A non-invasive method for detecting skin pressure injuries
By converting skin video images to YIQ color space and using Euler video amplification algorithm to extract brightness signals, combining image pyramid and Fourier transform analysis to identify potential pressure damage areas, solving the early recognition accuracy in the prior art Problem, more efficient pressure ulcer detection is achieved.
Patent Information
- Application Number
- CN202411116684.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-08-15
AI Technical Summary
The prior art has low early recognition accuracy in pressure ulcer detection, especially in complex contexts, which may be ignored or misjudged, resulting in insufficient detection efficiency and accuracy.
Using non-invasive skin pressure damage detection method, the skin brightness signal is extracted by extracting skin video images and converting them to YIQ color space, and the skin brightness signal is extracted using Euler video amplification algorithm to construct an image pyramid for amplification processing, calculate the power spectrum density of the skin area, and combine the correlation. and Fourier transform analysis to identify potential pressure damage areas.
It improves the accuracy and reliability of pressure ulcer detection, can more clearly observe slight changes in skin pressure damage, enhances the identification of specific patterns related to skin pressure damage, and improves the accuracy and efficiency of detection.
Smart Images

Figure CN118657757B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and in particular to a non-invasive skin pressure injury detection method. Background Art
[0002] Pressure sores, medically known as pressure injuries, are a pathological condition caused by continuous local pressure or pressure and shear force acting on the skin and subcutaneous tissue, which leads to restricted blood circulation, thus causing cell dysfunction and even tissue necrosis. Traditional clinical detection of pressure sores mainly relies on direct observation and tactile perception by medical staff, that is, medical staff judge the presence and severity of pressure sores by observing skin color changes, swelling, and feeling the hardness of the skin with their hands. This detection method is highly dependent on the professional knowledge, experience and intuitive judgment of medical staff, has certain limitations and low detection efficiency;
[0003] With the development of image recognition technology, professional researchers in the field of skin detection are exploring new detection methods that combine image recognition technology to solve the limitations and inefficiency of pressure ulcer detection. Existing technologies integrate deep learning models and computer vision algorithms. The system can automatically identify thermal imaging images of pressure injuries from high-resolution images, thereby assisting in the diagnosis of pressure ulcers and improving the efficiency of pressure ulcer detection.
[0004] However, in clinical practice, early identification of pressure ulcers is particularly critical, as the initial stage may only manifest as slight changes in skin color or slight differences in texture. These changes are often difficult to identify using existing image recognition methods and may be ignored or misjudged in complex backgrounds, resulting in low recognition accuracy. To this end, we propose a noninvasive skin pressure injury detection method. Summary of the invention
[0005] In order to improve the accuracy and efficiency of pressure sore identification, the purpose of the present invention is to provide a non-invasive skin pressure injury detection method, and the technical solution adopted is as follows:
[0006] A non-invasive skin pressure injury detection method, the method comprising:
[0007] Extract skin video images and convert them to YIQ color space;
[0008] The pre-processed skin video image is enlarged based on the Euler video enlargement algorithm to extract the skin brightness signal;
[0009] Correlation analysis of skin brightness signals to identify potential pressure-damaged skin areas;
[0010] The transfer function is used to calculate the power spectrum density of the potential pressure-injured skin area and the normal skin area to obtain the detection results.
[0011] Further, the pre-processed skin video image is magnified based on the Euler video magnification algorithm, and the skin brightness signal is extracted, including:
[0012] Construct an image downsampling pyramid and extract a high-level skin image by weighted averaging of adjacent pixels;
[0013] Construct an image upsampling pyramid to restore high-level skin images to their original resolution;
[0014] The brightness signals of normal skin and pressure-damaged skin are extracted from the upsampled skin images.
[0015] Furthermore, an image downsampling pyramid is constructed to extract high-level skin images by weighted averaging of adjacent pixels, including:
[0016] Select the coordinates of the center pixel of the skin image of the downsampling pyramid target layer and scale it to obtain the corresponding pixel position in the skin image of the next layer of the downsampling pyramid target layer;
[0017] Considering the pixels within a preset variation range around the central pixel of the target layer skin image, a weighted sum is performed using a weight function to obtain the pixel value at the corresponding pixel position in the next layer of skin image of the target layer.
[0018] Furthermore, an image upsampling pyramid is constructed to restore the high-level skin image to the original resolution, including:
[0019] Scaling is performed based on the pixel position of the downsampled skin image to obtain the pixel position of the upsampled pyramid target layer skin image;
[0020] Considering the pixels within a preset variation range around the pixel position of the skin image of the up-sampled pyramid target layer, a weighted sum is performed using a weight function to obtain the pixel value at the pixel position of the skin image of the up-sampled pyramid target layer.
[0021] Further, the preprocessed skin video image is magnified based on the Euler video magnification algorithm, and the skin brightness signal is extracted, which further includes:
[0022] Using a bandpass filter to separate frequency components related to pressure-injured skin from the downsampled skin image according to a preset frequency;
[0023] The signal output by the bandpass filter is spatially filtered and amplified to extract the brightness signals of normal skin and pressure-injured skin.
[0024] Furthermore, correlation analysis is performed on the skin brightness signal to identify potential pressure-damaged skin areas, including:
[0025] The brightness signals of normal skin and pressure-damaged skin were processed to 0 mean;
[0026] The brightness signals of normal skin and pressure-injured skin were subjected to covariance analysis to obtain the correlation coefficient and covariance of the brightness signals of normal skin and pressure-injured skin.
[0027] Furthermore, correlation analysis is performed on the skin brightness signal to identify potential pressure-damaged skin areas, including:
[0028] Normalization of brightness signals of normal skin and pressure-injured skin;
[0029] The normalized brightness signals of normal skin and pressure-injured skin were analyzed by difference analysis.
[0030] Furthermore, performing difference analysis on the normalized brightness signals of normal skin and pressure-damaged skin includes:
[0031] Calculate the root mean square error of the brightness signal of normal skin and pressure-injured skin;
[0032] The normalized brightness signals of normal skin and pressure-injured skin were subjected to fast Fourier transform, and the signals after fast Fourier transform processing were extracted.
[0033] Furthermore, the transfer function is used to calculate the power spectrum density of the potential pressure-injured skin area and the normal skin area, and the detection results are obtained:
[0034] Calculate the power spectral density based on the signal processed by fast Fourier transform;
[0035] The power spectrum density of normal skin and pressure-injured skin is divided to obtain a correlation spectrum and output the detection result.
[0036] Furthermore, the method also includes: extracting intrinsic mode functions from the potential pressure injury area based on empirical mode decomposition.
[0037] The present invention has the following beneficial effects:
[0038] The non-invasive skin pressure injury detection method provided in the present application extracts skin video images and converts them into YIQ color space. In the YIQ color space, brightness information is separated from hue and saturation information, so that the skin image processing process can focus more on the changes in skin brightness signals; then the Euler video amplification algorithm is used to enhance subtle changes in skin brightness signals. Through amplification processing, small changes related to skin pressure injuries can be more clearly observed; by calculating the correlation coefficient of normal skin and potential pressure injury skin brightness signals, the linear relationship between the two is evaluated, and areas where pressure injuries may exist are preliminarily screened out; finally, the transfer function is used to calculate the power spectral density of the potential pressure injury skin area and the normal skin area respectively, and the nonlinear characteristics of the brightness signal of the potential pressure injury area are further analyzed to identify specific patterns related to skin pressure injuries. The power spectral density helps to identify specific frequency components related to skin pressure injuries, thereby improving the accuracy and reliability of skin pressure injury detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0040] Figure 1 A schematic flow chart of a non-invasive skin pressure injury detection method provided by an embodiment of the present invention;
[0041] Figure 2 A schematic diagram of a process for extracting skin brightness signals provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0042] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, characteristics and effects of a non-invasive skin pressure injury detection method proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0043] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0044] A specific scheme of a non-invasive skin pressure injury detection method provided by the present invention is described in detail below with reference to the accompanying drawings.
[0045] See also Figure 1 , which shows a flow chart of a non-invasive skin pressure injury detection method provided by an embodiment of the present invention, the method comprising:
[0046] Step S100: extracting skin video images and converting them into YIQ color space;
[0047] Specifically, a real-time video image of the patient's skin is collected through an image video sensor. To ensure that the collected image is clear and the illumination is uniform, a high-resolution camera is preferably used to collect the skin image and combined with a light source for auxiliary illumination. The image collection angle and position can also be previewed and adjusted through an interactive display to ensure the accuracy of the collected skin image, and the collected video information is transmitted to the PC host for image processing; the PC host needs to pre-process the collected video image, and the pre-processing includes at least image denoising, brightness equalization and signal amplification operations to improve the accuracy of subsequent processing; in this embodiment, the YIQ color space specifically divides the skin image information into a brightness channel (Y) and chromaticity information (I and Q), extracts the skin area in the video image, and converts the image in the RGB color space into the YIQ color space, and the brightness channel contains the brightness information in the skin image; in the YIQ color space, the subsequent Euler amplification algorithm can enhance the small movements in the image without introducing significant color distortion by only amplifying the changes in the brightness channel, thereby improving the accuracy of detecting skin pressure injuries;
[0048] Step S200: Amplifying the preprocessed skin video image based on the Euler video amplification algorithm to extract the skin brightness signal; specifically, for the skin image converted to the YIQ color space, the Euler video amplification algorithm can amplify the slight changes in the skin image so as to more easily observe the skin pressure damage;
[0049] See also Figure 2 , step S200 specifically includes:
[0050] Step S210: constructing an image downsampling pyramid to extract a high-level skin image by weighted averaging of adjacent pixels; the purpose is to reduce the resolution of the image while maintaining the main features of the image; by constructing a downsampling pyramid, a multi-level representation of the skin image is extracted to provide multi-scale information for further processing;
[0051] Step S210 specifically includes:
[0052] Step S211: Select the coordinates of the center pixel of the skin image of the target layer of the downsampling pyramid and scale it to obtain the corresponding pixel position in the skin image of the next layer of the target layer of the downsampling pyramid; since the resolution of the image is halved with each layer down in the downsampling pyramid, it is necessary to scale the first layer. The coordinates of the first layer of skin image are halved and scaled to obtain the corresponding pixel position in the next layer of image;
[0053] Step S212: Considering the pixels within a preset variation range around the central pixel of the target layer skin image, weighted summation is performed using a weight function to obtain the pixel value at the corresponding pixel position in the skin image of the next layer of the target layer.
[0054] The expression of the downsampling pyramid is:
[0055] (1)
[0056] In formula (1), represents the downsampling pyramid Layer skin image coordinates The pixel value at The skin layer image is obtained by downsampling; represents the weight function, which determines the degree of influence of the pixels in the original skin image on the pixel values in the new skin image during the downsampling process. It is usually a Gaussian weight and is used to smooth the image. represents the downsampling pyramid Layer skin image coordinates The pixel value at ; Indicates the horizontal offset. Indicates the vertical offset, ranging from -2 to 2, that is, considering a 5×5 pixel area around the center pixel; : means to The coordinates of the layer pyramid image are scaled to obtain the downsampling pyramid layer. The pixel values of the layer image; by constructing the downsampling pyramid in this way, we can effectively reduce the resolution of the image while maintaining the main features of the image. In each downsampling process, we perform weighted averaging by considering the 5x5 neighborhood around the center pixel, which helps to smooth the skin image while retaining the key structural information in the skin image, thereby effectively identifying subtle skin changes;
[0057] Step S220: constructing an image upsampling pyramid to restore the high-level skin image to the original resolution, specifically including:
[0058] Step S221: scaling based on the pixel position of the down-sampled skin image to obtain the pixel position of the up-sampled pyramid target layer skin image;
[0059] Step S222: Considering the pixels within a preset variation range around the pixel position of the up-sampled pyramid target layer skin image, a weighted sum is performed using a weight function to obtain a pixel value at the pixel position of the up-sampled pyramid target layer skin image;
[0060] The expression of upsampling pyramid is:
[0061] (2)
[0062] In formula (2), represents the upsampled quinta Layer skin map coordinates The upsampling pyramid can restore the high-level skin image in the downsampling pyramid to the original resolution. The upsampling pyramid helps to maintain the detail information in the skin image and retain the key structural information even in the resolution restoration process, further improving the accuracy of skin pressure injury detection.
[0063] Step S230: extract the brightness signals of normal skin and pressure-damaged skin from the upsampled skin image, and amplify the skin image preprocessed in the above steps S210 and S220 based on the Euler video amplification algorithm. The YIQ value of each pixel in the skin image can be used , and To represent, the Euler video magnification algorithm is used to amplify the changes in the brightness (Y) channel to enhance the slight motion in the image, and the single frame sequence of each pixel position in the skin image is By extracting, we can get the brightness signal of the pixel , whose expression is:
[0064] (3)
[0065] In formula (3), express Skin image pixels after upsampling at each moment The brightness signal; Represents pixel The brightness value of express The brightness change at each moment is used to amplify the brightness change. The larger the value, the more obvious the amplification effect. Indicates the rate of change of brightness over time; the brightness channel (Y) in the YIQ color space contains important brightness information, and the slight changes related to skin pressure injuries can be observed more clearly through amplification processing; in the YIQ color space, by only amplifying the changes in the brightness (Y) channel, significant color distortion can be avoided, which is particularly important for the detection of skin pressure injuries. It can maintain the true color information of the skin and avoid misjudgment; in summary, the Euler video magnification algorithm can effectively amplify slight movements in skin images. Since skin pressure injuries are often accompanied by slight skin changes, this is particularly critical for improving the accuracy of detecting skin pressure injuries;
[0066] Step S230 also includes:
[0067] Step S231: using a bandpass filter to separate the frequency components related to the pressure-injured skin from the downsampled skin image according to a preset frequency; specifically, the frequency components related to the pressure-injured skin are preferably the blood flow and tissue movement under the skin, and the expression is:
[0068] (4)
[0069] In formula (4), Represents the input signal, that is, the brightness time domain signal of the video stream composed of image frames, with time t as the horizontal coordinate and the pixel brightness value as the vertical coordinate, which is a discrete one-dimensional array; represents the signal after bandpass filtering; is a bandpass filter; wherein the low-frequency and high-frequency cutoff frequencies of the bandpass filter are determined according to the frequency range of the blood flow and tissue movement to be detected, and an ideal bandpass filter is used, and its expression is:
[0070] (5)
[0071] In formula (5), , are the low and high frequency cutoff frequencies of the bandpass filter; represents the distance in frequency space; represents the lateral frequency domain component; Represents the longitudinal frequency domain component; specifically, the low-frequency and high-frequency cutoff frequencies of the bandpass filter are determined according to the frequency range of blood flow and tissue motion to be detected , , use a bandpass filter to filter the brightness signal and separate the signal within a specific frequency range; an ideal bandpass filter will pass , The frequency components between the two bands are filtered out, while other frequency components are filtered out; the frequency components related to skin pressure injuries, such as blood flow and tissue movement under the skin, can be effectively separated by a bandpass filter. The separated frequency components help to detect skin pressure injuries more accurately;
[0072] Step S232: Perform spatial filtering and amplification processing on the signal output by the bandpass filter, extract the brightness signal of normal skin and pressure-injured skin, and apply a spatial filter to process the signal after bandpass filtering, and the expression is:
[0073] (6)
[0074] In formula (6), represents the signal after spatial filtering; represents an ideal bandpass filter; Represents the signal after bandpass filtering; step S232 further extracts the brightness signals of normal skin and pressure-injured skin by performing spatial filtering and amplification processing on the signal output by the bandpass filter. The spatial filtering and amplification processing can enhance the subtle changes related to pressure injuries, further observe the skin pressure injuries more clearly, and improve the accuracy of detecting skin pressure injuries.
[0075] Step S300: performing correlation analysis on skin brightness signals to identify potential pressure-damaged skin areas;
[0076] Step S300 specifically includes:
[0077] Step S310: Perform zero mean processing on the brightness signals of normal skin and pressure-injured skin. Specifically, for the brightness signals of normal skin and pressure-injured skin, calculate the average value of the brightness signal of each pixel, make the average value of the brightness signal 0, and eliminate the DC component in the signal. The expression is:
[0078] (7)
[0079] In formula (7), is the average value of the brightness signal; Represents the original brightness signal; Represents the brightness signal after zero mean processing; step S310 eliminates the DC component in the signal by calculating the average value of the brightness signal and performing zero mean processing. This step can help simplify the subsequent correlation analysis for skin pressure injury detection and improve the accuracy of detection.
[0080] Step S320: Perform covariance analysis on the brightness signals of normal skin and pressure-injured skin to obtain the correlation coefficient and covariance of the brightness signals of normal skin and pressure-injured skin; since the time domain signal input after processing in step S200 is a discrete one-dimensional array, the data needs to be further processed as a reference indicator for the detection result; in this embodiment, the brightness signals of normal skin and pressure-injured skin are respectively expressed as and , can be obtained by processing in step S200; and Calculate their mean and , and the standard deviation and ; The formula for calculating covariance is:
[0081] (8)
[0082] In formula (8), represents covariance; and These are the brightness signals of normal skin and pressure-damaged skin, respectively; and is the mean of the brightness signals of normal skin and pressure-damaged skin; Represents the number of samples, that is, the number of elements in a discrete one-dimensional array; Indicates Normal skin brightness signal; Indicates A pressure-induced skin brightness signal;
[0083] and is their standard deviation, is the correlation coefficient
[0084] (9)
[0085] In formula (9), Represents the standard deviation of normal skin brightness signal; Represents the standard deviation of the brightness signal of pressure-injured skin; represents the correlation coefficient;
[0086] In this embodiment, by calculating the correlation coefficient, the linear relationship strength between the brightness signals of normal skin and pressure-injured skin can be evaluated; for example, if the correlation coefficient is close to 0, it indicates that there is no obvious linear relationship between the brightness signals of the two types of skin, and potential pressure injuries may exist; by evaluating the correlation coefficient, it is helpful to distinguish between normal skin and potential pressure-injured skin areas, and the accuracy of detecting skin pressure injuries can be improved; calculating the covariance can help quantify the difference between the brightness signals of normal skin and pressure-injured skin;
[0087] Step S330: Normalize the brightness signals of normal skin and pressure-injured skin; that is, normalize the brightness signals of normal skin and pressure-injured skin. and , calculate their maximum and minimum values respectively, and normalize the brightness signal to the range of [0-1]. Through normalization, the range of the brightness signal can be unified, which is convenient for subsequent analysis;
[0088] Step S340: performing difference analysis on the normalized brightness signals of normal skin and pressure-injured skin, specifically:
[0089] Step S341: Calculate the root mean square error of the brightness signal of normal skin and pressure-injured skin.
[0090] (10)
[0091] In formula (10), Represents the root mean square error of the normalized brightness signal; represents the normalized brightness signal; Represents the number of samples, that is, the number of elements in a discrete one-dimensional array;
[0092] Step S342: performing fast Fourier transform on the normalized brightness signals of normal skin and pressure-injured skin, and extracting the signals processed by fast Fourier transform;
[0093] (11)
[0094] In formula (11), represents the signal after fast Fourier transform processing; Indicates the amplitude of the brightness signal; Indicates the signal length; Indicates frequency; represents an imaginary unit; step S340 further quantifies the difference between the brightness signals of normal skin and pressure-injured skin by performing root mean square error calculation and fast Fourier transform on the normalized brightness signal, and extracts the transformed signal, which can help quantify the difference between normal skin and pressure-injured skin, and identify specific frequency components related to skin pressure injury through frequency domain analysis, thereby improving the accuracy of detection;
[0095] As an optional technical solution, the method also includes: extracting intrinsic mode functions from potential pressure injury areas based on empirical mode decomposition; specifically, performing empirical mode (EMD) analysis on the brightness signal and extracting intrinsic mode functions (IMFs) to identify the inherent modes and characteristics of the signal, which is to reveal the nonlinear characteristics and time-frequency characteristics of the signal, so as to have a deeper understanding of the dynamic process of skin changes; EMD decomposes the original signal into a series of IMFs, each IMF represents an inherent oscillation mode of the signal, and has the characteristics of instantaneous frequency and amplitude. For the detection of skin pressure injuries, extracting IMFs means decomposing the brightness signal into several modes reflecting different physiological states of the skin; IMFs can reveal specific frequency modes or oscillation characteristics related to pressure sores in the brightness signal, which may not be so obvious in correlation analysis. For example, a certain IMF may contain frequency information specific to skin microcirculation disorders or tissue damage, which can be used for early identification of pressure sores;
[0096] Step S400: using the transfer function to calculate the power spectrum density of the potential pressure-injured skin area and the normal skin area respectively, to obtain the detection result;
[0097] Step S400 specifically includes:
[0098] Step S410: Calculating power spectrum density based on the signal processed by fast Fourier transform;
[0099] (12)
[0100] In formula (12), represents the power spectral density; represents the signal after fast Fourier transform processing;
[0101] Step S420: Divide the power spectrum density of normal skin and pressure-injured skin to obtain a correlation spectrum, and output the detection result;
[0102] (13)
[0103] In formula (13), represents the transfer function; represents the power spectral density of normal skin; Indicates the power spectral density of the pressure-injured skin; Step S400 calculates the power spectral density of the potential pressure-injured skin area and the normal skin area, and obtains the transfer function by calculating the ratio of the power spectral density, thereby obtaining the detection result; The detection result can be used to assess the severity of the pressure injury, output the possible pressure injury risk level, and help identify the characteristic frequencies associated with skin pressure injuries, for example, if within certain specific frequency ranges If significant changes are shown, these frequency ranges may be associated with pressure injuries, according to The degree of change can be used to assess the potential risk level of pressure injury. The test results are finally output in the form of charts and reports on an interactive display, including the detected abnormal areas, possible pressure injury risk levels, etc., and early warnings are issued to provide corresponding nursing advice. The test results can also be stored in a database for medical staff to query and compare historical data at any time. These functions provide a basis for early diagnosis and intervention in clinical applications.
[0104] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0105] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.
Claims
1. A non-invasive skin pressure injury detection method, characterized in that: The method comprises: Extract skin video images and convert them to YIQ color space; The pre-processed skin video image is enlarged based on the Euler video enlargement algorithm to extract the skin brightness signal; Correlation analysis of skin brightness signals to identify potential pressure-damaged skin areas; Normalization of brightness signals of normal skin and pressure-injured skin; Perform difference analysis on the normalized brightness signals of normal skin and pressure-damaged skin, including: Calculate the root mean square error of the brightness signal of normal skin and pressure-injured skin; Perform fast Fourier transform on the normalized brightness signals of normal skin and pressure-injured skin, and extract the signals processed by fast Fourier transform; The power spectrum density of the potential pressure-injured skin area and the normal skin area is calculated to obtain the transfer function, and the detection results are output, including: Calculate the power spectral density based on the signal processed by fast Fourier transform; Divide the power spectrum density of normal skin and pressure-injured skin to obtain the correlation spectrum and output the test result, which is expressed as: In the formula, represents the transfer function; represents the power spectral density of normal skin; Represents the power spectral density of pressure-injured skin; calculates the power spectral density of potential pressure-injured skin area and normal skin area, and obtains the transfer function by calculating the ratio of the power spectral density, and outputs the test result; the test result is used to evaluate the severity of pressure injury, output the pressure injury risk level, and identify specific frequencies associated with skin pressure injury.
2. The non-invasive skin pressure injury detection method according to claim 1, characterized in that: The pre-processed skin video image is enlarged based on the Euler video enlargement algorithm, and the skin brightness signal is extracted including: Construct an image downsampling pyramid and extract a high-level skin image by weighted averaging of adjacent pixels; Construct an image upsampling pyramid to restore high-level skin images to their original resolution; The brightness signals of normal skin and pressure-damaged skin are extracted from the upsampled skin images.
3. The non-invasive skin pressure injury detection method according to claim 2, characterized in that: Construct an image downsampling pyramid and extract a high-level skin image by weighted averaging of adjacent pixels, including: Select the coordinates of the center pixel of the skin image of the target layer of the downsampling pyramid and scale it to obtain the corresponding pixel position in the skin image of the next layer of the target layer of the downsampling pyramid; Considering the pixels within a preset variation range around the central pixel of the target layer skin image, a weighted sum is performed using a weight function to obtain the pixel value at the corresponding pixel position in the next layer of skin image of the target layer.
4. The non-invasive skin pressure injury detection method according to claim 3, characterized in that: Constructing an image upsampling pyramid to restore the high-level skin image to the original resolution includes: Scaling is performed based on the pixel position of the downsampled skin image to obtain the pixel position of the upsampled pyramid target layer skin image; Considering the pixels within a preset variation range around the pixel position of the skin image of the up-sampled pyramid target layer, a weighted sum is performed using a weight function to obtain the pixel value at the pixel position of the skin image of the up-sampled pyramid target layer.
5. The non-invasive skin pressure injury detection method according to claim 4, characterized in that: The pre-processed skin video image is enlarged based on the Euler video enlargement algorithm, and the skin brightness signal is extracted, which also includes: Using a bandpass filter to separate frequency components related to pressure-injured skin from the downsampled skin image according to a preset frequency; The signal output by the bandpass filter is spatially filtered and amplified to extract the brightness signals of normal skin and pressure-injured skin.
6. The non-invasive skin pressure injury detection method according to claim 1, characterized in that: Correlation analysis of skin brightness signals was performed to identify potential pressure-damaged skin areas including: The brightness signals of normal skin and pressure-damaged skin were processed to 0 mean; The brightness signals of normal skin and pressure-injured skin were subjected to covariance analysis to obtain the correlation coefficient and covariance of the brightness signals of normal skin and pressure-injured skin.
7. The non-invasive skin pressure injury detection method according to claim 1, characterized in that: The method further includes extracting intrinsic mode functions from the potential pressure injury region based on empirical mode decomposition.
