Wound healing state evaluation system based on multi-modal image mode recognition

Abstract

The invention discloses a wound healing state evaluation system based on multi-modal image mode recognition, relates to the technical field of mode recognition, and aims to construct a multi-modal and three-dimensional wound monitoring network structure and improve the accuracy and coverage dimension of wound tissue distribution monitoring, so that efficient and accurate healing state evaluation decision is realized. The healing process and patient individual differences are considered, the monitoring precision and diagnosis response speed of wound healing state evaluation are improved, the healing efficiency and rehabilitation effect in the wound treatment process are remarkably improved, the skin gray threshold value and the deformation index threshold value can be dynamically adjusted according to the individual characteristics of different patients, and the healing effect is improved. The continuous influence of tissue state misjudgment on healing evaluation is effectively prevented, evaluation deviation and treatment delay caused by subjective judgment are avoided, differential evaluation can be performed according to different skin types and healing characteristics, and the intelligent level of wound healing state evaluation and the precision of clinical application are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of pattern recognition technology, and more specifically to a wound healing status assessment system based on multimodal image pattern recognition. Background Technology

[0002] With the increasing demand for precision and personalized medical treatment and the continuous improvement of chronic wound management requirements, wound healing status assessment is a key link in clinical treatment decision-making. Its accuracy and real-time performance directly affect patient rehabilitation and the efficiency of medical resource allocation. Therefore, it is essential to build an intelligent wound healing monitoring system based on multimodal image pattern recognition technology and to conduct precise tissue feature analysis and dynamic healing assessment accordingly.

[0003] Existing technologies lack the ability to jointly collect, analyze, and evaluate multi-dimensional data such as the distribution of tissues within the wound area, the characteristics of various tissue types, and changes in the healing process. They also lack a precise mechanism for identifying wound boundaries, often relying on the subjective experience of medical staff to judge the healing status. This results in shortcomings such as inconsistent evaluation standards, insufficient objectivity, and low efficiency.

[0004] In terms of perceiving the state of wound tissue and the pattern of healing distribution, existing technologies mainly rely on single visual observation or simple image recording. They cannot capture the changes in tissue characteristics and uneven distribution of healing in complex wound structures, resulting in a significant gap between healing assessment and actual tissue state. Traditional wound monitoring methods usually only make a rough observation of the wound as a whole, which cannot reflect the local tissue abnormalities and differences in healing efficiency inside the wound, and it is difficult to accurately identify the specific location and extent of tissues at different healing stages.

[0005] Most existing wound assessment technologies adopt a static observation strategy, lacking the ability to perform differentiated analysis for different tissue types. They cannot dynamically adjust assessment parameters based on tissue characteristics such as granulation tissue, necrotic tissue, eschar, and epithelium. Furthermore, they lack adaptive strategies for individual patient differences and cannot dynamically adjust monitoring and assessment mechanisms based on changes in skin type and healing characteristics. These technological limitations severely restrict the level of intelligence and accuracy of wound healing status assessment. Summary of the Invention

[0006] The purpose of this invention is to provide a wound healing status assessment system based on multimodal image pattern recognition, which solves the problems existing in the background art.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a wound healing status assessment system based on multimodal image pattern recognition, including: a wound color image recognition module, used to perform color image scanning on the patient's arm, distinguish the wound undetermined area, the normal skin undetermined area and the boundary buffer area of ​​the patient's arm, and divide the boundary buffer area to obtain the wound contour area and the normal skin area of ​​the patient's arm.

[0008] The wound secondary image recognition module is used to extract wound image features from the wound contour region of the patient's arm, thereby determining the corresponding tissue feature regions of the wound contour region of the patient's arm, and performing simulated image mapping to obtain the appearance modeling of the wound contour region of the patient's arm and the initial feature quantities of each target modeling.

[0009] The wound cycle image analysis module is used to perform periodic dynamic physiological detection on the tissue feature areas corresponding to the normal skin area and wound contour area of ​​the patient's arm, and to evaluate the wound deformation index of the patient's arm by combining the periodic changes of the appearance model of the wound contour area of ​​the patient's arm.

[0010] The wound healing assessment module is used to analyze the wound healing recovery level of the patient's arm based on the wound deformation index and send the results to the patient and the patient's attending physician.

[0011] The beneficial effects of this invention are as follows: This invention constructs a multimodal, three-dimensional wound monitoring network structure, improving the accuracy and coverage dimension of wound tissue distribution monitoring. It can capture complex tissue feature changes and differences in healing efficiency within the wound, thereby achieving efficient and accurate healing status assessment decisions. Considering the healing process and individual patient differences, it improves the monitoring accuracy and diagnostic response speed of wound healing status assessment, significantly enhancing healing efficiency and rehabilitation effects during wound treatment. Simultaneously, it constructs a mechanism for predicting tissue feature regions, which can dynamically adjust the skin grayscale threshold and deformation index threshold according to different individual patient characteristics, effectively preventing the continuous impact of tissue status misjudgment on healing assessment, avoiding assessment bias and treatment delays caused by subjective judgment. It can also capture microscopic changes in the wound healing process in real time, significantly improving the continuity and objectivity of monitoring. It can perform differentiated assessments according to different skin types and healing characteristics, effectively improving the intelligence level of wound healing status assessment and the accuracy of clinical applications. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic diagram of the system modules of the present invention. Detailed Implementation

[0014] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Reference Figure 1 As shown, the present invention provides a wound healing status assessment system based on multimodal image pattern recognition, including: a wound color image recognition module, a wound secondary image recognition module, a wound periodic image analysis module, and a wound periodic image analysis module.

[0016] It should be noted that the wound color image recognition module is connected to the wound secondary image recognition module, the wound secondary image recognition module is connected to the wound periodic image analysis module, and the wound periodic image analysis module is connected to the wound periodic image analysis module.

[0017] The wound color image recognition module is used to perform color image scanning on the patient's arm, distinguish the undetermined wound area, the undetermined normal skin area, and the boundary buffer area of ​​the patient's arm, and divide the boundary buffer area to obtain the wound outline area and normal skin area of ​​the patient's arm.

[0018] In a specific embodiment of the present invention, the method for performing color image scanning to distinguish the undetermined area of ​​the wound, the undetermined area of ​​normal skin, and the boundary buffer area of ​​the patient's arm is as follows: acquiring a color image of the patient's arm, performing grayscale transformation to obtain the grayscale value of each pixel in the color image of the patient's arm, and performing an average calculation to obtain the average grayscale value of the pixels in the color image of the patient's arm.

[0019] In one specific embodiment, the method for acquiring the color image of the patient's arm is as follows: acquiring the color image of the patient's arm using an RGB color ultra-high-definition camera.

[0020] Based on the average gray value of the pixels in the color image of the patient's arm, and combined with the gray value of each pixel in the color image of the patient's arm, the gray deviation coefficient of each pixel in the color image of the patient's arm is calculated, and the first-level gray set, the second-level gray set, and the third-level gray set of the patient's arm are obtained accordingly.

[0021] In one specific embodiment, the calculation method for the grayscale deviation coefficient of the color image of the patient's arm at each pixel is as follows: based on the average grayscale value A of the pixels in the color image of the patient's arm, and the grayscale value of the color image of the patient's arm at each pixel... , where x represents the number of each pixel. Let y be a positive integer greater than 2. Calculate the grayscale deviation coefficient at each pixel in the color image of the patient's arm. .

[0022] Smoothness and connectivity tests were performed on each primary grayscale set, secondary grayscale set, and tertiary grayscale set of the patient's arm to determine each undetermined set of wounds, each undetermined set of normal skin, and each boundary buffer set of the patient's arm.

[0023] Each set of undetermined wounds on the patient's arm is designated as the undetermined wound region, each set of undetermined normal skin on the patient's arm is designated as the undetermined normal skin region, and each set of boundary buffers on the patient's arm is designated as the boundary buffer region.

[0024] In a specific embodiment of the present invention, the method for obtaining the first-level grayscale set, the second-level grayscale set, and the third-level grayscale set of the patient's arm is as follows: based on the grayscale value of each pixel in the color image of the patient's arm, the median and mode of the pixel grayscale value of the color image of the patient's arm are extracted, and the estimated skin grayscale value coefficient of the color image of the patient's arm is evaluated accordingly.

[0025] In one specific embodiment, the method for evaluating the estimated skin grayscale coefficient of the color image of the patient's arm is as follows: The estimated skin grayscale coefficient of the color image of the patient's arm is calculated based on the median B and mode C of the pixel grayscale values ​​of the color image of the patient's arm. .

[0026] The primary grayscale deviation coefficient threshold and the secondary grayscale deviation coefficient threshold corresponding to each skin grayscale value evaluation coefficient range are obtained from the local database. Based on the estimated skin grayscale value coefficient of the color image of the patient's arm, the primary grayscale deviation coefficient threshold and the secondary grayscale deviation coefficient threshold of the color image of the patient's arm are mapped to obtain.

[0027] It should be noted that the local database is used to store the first-level grayscale deviation coefficient threshold and the second-level grayscale deviation coefficient threshold, the first-level pixel number threshold, the skin smoothness comparison value corresponding to the first-level grayscale average skin smoothness, the normal skin segmentation percentage corresponding to each estimated skin grayscale coefficient interval, the U-net wound tissue segmentation model obtained by data training, and the first-level wound deformation index threshold and the second-level wound deformation index threshold corresponding to each water content interval.

[0028] It should also be noted that the larger the skin grayscale value evaluation coefficient, the darker the skin, and the smaller the corresponding first-level grayscale deviation coefficient threshold and the corresponding second-level grayscale deviation coefficient threshold. This value is specifically divided by researchers through experiments. The larger the difference between the first-level grayscale deviation coefficient threshold and the second-level grayscale deviation coefficient threshold, the more buffer area in the middle of normal skin will be reserved. Its main function is to initially divide the arm skin into normal skin, wound, and buffer area between wound and normal skin.

[0029] Based on the grayscale deviation coefficient of each pixel in the color image of the patient's arm, it is compared with the first-level grayscale deviation coefficient threshold and the second-level grayscale deviation coefficient threshold to obtain each first-level pixel, each second-level pixel and each third-level pixel in the color image of the patient's arm.

[0030] In one specific embodiment, the method for obtaining the first-level pixels, second-level pixels, and third-level pixels of the color image of the patient's arm is as follows: based on the grayscale deviation coefficient of each pixel in the color image of the patient's arm, the grayscale deviation coefficient is compared with the first-level grayscale deviation coefficient threshold and the second-level grayscale deviation coefficient threshold. If the grayscale deviation coefficient of a pixel in the color image of the patient's arm is less than the first-level grayscale deviation coefficient threshold, the pixel is marked as a first-level pixel. If the grayscale deviation coefficient of a pixel in the color image of the patient's arm is greater than the first-level grayscale deviation coefficient threshold and less than the second-level grayscale deviation coefficient threshold, the pixel is marked as a second-level pixel. If the grayscale deviation coefficient of a pixel in the color image of the patient's arm is greater than the second-level grayscale deviation coefficient threshold, the pixel is marked as a third-level pixel, thereby filtering the first-level pixels, second-level pixels, and third-level pixels of the color image of the patient's arm.

[0031] The color image of the patient's arm is divided into primary, secondary, and tertiary pixels to obtain primary, secondary, and tertiary grayscale sets for the patient's arm.

[0032] In one specific embodiment, the pixel aggregation and division to obtain the first-level grayscale sets, second-level grayscale sets, and third-level grayscale sets of the patient's arm is performed as follows: If a pixel of a certain level in the color image of the patient's arm is adjacent to another pixel of the same level, then the first-level pixel and the other pixel of the same level are combined into a first-level filtering set. If another pixel is adjacent to this first-level filtering set, then this pixel is merged into this first-level filtering set, thereby obtaining the number of pixels in each first-level filtering set of the patient's arm. A threshold for the number of first-level pixels is obtained from the local database. If the number of pixels in a certain first-level filtering set of the patient's arm is greater than the threshold for the number of first-level pixels, then the first-level filtering set is marked as a first-level grayscale set, thereby filtering each first-level grayscale set of the patient's arm. Similarly, each second-level grayscale set and each third-level grayscale set of the patient's arm are filtered.

[0033] In a specific embodiment of the present invention, the method for performing smoothness detection and connectivity detection to determine the undetermined sets of wounds, undetermined sets of normal skin, and undetermined sets of boundary buffers on the patient's arm is as follows: obtaining the skin smoothness of each first-level grayscale set of the patient's arm and performing an average calculation to obtain the first-level grayscale average skin smoothness of the patient's arm; obtaining the skin smoothness comparison value corresponding to the first-level grayscale average skin smoothness from the local database; and mapping the skin smoothness comparison value of the patient's arm.

[0034] In one specific embodiment, the method for obtaining the skin smoothness of each first-level grayscale set of the patient's arm is as follows: obtaining the skin smoothness of each first-level grayscale set of the patient's arm through a skin detector, and other methods of obtaining skin smoothness in this invention are consistent with this method.

[0035] It should be noted that the relationship between the average skin smoothness of the first-level grayscale and the skin smoothness comparison value is a percentage relationship. For example, the skin smoothness comparison value is 90% of the average skin smoothness of the first-level grayscale. This value can be set by the researchers themselves, as long as the skin smoothness comparison value is less than the average skin smoothness of the first-level grayscale.

[0036] The skin smoothness of each tertiary grayscale set of the patient's arm is obtained and compared with the skin smoothness comparison value of the patient's arm. If the skin smoothness of a certain tertiary grayscale set of the patient's arm is less than the skin smoothness comparison value, then the tertiary grayscale set is marked as a wound undetermined set, thereby filtering each wound undetermined set of the patient's arm. If the skin smoothness of a certain tertiary grayscale set of the patient's arm is greater than the skin smoothness comparison value, then the tertiary grayscale set is marked as a normal skin undetermined set, thereby filtering each normal skin undetermined set of the patient's arm, and each tertiary grayscale set of the patient's arm is used as a normal skin undetermined set.

[0037] Based on the positional relationship between each secondary grayscale set of the patient's arm and other sets, each secondary grayscale set of the patient's arm is divided into each undetermined set of wounds, each undetermined set of normal skin, and each boundary buffer set.

[0038] In one specific embodiment, the method for dividing the secondary grayscale sets of the patient's arm into sets of undetermined wounds, sets of undetermined normal skin, and sets of boundary buffers is as follows: if a secondary grayscale set of the patient's arm is only adjacent to a set of undetermined normal skin, then the secondary grayscale set is marked as a set of undetermined normal skin; if a secondary grayscale set of the patient's arm is only adjacent to a set of undetermined wounds, then the secondary grayscale set is marked as a set of undetermined wounds; if a secondary grayscale set of the patient's arm is adjacent to both a set of undetermined wounds and a set of undetermined normal skin, then the secondary grayscale set is marked as a set of boundary buffers, thereby dividing the secondary grayscale sets of the patient's arm into sets of undetermined wounds, sets of undetermined normal skin, and sets of boundary buffers.

[0039] By summarizing the above data, we determined the undetermined sets of wounds, normal skin, and boundary buffers on the patient's arm.

[0040] In a specific embodiment of the present invention, the method for dividing the boundary buffer region to obtain the wound contour region and normal skin region of the patient's arm is as follows: based on the estimated skin grayscale value coefficient of the color image of the patient's arm, the percentage of normal skin segmentation of the boundary buffer region of the patient's arm is analyzed, and the adjacent points of the boundary buffer region of the patient's arm with the undetermined normal skin region and the adjacent points of the boundary buffer region with the undetermined wound region are matched, and the boundary buffer region of the patient's arm is segmented into the target normal skin region and the target wound contour region accordingly.

[0041] In one specific embodiment, the analysis obtains the percentage of normal skin segmentation in the boundary buffer area of ​​the patient's arm. The specific method is as follows: obtain the percentage of normal skin segmentation corresponding to each estimated skin grayscale value coefficient range from the local database; based on the estimated skin grayscale value coefficient of the color image of the patient's arm, map the percentage of normal skin segmentation in the boundary buffer area of ​​the patient's arm to obtain the percentage of normal skin segmentation in the boundary buffer area of ​​the patient's arm. The larger the estimated skin grayscale value coefficient, the smaller the corresponding percentage of normal skin segmentation. This value can be set by researchers. It needs to meet the requirement that the darker the skin, that is, the larger the estimated skin grayscale value coefficient, the more the boundary buffer area needs to be classified as a wound, that is, the smaller the percentage of normal skin segmentation, thereby improving the confidence level.

[0042] In one specific embodiment, the method of matching adjacent points between the boundary buffer region of the patient's arm and the normal skin undetermined region, and adjacent points between the boundary buffer region and the wound undetermined region, and thereby segmenting the boundary buffer region of the patient's arm into a target normal skin region and a target wound contour region, is as follows: Each adjacent point between the boundary buffer region of the patient's arm and the normal skin undetermined region is extracted and used as the adjacent points of the patient's arm's skin; each adjacent point between the boundary buffer region of the patient's arm and the wound undetermined region is extracted and used as the adjacent points of the patient's arm's wound; the nearest wound adjacent point to each skin adjacent point in the patient's arm is selected, and the two points are connected by a straight line to obtain the straight lines of each skin adjacent point in the patient's arm; based on the percentage of normal skin segmentation in the boundary buffer region of the patient's arm, feature points are marked on the straight lines of each skin adjacent point in the patient's arm to obtain the feature points of the straight lines of each skin adjacent point in the patient's arm; and the feature points are sequentially connected to other nearest feature points, thereby dividing the boundary buffer region of the patient's arm into two parts, with the part adjacent to the normal skin undetermined region being used as the target normal skin region, and the other part being used as the target wound contour region.

[0043] It should be noted that the percentage of the distance between each adjacent point of the patient's arm skin and the characteristic point of the straight line, relative to the total length of the straight line between each adjacent point of the patient's arm skin, is the normal skin segmentation percentage.

[0044] The target normal skin area and the undetermined normal skin area on the patient's arm are collectively referred to as the normal skin area.

[0045] The undetermined area of ​​the wound on the patient's arm and the target wound contour area are unified as the wound contour area.

[0046] The wound secondary image recognition module is used to extract wound image features from the wound contour region of the patient's arm, thereby determining the tissue feature regions corresponding to the wound contour region of the patient's arm, and performing simulated image mapping to obtain the appearance modeling of the wound contour region of the patient's arm and the initial feature quantities of each target modeling.

[0047] In a specific embodiment of the present invention, the method for determining the tissue feature regions corresponding to the wound contour region of the patient's arm is as follows: performing ROI tissue detection and segmentation on the color image of the wound contour region of the patient's arm to obtain the ROI image of the wound contour region of the patient's arm.

[0048] In one specific embodiment, the method for performing ROI tissue detection and segmentation to obtain the ROI image of the wound contour region of the patient's arm is as follows: using ROI segmentation technology, the wound contour region is segmented into tissues of interest, such as granulation tissue, necrotic tissue, eschar, and epithelium, which are then further segmented from the wound contour region to obtain the ROI image of the wound contour region of the patient's arm.

[0049] The U-net wound tissue segmentation model, trained using data from a local database, is used to input the ROI image of the wound contour region of the patient's arm into the U-net wound tissue segmentation model, thereby obtaining the tissue feature regions corresponding to the wound contour region of the patient's arm.

[0050] It should be noted that the U-net wound tissue segmentation model obtained by training the data was obtained by training and validating on a large number of wound tissue images, and continuously optimized by combining Dice Loss or Focal Loss loss functions.

[0051] It should also be noted that the aforementioned tissue characteristic regions include areas occupied by tissues such as granulation tissue, necrotic tissue, eschar, and epithelium.

[0052] In a specific embodiment of the present invention, the method for obtaining the appearance modeling of the wound contour region of the patient's arm and the initial feature quantities of each target modeling is as follows: the appearance modeling of the wound contour region of the patient's arm is obtained by using a 3D depth sensor and combined with an infrared sensor, and the appearance modeling is divided to obtain the appearance modeling volume of each tissue feature region corresponding to the wound contour region of the patient's arm, and is used as the initial feature quantities of each target modeling of the wound contour region of the patient's arm.

[0053] The wound periodic image analysis module is used to perform periodic dynamic physiological detection on the tissue feature regions corresponding to the normal skin area and wound contour area of ​​the patient's arm, and to evaluate the wound deformation index of the patient's arm by combining the periodic changes of the appearance model of the wound contour area of ​​the patient's arm.

[0054] In a specific embodiment of the present invention, the method of performing periodic dynamic physiological detection and evaluating the wound deformation index of the patient's arm by combining the periodic changes in the appearance model of the wound contour area of ​​the patient's arm is as follows: each tissue feature area corresponding to the wound contour area of ​​the patient's arm is marked as a target area.

[0055] Based on the periodic dynamic physiological test results of the normal skin area of ​​the patient's arm, the comparative values ​​of various physiological characteristics of the patient's arm were obtained.

[0056] It should be noted that the physiological characteristics mentioned include: temperature, blood flow rate, and other physiological characteristics.

[0057] The physiological characteristic data of each target area of ​​the patient's arm at each heart rate cycle detection point are obtained, and the physiological characteristic variation hazard coefficient of each target area of ​​the patient's arm is calculated by combining the comparison values ​​of each physiological characteristic data of the patient's arm.

[0058] It should be noted that the heart rate cycle detection point is the detection point at each time heart rate data is received, that is, a detection is performed every time the heart beats.

[0059] In one specific embodiment, the method for obtaining the physiological feature data detection values ​​of each target area of ​​the patient's arm at each heart rate cycle detection point is as follows: obtaining the physiological feature data detection values ​​of each target area of ​​the patient's arm at each heart rate cycle detection point through the corresponding physiological feature sensor.

[0060] In one specific embodiment, the method for calculating the physiological characteristic variation hazard coefficient of each target region of the patient's arm is as follows: based on the physiological characteristic data detection values ​​of each target region of the patient's arm at each heart rate cycle detection point. , where p represents the number of each target region. q is a positive integer greater than 2, and i represents the number of each heart rate monitoring point. j is a positive integer greater than 2, and n is the number of each physiological characteristic. m is a positive integer greater than 2, and is based on the comparison values ​​of various physiological characteristics of the patient's arm. Calculate the risk coefficient of physiological characteristic variation in each target area of ​​the patient's arm. , where j represents the number of heart rate cycle detection points.

[0061] The real-time feature quantities of the wound contour region of the patient's arm at each heart rate cycle detection point are obtained, and the real-time appearance model of the wound contour region of the patient's arm at each heart rate cycle detection point is obtained. Based on the initial feature quantities of the wound contour region of the patient's arm, the appearance deformation coefficient of the wound of the patient's arm is analyzed.

[0062] In one specific embodiment, the method for obtaining the real-time target modeling features of the wound contour region of the patient's arm at each heart rate cycle detection point and obtaining the real-time appearance modeling of the wound contour region of the patient's arm at each heart rate cycle detection point is as follows: using a 3D depth sensor and in combination with an infrared sensor to obtain the real-time target modeling features of the wound contour region of the patient's arm at each heart rate cycle detection point and to obtain the appearance modeling of the wound contour region of the patient's arm at each heart rate cycle detection point.

[0063] In one specific embodiment, the method for analyzing the appearance deformation coefficient of the patient's arm wound is as follows: using existing modeling similarity comparison technology, the real-time appearance modeling of the wound contour region of the patient's arm at each heart rate cycle detection point is compared with the appearance modeling of the wound contour region of the patient's arm, thereby obtaining the appearance modeling similarity of the wound contour region of the patient's arm at each heart rate cycle detection point. Based on the real-time feature quantities of the wound contour area of ​​the patient's arm at each heart rate cycle detection point, the real-time appearance modeling volume of each target area of ​​the patient's arm at each heart rate cycle detection point is obtained. , where t represents the number of each target region. s is a positive integer greater than 2, and based on the initial feature quantities of each target region in the wound contour region of the patient's arm, the appearance modeling volume of each target region of the patient's arm is obtained. Calculate the deformation coefficient of the wound appearance on the patient's arm. .

[0064] The wound deformation index of the patient's arm was calculated based on the appearance deformation coefficient of the patient's arm and the physiological characteristic variation hazard coefficient of each target area.

[0065] In one specific embodiment, the method for calculating the wound deformation index of the patient's arm is as follows: based on the appearance deformation coefficient of the patient's arm and the physiological characteristic variation hazard coefficient of each target area, the wound deformation index of the patient's arm is calculated. .

[0066] In a specific embodiment of the present invention, the method for obtaining the comparison values ​​of various physiological feature data of the patient's arm based on the periodic dynamic physiological detection results of the normal skin area of ​​the patient's arm is as follows: based on the positional relationship between the normal skin area of ​​the patient's arm and the wound contour area, the blood vessel area in the normal skin area of ​​the patient's arm that is associated with the wound contour area is selected and marked as the normal comparison area of ​​the patient's arm.

[0067] In one specific embodiment, the method for selecting the vascular region associated with the wound contour region in the normal skin region of the patient's arm is as follows: the blood vessels involved in the wound contour region of the patient's arm are obtained by laser transmission technology, and the position of the corresponding blood vessel in the normal skin region of the patient's arm is obtained according to the blood vessel flow path, and this position is taken as the vascular region associated with the wound contour region in the normal skin region of the patient's arm.

[0068] The normal comparison area of ​​the patient's arm is obtained at each heart rate cycle detection point, and the average value is calculated to obtain the average value of the normal comparison area of ​​the patient's arm within the cycle, which is then used as the comparison value of the physiological characteristics of the patient's arm.

[0069] The wound healing assessment module is used to analyze the wound healing recovery level of the patient's arm based on the wound deformation index of the patient's arm, and send it to the patient and the patient's attending physician.

[0070] In a specific embodiment of the present invention, the method for analyzing the wound healing recovery level of the patient's arm is as follows: obtain the water content of the normal skin area of ​​the patient's arm, obtain the primary wound deformation index threshold and the secondary wound deformation index threshold corresponding to each water content interval from the local database, and map them to obtain the primary wound deformation index threshold and the secondary wound deformation index threshold of the patient's arm.

[0071] In one specific embodiment, the method for obtaining the water content of the normal skin area of ​​the patient's arm is as follows: the water content of the normal skin area of ​​the patient's arm is obtained by using a skin analyzer.

[0072] It should be noted that the higher the water content, the larger the corresponding threshold for the primary wound deformation index and the larger the corresponding threshold for the secondary wound deformation index. This value is set by researchers. The higher the water content, the higher the skin elasticity, and the greater the allowable degree of changes in wound deformation, temperature, etc.

[0073] The wound healing recovery level of the patient's arm is obtained by comparing the wound deformation index of the patient's arm with the primary and secondary wound deformation index thresholds.

[0074] In one specific embodiment, the method for obtaining the wound healing recovery level of the patient's arm is as follows: if the wound deformation index of the patient's arm is less than the first-level wound deformation index threshold, the wound healing recovery level of the patient's arm is recorded as level one; if the wound deformation index of the patient's arm is greater than the first-level wound deformation index threshold and less than the second-level wound deformation index threshold, the wound healing recovery level of the patient's arm is recorded as level two; if the wound deformation index of the patient's arm is greater than the second-level wound deformation index threshold, the wound healing recovery level of the patient's arm is recorded as level three.

[0075] It should be noted that wound healing recovery levels include: Level 1, Level 2, and Level 3, with the healing effect ranked from best to worst as follows: Level 1, Level 2, and Level 3.

[0076] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0077] This invention constructs a multimodal, three-dimensional wound monitoring network structure, improving the accuracy and coverage of wound tissue distribution monitoring. It can capture complex tissue feature changes and differences in healing efficiency within the wound, thereby achieving efficient and accurate healing status assessment decisions. Considering the healing process and individual patient differences, it improves the monitoring accuracy and diagnostic response speed of wound healing status assessment, significantly enhancing healing efficiency and rehabilitation outcomes during wound treatment. Simultaneously, it constructs a mechanism for identifying predicted tissue feature regions, dynamically adjusting skin grayscale and deformation index thresholds based on different patient characteristics. This effectively prevents the continuous impact of tissue status misjudgment on healing assessment, avoiding assessment bias and treatment delays caused by subjective judgment. Furthermore, it can capture microscopic changes during the wound healing process in real time, significantly improving the continuity and objectivity of monitoring. It can perform differentiated assessments based on different skin types and healing characteristics, effectively enhancing the intelligence level of wound healing status assessment and the accuracy of clinical applications.

[0078] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications or additions should fall within the protection scope of the present invention.

Claims

1. A wound healing status assessment system based on multimodal image pattern recognition, characterized in that, include: The wound color image recognition module is used to scan the patient's arm in color, distinguish the wound undetermined area, the normal skin undetermined area, and the boundary buffer area of ​​the patient's arm, and divide the boundary buffer area to obtain the wound outline area and normal skin area of ​​the patient's arm. The wound secondary image recognition module is used to extract wound image features from the wound contour area of ​​the patient's arm, thereby determining the tissue feature areas corresponding to the wound contour area of ​​the patient's arm, and performing simulated image mapping to obtain the appearance modeling of the wound contour area of ​​the patient's arm and the initial feature quantities of each target modeling. The wound cycle image analysis module is used to perform periodic dynamic physiological detection on the tissue feature areas corresponding to the normal skin area and wound contour area of ​​the patient's arm, and to evaluate the wound deformation index of the patient's arm by combining the periodic changes of the appearance model of the wound contour area of ​​the patient's arm. The wound healing assessment module is used to analyze the wound healing recovery level of the patient's arm based on the wound deformation index and send the results to the patient and the patient's attending physician.

2. The wound healing status assessment system based on multimodal image pattern recognition according to claim 1, characterized in that, The specific method for performing color image scanning to distinguish the undetermined wound area, the undetermined normal skin area, and the boundary buffer area on the patient's arm is as follows: A color image of the patient's arm is acquired and subjected to grayscale transformation to obtain the grayscale value of each pixel in the color image of the patient's arm. The average grayscale value of each pixel in the color image of the patient's arm is then calculated. Based on the average gray value of the pixels in the color image of the patient's arm, and combined with the gray value of each pixel in the color image of the patient's arm, the gray deviation coefficient of each pixel in the color image of the patient's arm is calculated, and the first-level gray set, the second-level gray set, and the third-level gray set of the patient's arm are obtained accordingly. Smoothness and connectivity tests were performed on each primary grayscale set, secondary grayscale set, and tertiary grayscale set of the patient's arm to determine each undetermined set of wounds, each undetermined set of normal skin, and each boundary buffer set of the patient's arm. Each set of undetermined wounds on the patient's arm is designated as the undetermined wound region, each set of undetermined normal skin on the patient's arm is designated as the undetermined normal skin region, and each set of boundary buffers on the patient's arm is designated as the boundary buffer region.

3. The wound healing status assessment system based on multimodal image pattern recognition according to claim 2, characterized in that, The specific method for obtaining the primary grayscale sets, secondary grayscale sets, and tertiary grayscale sets of the patient's arm is as follows: Based on the grayscale values ​​of each pixel in the color image of the patient's arm, the median and mode of the pixel grayscale values ​​in the color image of the patient's arm are extracted, and the predicted skin grayscale value coefficient of the color image of the patient's arm is evaluated accordingly. The first-level gray-scale deviation coefficient threshold and the second-level gray-scale deviation coefficient threshold corresponding to each skin gray-scale value evaluation coefficient range are obtained from the local database. Based on the estimated skin gray-scale value coefficient of the color image of the patient's arm, the first-level gray-scale deviation coefficient threshold and the second-level gray-scale deviation coefficient threshold of the color image of the patient's arm are mapped to obtain the color image of the patient's arm. Based on the grayscale deviation coefficient of each pixel in the color image of the patient's arm, it is compared with the first-level grayscale deviation coefficient threshold and the second-level grayscale deviation coefficient threshold to obtain each first-level pixel, each second-level pixel and each third-level pixel in the color image of the patient's arm. The color image of the patient's arm is divided into primary, secondary, and tertiary pixels to obtain primary, secondary, and tertiary grayscale sets for the patient's arm.

4. The wound healing status assessment system based on multimodal image pattern recognition according to claim 2, characterized in that, The method for performing smoothness and connectivity tests to determine the undetermined sets of wounds, normal skin, and boundary buffers on the patient's arm is as follows: The skin smoothness of each first-level grayscale set of the patient's arm is obtained and averaged to obtain the average skin smoothness of the first-level grayscale of the patient's arm. The skin smoothness comparison value corresponding to the average skin smoothness of the first-level grayscale is obtained from the local database and mapped to obtain the skin smoothness comparison value of the patient's arm. The skin smoothness of each tertiary grayscale set of the patient's arm is obtained and compared with the skin smoothness comparison value of the patient's arm. If the skin smoothness of a certain tertiary grayscale set of the patient's arm is less than the skin smoothness comparison value, then the tertiary grayscale set is marked as a wound undetermined set, thereby filtering each wound undetermined set of the patient's arm. If the skin smoothness of a certain tertiary grayscale set of the patient's arm is greater than the skin smoothness comparison value, then the tertiary grayscale set is marked as a normal skin undetermined set, thereby filtering each normal skin undetermined set of the patient's arm, and each tertiary grayscale set of the patient's arm is used as each normal skin undetermined set. Based on the positional relationship between each secondary grayscale set of the patient's arm and other sets, each secondary grayscale set of the patient's arm is divided into each undetermined set of wounds, each undetermined set of normal skin, and each boundary buffer set. By summarizing the above data, we determined the undetermined sets of wounds, normal skin, and boundary buffers on the patient's arm.

5. The wound healing status assessment system based on multimodal image pattern recognition according to claim 3, characterized in that, The specific method for dividing the boundary buffer zone to obtain the wound outline area and normal skin area of ​​the patient's arm is as follows: Based on the estimated skin grayscale coefficients of the color image of the patient's arm, the percentage of normal skin segmentation in the boundary buffer area of ​​the patient's arm is analyzed. The adjacent points of the boundary buffer area of ​​the patient's arm with the undetermined normal skin area and the adjacent points of the boundary buffer area with the undetermined wound area are matched. Based on this, the boundary buffer area of ​​the patient's arm is segmented into the target normal skin area and the target wound contour area. The target normal skin area and the undetermined normal skin area on the patient's arm are collectively defined as the normal skin area. The undetermined area of ​​the wound on the patient's arm and the target wound contour area are unified as the wound contour area.

6. The wound healing status assessment system based on multimodal image pattern recognition according to claim 1, characterized in that, The specific method for determining the tissue feature regions corresponding to the wound contour area of ​​the patient's arm is as follows: ROI tissue detection and segmentation were performed on the color image of the wound contour region of the patient's arm to obtain the ROI image of the wound contour region of the patient's arm. The U-net wound tissue segmentation model, trained using data from a local database, is used to input the ROI image of the wound contour region of the patient's arm into the U-net wound tissue segmentation model, thereby obtaining the tissue feature regions corresponding to the wound contour region of the patient's arm.

7. The wound healing status assessment system based on multimodal image pattern recognition according to claim 1, characterized in that, The specific method for obtaining the appearance modeling of the wound contour region of the patient's arm and the initial feature quantities of each target modeling is as follows: The appearance model of the wound contour area of ​​the patient's arm is obtained by using a 3D depth sensor and an infrared sensor. The appearance model is then divided to obtain the appearance model volume of each tissue feature area corresponding to the wound contour area of ​​the patient's arm, and this volume is used as the initial feature quantity for each target modeling of the wound contour area of ​​the patient's arm.

8. The wound healing status assessment system based on multimodal image pattern recognition according to claim 7, characterized in that, The method for performing periodic dynamic physiological monitoring and assessing the wound deformation index of the patient's arm by combining periodic changes in the appearance model of the wound contour area is as follows: Mark the tissue feature regions corresponding to the wound contour area of ​​the patient's arm as each target region; Based on the periodic dynamic physiological test results of the normal skin area of ​​the patient's arm, the comparative values ​​of various physiological characteristics of the patient's arm were obtained; The physiological characteristic data of each target area of ​​the patient's arm at each heart rate cycle detection point are obtained, and the physiological characteristic variation hazard coefficient of each target area of ​​the patient's arm is calculated by combining the comparison value of each physiological characteristic data of the patient's arm. The real-time feature quantities of the wound contour region of the patient's arm at each heart rate cycle detection point are obtained, and the real-time appearance model of the wound contour region of the patient's arm at each heart rate cycle detection point is obtained. Based on the initial feature quantities of the wound contour region of the patient's arm, the appearance deformation coefficient of the wound of the patient's arm is analyzed. The wound deformation index of the patient's arm was calculated based on the appearance deformation coefficient of the patient's arm and the physiological characteristic variation hazard coefficient of each target area.

9. The wound healing status assessment system based on multimodal image pattern recognition according to claim 8, characterized in that, The method for obtaining comparative values ​​of various physiological characteristics of the patient's arm based on the periodic dynamic physiological detection results of the normal skin area of ​​the patient's arm is as follows: Based on the positional relationship between the normal skin area and the wound contour area of ​​the patient's arm, the blood vessel area in the normal skin area of ​​the patient's arm that is associated with the wound contour area is selected and marked as the normal comparison area of ​​the patient's arm. The normal comparison area of ​​the patient's arm is obtained at each heart rate cycle detection point, and the average value is calculated to obtain the average value of the normal comparison area of ​​the patient's arm within the cycle, which is then used as the comparison value of the physiological characteristics of the patient's arm.

10. The wound healing status assessment system based on multimodal image pattern recognition according to claim 1, characterized in that, The specific method for analyzing the wound healing recovery level of the patient's arm is as follows: The water content of the normal skin area of ​​the patient's arm is obtained, and the primary wound deformation index threshold and secondary wound deformation index threshold corresponding to each water content range are obtained from the local database. The primary wound deformation index threshold and secondary wound deformation index threshold of the patient's arm are then mapped to obtain the primary wound deformation index threshold and secondary wound deformation index threshold of the patient's arm. The wound healing recovery level of the patient's arm is obtained by comparing the wound deformation index of the patient's arm with the primary and secondary wound deformation index thresholds.

Images