Human body scar rapid classification method and system based on visual calculation
Through visual computing technology, scar morphology and vascular distribution information are extracted and processed, and a classification model is established, which solves the problems of multidimensionality and low classification efficiency in human scar evaluation in the existing technology, and achieves rapid and effective scar classification evaluation. and auxiliary medical support.
Patent Information
- Application Number
- CN202411067848.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the prior art, in human scar assessment, it is difficult to quickly and effectively obtain and process multi-dimensional data information, resulting in inconvenient classification and grading process.
Using a rapid human scar classification method based on visual calculation, scar morphology data is obtained and preprocessed through image processing technology, morphological recognition features are extracted, and classification models are trained in neural network models, and cross-evaluation is carried out in combination with vascular distribution information. , optimize the model using random grouping and cross-validation.
A rapid classification evaluation of human scars is realized, which avoids feature calculation weight conflicts and imbalances, improves the rapidity and effectiveness of classification evaluation, and supports auxiliary medical evaluation.
Smart Images

Figure CN119006906A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to data image recognition related fields, and in particular to a human scar rapid classification method and system based on visual computing. Background Art
[0002] When the skin is damaged, scars usually form. Scar formation is a natural physiological process that repairs damaged skin tissue through a series of cellular and molecular interactions. Generally speaking, scars include hypertrophic scars, keloid scars, atrophic scars, acne scars and other types, and have different characteristic manifestations in different periods, including the inflammatory stage, proliferative stage and mature stage.
[0003] At present, human scar assessment is mostly based on four main indicators: color, thickness, vascular distribution, and softness. Color and vascular distribution are natural visual information, while thickness and softness require additional or special devices for measurement. The acquisition and corresponding calculation of information from these four different data dimensions and modalities make it inconvenient to quickly implement scar classification and grading. Summary of the invention
[0004] The purpose of the present invention is to provide a method and system for rapid classification of human scars based on visual computing, so as to solve the problems raised in the above background technology.
[0005] To achieve the above object, the present invention provides the following technical solutions:
[0006] A rapid classification method for human scars based on visual computing, including:
[0007] Acquisition and preprocessing of scar morphological data; acquiring an array containing multiple groups of image data sets of different types of scars based on a medical record database, and performing homogenization preprocessing on the image data, wherein the homogenization preprocessing includes noise filtering, size standardization, and image edge feature enhancement;
[0008] Extract identification features based on scar morphological data; extract morphological identification features of scars through image processing technology, and classify the morphological identification features according to the corresponding scar record categories, wherein the morphological identification features include scar area, perimeter, concavity and convexity, and surface texture distribution characteristics;
[0009] The scar classification model is trained according to the data set; the morphological recognition feature data with category labels in the array are classified and trained in a preset neural network model to establish the scar classification model;
[0010] Importing, judging and classifying the objects to be identified; obtaining the data of the objects to be classified, preprocessing and extracting identification features, obtaining the features to be identified for classification and identification, classifying and identifying the features to be identified through the scar classification model, and obtaining the corresponding scar category.
[0011] As a further solution of the present invention, it also includes the step of performing scar category cross-evaluation based on the vascular visualization image, specifically including:
[0012] Performing a tomographic scan of the blood vessel distribution on the scar, obtaining multiple groups of blood vessel distribution layers and performing spatial stacking to obtain a blood vessel spatial distribution model of the scar;
[0013] Performing spatial feature vectorization on the vascular spatial distribution feature to establish the vascular distribution feature of the scar;
[0014] Based on the medical record database, the spatial distribution characteristics of blood vessels corresponding to different scar categories are obtained, and the spatial characteristics are vectorized to obtain a blood vessel distribution mapping model corresponding to the scar category;
[0015] The vascular distribution features are classified and matched based on the vascular distribution mapping model, and the scar category of the object data to be classified is cross-validated and judged based on the matching rate.
[0016] As a further solution of the present invention: the step of training the scar classification model according to the data set specifically includes:
[0017] Randomly grouping the data set to establish multiple training groups, and training the scar classification model based on the training groups, each of the training groups including multiple different types of scars;
[0018] Cross-validating the plurality of randomly grouped image data sets with the plurality of scar classification models obtained through training, respectively, to obtain a plurality of cross-validation result sets corresponding to the random groups;
[0019] Calculating the overall deviation value of the cross-validation result set based on the actual scar records corresponding to the scar morphology data in the random group, and simulating the point cloud distribution of the scar classification model corresponding to the random group based on multiple calculation results;
[0020] Based on the point cloud distribution, the parameters of multiple scar classification models with small deviation values were averaged to establish the optimal scar classification model.
[0021] As a further solution of the present invention, it also includes the step of judging the scar cycle:
[0022] Acquiring morphological features and color features corresponding to the scar, as well as the vascular distribution and active state of the scar based on the image data, so as to generate a cycle judgment data set;
[0023] A feature matching library of the corresponding scar category at different stages is obtained and used to match the cycle judgment data set to obtain the cycle stage of the scar.
[0024] As a further solution of the present invention: it also includes an inducement screening step:
[0025] Obtain scar categories based on scar classification model matching, match several triggers of the corresponding scar categories, generate trigger-guided interactive links and output them;
[0026] User interaction feedback is obtained to record the cause information corresponding to the scar.
[0027] The embodiment of the present invention aims to provide a human scar rapid classification system based on visual computing, comprising:
[0028] A data preprocessing module is used for obtaining and preprocessing scar morphological data; based on the medical record database, an array of image data sets containing multiple groups of scars of different categories is obtained, and the image data is preprocessed for homogenization, wherein the homogenization preprocessing includes noise filtering, size standardization, and image edge feature enhancement;
[0029] A feature extraction module is used to extract identification features based on scar morphological data; the morphological identification features of scars are extracted by image processing technology, and the morphological identification features are categorized according to the corresponding scar record categories, wherein the morphological identification features include scar area, perimeter, concavity and convexity, and surface texture distribution features;
[0030] The model training module is used to train the scar classification model according to the data set; the morphological recognition feature data with category labels in the array are classified and trained in a preset neural network model to establish the scar classification model;
[0031] The scar matching module is used for importing, judging and classifying the objects to be identified; obtaining the data of the objects to be classified, preprocessing and extracting identification features, obtaining the features to be identified for classification and identification, classifying and identifying the features to be identified through the scar classification model, and obtaining the corresponding scar category.
[0032] As a further solution of the present invention, it also includes a vascular intersection assessment module, specifically including:
[0033] A blood vessel data acquisition unit, used for performing a tomographic scan of the blood vessel distribution of the scar, acquiring multiple groups of blood vessel distribution layers and performing spatial stacking to acquire a blood vessel spatial distribution model of the scar;
[0034] A feature vectorization unit, used for performing spatial feature vectorization on the vascular spatial distribution feature to establish the vascular distribution feature of the scar;
[0035] A mapping model building unit is used to obtain the spatial distribution characteristics of blood vessels corresponding to different scar categories based on the medical record database, and to perform spatial feature vectorization to obtain a blood vessel distribution mapping model corresponding to the scar category;
[0036] A cross-matching unit is used to classify and match the blood vessel distribution features based on the blood vessel distribution mapping model, and to cross-validate and judge the scar category of the object data to be classified based on the matching rate.
[0037] As a further solution of the present invention: the model training module specifically includes:
[0038] A data set grouping unit, used for randomly grouping the data set to establish a plurality of training groups, and training the scar classification model based on the training groups, each of the training groups including a plurality of scars of different categories;
[0039] A cross-validation unit, used to cross-validate the plurality of randomly grouped image data sets with the plurality of scar classification models obtained through training, respectively, to obtain a plurality of cross-validation result sets corresponding to the random groups;
[0040] a deviation calculation unit, configured to calculate an overall deviation value of the cross-validation result set based on actual scar records corresponding to the scar morphology data in the random group, and perform point cloud distribution simulation on the scar classification model corresponding to the random group based on multiple calculation results;
[0041] The cross correction unit is used to perform parameter averaging processing on multiple scar classification models with smaller deviation values based on point cloud distribution, so as to establish an optimized scar classification model.
[0042] As a further solution of the present invention: it also includes a period determination module, specifically including:
[0043] A judgment data acquisition unit, used to acquire morphological features and color features corresponding to the scar, as well as the blood vessel distribution and active state of the scar based on the image data, so as to generate a period judgment data set;
[0044] The cycle stage judgment unit is used to obtain the feature matching library of the corresponding scar category at different stages, and to match the cycle judgment data set to obtain the cycle stage of the scar.
[0045] As a further solution of the present invention: it also includes an inducement screening module, specifically including:
[0046] The inducement interaction unit is used to obtain the scar category based on the scar classification model matching, match several inducements corresponding to the scar category, generate inducement-guided interaction links and output them;
[0047] The inducement establishing unit is used to obtain user interaction feedback to record inducement information corresponding to the scar.
[0048] Compared with the prior art, the beneficial effects of the present invention are: by performing morphological characterization processing on scars and digitizing the morphological features, the establishment and use of a morphological classification model is realized. Therefore, by using less data fuzzy information compared to the prior art, a faster and more effective classification evaluation effect and performance can be obtained, and the conflict or imbalance of feature calculation weights in the prior art can be effectively avoided, so that the purpose of auxiliary medical evaluation can be more effectively achieved. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 This is a flowchart of a rapid human scar classification method based on visual computing.
[0050] Figure 2 A flowchart of the steps for cross-evaluation of scar categories in a rapid human scar classification method based on visual computing.
[0051] Figure 3 This is a block diagram of the rapid human scar classification system based on visual computing. DETAILED DESCRIPTION
[0052] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0053] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.
[0054] like Figure 1 The method for rapid classification of human scars based on visual computing provided in one embodiment of the present invention comprises the following steps:
[0055] S10, acquisition and preprocessing of scar morphological data; acquiring an array containing multiple groups of image data sets of different types of scars based on the medical record database, and performing homogenization preprocessing on the image data, wherein the homogenization preprocessing includes noise filtering, size standardization, and image edge feature enhancement.
[0056] S20, extracting identification features based on scar morphological data; extracting morphological identification features of scars by image processing technology, and classifying the morphological identification features according to corresponding scar record categories, wherein the morphological identification features include scar area, perimeter, concavity and convexity, and surface texture distribution features.
[0057] S30, training a scar classification model according to the data set; classifying and training the morphological recognition feature data with category labels in a preset neural network model to establish a scar classification model.
[0058] S40, importing, judging and classifying the object to be identified; obtaining data of the object to be classified, preprocessing and extracting identification features, obtaining features to be identified for classification and identification, classifying and identifying the features to be identified through a scar classification model, and obtaining the corresponding scar category.
[0059] In this embodiment, a method for rapid classification of human scars based on visual computing is provided. By performing morphological characterization processing on scars and digitizing morphological features, the establishment and use of a morphological classification model is realized. Therefore, by using less data fuzzy information compared with the prior art, a faster and more effective classification evaluation effect and performance can be obtained, and the conflict or imbalance of feature calculation weights in the prior art can be effectively avoided, so as to more effectively achieve the purpose of auxiliary medical evaluation. In specific use, a large number of image data sets containing different types of scars are first obtained through a medical record database, and the data sets are ensured to be Each sample has a clear scar type label, and the data is preprocessed. Here, denoising can be performed by Gaussian filtering and median filtering, and edge and line enhancement can be achieved by image algorithms such as Canny edge detection. In the training of the scar classification model, in addition to the neural network model, classification algorithms such as random forest and support vector machine can also be used for model training to obtain a model that can be used for scar classification. Step S40 is used for the actual operation of the model. During the recognition process, the feature processing and extraction steps of steps S10 and S20 are required to obtain feature data for model evaluation and classification.
[0060] like Figure 2 As shown, as another preferred embodiment of the present invention, it also includes the step of performing scar category cross-evaluation based on the vascular visualization image, specifically including:
[0061] S51, performing a tomographic scanning of the blood vessel distribution of the scar, obtaining multiple groups of blood vessel distribution layers and performing spatial stacking to obtain a blood vessel spatial distribution model of the scar;
[0062] S52, performing spatial feature vectorization on the vascular spatial distribution feature to establish the vascular distribution feature of the scar;
[0063] S53, obtaining spatial distribution characteristics of blood vessels corresponding to different scar categories based on the medical record database, and performing spatial feature vectorization to obtain a blood vessel distribution mapping model corresponding to the scar category;
[0064] S54, classifying and matching the blood vessel distribution features based on the blood vessel distribution mapping model, and cross-validating and judging the scar category of the object data to be classified based on the matching rate.
[0065] In this embodiment, a step of auxiliary cross-validation of scar classification based on the vascular distribution of the scar is added. Compared with the method of directly fusing with morphological features for category judgment, the classification method of independent and cross-validated two visual information fusion calculations can achieve the original or better classification evaluation effect and performance with less data modal information and in a faster and more effective way. Furthermore, the scar area is automatically segmented based on deep learning, and the color score is evaluated according to the average pixel value of the area; the vascular distribution score is evaluated based on the average value of the skin color pixels around the scar, and the continuity and perfection of the topological structure characteristics of the vascular distribution are combined to automatically give a score. The two visual information are used as input for cross-fusion evaluation to obtain the final rapid classification result. The algorithm calculation logic driven by big data further ensures the rationality of the sample distribution statistics, making the method and system in this patent more realistic and effective.
[0066] As another preferred embodiment of the present invention, the step of training the scar classification model according to the data set specifically further includes:
[0067] Randomly grouping the data set to establish multiple training groups, and training the scar classification model based on the training groups, each of the training groups including multiple different types of scars;
[0068] Cross-validating the plurality of randomly grouped image data sets with the plurality of scar classification models obtained through training, respectively, to obtain a plurality of cross-validation result sets corresponding to the random groups;
[0069] Calculating the overall deviation value of the cross-validation result set based on the actual scar records corresponding to the scar morphology data in the random group, and simulating the point cloud distribution of the scar classification model corresponding to the random group based on multiple calculation results;
[0070] Based on the point cloud distribution, the parameters of multiple scar classification models with small deviation values were averaged to establish the optimal scar classification model.
[0071] In this embodiment, the steps of training the scar classification model are supplemented, mainly in the grouping and cross-validation of the data set, that is, the entire data sample used for training is divided into multiple data sets and independently trained respectively, and the trained model is cross-classified and verified with other data sets. The deviation values of multiple models are judged according to the classification results of multiple models, and the models with excessively large deviation values are screened out, and multiple low-deviation models that are relatively clustered in the point cloud distribution are retained and averaged, thereby optimizing the accuracy of the model.
[0072] As another preferred embodiment of the present invention, the step of judging the scar cycle is also included:
[0073] Acquiring morphological features and color features corresponding to the scar, as well as the vascular distribution and active state of the scar based on the image data, so as to generate a cycle judgment data set;
[0074] A feature matching library of the corresponding scar category at different stages is obtained and used to match the cycle judgment data set to obtain the cycle stage of the scar.
[0075] In this embodiment, the color and texture of the scar surface will change significantly in different cycles, and because the inflammatory reaction and proliferative reaction inside the scar at different stages will lead to changes in blood vessel distribution and activity (tissue clearance, repair and growth, and changes in the demand for nutrients and oxygen, etc.), the current stage of the scar can be effectively assessed based on this, facilitating the implementation of medical treatment.
[0076] As another preferred embodiment of the present invention, it also includes an inducement screening step:
[0077] Obtain scar categories based on scar classification model matching, match several triggers of the corresponding scar categories, generate trigger-guided interactive links and output them;
[0078] User interaction feedback is obtained to record the cause information corresponding to the scar.
[0079] In this embodiment, the causes of scar formation and the causes of scar changes (such as failure to protect the scar and secondary damage affecting the scar stage change, etc.) are different for different users, but the overall causes can be classified. Therefore, the actual causes can be quickly determined by interacting with the user, reducing the large amount of time wasted in doctor-patient inquiry and screening interactions, thereby improving the efficiency of medical workers.
[0080] like Figure 3 As shown, the present invention also provides a human scar rapid classification system based on visual computing, which comprises:
[0081] The data preprocessing module 100 is used for obtaining and preprocessing scar morphological data; based on the medical record database, an array of image data sets containing multiple groups of scars of different categories is obtained, and the image data is preprocessed for homogenization, wherein the homogenization preprocessing includes noise filtering, size standardization and image edge feature enhancement.
[0082] The feature extraction module 200 is used to extract identification features based on scar morphological data; the morphological identification features of the scar are extracted by image processing technology, and the morphological identification features are categorized according to the corresponding scar record categories. The morphological identification features include scar area, perimeter, concavity and convexity, and surface texture distribution features.
[0083] The model training module 300 is used to train the scar classification model according to the data set; the morphological recognition feature data with category labels are classified and trained in a preset neural network model to establish the scar classification model.
[0084] The scar matching module 400 is used for importing, judging and classifying the objects to be identified; obtaining the data of the objects to be classified, preprocessing and extracting identification features, obtaining the features to be identified for classification and identification, classifying and identifying the features to be identified through the scar classification model, and obtaining the corresponding scar category.
[0085] As another preferred embodiment of the present invention, it also includes a vascular intersection assessment module, specifically including:
[0086] A blood vessel data acquisition unit, used for performing a tomographic scan of the blood vessel distribution of the scar, acquiring multiple groups of blood vessel distribution layers and performing spatial stacking to acquire a blood vessel spatial distribution model of the scar;
[0087] A feature vectorization unit, used for performing spatial feature vectorization on the vascular spatial distribution feature to establish the vascular distribution feature of the scar;
[0088] A mapping model building unit is used to obtain the spatial distribution characteristics of blood vessels corresponding to different scar categories based on the medical record database, and to perform spatial feature vectorization to obtain a blood vessel distribution mapping model corresponding to the scar category;
[0089] A cross-matching unit is used to classify and match the blood vessel distribution features based on the blood vessel distribution mapping model, and to cross-validate and judge the scar category of the object data to be classified based on the matching rate.
[0090] As another preferred embodiment of the present invention, the model training module specifically includes:
[0091] A data set grouping unit, used for randomly grouping the data set to establish a plurality of training groups, and training the scar classification model based on the training groups, each of the training groups including a plurality of scars of different categories;
[0092] A cross-validation unit, used to cross-validate the plurality of randomly grouped image data sets with the plurality of scar classification models obtained through training, respectively, to obtain a plurality of cross-validation result sets corresponding to the random groups;
[0093] a deviation calculation unit, configured to calculate an overall deviation value of the cross-validation result set based on actual scar records corresponding to the scar morphology data in the random group, and perform point cloud distribution simulation on the scar classification model corresponding to the random group based on multiple calculation results;
[0094] The cross correction unit is used to perform parameter averaging processing on multiple scar classification models with smaller deviation values based on point cloud distribution, so as to establish an optimized scar classification model.
[0095] As another preferred embodiment of the present invention, it also includes a period determination module, specifically including:
[0096] A judgment data acquisition unit, used to acquire morphological features and color features corresponding to the scar, as well as the blood vessel distribution and active state of the scar based on the image data, so as to generate a period judgment data set;
[0097] The cycle stage judgment unit is used to obtain the feature matching library of the corresponding scar category at different stages, and to match the cycle judgment data set to obtain the cycle stage of the scar.
[0098] As another preferred embodiment of the present invention, it also includes an inducement screening module, specifically including:
[0099] The inducement interaction unit is used to obtain the scar category based on the scar classification model matching, match several inducements corresponding to the scar category, generate inducement-guided interaction links and output them;
[0100] The inducement establishing unit is used to obtain user interaction feedback to record inducement information corresponding to the scar.
[0101] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0102] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the disclosure in the specification and examples. This application is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.
[0103] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims
1. A rapid classification method for human scars based on visual computing, characterized in that: Include: Acquisition and preprocessing of scar morphological data; acquiring an array containing multiple groups of image data sets of different types of scars based on a medical record database, and performing homogenization preprocessing on the image data, wherein the homogenization preprocessing includes noise filtering, size standardization, and image edge feature enhancement; Extraction of identification features based on scar morphological data; Extracting morphological identification features of scars by image processing technology, and marking the morphological identification features according to the corresponding scar record categories, wherein the morphological identification features include scar area, perimeter, concavity and convexity, and surface texture distribution features; The scar classification model is trained based on the data set; Classify and train the morphological recognition feature data with category labels in a preset neural network model to establish a scar classification model; Importing, judging and classifying the objects to be identified; obtaining the data of the objects to be classified, preprocessing and extracting identification features, obtaining the features to be identified for classification and identification, classifying and identifying the features to be identified through the scar classification model, and obtaining the corresponding scar category.
2. The method for rapid classification of human scars based on visual computing according to claim 1, characterized in that: The method also includes the steps of cross-evaluating scar categories based on the vascular visualization images, specifically including: Performing a tomographic scan of the blood vessel distribution on the scar, obtaining multiple groups of blood vessel distribution layers and performing spatial stacking to obtain a blood vessel spatial distribution model of the scar; Performing spatial feature vectorization on the vascular spatial distribution feature to establish the vascular distribution feature of the scar; Based on the medical record database, the spatial distribution characteristics of blood vessels corresponding to different scar categories are obtained, and the spatial characteristics are vectorized to obtain a blood vessel distribution mapping model corresponding to the scar category; The vascular distribution features are classified and matched based on the vascular distribution mapping model, and the scar category of the object data to be classified is cross-validated and judged based on the matching rate.
3. The method for rapid classification of human scars based on visual computing according to claim 2, characterized in that: The step of training the scar classification model according to the data set specifically includes: Randomly grouping the data set to establish multiple training groups, and training the scar classification model based on the training groups, each of the training groups including multiple different types of scars; Cross-validating the plurality of randomly grouped image data sets with the plurality of scar classification models obtained through training, respectively, to obtain a plurality of cross-validation result sets corresponding to the random groups; Calculating the overall deviation value of the cross-validation result set based on the actual scar records corresponding to the scar morphology data in the random group, and simulating the point cloud distribution of the scar classification model corresponding to the random group based on multiple calculation results; Based on the point cloud distribution, the parameters of multiple scar classification models with small deviation values were averaged to establish the optimal scar classification model.
4. The method for rapid classification of human scars based on visual computing according to claim 3, characterized in that: The step of determining the scar cycle is also included: Acquiring morphological features and color features corresponding to the scar, as well as the vascular distribution and active state of the scar based on the image data, so as to generate a cycle judgment data set; A feature matching library of the corresponding scar category at different stages is obtained and used to match the cycle judgment data set to obtain the cycle stage of the scar.
5. The method for rapid classification of human scars based on visual computing according to claim 1, characterized in that: It also includes the following steps: Obtain scar categories based on scar classification model matching, match several triggers of the corresponding scar categories, generate trigger-guided interactive links and output; User interaction feedback is obtained to record the cause information corresponding to the scar.
6. A rapid human scar classification system based on visual computing, characterized in that: Include: A data preprocessing module is used for obtaining and preprocessing scar morphological data; based on the medical record database, an array of image data sets containing multiple groups of scars of different categories is obtained, and the image data is preprocessed for homogenization, wherein the homogenization preprocessing includes noise filtering, size standardization, and image edge feature enhancement; A feature extraction module is used to extract identification features based on scar morphological data; the morphological identification features of scars are extracted by image processing technology, and the morphological identification features are categorized according to the corresponding scar record categories, wherein the morphological identification features include scar area, perimeter, concavity and convexity, and surface texture distribution features; The model training module is used to train the scar classification model according to the data set; the morphological recognition feature data with category labels in the array are classified and trained in a preset neural network model to establish the scar classification model; The scar matching module is used for importing, judging and classifying the objects to be identified; obtaining the data of the objects to be classified, preprocessing and extracting identification features, obtaining the features to be identified for classification and identification, classifying and identifying the features to be identified through the scar classification model, and obtaining the corresponding scar category.
7. The human scar rapid classification system based on visual computing according to claim 6 is characterized in that: Also included is a vascular cross-assessment module, specifically including: A blood vessel data acquisition unit, used for performing a tomographic scan of the blood vessel distribution of the scar, acquiring multiple groups of blood vessel distribution layers and performing spatial stacking to acquire a blood vessel spatial distribution model of the scar; A feature vectorization unit, used for performing spatial feature vectorization on the vascular spatial distribution feature to establish the vascular distribution feature of the scar; A mapping model building unit is used to obtain the spatial distribution characteristics of blood vessels corresponding to different scar categories based on the medical record database, and to perform spatial feature vectorization to obtain a blood vessel distribution mapping model corresponding to the scar category; A cross-matching unit is used to classify and match the blood vessel distribution features based on the blood vessel distribution mapping model, and to cross-validate and judge the scar category of the object data to be classified based on the matching rate.
8. The human scar rapid classification system based on visual computing according to claim 7 is characterized in that: The model training module specifically includes: A data set grouping unit, used for randomly grouping the data set to establish a plurality of training groups, and training the scar classification model based on the training groups, each of the training groups including a plurality of scars of different categories; A cross-validation unit, used to cross-validate the plurality of randomly grouped image data sets with the plurality of scar classification models obtained through training, respectively, to obtain a plurality of cross-validation result sets corresponding to the random groups; a deviation calculation unit, configured to calculate an overall deviation value of the cross-validation result set based on actual scar records corresponding to the scar morphology data in the random group, and perform point cloud distribution simulation on the scar classification model corresponding to the random group based on multiple calculation results; The cross correction unit is used to perform parameter averaging processing on multiple scar classification models with smaller deviation values based on point cloud distribution, so as to establish an optimized scar classification model.
9. The human scar rapid classification system based on visual computing according to claim 8, characterized in that: It also includes a cycle judgment module, which specifically includes: A judgment data acquisition unit, used to acquire morphological features and color features corresponding to the scar, as well as the blood vessel distribution and active state of the scar based on the image data, so as to generate a period judgment data set; The cycle stage judgment unit is used to obtain the feature matching library of the corresponding scar category at different stages, and to match the cycle judgment data set to obtain the cycle stage of the scar.
10. The human scar rapid classification system based on visual computing according to claim 6, characterized in that: It also includes a factor screening module, including: The inducement interaction unit is used to obtain the scar category based on the scar classification model matching, match several inducements corresponding to the scar category, generate inducement-guided interaction links and output them; The inducement establishing unit is used to obtain user interaction feedback to record inducement information corresponding to the scar.
Citation Information
Cited By
Orthopedic surgery postoperative scar recovery dynamic evaluation system based on deep learning
CN121120616A
Orthopedic postoperative scar recovery dynamic evaluation system based on deep learning
CN121120616B
