicates decreased skin surface uniformity, which may indicate localized scaling or irregular texture changes.

[0045] The effectiveness of texture feature extraction technology is reflected in its ability to identify early signs of pressure ulcers. In clinical validation, by continuously monitoring skin texture indices in high-risk patients, roughness indices were identified. and uniformity index Abnormal changes typically appear 24 to 48 hours earlier than visible skin changes, giving caregivers a valuable window of opportunity to take preventative measures.

[0046] In a further preferred embodiment of the present invention, texture feature extraction further includes local binary pattern analysis as a supplement to the gray-level co-occurrence matrix (GLCM). Local binary pattern analysis is a texture description method based on comparison of local neighboring pixels, generating feature codes by comparing the gray-level relationship between the central pixel and its surrounding pixels. Compared to the GLCM, local binary pattern analysis is more robust to illumination changes and can capture the fine texture structure of the skin surface. Specifically, a variant of uniform local binary pattern is used, comparing eight pixels within a 3×3 neighborhood with the central pixel to generate an 8-bit binary code, and statistically analyzing the frequency of each coding pattern to form a histogram feature. The combined use of local binary pattern features and GLCM features improves the comprehensiveness and accuracy of texture analysis.

[0047] Furthermore, this invention also designs a multi-scale texture analysis strategy. Abnormal changes in skin texture can occur at different spatial scales. Changes at the microscale reflect changes in skin surface roughness, changes at the mesoscale reflect changes in the direction and distribution of skin texture, and changes at the macroscale reflect changes in the overall state of the skin region. By extracting texture features at multiple scales separately and then performing scale fusion, a more comprehensive texture representation can be obtained. In this embodiment, a Gaussian pyramid is used to perform three-layer downsampling of the image, extracting texture features at three scales: the original resolution, half the resolution, and one-quarter the resolution. Finally, the features from the three scales are weighted and fused with weights of 0.5, 0.3, and 0.2, respectively, prioritizing details at the microscale.

[0048] The threshold for abnormal texture features was determined based on large-scale clinical data statistics. Through statistical analysis of the texture features of 500 normal skin samples and 500 samples of skin showing early signs of pressure ulcers, critical thresholds for roughness and uniformity indices were determined. When the roughness index... Exceeding 0.55 or uniformity index When the value falls below 0.40, the system issues an early warning of abnormal skin texture. These thresholds were selected to balance sensitivity and specificity, achieving a sensitivity of 88% and a specificity of 85%.

[0049] Step S4: Temperature thermal imaging fusion step.

[0050] In one embodiment of the present invention, the temperature thermal imaging fusion step utilizes infrared thermal imaging technology to detect local circulatory disturbances. Circulatory disturbances are one of the core pathological mechanisms of pressure ulcer formation; reduced blood flow to the pressure site leads to local temperature changes, which often precede visible changes in skin appearance. Therefore, thermal imaging fusion analysis can provide information on deep tissue conditions that cannot be obtained from visible light images.

[0051] First, an infrared thermal image is acquired synchronously with the visible light image. In a preferred embodiment of the invention, the infrared thermal imaging camera has a wavelength range of 8μm to 14μm, belonging to the long-wave infrared band, which can effectively detect thermal radiation from the human skin surface. The thermal imaging resolution is 640×480 pixels or higher, the temperature resolution is better than 0.05℃, and the temperature measurement range is 25℃ to 42℃, covering the range of normal body temperature and abnormal temperature changes. The infrared thermal imaging camera and the visible light camera adopt a coaxial integrated design or are installed in a fixed relative position to ensure that the two images can be accurately registered.

[0052] Geometric registration of visible light images and infrared thermal images is a prerequisite for fusion analysis. Due to the different imaging principles and field-of-view characteristics of the two types of sensors, direct superposition will result in misalignment. This invention employs an affine transformation method based on feature points for geometric registration. The specific process is as follows: First, feature points are detected in both the visible light image and the infrared thermal image. These feature points are typically located at skin edges or anatomical landmarks. Then, feature point matching is performed to establish the point correspondence between the two images. Finally, affine transformation parameters are calculated based on the matched feature points to transform the infrared thermal image to the same coordinate system as the visible light image. The residual error after registration should be controlled within 3 pixels to ensure the spatial accuracy of the correspondence between temperature information and visible light information.

[0053] In the registered and fused image, each pixel location simultaneously contains visible light information. and temperature information This invention designs an algorithm for detecting abnormal temperature regions, used to identify localized temperature anomalies caused by impaired blood circulation.

[0054] The basic principle of temperature anomaly detection is to compare the temperature difference between the target area and the surrounding normal skin area. Specifically, for each pixel location within the area to be evaluated... Calculate its temperature value Average temperature of the surrounding reference area Difference values: ,in, For position The infrared thermal image temperature value, in °C; The average temperature of the surrounding normal skin reference area is used. The selection method of the reference area is consistent with the selection principle of the normal skin area in step S2. The value range is usually 32℃ to 37℃, reflecting the surface temperature of normal skin.

[0055] Temperature anomaly threshold Used to determine the presence of temperature anomalies, the value ranges from 1.5℃ to 2.5℃, with a preferred value of 2.0℃. A temperature anomaly zone is determined when the temperature difference meets the following conditions: When this is detected, it is considered an abnormal temperature. It's important to note that abnormal temperatures include both elevated and decreased temperatures. Elevated temperature typically indicates an inflammatory response or a risk of infection. This suggests local congestion or inflammatory exudation; decreased temperature usually indicates impaired blood circulation or tissue ischemia. This indicates insufficient local blood perfusion, a direct cause of pressure ulcer formation. This invention simultaneously detects these two types of temperature abnormalities and labels and quantifies them separately.

[0056] The degree of temperature anomaly is quantified using the following formula: ,in, The total number of pixels in the region of interest; This is an indicator function that takes the value 1 when the condition is met, and 0 otherwise; This represents the temperature anomaly threshold. It reflects the severity of temperature anomalies. Theoretically, the value range starts from 0 with no upper limit, but it is normalized to the 0 to 1 range through saturation processing to facilitate subsequent multi-feature fusion calculations.

[0057] The output of the thermal imaging fusion step includes: a spatial distribution mask of temperature anomaly regions, a type marker of the temperature anomaly (increase / decrease), and a quantification value of the degree of temperature anomaly. This information provides an important supplementary dimension for the comprehensive assessment of pressure ulcer risk. In clinical validation, the introduction of thermal imaging fusion analysis improved the detection rate of early signs of deep tissue injury by 40%, significantly compensating for the limitations of single visible light image analysis.

[0058] In a further preferred embodiment of the present invention, thermal imaging fusion further includes temperature gradient analysis. In addition to detecting temperature anomaly regions, the rate of change of temperature gradient at the edge of the anomaly region is calculated. The temperature gradient is defined as the amount of temperature change per unit distance from the center of the anomaly region to its edge, expressed in °C / cm. Steep temperature gradients typically indicate well-defined local circulatory disturbances, while gentle temperature gradients may indicate progressive tissue changes. The temperature gradient is calculated as follows: temperature values ​​are sampled every 1cm along the direction radiating outwards from the center of the anomaly region; the temperature difference between adjacent sampling points is linearly fitted, and the slope of the fitted line is the temperature gradient value. When the absolute value of the temperature gradient exceeds 0.8 °C / cm, it indicates the presence of a clear boundary of a local circulatory disturbance.

[0059] Furthermore, this invention incorporates a time-series analysis function for thermal imaging. By comparing multiple consecutively acquired infrared thermal images, the evolution trend of abnormal temperature regions can be observed. Changes in the area of ​​abnormal temperature regions, changes in temperature difference values, and changes in temperature gradients are all important dynamic indicators. If the abnormal temperature region continues to expand or the temperature difference value continues to increase, it indicates a deterioration in local blood circulation, requiring close attention. Conversely, if the abnormal temperature region gradually shrinks or the temperature difference value gradually decreases, it suggests that nursing interventions are taking effect. This time-series analysis provides an objective basis for evaluating nursing effectiveness.

[0060] In practical applications, infrared thermal imaging also needs to consider the elimination of interference factors. The evaporation of moisture from the skin surface carries away heat, causing a local temperature decrease. This decrease needs to be distinguished from the temperature decrease caused by impaired blood circulation. This invention reduces the impact of surface moisture evaporation on temperature measurement results by gently wiping the skin surface before acquiring infrared thermal images and waiting for a 30-60 second equilibration period. Furthermore, ambient airflow can also lead to uneven distribution of skin surface temperature; therefore, acquisition should be avoided near ventilation openings or air conditioning vents, and a shield should be used if necessary to reduce airflow interference.

[0061] Step S5: Dynamic tracking and comparison step.

[0062] In one embodiment of the present invention, the dynamic tracking and comparison step achieves quantitative analysis of skin condition change trends by registering and comparing the currently acquired skin image with historical images. Compared with static single assessment, dynamic tracking can capture gradual changes in skin condition, and even if the result of a single assessment has not yet reached the abnormal threshold, the continuous change trend can provide an important reference for risk warning.

[0063] First, historical images of the same area from the current patient are retrieved from the database. In a preferred embodiment of the invention, the acquisition time interval for historical images is 4 to 24 hours, and the number of historical images stored is the most recent 3 to 7 acquisitions. The time interval and the number of stored images can be adjusted according to the patient's risk level: shorter intervals and more historical images are used for high-risk patients, while the intervals can be appropriately relaxed for low-risk patients.

[0064] Registration of the current image with historical images is a crucial step in dynamic tracking. Since patient positioning may change slightly, images acquired at different times may exhibit spatial displacement and rotation. This invention employs a feature point matching method based on scale-invariant feature transformation for image registration. This method is robust to image scaling, rotation, and some viewpoint changes. The registration process is as follows: First, feature points and their descriptors are extracted from both the current and historical images; then, feature point matching is performed, eliminating mismatched point pairs; finally, the perspective transformation matrix is ​​calculated based on the matched point pairs to transform the historical image to align with the current image.

[0065] After registration, the changes in each feature dimension between the current image and historical images are calculated. The feature dimensions tracked in this invention include: color features (redness index). Percentage of erythema area Texture features (roughness index) Uniformity index ) and temperature characteristics (degree of temperature anomaly) ).

[0066] Skin condition change trend The comprehensive calculation formula is as follows: ,in, The number of historical records to be included in the calculation, with a value ranging from 3 to 7; For time point indexing, This represents the earliest historical record. Indicates the current time point; This is a time decay factor, with a value ranging from 0.7 to 0.9, and a preferred value of 0.8, used to give higher weight to changes at more recent time points; Indicates the first The time point relative to the first Changes in redness level at specific time points, other The items have similar meanings; to These are the weighting coefficients for each feature variation, with default values ​​of 0.25, 0.20, 0.20, 0.15, and 0.20, respectively. They can be optimized and adjusted based on clinical validation results.

[0067] Theoretically, the value range starts from 0 with no upper limit. In practical applications, it is mapped to a scoring range of 0 to 100 through normalization. The higher the value, the more drastic the changes in skin condition and the higher the risk. In particular, when redness continues to increase, roughness continues to increase, or evenness continues to decrease, the system will issue a risk warning based on the trend of change, even if the individual indicator has not exceeded the warning threshold.

[0068] The advantages of dynamic tracking and comparison technology are reflected in the following aspects: First, it can identify progressive deterioration of skin condition, ensuring that even small changes are not overlooked; second, the time decay mechanism makes the system more sensitive to recent changes, enabling timely detection of acute risks; third, multi-feature joint tracking can discover complex change patterns that may be missed by single-feature tracking. In clinical validation, the dynamic tracking and comparison mechanism increased the average lead time for pressure ulcer risk warning by 48 hours.

[0069] In a further preferred embodiment of the present invention, the dynamic tracking and comparison also includes a change pattern recognition function. By analyzing the combination patterns of changes in various feature dimensions, typical pressure ulcer risk evolution patterns are identified. The first pattern is a single-feature continuous deterioration pattern, characterized by a certain feature index showing a monotonically increasing or decreasing trend in multiple consecutive assessments, while other features remain relatively stable. This pattern usually appears in the early stages of pressure ulcer risk; for example, a continuous increase in roughness index indicates a gradual worsening of skin dryness. The second pattern is a multi-feature synergistic deterioration pattern, characterized by multiple feature indicators showing a simultaneous deterioration trend; for example, increased redness accompanied by abnormal temperature. This pattern indicates that the pressure ulcer risk is rapidly escalating and requires immediate intervention. The third pattern is a fluctuating change pattern, characterized by feature indicators repeatedly fluctuating between normal and abnormal ranges. This pattern may be related to the intermittent implementation of nursing measures, indicating a need to strengthen the continuity and standardization of nursing care.

[0070] Furthermore, this invention also designs a graded early warning mechanism for abnormal changes. Based on the magnitude and rate of change of the characteristic, the early warning is divided into three levels. Level 1 is the attention level, triggered when the change in a single characteristic in a single assessment exceeds 1.5 times the normal fluctuation range of that characteristic, indicating the need for closer observation. Level 2 is the warning level, triggered when the changes in two or more characteristics simultaneously exceed 1.5 times the normal fluctuation range in a single assessment, or when the change in a single characteristic exceeds 2.5 times the normal fluctuation range, indicating the need for preventative measures. Level 3 is the emergency level, triggered when the change in the same characteristic in two consecutive assessments exceeds 2 times the normal fluctuation range, or when a multi-characteristic synergistic deterioration pattern appears, indicating the need for immediate professional assessment and intervention. The early warning level is directly linked to nursing recommendations, providing tiered guidance for clinical decision-making.

[0071] Dynamic tracking and comparison also supports trend extrapolation and prediction. Based on historical trends, it uses linear extrapolation or exponential smoothing methods to predict the possible range of values ​​for each characteristic indicator within a future assessment period. If the predicted value exceeds the risk threshold, a predictive warning is issued in advance, enabling caregivers to take proactive intervention measures. This predictive warning function transforms risk identification from a passive response to proactive prevention, further improving the effectiveness of pressure ulcer prevention efforts.

[0072] Step S6: Risk assessment output steps.

[0073] In one embodiment of the present invention, the risk assessment output step weighted and fused the multidimensional features extracted in the preceding steps to calculate a comprehensive pressure ulcer risk score, and based on this score, outputs pressure ulcer staging predictions and nursing intervention recommendations. This step is the core of the entire assessment method and directly relates to its clinical application value.

[0074] Multi-feature weighted fusion employs an adaptive weight allocation strategy. Compared to fixed-weight schemes, adaptive weights can be dynamically adjusted based on the reliability and importance of each feature, improving the accuracy and robustness of the evaluation results.

[0075] Pressure ulcer risk assessment The calculation formula is as follows: The meanings and value ranges of each feature term are as follows: This is an indicator of redness, with a value ranging from 0 to 1, reflecting the severity of redness in the erythematous area; The percentage of erythema area ranges from 0 to 100%, and is normalized to the range of 0 to 1 after being divided by 100. It is a roughness index, with a value ranging from 0 to 1, reflecting the roughness of the skin surface; As the complement of the uniformity index, since A higher value indicates healthier skin, therefore the supplementary value should be positively correlated with risk. The degree of temperature anomaly, after normalization, ranges from 0 to 1; The value represents the trend of skin condition changes, ranging from 0 to 100, and is normalized to the 0 to 1 range after being divided by 100.

[0076] Weighting coefficients to The range of values ​​and preferred values ​​are as follows: The weight of redness is 0.20 to 0.30, with a preferred value of 0.25. Redness is one of the most obvious signs of early pressure ulcers, so it is given a higher weight. The weight of erythema area ranges from 0.15 to 0.25, with a preferred value of 0.18. The erythema area reflects the size of the affected area. The roughness weight ranges from 0.10 to 0.20, with a preferred value of 0.15. Changes in skin texture are an early precursor. The (uniformity weight) value ranges from 0.10 to 0.15, with a preferred value of 0.12, which complements the roughness and reflects the texture state; The (temperature anomaly weight) ranges from 0.15 to 0.25, with a preferred value of 0.18. Temperature anomalies directly reflect the state of blood circulation. The (trend weight) ranges from 0.10 to 0.20, with an optimal value of 0.12. Dynamic trends are crucial for early warning. The sum of all weight coefficients is constrained to 1 to ensure the normalization of the score.

[0077] The adaptive adjustment mechanism for weight coefficients follows these principles: when the detection confidence of a feature is low, its weight is reduced to mitigate the impact of uncertainty; when a feature exhibits significant anomalies, its weight is appropriately increased to highlight key risk factors. For example, when the thermal imaging image quality is poor or occlusion exists, the weight is automatically reduced. The weighting; when the redness index rises sharply, the weighting should be appropriately increased. The weight.

[0078] Calculated The score range is from 0 to 100. The risk levels are divided according to the score range as follows: 0 to 30 points correspond to normal condition, with good skin condition and no special intervention required; 31 to 50 points correspond to low risk, with slight abnormal signals, and it is recommended to strengthen observation and basic care; 51 to 70 points correspond to medium risk, with multiple indicators showing abnormalities, and it is recommended to take preventive measures and increase the monitoring frequency; 71 to 100 points correspond to high risk, with significant deterioration of skin condition, and it is recommended to take stress reduction measures and professional care intervention immediately.

[0079] Based on the risk level, the system outputs corresponding nursing intervention recommendations. Recommendations for normal conditions include maintaining regular turning frequency and keeping the skin clean and dry. Recommendations for low-risk conditions include turning every 2 to 3 hours, applying moisturizing cream, and paying attention to pressure areas. Recommendations for medium-risk conditions include turning every 1 to 2 hours, using pressure-relief pads or dressings, and nutritional support assessment. Recommendations for high-risk conditions include continuous pressure relief, use of specialized pressure ulcer prevention equipment, consultation with a wound care specialist, and assessment by a nutritionist. In addition, the system predicts the potential stage of pressure ulcer progression based on the scoring trend, providing a prospective reference for clinical nursing decisions.

[0080] The technical effectiveness of the risk assessment model has been confirmed in large-scale clinical validation. In a multicenter clinical study involving 3,500 assessments, the method of this invention achieved a prediction accuracy of 92.3% for pressure ulcer risk, with a sensitivity of 94.1% and a specificity of 89.7%. Compared to the traditional Braden scale assessment method, the sensitivity was improved by 15 percentage points and the specificity by 12 percentage points. More importantly, this invention achieves objectivity and standardization of the assessment process, eliminating the inconsistency caused by differences in the subjective judgment of nursing staff.

[0081] See Figure 2As shown, this embodiment of the invention provides a patient skin pressure ulcer risk image assessment system. Corresponding to the aforementioned method embodiment, this system includes six functional modules: image acquisition module 1, color feature analysis module 2, texture feature extraction module 3, thermal imaging fusion module 4, dynamic tracking module 5, and risk assessment module 6. These modules form a data flow-driven collaborative architecture, realizing a complete processing chain from image acquisition to risk assessment.

[0082] Image acquisition module 1 is the system's data input front end, responsible for acquiring high-quality skin image data. This module integrates two sub-units: a visible light imaging unit and an infrared thermal imaging unit. The visible light imaging unit is equipped with a high-resolution camera sensor, a ring-shaped uniform light source, and a color chart calibration plate for acquiring visible light images of the skin surface; the infrared thermal imaging unit is equipped with a long-wave infrared detector for simultaneously acquiring temperature distribution images of the skin surface. The two sub-units employ a coaxial or fixed relative position mechanical structure design to ensure accurate registration of the acquired images. Furthermore, the image acquisition module includes a color correction processing unit, which performs real-time color correction on the visible light images based on the color chart calibration plate information, outputting skin images with guaranteed color consistency. The output of the image acquisition module includes corrected visible light skin images and infrared thermal images, which are synchronously transmitted to subsequent processing modules and stored in a historical image database.

[0083] The color feature analysis module 2 implements the function of step S2 in the method embodiment. This module receives the corrected skin image output by the image acquisition module, performs RGB to Lab color space conversion, extracts the pixel value distribution of the a channel, calculates the outlier threshold based on the statistical characteristics of the normal skin area, identifies erythema areas, and calculates the degree of redness index and the proportion of erythema area. The processing flow and algorithm details of the color feature analysis module are as described in step S2 of the method embodiment, and will not be repeated here. This module outputs the degree of redness index. and the percentage of erythema area Two characteristic parameters are transmitted to the risk assessment module.

[0084] The texture feature extraction module 3 implements the function of step S3 in the method embodiment. This module receives a skin image and performs grayscale conversion, grayscale co-occurrence matrix construction, texture statistics calculation, and comprehensive calculation of roughness and uniformity indices. The processing flow and algorithm details of the texture feature extraction module are as described in step S3 of the method embodiment. This module outputs the roughness index. and uniformity index Two characteristic parameters are transmitted to the risk assessment module.

[0085] The thermal imaging fusion module 4 implements the function of step S4 in the method embodiment. This module receives the visible light image and infrared thermal image output by the image acquisition module, and performs geometric registration, overlay fusion, and detection and quantization of temperature anomaly regions. The processing flow and algorithm details of the thermal imaging fusion module are as described in step S4 of the method embodiment. This module outputs a quantized value of the degree of temperature anomaly. The spatial distribution information of areas with abnormal temperatures is transmitted to the risk assessment module.

[0086] The dynamic tracking module 5 implements the function of step S5 in the method embodiment. This module connects to the historical image database, retrieves historical images of the same area of ​​the current patient, compares them with the currently acquired image, calculates the temporal changes in each feature dimension, and comprehensively quantifies the trend of skin condition changes. The processing flow and algorithm details of the dynamic tracking module are as described in step S5 of the method embodiment. This module outputs a quantified value of the skin condition change trend. The data is then transmitted to the risk assessment module.

[0087] Risk assessment module 6 is the core of the system's decision-making process, implementing the function of step S6 in the method embodiment. This module receives the characteristic parameters output by the aforementioned modules, including... , , , , and A multi-feature weighted fusion calculation is performed to obtain a comprehensive pressure ulcer risk score. The system categorizes risk levels based on scores, generates pressure ulcer staging predictions, and provides nursing intervention recommendations. The output of the risk assessment module is presented to nursing staff through a human-computer interaction interface and simultaneously stored in the patient's health record system for traceability and statistical analysis.

[0088] The overall workflow of the system is as follows: Nursing staff use image acquisition module 1 to photograph the common pressure ulcer sites on patients. The acquired images, after color correction, are simultaneously transmitted to color feature analysis module 2, texture feature extraction module 3, and thermal imaging fusion module 4 for parallel processing. Simultaneously, dynamic tracking module 5 retrieves historical images from the database for comparative analysis. After processing by each module, the feature parameters are summarized and fused into risk assessment module 6 for calculation. Finally, risk assessment module 6 outputs a comprehensive pressure ulcer risk score, risk level, staging prediction, and nursing recommendations, completing a full assessment cycle. The response time of the entire process is controlled within 5 seconds, meeting the requirements of real-time clinical applications.

[0089] This invention's system can be deployed in hospital ward nursing stations or mobile nursing terminals, and interfaces with the hospital's information system to achieve automatic recording and statistical analysis of assessment data. The system also supports remote consultation, allowing nursing staff to upload assessment results and skin images to a cloud platform for remote review and guidance by wound care experts.

[0090] In summary, the patient skin pressure ulcer risk image assessment method and system provided by this invention, through Lab color space erythema analysis, gray-level co-occurrence matrix texture extraction, visible light and infrared thermal imaging fusion, dynamic tracking comparison, and multi-feature weighted fusion, achieves objective, quantitative, and intelligent assessment of pressure ulcer risk. It can effectively improve the early identification ability of pressure ulcers, provide a scientific basis for clinical nursing decisions, and has important clinical application value and promotion prospects.

[0091] The embodiments of the present invention are not limited to the specific embodiments described above. Those skilled in the art can make various equivalent changes or substitutions based on the technical solutions of the present invention, and all such changes or substitutions should be included within the protection scope of the present invention.

Claims

1. A method for image-based assessment of patient pressure ulcer risk, characterized in that, Includes the following steps: S1. Image acquisition and preprocessing steps: A standardized acquisition device equipped with a ring-shaped uniform light source and a color calibration plate is used to periodically photograph the skin images of the patient's pressure ulcer-prone areas. The skin images are then color-corrected to ensure consistent imaging colors. S2, Color Feature Analysis Steps: Convert the corrected skin image from RGB color space to Lab color space, extract the pixel value distribution of the skin region in the a channel, determine the outlier threshold based on the statistical characteristics of the a channel of the normal skin region, identify the erythema region and calculate the redness degree index and erythema area ratio; S3. Skin texture feature extraction steps: Convert the skin image into a grayscale image, calculate the texture statistics of the skin surface based on the gray-level co-occurrence matrix, and calculate the roughness index and uniformity index based on the texture statistics to identify early signs of pressure ulcers such as dry skin and desquamation. S4. Temperature thermal imaging fusion step: Acquire infrared thermal images that are acquired synchronously with visible light images, perform geometric registration and superposition fusion of visible light images and infrared thermal images, detect abnormal temperature areas caused by local blood circulation disorders and quantify the degree of abnormality. S5. Dynamic tracking and comparison step: Register and compare the currently acquired skin image with the stored historical images, calculate the temporal changes in skin color features and texture features, and quantify the trend of skin condition changes. S6. Risk assessment output steps: The redness level index, erythema area ratio, roughness index, uniformity index, temperature abnormality degree and skin condition change trend are weighted and fused to calculate the pressure ulcer risk comprehensive score, and output pressure ulcer stage prediction and nursing intervention suggestions based on the score.

2. The method for image-based assessment of patient pressure ulcer risk according to claim 1, characterized in that, In step S1, the common sites of pressure ulcers include the sacrum, coccyx, heel, scapula, hip, and occipital region; the color temperature range of the annular uniform light source is 5000K to 6500K, and the color rendering index is not less than 95; the color card calibration plate contains at least 24 standard color blocks.

3. The method for image-based assessment of patient pressure ulcer risk according to claim 1, characterized in that, In step S2, the outlier determination threshold is determined as follows: the mean and standard deviation of the pixel values ​​of the a channel in the normal skin area are calculated, and the mean is added to 2.5 to 3.5 times the standard deviation as the outlier determination threshold; when the a channel value of a pixel in the skin area exceeds the outlier determination threshold, the pixel is determined to belong to the erythema area.

4. The method for image-based assessment of patient pressure ulcer risk according to claim 1, characterized in that, In step S2, the redness index is calculated as follows: the difference between the a-channel value of each pixel in the erythema region and the outlier determination threshold is accumulated and divided by the total number of pixels in the erythema region and the preset normalization coefficient; the erythema area ratio is the ratio of the number of pixels in the erythema region to the total number of pixels in the skin analysis region.

5. The method for image-based assessment of patient pressure ulcer risk according to claim 1, characterized in that, In step S3, the construction parameters of the gray-level co-occurrence matrix include: the pixel spacing is set to 1 to 5 pixels, and the calculation direction includes four directions: 0 degrees, 45 degrees, 90 degrees and 135 degrees; the texture statistics include contrast, entropy, energy and homogeneity.

6. The method for image-based assessment of patient pressure ulcer risk according to claim 5, characterized in that, The roughness index is calculated by weighting and averaging the contrast and entropy values ​​in the four directions and then normalizing them to the range of 0 to 1. The uniformity index is calculated by weighting and averaging the energy and homogeneity values ​​in the four directions and then normalizing them to the range of 0 to 1.

7. The method for image-based assessment of patient pressure ulcer risk according to claim 1, characterized in that, In step S4, the geometric registration adopts an affine transformation method based on feature points, and the registration error does not exceed 3 pixels; the method for determining the temperature abnormal area is: calculate the temperature difference between the target area and the surrounding normal skin area, and when the absolute value of the temperature difference exceeds 1.5℃ to 2.5℃, it is determined to be a temperature abnormal area.

8. The method for image-based assessment of patient pressure ulcer risk according to claim 1, characterized in that, In step S5, the acquisition time interval of the historical records images is 4 to 24 hours, and the number of historical records images stored is the images acquired in the most recent 3 to 7 times; the registration and comparison adopts a feature point matching method based on scale-invariant feature transformation.

9. The method for image-based assessment of patient pressure ulcer risk according to claim 1, characterized in that, In step S6, the multi-feature weighted fusion adopts an adaptive weight allocation strategy, wherein the weight of the redness index is 0.20 to 0.30, the weight of the erythema area ratio is 0.15 to 0.25, the weight of the roughness index is 0.10 to 0.20, the weight of the uniformity index is 0.10 to 0.15, the weight of the temperature abnormality is 0.15 to 0.25, and the weight of the skin condition change trend is 0.10 to 0.20; the comprehensive pressure ulcer risk score ranges from 0 to 100 points, where 0 to 30 points correspond to a normal state, 31 to 50 points correspond to low risk, 51 to 70 points correspond to medium risk, and 71 to 100 points correspond to high risk.

10. A patient pressure ulcer risk image assessment system, used to implement the patient pressure ulcer risk image assessment method according to any one of claims 1-9, characterized in that, include: The image acquisition module is equipped with a ring-shaped uniform light source, a color calibration plate, and an infrared thermal imaging sensor. It is used to periodically capture visible light skin images and infrared thermal images of areas prone to pressure ulcers in patients, and to perform color correction processing on the visible light skin images. The color feature analysis module is used to convert the corrected skin image from the RGB color space to the Lab color space, extract the outlier distribution of the a channel, and calculate the redness index and the proportion of erythema area. The texture feature extraction module is used to convert skin images into grayscale images and calculate texture statistics based on the gray-level co-occurrence matrix. Roughness index and uniformity index are calculated based on the texture statistics. The thermal imaging fusion module is used to geometrically register and overlay visible light images with infrared thermal images to detect and quantify areas of abnormal temperature. The dynamic tracking module is used to register and compare the current skin image with historical images to quantify the trend of skin condition changes; The risk assessment module is used to calculate a comprehensive pressure ulcer risk score by weighting and fusing various characteristic indicators, and outputs pressure ulcer stage prediction and nursing intervention suggestions.

Images