Wound recognition and detection method, system and cloud platform based on deep reinforcement learning

Abstract

The present invention relates to the field of artificial intelligence technology, and specifically discloses a wound recognition detection method, system and cloud platform based on deep reinforcement learning, including: based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained; based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, all construction straight lines of all reference wound images of each wound image to be detected are obtained; based on the construction straight line of each reference wound image of each wound image to be detected, the optimal doctor's prescription drug dosage of each wound image to be detected is obtained. The present invention comprehensively considers the degree of wound tissue damage and the wound area, and accurately calculates the optimal doctor's prescription drug dosage of each wound image to be detected, so as to better meet the specific needs of patients.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a wound recognition and detection method, system and cloud platform based on deep reinforcement learning. Background Art

[0002] At present, with the continuous development of technologies such as computer vision and deep reinforcement learning, it is possible to automatically identify and quantitatively analyze wounds. These technologies can realize the precise processing and analysis of wound images, thereby assisting doctors to make more accurate diagnosis and treatment decisions. Accurate wound treatment and personalized treatment plans can significantly improve treatment effects, reduce the occurrence of complications, and improve patient satisfaction and comfort. However, wound treatment is a complex and changeable process in medical practice. The size, depth, shape, and degree of tissue damage of different wounds will affect the choice of drug dosage. How to comprehensively consider multiple factors to achieve accurate calculation of drug dosage is currently a difficulty in the field of wound image recognition and detection.

[0003] However, the existing wound recognition detection methods, systems and cloud platforms based on deep reinforcement learning only determine the dosage of nursing drugs by the size of the wound and the type of nursing drugs, so as to assist medical staff in providing fast and accurate care to patients. However, this method only considers the relationship between wound area and drug dosage, and does not comprehensively consider the degree of tissue damage and wound area of ​​the wound, and accurately calculate the optimal doctor's prescription drug dosage for each wound image to be detected, so as to better meet the specific needs of patients. For example, the patent with the publication number of "CN114494189A" and the patent name of "A method and device for wound image recognition detection for nursing care" includes the following steps: pre-establishing a mapping relationship between the dosage of various drugs and the wound area; obtaining the wound area image of the user; performing image recognition on the wound area image and calculating the wound area; obtaining the type of drug required for the wound, and determining the dosage of the required drug according to the preset mapping relationship in combination with the wound area. First, a mapping between the types and dosages of various drugs and the wound area is established, and then the wound image of each patient is identified to determine the area of ​​the patient's wound. Finally, the dosage of the current drug is automatically determined based on the area of ​​the patient's wound and the type of drug provided by the medical staff. The above patent can intelligently determine the dosage of nursing drugs according to the size of the wound and the type of nursing drugs, so as to assist medical staff in providing fast and accurate care to patients. However, this patent only determines the dosage of nursing drugs according to the size of the wound and the type of nursing drugs, so as to assist medical staff in providing fast and accurate care to patients. However, this method only considers the relationship between the wound area and the dosage of drugs, and does not comprehensively consider the degree of tissue damage of the wound and the area of ​​the wound, and accurately calculates the optimal doctor's prescription dosage of each wound image to be detected, so as to better meet the specific needs of patients.

[0004] Therefore, the present invention proposes a wound recognition and detection method, system and cloud platform based on deep reinforcement learning. Summary of the invention

[0005] The present invention provides a wound recognition and detection method, system and cloud platform based on deep reinforcement learning, which are used to obtain all reference wound images of each wound image to be detected according to a preset database and the tissue damage assignment results of all wound-like extracted information of each wound image to be detected, and realize the analysis of the optimal doctor-ordered drug dosage of each wound image to be detected by accurately selecting some wound images from the preset database, and then obtain the comprehensive tissue damage value of each wound image to be detected according to the tissue damage assignment results of all wound-like extracted information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, and realize the comprehensive wound edge, wound structure and wound shape. The wound tissue damage degree of each wound image to be detected is quantified, and the comprehensive combination of all reference wound images of each wound image to be detected is used. The wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained according to the comprehensive tissue damage value of each wound image to be detected, the wound area and the construction straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained, and finally, the optimal doctor-prescribed drug dosage for each wound image to be detected is accurately obtained according to the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, and the optimal doctor-prescribed drug dosage for each wound image to be detected is accurately obtained in terms of the degree of wound tissue damage and the wound area, so as to better meet the specific needs of patients and improve the treatment effect.

[0006] The present invention provides a wound recognition and detection method based on deep reinforcement learning, comprising:

[0007] S1: Based on each wound image to be detected and a preset deep learning model, all wound-like extraction information of each wound image to be detected is obtained, and based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, tissue damage assignment results of all wound-like extraction information of each wound image to be detected are obtained;

[0008] S2: based on the preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained;

[0009] S3: based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, a comprehensive tissue damage value of each wound image to be detected is obtained, and based on the comprehensive tissue damage values ​​of all reference wound images of each wound image to be detected, a construction straight line of all reference wound images of each wound image to be detected is obtained;

[0010] S4: Based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the constructed straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained, and based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the optimal prescribed drug dosage for each wound image to be detected is obtained.

[0011] Preferably, the wound recognition and detection method based on deep reinforcement learning, S1: based on each wound image to be detected and a preset deep learning model, all wound-like extraction information of each wound image to be detected is obtained, and based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, the tissue damage assignment result of all wound-like extraction information of each wound image to be detected is obtained, including:

[0012] Based on each wound image to be detected and a preset deep learning model, all wound-like extraction information of each wound image to be detected is obtained, wherein all wound-like extraction information includes wound edge feature information, wound structure feature information, and wound shape feature information;

[0013] Based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, the tissue damage assignment result of all wound-like extraction information of each wound image to be detected is obtained.

[0014] Preferably, the wound recognition and detection method based on deep reinforcement learning obtains the tissue damage assignment result of all wound-like extraction information of each wound image to be detected based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, including:

[0015] Based on the wound edge feature information of each wound image to be detected and the first preset damage assignment model, obtaining a tissue damage assignment result of the wound edge feature information of each wound image to be detected;

[0016] Based on the wound structure feature information of each wound image to be detected and the second preset damage assignment model, obtaining a tissue damage assignment result of the wound structure feature information of each wound image to be detected;

[0017] Based on the wound shape feature information of each wound image to be detected and the third preset damage assignment model, a tissue damage assignment result of the wound shape feature information of each wound image to be detected is obtained.

[0018] Preferably, in the wound recognition and detection method based on deep reinforcement learning, S2: based on the preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained, including:

[0019] If the difference between the tissue damage assignment result of any type of wound extraction information among all types of wound extraction information of each wound image stored in the preset database and the tissue damage assignment result of the corresponding type of wound extraction information of each wound image to be detected is less than a preset threshold, the corresponding wound image will be used as the reference wound image of the corresponding wound image to be detected.

[0020] Preferably, the wound recognition and detection method based on deep reinforcement learning obtains the comprehensive tissue damage value of each wound image to be detected based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, including:

[0021] Obtaining tissue damage assignment results of all wound-like extraction information of all reference wound images of each wound image to be detected;

[0022] Based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the comprehensive tissue damage value of each wound image to be detected is obtained, which is:

[0023] ;

[0024] in, is the comprehensive tissue damage value of the wound image to be detected currently calculated, z is the tissue damage value assignment result of the wound edge feature information of the wound image to be detected currently calculated, c is the tissue damage value assignment result of the wound structure feature information of the wound image to be detected currently calculated, k is the tissue damage value assignment result of the wound shape feature information of the wound image to be detected currently calculated, is the sum of the tissue damage assignment results of the wound edge feature information of all reference wound images of the wound image to be detected that is currently calculated, is the sum of the tissue damage assignment results of the wound structure feature information of all reference wound images of the wound image to be detected that is currently calculated, is the sum of the tissue damage assignment results of the wound shape feature information of all reference wound images of the wound image to be detected that is currently calculated, is the standard deviation of the tissue damage assignment results of the wound edge feature information of all reference wound images of the wound image to be detected, is the standard deviation of the tissue damage assignment results of the wound structure feature information of all reference wound images of the wound image to be detected, is the standard deviation of the tissue damage assignment results of the wound shape feature information of all reference wound images of the wound image to be detected currently calculated, ln is the natural logarithm, and the value of the natural constant e is 2.718.

[0025] Preferably, the wound recognition and detection method based on deep reinforcement learning obtains a construction straight line of all reference wound images of each wound image to be detected based on the comprehensive tissue damage value of all reference wound images of each wound image to be detected, including:

[0026] Obtaining wound areas and prescribed dosages of all reference wound images for each wound image to be detected;

[0027] The quotient of the comprehensive tissue damage value of each reference wound image of each wound image to be detected and the sum of the comprehensive tissue damage values ​​of all reference wound images corresponding to the wound image to be detected is regarded as the first coefficient of the corresponding reference wound image corresponding to the wound image to be detected;

[0028] Taking the quotient of the wound area of ​​each reference wound image of each wound image to be detected and the sum of the wound areas of all reference wound images corresponding to the wound image to be detected as the second coefficient of the corresponding reference wound image corresponding to the wound image to be detected;

[0029] The quotient of the prescribed medication dosage of each reference wound image of each wound image to be detected and the sum of the prescribed medication dosage of all reference wound images corresponding to the wound image to be detected is regarded as the third coefficient of the corresponding reference wound image corresponding to the wound image to be detected;

[0030] Taking the first coefficient of each reference wound image of each wound image to be detected as the abscissa value and the third coefficient of the corresponding reference wound image of the wound image to be detected as the ordinate value, a first positioning point of each reference wound image of each wound image to be detected is obtained;

[0031] Taking the second coefficient of each reference wound image of each wound image to be detected as the abscissa value and the third coefficient of the corresponding reference wound image of the wound image to be detected as the ordinate value, obtaining the second positioning point of each reference wound image of each wound image to be detected;

[0032] Connecting the first positioning point and the second positioning point of each reference wound image of each wound image to be detected to obtain a positioning line segment of each reference wound image of each wound image to be detected;

[0033] The straight line formed by connecting the midpoint of the positioning line segment of each reference wound image of each wound image to be detected and the coordinate origin is regarded as the construction straight line of the corresponding reference wound image corresponding to the wound image to be detected.

[0034] Preferably, the wound recognition and detection method based on deep reinforcement learning obtains the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the constructed straight line of each reference wound image, including:

[0035] Acquire the wound area of ​​each wound image to be detected, and use the quotient of the comprehensive tissue damage value of each wound image to be detected and the sum of the comprehensive tissue damage values ​​of all reference wound images corresponding to the wound image to be detected as the first coefficient of the wound image to be detected;

[0036] The quotient of the wound area of ​​each wound image to be detected and the sum of the wound areas of all reference wound images corresponding to the wound image to be detected is regarded as the second coefficient corresponding to the wound image to be detected;

[0037] Based on the first coefficient, the second coefficient of each wound image to be detected and the construction straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained, that is:

[0038] ;

[0039] in, is the wound similarity value between the currently calculated wound image to be detected and the currently calculated reference wound image corresponding to the wound image to be detected, is the first coefficient of the wound image to be detected currently calculated, is the first coefficient of the currently calculated reference wound image of the currently calculated wound image to be detected, is the maximum value of the first coefficients of all reference wound images of the wound image to be detected currently calculated, is the second coefficient of the wound image to be detected currently calculated, is the second coefficient of the currently calculated reference wound image of the currently calculated wound image to be detected, is the maximum value of the second coefficients of all reference wound images of the currently calculated wound image to be detected, and p is the slope of the construction straight line of the currently calculated reference wound image of the currently calculated wound image to be detected.

[0040] Preferably, the wound recognition and detection method based on deep reinforcement learning obtains the optimal dosage of the prescribed medication for each wound image to be detected based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, including:

[0041] The reference wound image with the largest wound similarity value with the corresponding wound image to be detected among all the reference wound images of each wound image to be detected is used as the analysis wound image of the corresponding wound image to be detected;

[0042] The doctor's prescribed dosage of the analyzed wound image of each wound image to be detected is regarded as the optimal doctor's prescribed dosage of the corresponding wound image to be detected.

[0043] The present invention provides a wound recognition and detection system based on deep reinforcement learning, which is used to execute any of the wound recognition and detection methods based on deep reinforcement learning in embodiments 1 to 8, including:

[0044] An assignment module, for obtaining all wound-like extraction information of each wound image to be detected based on each wound image to be detected and a preset deep learning model, and obtaining tissue damage assignment results of all wound-like extraction information of each wound image to be detected based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms;

[0045] An extraction module, used for obtaining all reference wound images of each wound image to be detected based on a preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected;

[0046] A construction module, for obtaining a comprehensive tissue damage value of each wound image to be detected based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, and obtaining a construction straight line of all reference wound images of each wound image to be detected based on the comprehensive tissue damage values ​​of all reference wound images of each wound image to be detected;

[0047] A calculation module is used to obtain a wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the constructed straight line of each reference wound image, and to obtain the optimal doctor-prescribed drug dosage for each wound image to be detected based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected.

[0048] The present invention provides a cloud platform, which includes a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to execute any wound recognition and detection method based on deep reinforcement learning in Examples 1-8.

[0049] The beneficial effects of the present invention compared with the prior art are as follows: according to the preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained, and the optimal doctor's prescription dosage of each wound image to be detected is analyzed by accurately selecting some wound images from the preset database, and then the comprehensive tissue damage value of each wound image to be detected is obtained according to the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, and the wound tissue damage degree of each wound image to be detected is quantified by comprehensively considering the wound edge, wound structure and wound shape, and according to the comprehensive tissue damage value of all reference wound images of each wound image to be detected, Obtaining a construction straight line of all reference wound images of each wound image to be detected facilitates the subsequent calculation of wound similarity values ​​between each wound image to be detected and all reference wound images corresponding to the wound image to be detected. According to the comprehensive tissue damage value of each wound image to be detected, the wound area and the construction straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained. Finally, according to the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the optimal prescribed drug dosage for each wound image to be detected is accurately obtained. In terms of the comprehensive wound tissue damage degree and wound area, the optimal prescribed drug dosage for each wound image to be detected is accurately obtained to better meet the specific needs of patients and improve the treatment effect.

[0050] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structure specifically pointed out in the written application document.

[0051] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0053] Figure 1 This is a flow chart of a wound recognition and detection method based on deep reinforcement learning in an embodiment of the present invention;

[0054] Figure 2 Schematic diagram of a wound recognition and detection system based on deep reinforcement learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0056] Embodiment 1: The present invention provides a wound recognition and detection method based on deep reinforcement learning, referring to Figure 1 ,include:

[0057] S1: Based on each wound image to be detected and a preset deep learning model, all wound-like extraction information of each wound image to be detected is obtained, and based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, tissue damage assignment results of all wound-like extraction information of each wound image to be detected are obtained;

[0058] S2: based on the preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained;

[0059] S3: based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, a comprehensive tissue damage value of each wound image to be detected is obtained, and based on the comprehensive tissue damage values ​​of all reference wound images of each wound image to be detected, a construction straight line of all reference wound images of each wound image to be detected is obtained;

[0060] S4: Based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the constructed straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained, and based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the optimal prescribed drug dosage for each wound image to be detected is obtained.

[0061] In this embodiment, the wound image to be detected is an image for performing wound recognition detection and obtaining the optimal dosage of medically prescribed medicines using a wound recognition detection system based on deep reinforcement learning.

[0062] In this embodiment, the preset deep learning model is a pre-set model used to extract all types of wound extraction information from each wound image to be detected. In this embodiment, the preset deep learning model is an existing convolutional neural network (CNN) model.

[0063] In this embodiment, the preset reinforcement learning algorithm is a pre-set existing reinforcement learning algorithm used to construct the first preset damage assignment model, the second preset damage assignment model, and the third preset damage assignment model, such as the Q-learning algorithm used in this embodiment.

[0064] In this embodiment, the tissue damage assignment result is a numerical value obtained based on all types of wound extraction information of each wound image to be detected and all preset reinforcement learning algorithms, which can characterize the degree of wound tissue damage reflected by each type of wound extraction information of each wound image to be detected.

[0065] In this embodiment, the preset database is a pre-set database for storing all relevant information of a large number of wound images (such as all types of wound extraction information, doctor's prescription drug dosage, wound area, etc.), and the wound images in the preset database will not increase or decrease subsequently.

[0066] In this embodiment, all reference wound images of each wound image to be detected are partial wound images selected from a preset database required for analyzing the optimal doctor-ordered medication dosage of each wound image to be detected.

[0067] In this embodiment, the comprehensive tissue damage value of each wound image to be detected is a numerical value that quantifies the degree of wound tissue damage of each wound image to be detected by comprehensively considering the wound edge, wound structure and wound shape.

[0068] In this embodiment, the constructed straight line is a straight line constructed based on the comprehensive tissue damage value of all reference wound images of each wound image to be detected, and is used to calculate the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected.

[0069] In this embodiment, the wound area is the area of ​​the wound in the wound image to be detected or the reference image, and existing technology can be used, such as professional image analysis software ImageJ.

[0070] In this embodiment, the wound similarity value is a numerical value that quantifies the similarity between each wound image to be detected and each reference wound image corresponding to the wound image to be detected in terms of the comprehensive wound tissue damage degree and wound area.

[0071] In this embodiment, the optimal prescribed medication dosage for each wound image to be detected is the optimal medication dosage for the wound in each wound image to be detected, obtained based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected.

[0072] The beneficial effects of the above technology are as follows: according to the preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained, and the optimal medical prescription dosage of each wound image to be detected is analyzed by accurately selecting some wound images from the preset database, and then the comprehensive tissue damage value of each wound image to be detected is obtained according to the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, which realizes the quantification of the degree of wound tissue damage of each wound image to be detected in terms of comprehensive wound edge, wound structure and wound shape, and according to the comprehensive tissue damage values ​​of all reference wound images of each wound image to be detected, the wound image of each wound image to be detected is obtained. The construction straight line of all reference wound images of the wound image to be detected facilitates the subsequent calculation of the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected. According to the comprehensive tissue damage value of each wound image to be detected, the wound area and the construction straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained. Finally, according to the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the optimal doctor-prescribed drug dosage for each wound image to be detected is accurately obtained. In terms of the comprehensive wound tissue damage degree and wound area, the optimal doctor-prescribed drug dosage for each wound image to be detected is accurately obtained to better meet the specific needs of patients and improve the treatment effect.

[0073] Embodiment 2: Based on Embodiment 1, a wound recognition and detection method based on deep reinforcement learning, S1: Based on each wound image to be detected and a preset deep learning model, all wound-like extraction information of each wound image to be detected is obtained, and based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, the tissue damage assignment result of all wound-like extraction information of each wound image to be detected is obtained, including:

[0074] Based on each wound image to be detected and a preset deep learning model, all wound-like extraction information of each wound image to be detected is obtained, wherein all wound-like extraction information includes wound edge feature information, wound structure feature information, and wound shape feature information;

[0075] Based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, the tissue damage assignment result of all wound-like extraction information of each wound image to be detected is obtained.

[0076] In this embodiment, the wound edge feature information is the wound edge texture information extracted from each wound image to be detected according to a preset deep learning model, such as whether the outline of the wound is clear.

[0077] In this embodiment, the wound structure feature information is the wound structure information extracted from each wound image to be detected according to a preset deep learning model, such as whether the wound penetrates deep into the subcutaneous tissue, whether there is surrounding tissue damage, etc.

[0078] In this embodiment, the wound shape feature information is the shape information of the wound extracted from each wound image to be detected according to a preset deep learning model, such as the shape of the wound (circular or elliptical, etc.).

[0079] The beneficial effects of the above technology are: clarifying the specific parameter items of all types of wound extraction information for each wound image to be detected, and then obtaining the tissue damage assignment results of all types of wound extraction information for each wound image to be detected based on all types of wound extraction information for each wound image to be detected and all preset reinforcement learning algorithms.

[0080] Embodiment 3: Based on Embodiment 2, a wound recognition and detection method based on deep reinforcement learning obtains tissue damage assignment results of all wound-like extraction information of each wound image to be detected based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, including:

[0081] Based on the wound edge feature information of each wound image to be detected and the first preset damage assignment model, obtaining a tissue damage assignment result of the wound edge feature information of each wound image to be detected;

[0082] Based on the wound structure feature information of each wound image to be detected and the second preset damage assignment model, obtaining a tissue damage assignment result of the wound structure feature information of each wound image to be detected;

[0083] Based on the wound shape feature information of each wound image to be detected and the third preset damage assignment model, a tissue damage assignment result of the wound shape feature information of each wound image to be detected is obtained.

[0084] In this embodiment, the first preset damage assignment model is constructed according to a preset reinforcement learning algorithm (the Q-learning algorithm is used in this application document, and the wound edge feature information of a large number of wound images extracted by a deep learning network (CNN) is regarded as the state space, and the numerical adjustment of the assignment results of the tissue damage of a large number of wound images is regarded as the action space, and the accuracy of the assignment results is regarded as the reward. The Q value table is initialized, and the Q-value of each state-action pair is updated according to the Q-learning algorithm. The Q-value represents the expected cumulative reward for taking a certain action in a given state. By iteratively updating the Q-value, the algorithm will gradually learn the optimal strategy - that is, the best action to be taken in each state). The model can input the wound edge feature information of each wound image to be detected, and can output the assignment result of the degree of wound tissue damage reflected by the wound edge feature information of each wound image to be detected.

[0085] In this embodiment, the second preset damage assignment model is constructed according to a preset reinforcement learning algorithm (the Q-learning algorithm is used in this application document, and the wound structure feature information of a large number of wound images extracted by a deep learning network (CNN) is regarded as the state space, and the numerical adjustment of the assignment results of the tissue damage of a large number of wound images is regarded as the action space, and the accuracy of the assignment results is regarded as the reward. The Q value table is initialized, and the Q-value of each state-action pair is updated according to the Q-learning algorithm. The Q-value represents the expected cumulative reward for taking a certain action in a given state. By iteratively updating the Q-value, the algorithm will gradually learn the optimal strategy - that is, the best action to be taken in each state), which can input the wound structure feature information of each wound image to be detected, and can output the assignment result of the degree of wound tissue damage reflected by the wound structure feature information of each wound image to be detected.

[0086] In this embodiment, the third preset damage assignment model is constructed according to a preset reinforcement learning algorithm (the Q-learning algorithm is used in this application document, and the wound shape feature information of a large number of wound images extracted by a deep learning network (CNN) is regarded as the state space, and the numerical adjustment of the assignment results of the tissue damage of a large number of wound images is regarded as the action space, and the accuracy of the assignment results is regarded as the reward. The Q value table is initialized, and the Q-value of each state-action pair is updated according to the Q-learning algorithm. The Q-value represents the expected cumulative reward for taking a certain action in a given state. By iteratively updating the Q-value, the algorithm will gradually learn the optimal strategy - that is, the best action to be taken in each state). The model can input the wound shape feature information of each wound image to be detected, and can output the assignment result of the degree of wound tissue damage reflected by the wound shape feature information of each wound image to be detected.

[0087] The beneficial effect of the above technology is: based on all types of wound extraction information of each wound image to be detected and all preset reinforcement learning algorithms, the tissue damage assignment results of all types of wound extraction information of each wound image to be detected are obtained, and the degree of wound tissue damage reflected by each type of wound extraction information of each wound image to be detected is accurately quantified, which facilitates the subsequent screening of reference wound images for each wound image to be detected.

[0088] Embodiment 4: Based on Embodiment 1, a wound recognition and detection method based on deep reinforcement learning, S2: based on a preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained, including:

[0089] If the difference between the tissue damage assignment result of any type of wound extraction information among all types of wound extraction information of each wound image stored in the preset database and the tissue damage assignment result of the corresponding type of wound extraction information of each wound image to be detected is less than a preset threshold, the corresponding wound image will be used as the reference wound image of the corresponding wound image to be detected.

[0090] In this embodiment, the preset threshold is a threshold preset to obtain all reference wound images of each wound image to be detected.

[0091] The beneficial effects of the above technology are: based on the preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained, and the optimal medical prescription drug dosage of each wound image to be detected is analyzed by accurately selecting some wound images from the preset database, which facilitates the subsequent calculation of the comprehensive tissue damage value of each wound image to be detected.

[0092] Embodiment 5: Based on Embodiment 1, a wound recognition and detection method based on deep reinforcement learning obtains a comprehensive tissue damage value for each wound image to be detected based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, including:

[0093] Obtaining tissue damage assignment results of all wound-like extraction information of all reference wound images of each wound image to be detected;

[0094] Based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the comprehensive tissue damage value of each wound image to be detected is obtained, which is:

[0095] ;

[0096] in, is the comprehensive tissue damage value of the wound image to be detected currently calculated, z is the tissue damage value assignment result of the wound edge feature information of the wound image to be detected currently calculated, c is the tissue damage value assignment result of the wound structure feature information of the wound image to be detected currently calculated, k is the tissue damage value assignment result of the wound shape feature information of the wound image to be detected currently calculated, is the sum of the tissue damage assignment results of the wound edge feature information of all reference wound images of the wound image to be detected that is currently calculated, is the sum of the tissue damage assignment results of the wound structure feature information of all reference wound images of the wound image to be detected that is currently calculated, is the sum of the tissue damage assignment results of the wound shape feature information of all reference wound images of the wound image to be detected that is currently calculated, is the standard deviation of the tissue damage assignment results of the wound edge feature information of all reference wound images of the wound image to be detected, is the standard deviation of the tissue damage assignment results of the wound structure feature information of all reference wound images of the wound image to be detected, is the standard deviation of the tissue damage assignment results of the wound shape feature information of all reference wound images of the wound image to be detected currently calculated, ln is the natural logarithm, and the value of the natural constant e is 2.718.

[0097] The beneficial effect of the above technology is: based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the comprehensive tissue damage value of each wound image to be detected is obtained. This embodiment gives in detail a specific method that can comprehensively consider the wound edge, wound structure and wound shape to quantify the degree of wound tissue damage of each wound image to be detected.

[0098] Example 6: Based on Example 1, a wound recognition and detection method based on deep reinforcement learning obtains a construction straight line of all reference wound images of each wound image to be detected based on the comprehensive tissue damage value of all reference wound images of each wound image to be detected, including:

[0099] Obtaining wound areas and prescribed dosages of all reference wound images for each wound image to be detected;

[0100] The quotient of the comprehensive tissue damage value of each reference wound image of each wound image to be detected and the sum of the comprehensive tissue damage values ​​of all reference wound images corresponding to the wound image to be detected is regarded as the first coefficient of the corresponding reference wound image corresponding to the wound image to be detected;

[0101] Taking the quotient of the wound area of ​​each reference wound image of each wound image to be detected and the sum of the wound areas of all reference wound images corresponding to the wound image to be detected as the second coefficient of the corresponding reference wound image corresponding to the wound image to be detected;

[0102] The quotient of the prescribed medication dosage of each reference wound image of each wound image to be detected and the sum of the prescribed medication dosage of all reference wound images corresponding to the wound image to be detected is regarded as the third coefficient of the corresponding reference wound image corresponding to the wound image to be detected;

[0103] Taking the first coefficient of each reference wound image of each wound image to be detected as the abscissa value and the third coefficient of the corresponding reference wound image of the wound image to be detected as the ordinate value, a first positioning point of each reference wound image of each wound image to be detected is obtained;

[0104] Taking the second coefficient of each reference wound image of each wound image to be detected as the abscissa value and the third coefficient of the corresponding reference wound image of the wound image to be detected as the ordinate value, obtaining the second positioning point of each reference wound image of each wound image to be detected;

[0105] Connecting the first positioning point and the second positioning point of each reference wound image of each wound image to be detected to obtain a positioning line segment of each reference wound image of each wound image to be detected;

[0106] The straight line formed by connecting the midpoint of the positioning line segment of each reference wound image of each wound image to be detected and the coordinate origin is regarded as the construction straight line of the corresponding reference wound image corresponding to the wound image to be detected.

[0107] In this embodiment, the amount of medicine prescribed is the amount of medicine prescribed by the doctor for each wound in the reference wound image, for example, 2 ml.

[0108] The beneficial effects of the above technology are: based on the comprehensive tissue damage value of all reference wound images of each wound image to be detected, the positioning line segment of each reference wound image of each wound image to be detected is obtained, and then based on the positioning line segment of each reference wound image of each wound image to be detected, the construction straight line of all reference wound images of each wound image to be detected is obtained, which facilitates the subsequent calculation of the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected. This embodiment gives in detail a specific method for constructing a construction straight line of all reference wound images of each wound image to be detected.

[0109] Embodiment 7: Based on Embodiment 6, a wound recognition and detection method based on deep reinforcement learning is provided, which obtains a wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the constructed straight line of each reference wound image, including:

[0110] Acquire the wound area of ​​each wound image to be detected, and use the quotient of the comprehensive tissue damage value of each wound image to be detected and the sum of the comprehensive tissue damage values ​​of all reference wound images corresponding to the wound image to be detected as the first coefficient of the wound image to be detected;

[0111] The quotient of the wound area of ​​each wound image to be detected and the sum of the wound areas of all reference wound images corresponding to the wound image to be detected is regarded as the second coefficient corresponding to the wound image to be detected;

[0112] Based on the first coefficient, the second coefficient of each wound image to be detected and the construction straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained, that is:

[0113] ;'

[0114] in, is the wound similarity value between the currently calculated wound image to be detected and the currently calculated reference wound image corresponding to the wound image to be detected, is the first coefficient of the wound image to be detected currently calculated, is the first coefficient of the currently calculated reference wound image of the currently calculated wound image to be detected, is the maximum value of the first coefficients of all reference wound images of the wound image to be detected currently calculated, is the second coefficient of the wound image to be detected currently calculated, is the second coefficient of the currently calculated reference wound image of the currently calculated wound image to be detected, is the maximum value of the second coefficients of all reference wound images of the currently calculated wound image to be detected, and p is the slope of the construction straight line of the currently calculated reference wound image of the currently calculated wound image to be detected.

[0115] The beneficial effects of the above technology are: based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the constructed straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained, thereby achieving the comprehensive wound tissue damage degree and wound area, and the similarity between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is accurately quantified.

[0116] Embodiment 8: Based on Embodiment 1, a wound recognition and detection method based on deep reinforcement learning is provided, which obtains the optimal dosage of a prescribed drug for each wound image to be detected based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, including:

[0117] The reference wound image with the largest wound similarity value with the corresponding wound image to be detected among all the reference wound images of each wound image to be detected is used as the analysis wound image of the corresponding wound image to be detected;

[0118] The doctor's prescribed dosage of the analyzed wound image of each wound image to be detected is regarded as the optimal doctor's prescribed dosage of the corresponding wound image to be detected.

[0119] The beneficial effects of the above technology are: based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the optimal doctor-prescribed drug dosage for each wound image to be detected is accurately obtained, and the degree of wound tissue damage and wound area are comprehensively considered to accurately obtain the optimal doctor-prescribed drug dosage for each wound image to be detected, so as to better meet the specific needs of patients and improve the treatment effect.

[0120] Embodiment 9: The present invention provides a wound recognition and detection system based on deep reinforcement learning, which is used to execute any wound recognition and detection method based on deep reinforcement learning in Embodiments 1 to 8, referring to Figure 2 ,include:

[0121] An assignment module, for obtaining all wound-like extraction information of each wound image to be detected based on each wound image to be detected and a preset deep learning model, and obtaining tissue damage assignment results of all wound-like extraction information of each wound image to be detected based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms;

[0122] An extraction module, used for obtaining all reference wound images of each wound image to be detected based on a preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected;

[0123] A construction module, for obtaining a comprehensive tissue damage value of each wound image to be detected based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, and obtaining a construction straight line of all reference wound images of each wound image to be detected based on the comprehensive tissue damage values ​​of all reference wound images of each wound image to be detected;

[0124] A calculation module is used to obtain a wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the constructed straight line of each reference wound image, and to obtain the optimal doctor-prescribed drug dosage for each wound image to be detected based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected.

[0125] The beneficial effects of the above technology are as follows: according to the preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained, and the optimal medical prescription dosage of each wound image to be detected is analyzed by accurately selecting some wound images from the preset database, and then the comprehensive tissue damage value of each wound image to be detected is obtained according to the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, which realizes the quantification of the degree of wound tissue damage of each wound image to be detected in terms of comprehensive wound edge, wound structure and wound shape, and according to the comprehensive tissue damage values ​​of all reference wound images of each wound image to be detected, the wound image of each wound image to be detected is obtained. The construction straight line of all reference wound images of the wound image to be detected facilitates the subsequent calculation of the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected. According to the comprehensive tissue damage value of each wound image to be detected, the wound area and the construction straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained. Finally, according to the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the optimal doctor-prescribed drug dosage for each wound image to be detected is accurately obtained. In terms of the comprehensive wound tissue damage degree and wound area, the optimal doctor-prescribed drug dosage for each wound image to be detected is accurately obtained to better meet the specific needs of patients and improve the treatment effect.

[0126] Example 10: The present invention provides a cloud platform, which includes a processor and a machine-readable storage medium, wherein the machine-readable storage medium and the processor are connected, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to execute any wound recognition and detection method based on deep reinforcement learning in Examples 1-8.

[0127] The beneficial effects of the above technology are as follows: This embodiment provides a cloud platform based on deep reinforcement learning to determine the optimal dosage of medical prescriptions for wounds in wound images.

[0128] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention, and the present invention is intended to include these changes and modifications.

Claims

1. A wound recognition and detection method based on deep reinforcement learning, characterized in that: include: S1: Based on each wound image to be detected and a preset deep learning model, all wound-like extraction information of each wound image to be detected is obtained, and based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, tissue damage assignment results of all wound-like extraction information of each wound image to be detected are obtained; S2: based on the preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained; S3: based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, a comprehensive tissue damage value of each wound image to be detected is obtained, and based on the comprehensive tissue damage values ​​of all reference wound images of each wound image to be detected, a construction straight line of all reference wound images of each wound image to be detected is obtained; S4: based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the constructed straight line of each reference wound image, obtaining a wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected, and based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, obtaining an optimal doctor-ordered dosage of each wound image to be detected; Wherein, based on the comprehensive tissue damage value of all reference wound images of each wound image to be detected, the construction straight line of all reference wound images of each wound image to be detected is obtained, including: Obtaining wound areas and prescribed dosages of all reference wound images for each wound image to be detected; The quotient of the comprehensive tissue damage value of each reference wound image of each wound image to be detected and the sum of the comprehensive tissue damage values ​​of all reference wound images corresponding to the wound image to be detected is regarded as the first coefficient of the corresponding reference wound image corresponding to the wound image to be detected; Taking the quotient of the wound area of ​​each reference wound image of each wound image to be detected and the sum of the wound areas of all reference wound images corresponding to the wound image to be detected as the second coefficient of the corresponding reference wound image corresponding to the wound image to be detected; The quotient of the prescribed medication dosage of each reference wound image of each wound image to be detected and the sum of the prescribed medication dosage of all reference wound images corresponding to the wound image to be detected is regarded as the third coefficient of the corresponding reference wound image corresponding to the wound image to be detected; Taking the first coefficient of each reference wound image of each wound image to be detected as the abscissa value and the third coefficient of the corresponding reference wound image of the wound image to be detected as the ordinate value, a first positioning point of each reference wound image of each wound image to be detected is obtained; Taking the second coefficient of each reference wound image of each wound image to be detected as the abscissa value and the third coefficient of the corresponding reference wound image of the wound image to be detected as the ordinate value, obtaining the second positioning point of each reference wound image of each wound image to be detected; Connecting the first positioning point and the second positioning point of each reference wound image of each wound image to be detected to obtain a positioning line segment of each reference wound image of each wound image to be detected; The straight line formed by connecting the midpoint of the positioning line segment of each reference wound image of each wound image to be detected and the coordinate origin is regarded as the construction straight line of the corresponding reference wound image corresponding to the wound image to be detected.

2. The wound recognition and detection method based on deep reinforcement learning according to claim 1, characterized in that: S1: Based on each wound image to be detected and a preset deep learning model, all wound-like extraction information of each wound image to be detected is obtained, and based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, tissue damage assignment results of all wound-like extraction information of each wound image to be detected are obtained, including: Based on each wound image to be detected and a preset deep learning model, all wound-like extraction information of each wound image to be detected is obtained, wherein all wound-like extraction information includes wound edge feature information, wound structure feature information, and wound shape feature information; Based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, the tissue damage assignment result of all wound-like extraction information of each wound image to be detected is obtained.

3. The wound recognition and detection method based on deep reinforcement learning according to claim 2 is characterized in that: Based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms, the tissue damage assignment results of all wound-like extraction information of each wound image to be detected are obtained, including: Based on the wound edge feature information of each wound image to be detected and the first preset damage assignment model, obtaining a tissue damage assignment result of the wound edge feature information of each wound image to be detected; Based on the wound structure feature information of each wound image to be detected and the second preset damage assignment model, obtaining a tissue damage assignment result of the wound structure feature information of each wound image to be detected; Based on the wound shape feature information of each wound image to be detected and the third preset damage assignment model, a tissue damage assignment result of the wound shape feature information of each wound image to be detected is obtained.

4. The wound recognition and detection method based on deep reinforcement learning according to claim 1, characterized in that: S2: Based on the preset database and the tissue damage assignment results of all wound-like extracted information of each wound image to be detected, all reference wound images of each wound image to be detected are obtained, including: If the difference between the tissue damage assignment result of any type of wound extraction information among all types of wound extraction information of each wound image stored in the preset database and the tissue damage assignment result of the corresponding type of wound extraction information of each wound image to be detected is less than a preset threshold, the corresponding wound image will be used as the reference wound image of the corresponding wound image to be detected.

5. The wound recognition and detection method based on deep reinforcement learning according to claim 1, characterized in that: Based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, a comprehensive tissue damage value of each wound image to be detected is obtained, including: Obtaining tissue damage assignment results of all wound-like extraction information of all reference wound images of each wound image to be detected; Based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the comprehensive tissue damage value of each wound image to be detected is obtained, which is: ; in, is the comprehensive tissue damage value of the wound image to be detected currently calculated, z is the tissue damage value assignment result of the wound edge feature information of the wound image to be detected currently calculated, c is the tissue damage value assignment result of the wound structure feature information of the wound image to be detected currently calculated, k is the tissue damage value assignment result of the wound shape feature information of the wound image to be detected currently calculated, is the sum of the tissue damage assignment results of the wound edge feature information of all reference wound images of the wound image to be detected that is currently calculated, is the sum of the tissue damage assignment results of the wound structure feature information of all reference wound images of the wound image to be detected that is currently calculated, is the sum of the tissue damage assignment results of the wound shape feature information of all reference wound images of the wound image to be detected that is currently calculated, is the standard deviation of the tissue damage assignment results of the wound edge feature information of all reference wound images of the wound image to be detected, is the standard deviation of the tissue damage assignment results of the wound structure feature information of all reference wound images of the wound image to be detected, is the standard deviation of the tissue damage assignment results of the wound shape feature information of all reference wound images of the wound image to be detected currently calculated, ln is the natural logarithm, and the value of the natural constant e is 2.

718.

6. The wound recognition and detection method based on deep reinforcement learning according to claim 1, characterized in that: Based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the construction straight line of each reference wound image, a wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained, including: Acquire the wound area of ​​each wound image to be detected, and use the quotient of the comprehensive tissue damage value of each wound image to be detected and the sum of the comprehensive tissue damage values ​​of all reference wound images corresponding to the wound image to be detected as the first coefficient of the wound image to be detected; The quotient of the wound area of ​​each wound image to be detected and the sum of the wound areas of all reference wound images corresponding to the wound image to be detected is regarded as the second coefficient corresponding to the wound image to be detected; Based on the first coefficient, the second coefficient of each wound image to be detected and the construction straight line of each reference wound image, the wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected is obtained, that is: ; in, is the wound similarity value between the currently calculated wound image to be detected and the currently calculated reference wound image corresponding to the wound image to be detected, is the first coefficient of the wound image to be detected currently calculated, is the first coefficient of the currently calculated reference wound image of the currently calculated wound image to be detected, is the maximum value of the first coefficients of all reference wound images of the wound image to be detected currently calculated, is the second coefficient of the wound image to be detected currently calculated, is the second coefficient of the currently calculated reference wound image of the currently calculated wound image to be detected, is the maximum value of the second coefficients of all reference wound images of the currently calculated wound image to be detected, and p is the slope of the construction straight line of the currently calculated reference wound image of the currently calculated wound image to be detected.

7. The wound recognition and detection method based on deep reinforcement learning according to claim 1, characterized in that: Based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected, the optimal doctor-ordered dosage of each wound image to be detected is obtained, including: The reference wound image with the largest wound similarity value with the corresponding wound image to be detected among all the reference wound images of each wound image to be detected is used as the analysis wound image of the corresponding wound image to be detected; The doctor's prescribed dosage of the analyzed wound image of each wound image to be detected is regarded as the optimal doctor's prescribed dosage of the corresponding wound image to be detected.

8. A wound recognition and detection system based on deep reinforcement learning, characterized in that: The method for wound recognition and detection based on deep reinforcement learning according to any one of claims 1 to 7 comprises: An assignment module, for obtaining all wound-like extraction information of each wound image to be detected based on each wound image to be detected and a preset deep learning model, and obtaining tissue damage assignment results of all wound-like extraction information of each wound image to be detected based on all wound-like extraction information of each wound image to be detected and all preset reinforcement learning algorithms; An extraction module, used for obtaining all reference wound images of each wound image to be detected based on a preset database and the tissue damage assignment results of all wound-like extraction information of each wound image to be detected; A construction module, for obtaining a comprehensive tissue damage value of each wound image to be detected based on the tissue damage assignment results of all wound-like extraction information of each wound image to be detected and all reference wound images corresponding to the wound image to be detected, and obtaining a construction straight line of all reference wound images of each wound image to be detected based on the comprehensive tissue damage values ​​of all reference wound images of each wound image to be detected; A calculation module is used to obtain a wound similarity value between each wound image to be detected and each reference wound image corresponding to the wound image to be detected based on the comprehensive tissue damage value of each wound image to be detected, the wound area and the constructed straight line of each reference wound image, and to obtain the optimal doctor-prescribed drug dosage for each wound image to be detected based on the wound similarity value between each wound image to be detected and all reference wound images corresponding to the wound image to be detected.

9. A cloud platform, characterized in that: The cloud platform includes a processor and a machine-readable storage medium, the machine-readable storage medium and the processor are connected, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to execute the wound recognition and detection method based on deep reinforcement learning as described in any one of claims 1-7.

Images