CN118552793B - A postoperative incision healing status recognition system based on artificial intelligence - Google Patents

Abstract

The application discloses an artificial intelligence-based postoperative incision healing state identification system, which relates to the field of intelligent identification, and is characterized in that an operation incision state image of a patient is acquired through a camera, and an image processing and analysis algorithm based on artificial intelligence and deep learning is introduced at the rear end to analyze the operation incision state image, so that hidden characteristics and multi-scale fusion information related to the operation incision state of the patient in the image are learned and identified, and the identification and detection of the operation incision healing state of the patient are carried out to judge whether abnormality exists. In this way, more scientific and intelligent support can be provided for the healing status identification and abnormality detection of a patient's post-operative wound based on artificial intelligence and machine learning techniques.

Description

A postoperative incision healing status recognition system based on artificial intelligence

Technical Field

The present application relates to the field of intelligent identification, and more specifically, to a postoperative incision healing status identification system based on artificial intelligence.

Background Art

Monitoring the healing status of surgical incisions is a key part of postoperative care and recovery. Postoperative incision healing status identification refers to observing and evaluating the surgical incision to determine whether the incision is healing normally. This process plays a vital role in the postoperative period and postoperative care. It can detect and deal with abnormal wound conditions in a timely manner, ensure good incision healing, and reduce the risk of infection and complications.

However, the traditional method of monitoring the healing status of postoperative incisions usually relies on regular doctor's inspections. The doctor observes the appearance characteristics of the incision, such as redness, exudation, swelling, etc., through naked eye observation and empirical judgment during ward rounds to determine the postoperative wound healing of the patient. This is not only time-consuming and labor-intensive, but may also result in inaccurate or in-time monitoring of the patient's wound healing status due to human factors, increasing the risk of wound infection and complications for the patient.

Therefore, an optimized postoperative incision healing status identification system is desired.

Summary of the invention

In order to solve the above technical problems, this application is proposed.

According to one aspect of the present application, a postoperative incision healing status recognition system based on artificial intelligence is provided, which comprises:

A surgical incision status image acquisition module, used to acquire a surgical incision status image acquired by a camera;

A surgical incision state HOG feature extraction module, used to extract HOG features from the surgical incision state image to obtain a surgical incision state HOG feature vector;

A surgical incision state feature boundary compensation module, used for performing boundary compensation on the surgical incision state HOG feature vector based on the surgical incision state feature to obtain a surgical incision state boundary compensation multi-scale feature map;

A surgical incision state feature multi-scale perception module, used for inputting the surgical incision state boundary compensation multi-scale feature map into a feature multi-scale perception enhancement module to obtain an enhanced surgical incision state boundary compensation multi-scale feature map;

A surgical incision state importance feature guided enhanced expression module, used for performing a cross-modal feature guided joint constraint expression on the enhanced surgical incision state boundary compensation multi-scale feature map based on the surgical incision state HOG feature vector to obtain a texture feature guided surgical incision state multi-scale fusion feature map;

The surgical incision state abnormality recognition module is used to guide the surgical incision state multi-scale fusion feature map based on the texture feature to determine the recognition result, and the recognition result is used to indicate whether there is an abnormality.

In the above-mentioned artificial intelligence-based postoperative incision healing status recognition system, the surgical incision status feature boundary compensation module is used to: input the surgical incision status image into the MBCNet based on the backbone network and the boundary feature extraction branch to obtain the surgical incision status boundary compensation multi-scale feature map.

In the above-mentioned artificial intelligence-based postoperative incision healing state recognition system, the surgical incision state feature multi-scale perception module includes:

A channel perception enhancement unit is used to perform channel perception enhancement processing on the surgical incision state boundary compensation multi-scale feature map on different branches to obtain a first surgical incision state boundary compensation multi-scale channel local activation feature vector, a second surgical incision state boundary compensation multi-scale channel compression feature map and a receptive field expansion surgical incision state boundary compensation global multi-scale activation feature matrix;

An incision state boundary compensation global multi-scale activation unit, used for performing position point multiplication of the receptive field expansion surgery incision state boundary compensation global multi-scale activation feature matrix and each feature matrix along the channel dimension of the second surgery incision state boundary compensation multi-scale channel compression feature map to obtain a second channel compression surgery incision state boundary compensation global multi-scale activation feature map;

An incision state boundary compensation multi-scale channel compression local activation unit is used to use each position feature value in the first surgical incision state boundary compensation multi-scale channel local activation feature vector as a weighted weight, and weight each feature matrix along the channel dimension of the second surgical incision state boundary compensation multi-scale channel compression feature map to obtain a second surgical incision state boundary compensation multi-scale channel compression local activation feature map;

An incision state boundary compensation multi-scale fusion unit, used for adding the second channel compressed surgery incision state boundary compensation global multi-scale activation feature map and the second surgery incision state boundary compensation multi-scale channel compressed local activation feature map according to position to obtain a second channel compressed surgery incision state boundary compensation multi-scale fusion activation feature map;

The enhanced surgical incision state boundary compensation multi-scale feature extraction unit is used to perform hole convolution encoding on the second channel compressed surgical incision state boundary compensation multi-scale fusion activation feature map to obtain an enhanced surgical incision state boundary compensation multi-scale feature map.

In the above-mentioned artificial intelligence-based postoperative incision healing state recognition system, the channel sensing enhancement unit is used to:

In the first branch, point convolution processing is performed on the surgical incision state boundary compensation multi-scale feature map to obtain a first surgical incision state boundary compensation multi-scale channel compression feature map; global mean pooling is performed on the first surgical incision state boundary compensation multi-scale channel compression feature map to obtain a first surgical incision state boundary compensation multi-scale channel compression feature vector; nonlinear activation processing is performed on the first surgical incision state boundary compensation multi-scale channel compression feature vector to obtain the first surgical incision state boundary compensation multi-scale channel local activation feature vector;

In the second branch, point convolution processing is performed on the surgical incision state boundary compensation multi-scale feature map to obtain the second surgical incision state boundary compensation multi-scale channel compression feature map;

In the third branch, the surgical incision state boundary compensation multi-scale feature map is subjected to hole convolution encoding to obtain a surgical incision state boundary compensation multi-scale receptive field expansion feature map; the surgical incision state boundary compensation multi-scale receptive field expansion feature map is subjected to point convolution processing to obtain a receptive field expansion surgical incision state boundary compensation global multi-scale feature matrix; the receptive field expansion surgical incision state boundary compensation global multi-scale feature matrix is subjected to nonlinear activation to obtain the receptive field expansion surgical incision state boundary compensation global multi-scale activation feature matrix.

In the above-mentioned artificial intelligence-based postoperative incision healing status recognition system, the surgical incision status importance feature-guided enhanced expression module is used to: input the enhanced surgical incision status boundary compensation multi-scale feature map and the surgical incision status HOG feature vector into the MetaNet-based cross-modal joint constraint encoder to obtain the texture feature-guided surgical incision status multi-scale fusion feature map.

In the above-mentioned artificial intelligence-based postoperative incision healing status recognition system, the surgical incision status abnormality recognition module is used to: input the texture feature-guided surgical incision status multi-scale fusion feature map into the classifier-based healing status recognition module to obtain the recognition result, and the recognition result is used to indicate whether there is an abnormality.

In the above-mentioned artificial intelligence-based postoperative incision healing status recognition system, the surgical incision status abnormality recognition module includes: an expansion unit, used to expand the texture feature-guided surgical incision status multi-scale fusion feature map into a classification feature vector based on a row vector or a column vector; a fully connected encoding unit, used to use multiple fully connected layers of the classifier to fully connect encode the classification feature vector to obtain an encoded classification feature vector; a classification result generation unit, used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the recognition result.

Compared with the prior art, the present application provides a postoperative incision healing state recognition system based on artificial intelligence, which collects the patient's surgical incision state image through a camera, and introduces an image processing and analysis algorithm based on artificial intelligence and deep learning in the back end to analyze the surgical incision state image, so as to learn and identify the hidden features and multi-scale fusion information about the patient's postoperative incision state in the image, so as to identify and detect the patient's postoperative incision healing state to determine whether there is an abnormality. In this way, it is possible to provide more scientific and intelligent support for the recognition of the healing state and abnormality detection of the patient's postoperative wound based on artificial intelligence and machine learning technology.

BRIEF DESCRIPTION OF THE DRAWINGS

By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

FIG1 is a block diagram of a postoperative incision healing status recognition system based on artificial intelligence according to an embodiment of the present application;

FIG2 is a system architecture diagram of a postoperative incision healing status recognition system based on artificial intelligence according to an embodiment of the present application;

Figure 3 is a block diagram of a multi-scale perception module of surgical incision state features in an artificial intelligence-based postoperative incision healing state recognition system according to an embodiment of the present application.

DETAILED DESCRIPTION

Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and/or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and/or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

Traditional postoperative wound healing status monitoring methods usually rely on regular doctor inspections, through the doctor's naked eye observation and experience to observe the appearance of the incision, such as redness, exudation, swelling, etc., to determine the postoperative wound healing status of patients. This is not only time-consuming and labor-intensive, but also may cause the patient's wound healing status to be not accurately or timely checked due to human factors, increasing the risk of wound infection and complications for patients. Therefore, an optimized postoperative incision healing status recognition system is desired.

With the rapid development of artificial intelligence technology, machine learning, especially deep learning, has made significant progress in image recognition, pattern recognition and other fields. Deep learning models can automatically learn features from large amounts of data and have natural advantages in processing and analyzing complex images. Artificial intelligence is increasingly used in the medical field, especially in postoperative management, where its accuracy and efficiency advantages have been fully demonstrated.

In the technical solution of the present application, a postoperative incision healing state recognition system based on artificial intelligence is proposed. FIG. 1 is a block diagram of a postoperative incision healing state recognition system based on artificial intelligence according to an embodiment of the present application. FIG. 2 is a system architecture diagram of a postoperative incision healing state recognition system based on artificial intelligence according to an embodiment of the present application. As shown in FIG. 1 and FIG. 2, a postoperative incision healing state recognition system 300 based on artificial intelligence according to an embodiment of the present application includes: a surgical incision state image acquisition module 310, for acquiring a surgical incision state image acquired by a camera; a surgical incision state HOG feature extraction module 320, for extracting HOG features from the surgical incision state image to obtain a surgical incision state HOG feature vector; a surgical incision state feature boundary compensation module 330, for performing boundary compensation on the surgical incision state HOG feature vector based on the surgical incision state feature to obtain a surgical incision state boundary compensated multi-scale feature map; and a surgical incision state feature multi-scale perception module 340. 0, used to input the surgical incision state boundary compensation multi-scale feature map into the feature multi-scale perception enhancement module to obtain an enhanced surgical incision state boundary compensation multi-scale feature map; a surgical incision state importance feature guided enhancement expression module 350, used to perform a cross-modal feature guided joint constraint expression on the enhanced surgical incision state boundary compensation multi-scale feature map based on the surgical incision state HOG feature vector to obtain a texture feature guided surgical incision state multi-scale fusion feature map; a surgical incision state abnormality recognition module 360, used to determine a recognition result based on the texture feature guided surgical incision state multi-scale fusion feature map, and the recognition result is used to indicate whether there is an abnormality.

In particular, the surgical incision state image acquisition module 310 is used to acquire a surgical incision state image acquired by a camera. It should be understood that the surgical incision state image refers to an image obtained by imaging technology (such as medical imaging) to display the state of the surgical incision. These images can provide detailed information about the surgical incision. The surgical incision state image plays a key role in identifying and detecting the healing state of the patient's postoperative incision. Medical personnel can judge the healing state by analyzing the degree of closure of the incision, inflammation, tissue reconstruction and other information through image analysis.

In particular, the surgical incision state HOG feature extraction module 320 is used to extract HOG features from the surgical incision state image to obtain a surgical incision state HOG feature vector. It should be understood that the Histogram of Oriented Gradients (HOG) is a feature descriptor widely used in computer vision. It constructs a feature vector by counting the gradient direction and amplitude of the local area in the image, and can effectively describe the edge and texture information in the image, thereby describing the local features in the image, so as to better represent the state characteristics of the surgical incision in the future. In addition, the HOG feature has a certain invariance to image transformations such as illumination changes, scale changes, and rotations, and can overcome the influence of these changes on feature extraction to a certain extent, and improve the robustness and stability of the system. Based on this, in the technical solution of the present application, HOG features are extracted from the surgical incision state image to obtain a surgical incision state HOG feature vector.

In particular, the surgical incision state feature boundary compensation module 330 is used to perform boundary compensation on the surgical incision state HOG feature vector based on the surgical incision state feature to obtain a surgical incision state boundary compensation multi-scale feature map. In particular, in a specific example of the present application, the surgical incision state image is input into the MBCNet based on the backbone network and the boundary feature extraction branch to obtain the surgical incision state boundary compensation multi-scale feature map. Considering that there are feature information in different aspects and scales in the surgical incision state image, these features are of great significance to the recognition and abnormality detection of the surgical incision state. Therefore, a convolutional neural network model with excellent performance in implicit feature extraction of images is used to extract features of the surgical incision state image. In particular, considering that MBCNet is a deep convolutional neural network for image semantic segmentation and understanding, it mainly solves the problem of boundary information loss caused by multiple convolutions and upsampling. The network adopts a multi-scale fusion boundary feature extraction branch, which can improve the accuracy of image semantic understanding. Therefore, the surgical incision state image is further input into the MBCNet based on the backbone network and the boundary feature extraction branch to obtain the surgical incision state boundary compensation multi-scale feature map. It is worth mentioning that here, the MBCNet contains two branches, one is the backbone network and the other is the boundary feature extraction branch. The backbone network is used to extract the global features of the surgical incision state in the surgical incision state image, while the boundary feature extraction branch is used to extract the boundary feature information of the surgical incision state, which helps to capture the key features in the surgical incision state image more comprehensively and accurately, so as to enhance the model's understanding of the surgical incision state.

In particular, the surgical incision state feature multi-scale perception module 340 is used to input the surgical incision state boundary compensation multi-scale feature map into the feature multi-scale perception enhancement module to obtain an enhanced surgical incision state boundary compensation multi-scale feature map. Considering that the healing state of the surgical incision exhibits different feature information at different feature scales, for example, fine texture changes may be more obvious at a smaller scale, while a larger scale may be more helpful in identifying the overall healing trend. Moreover, these multi-scale features of the incision state are of great significance for subsequent healing state perception and abnormal situation identification. Through the analysis and enhancement of multi-scale features, these subtle and macro features of the patient's postoperative wound healing state can be captured at the same time to help the recognition system understand the healing state of the incision more accurately. Based on this, in the technical solution of the present application, the surgical incision state boundary compensation multi-scale feature map is further input into the feature multi-scale perception enhancement module to obtain an enhanced surgical incision state boundary compensation multi-scale feature map. Through the multi-scale feature analysis and enhancement processing of the feature multi-scale perception enhancement module, the system can simultaneously focus on the wound status feature information at different scales, and learn and identify the correlation between these wound status features at different scales and the subsequent wound healing recognition task, thereby enhancing the expression of key features. In this way, more comprehensive surgical incision state healing features can be automatically learned and portrayed, improving the accuracy and robustness of wound healing state recognition. In particular, in a specific example of the present application, as shown in FIG3 , the surgical incision state feature multi-scale perception module 340 includes: a channel perception enhancement unit 341, which is used to perform channel perception enhancement processing on the surgical incision state boundary compensation multi-scale feature map on different branches to obtain a first surgical incision state boundary compensation multi-scale channel local activation feature vector, a second surgical incision state boundary compensation multi-scale channel compression feature map and a receptive field expansion surgical incision state boundary compensation global multi-scale activation feature matrix; an incision state boundary compensation global multi-scale activation unit 342, which is used to multiply the receptive field expansion surgical incision state boundary compensation global multi-scale activation feature matrix with each feature matrix along the channel dimension of the second surgical incision state boundary compensation multi-scale channel compression feature map by position point to obtain a second channel compression surgical incision state boundary compensation global multi-scale activation feature map; an incision state boundary compensation multi-scale channel compression local activation unit 343, used to use the various positional feature values in the first surgical incision state boundary compensation multi-scale channel local activation feature vector as weighted weights, and weighted the various feature matrices along the channel dimension of the second surgical incision state boundary compensation multi-scale channel compression feature map to obtain the second surgical incision state boundary compensation multi-scale channel compression local activation feature map; the incision state boundary compensation multi-scale fusion unit 344, used to add the second channel compression surgical incision state boundary compensation global multi-scale activation feature map and the second surgical incision state boundary compensation multi-scale channel compression local activation feature map by position to obtain the second channel compression surgical incision state boundary compensation multi-scale fusion activation feature map; the enhanced surgical incision state boundary compensation multi-scale feature extraction unit 345, used to perform hole convolution encoding on the second channel compression surgical incision state boundary compensation multi-scale fusion activation feature map to obtain the enhanced surgical incision state boundary compensation multi-scale feature map.

Specifically, the channel perception enhancement unit 341 is used to perform channel perception enhancement processing on the surgical incision state boundary compensation multi-scale feature map on different branches to obtain a first surgical incision state boundary compensation multi-scale channel local activation feature vector, a second surgical incision state boundary compensation multi-scale channel compression feature map and a receptive field expansion surgical incision state boundary compensation global multi-scale activation feature matrix. In an embodiment of the present application, in the first branch, the surgical incision state boundary compensation multi-scale feature map is point convolution processed to obtain a first surgical incision state boundary compensation multi-scale channel compression feature map; the first surgical incision state boundary compensation multi-scale channel compression feature map is global mean pooling to obtain a first surgical incision state boundary compensation multi-scale channel compression feature vector; the first surgical incision state boundary compensation multi-scale channel compression feature vector is nonlinearly activated to obtain the first surgical incision state boundary compensation multi-scale channel local activation feature vector; in the second branch, the surgical incision state boundary compensation multi-scale feature is processed. The image is point convolution processed to obtain the second surgical incision state boundary compensation multi-scale channel compression feature map; in the third branch, the surgical incision state boundary compensation multi-scale feature map is void convolution encoded to obtain the surgical incision state boundary compensation multi-scale receptive field expansion feature map; the surgical incision state boundary compensation multi-scale receptive field expansion feature map is point convolution processed to obtain the receptive field expansion surgical incision state boundary compensation global multi-scale feature matrix; the receptive field expansion surgical incision state boundary compensation global multi-scale feature matrix is nonlinearly activated to obtain the receptive field expansion surgical incision state boundary compensation global multi-scale activation feature matrix.

Specifically, the incision state boundary compensation global multi-scale activation unit 342 is used to perform position point multiplication of the receptive field expansion surgical incision state boundary compensation global multi-scale activation feature matrix and the second surgical incision state boundary compensation multi-scale channel compression feature map along the channel dimension to obtain the second channel compression surgical incision state boundary compensation global multi-scale activation feature map. It should be understood that by multiplying the global multi-scale activation feature map of the first channel with the second channel compression surgical incision state boundary compensation feature map by position point, information fusion at different levels and channels can be achieved. This helps to improve the model's ability to identify and compensate for the surgical incision state boundary, so that the model can more accurately capture the boundary information of the surgical incision and improve the ability to understand and judge the surgical incision state.

Specifically, the incision state boundary compensation multi-scale channel compression local activation unit 343 and the incision state boundary compensation multi-scale fusion unit 344 are used to use each position feature value in the first surgical incision state boundary compensation multi-scale channel local activation feature vector as a weighted weight, and weight each feature matrix along the channel dimension of the second surgical incision state boundary compensation multi-scale channel compression feature map to obtain a second surgical incision state boundary compensation multi-scale channel compression local activation feature map; and add the second channel compression surgical incision state boundary compensation global multi-scale activation feature map and the second surgical incision state boundary compensation multi-scale channel compression local activation feature map by position to obtain a second channel compression surgical incision state boundary compensation multi-scale fusion activation feature map. In this way, the model's ability to understand and recognize the surgical incision state can be improved, so that the model can more accurately capture the characteristics and boundary information of the surgical incision.

Specifically, the enhanced surgical incision state boundary compensation multi-scale feature extraction unit 345 is used to perform hole convolution coding on the second channel compressed surgical incision state boundary compensation multi-scale fusion activation feature map to obtain an enhanced surgical incision state boundary compensation multi-scale feature map. It should be understood that hole convolution coding can help the model better capture the detailed features of the surgical incision state, especially in the boundary compensation multi-scale feature map. This helps to enhance the representation ability of the surgical incision state in the feature map and improve the model's recognition and understanding of the surgical incision boundary. Among them, hole convolution is a special convolution operation in the convolutional neural network. By introducing holes (or called expansion rate) between the convolution kernels, the field of view of the convolution kernel can be increased, thereby capturing a wider range of contextual information.

In summary, in the above embodiment, the surgical incision state boundary compensation multi-scale feature map is input into the feature multi-scale perception enhancement module to obtain an enhanced surgical incision state boundary compensation multi-scale feature map, including: inputting the surgical incision state boundary compensation multi-scale feature map into the feature multi-scale perception enhancement module to process it with the following multi-scale perception enhancement formula to obtain the enhanced surgical incision state boundary compensation multi-scale feature map; wherein the multi-scale perception enhancement formula is:

;

in, A multi-scale feature map for compensating the surgical incision state boundary, and is a 3×3 dilated convolution operation with dilation numbers of 3 and 2. is a 1×1 convolution operation, Indicates that global mean pooling is performed on each feature matrix along the channel dimension in the feature map. For nonlinear activation processing, It means point multiplication by position. Indicates that the feature map is weighted multiplied along the channel dimension using the feature vector. It is the multi-scale feature map of the enhanced surgical incision state boundary compensation.

It is worth mentioning that in other specific examples of the present application, the surgical incision state boundary compensation multi-scale feature map can also be input into the feature multi-scale perception enhancement module in other ways to obtain an enhanced surgical incision state boundary compensation multi-scale feature map, for example: input the surgical incision state boundary compensation multi-scale feature map; perform multi-scale feature extraction on the input surgical incision state boundary compensation multi-scale feature map, which may include convolution kernels or pooling operations of different sizes to capture feature information at different scales; fuse the extracted multi-scale features, which can be done by splicing, weighted summation, etc., to integrate feature information at different scales; enhance the fused features, such as applying activation functions, normalization operations, etc., to enhance the representation ability of the features; output the enhanced surgical incision state boundary compensation multi-scale feature map.

In particular, the surgical incision state importance feature guided enhanced expression module 350 is used to perform a joint constraint expression based on cross-modal feature guidance on the enhanced surgical incision state boundary compensation multi-scale feature map based on the surgical incision state HOG feature vector to obtain a texture feature guided surgical incision state multi-scale fusion feature map. In particular, in a specific example of the present application, the enhanced surgical incision state boundary compensation multi-scale feature map and the surgical incision state HOG feature vector are input into a MetaNet-based cross-modal joint constraint encoder to obtain the texture feature guided surgical incision state multi-scale fusion feature map. It should be understood that the enhanced surgical incision state boundary compensation multi-scale feature map and the surgical incision state HOG feature vector respectively contain multi-scale enhanced features and semantic information about the surgical incision state, and the surgical incision state HOG feature vector contains feature information such as texture of the surgical incision state. These two features are different feature expressions about the patient's postoperative incision healing state, and both contain feature representations and information about the surgical incision state. Based on this, in order to integrate the postoperative incision state feature information from different analysis levels and methods, so that the system can comprehensively utilize these different categories of wound state features, and improve the expression ability and discrimination of the final wound healing state features, in the technical solution of the present application, the enhanced surgical incision state boundary compensation multi-scale feature map and the surgical incision state HOG feature vector are further input into the cross-modal joint constraint encoder based on MetaNet to obtain the texture feature-guided surgical incision state multi-scale fusion feature map. Through the processing of the cross-modal joint constraint encoder based on MetaNet, the surgical incision state HOG feature can be used to constrain the expression of the enhanced surgical incision state boundary compensation multi-scale feature map based on the channel dimension, so that the implicit features and key features of the patient's postoperative incision state are more prominent. In other words, by introducing texture features, the system can better capture the texture information in the surgical incision state image, which is of great significance to the recognition of the incision healing state. The texture features guide the generation of the surgical incision state multi-scale fusion feature map, which helps the system to more comprehensively and accurately characterize the characteristics of the surgical incision state and improve the performance of the recognition system.

In particular, the surgical incision state abnormality recognition module 360 is used to determine the recognition result based on the texture feature-guided surgical incision state multi-scale fusion feature map, and the recognition result is used to indicate whether there is an abnormality. In particular, in a specific example of the present application, the texture feature-guided surgical incision state multi-scale fusion feature map is input into a healing state recognition module based on a classifier to obtain the recognition result, and the recognition result is used to indicate whether there is an abnormality. That is, the texture feature of the patient's wound state is used to guide the fusion representation information of the multi-scale feature expression of the surgical incision state for classification processing, so as to identify and detect the healing state of the patient's postoperative incision to determine whether there is a healing abnormality. In this way, it is possible to provide more scientific and intelligent support for the recognition and abnormality detection of the healing state of the patient's postoperative wound based on artificial intelligence and machine learning technology. Specifically, the texture feature-guided multi-scale fusion feature map of the surgical incision state is input into a healing state recognition module based on a classifier to obtain the recognition result, and the recognition result is used to indicate whether there is an abnormality. The process includes: first, expanding the texture feature-guided multi-scale fusion feature map of the surgical incision state into a classification feature vector based on a row vector or a column vector; then using multiple fully connected layers of the classifier to fully connect encode the classification feature vector to obtain an encoded classification feature vector; and then passing the encoded classification feature vector through the Softmax classification function of the classifier to obtain the recognition result.

Preferably, inputting the texture feature-guided multi-scale fusion feature map of the surgical incision state into a classifier-based healing state recognition module to obtain a recognition result comprises:

Probabilistically converting each feature value of the texture feature-guided surgical incision state multi-scale fusion feature map based on a probability activation function, such as a sigmoid function or a softmax function, to obtain a probabilistic texture feature-guided surgical incision state multi-scale fusion feature map;

Obtaining the texture feature-guided surgical incision state multi-scale fusion feature map and inputting it into a healing state recognition module based on a classifier to obtain an abnormal probability value;

Determine a class recognition symbol value based on a comparison between each feature value of the probabilistic texture feature-guided surgical incision state multi-scale fusion feature map and the abnormal probability value, wherein the class recognition symbol value is equal to one, zero, and negative one in response to the feature value of the probabilistic texture feature-guided surgical incision state multi-scale fusion feature map being greater than, equal to, and less than the abnormal probability value, respectively;

Calculate the mean of all feature values of the probabilistic texture feature-guided surgical incision state multi-scale fusion feature map to obtain a class overall phase shift value;

After multiplying each eigenvalue of the probabilistic texture feature-guided surgical incision state multi-scale fusion feature map by the class cognitive symbol value and the class overall phase shift value, weighted difference calculation is performed, and the absolute value is taken to obtain the optimized eigenvalue of the probabilistic texture feature-guided surgical incision state multi-scale fusion feature map;

The optimized texture feature-guided surgical incision state multi-scale fusion feature map composed of the optimized feature values is input into a classifier-based healing state recognition module to obtain a recognition result.

Specifically expressed as:

;

;

and are the feature values before and after optimization of the probabilistic texture feature-guided multi-scale fusion feature map of the surgical incision state, is the abnormal probability value, is a quasi-cognitive symbolic function, and is a hyperparameter, It is the mean of all feature values of the probabilistic texture feature-guided multi-scale fusion feature map of the surgical incision state, that is, the overall phase shift value of the class.

Here, the enhanced surgical incision state boundary compensation multi-scale feature map expresses the backbone-boundary image semantic feature distribution of the surgical incision state image and the multi-scale perceptual enhanced image semantic feature distribution. Therefore, after the enhanced surgical incision state boundary compensation multi-scale feature map and the surgical incision state HOG feature vector are input into the cross-modal joint constraint encoder based on MetaNet, the channel distribution of the enhanced surgical incision state boundary compensation multi-scale feature map will be constrained based on the HOG feature distribution of the surgical incision state image expressed by the surgical incision state HOG feature vector, thereby resulting in the loss of the prior-posterior class probability causal association of the image semantic feature representation of the texture feature-guided surgical incision state multi-scale fusion feature map relative to the enhanced surgical incision state boundary compensation multi-scale feature map, affecting the accuracy of the classification results.

Therefore, in the above optimization process, the class cognitive phase shift response of the texture feature-guided multi-scale fusion feature map of the surgical incision state is obtained by comparing the probabilistic amplitude of the feature value of the texture feature-guided multi-scale fusion feature map with the class probability, and the feature distribution sequence invariance transformation based on the class differential distribution expansion is performed on the feature value of the texture feature-guided multi-scale fusion feature map of the surgical incision state relative to the overall class probability representation of the feature map, so as to realize the causal constraint of the posterior class probability of the texture feature-guided multi-scale fusion feature map on its prior feature distribution representation, so as to improve the accuracy of the recognition result obtained by the texture feature-guided multi-scale fusion feature map of the surgical incision state input into the healing state recognition module based on the classifier. In this way, the postoperative incision healing state of the patient can be more accurately identified and detected to determine whether there is an abnormal healing state.

As described above, a postoperative incision healing state recognition system 300 based on artificial intelligence according to an embodiment of the present application can be implemented in various wireless terminals, such as a server with a postoperative incision healing state recognition algorithm based on artificial intelligence. In a possible implementation, a postoperative incision healing state recognition system 300 based on artificial intelligence according to an embodiment of the present application can be integrated into a wireless terminal as a software module and/or a hardware module. For example, the postoperative incision healing state recognition system 300 based on artificial intelligence can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the postoperative incision healing state recognition system 300 based on artificial intelligence can also be one of the many hardware modules of the wireless terminal.

Alternatively, in another example, the artificial intelligence-based postoperative incision healing status identification system 300 and the wireless terminal may also be separate devices, and the artificial intelligence-based postoperative incision healing status identification system 300 may be connected to the wireless terminal via a wired and/or wireless network and transmit interactive information in accordance with an agreed data format.

The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims (5)

1. An artificial intelligence based post-operative incision healing state identification system, comprising:

the surgical incision state image acquisition module is used for acquiring a surgical incision state image acquired by the camera;

The surgical incision state HOG feature extraction module is used for extracting HOG features from the surgical incision state image to obtain a surgical incision state HOG feature vector;

The surgical incision state feature boundary compensation module is used for performing boundary compensation based on the surgical incision state feature on the surgical incision state HOG feature vector to obtain a surgical incision state boundary compensation multi-scale feature map;

The surgical incision state characteristic multi-scale sensing module is used for inputting the surgical incision state boundary compensation multi-scale characteristic map into the characteristic multi-scale sensing strengthening module to obtain a strengthened surgical incision state boundary compensation multi-scale characteristic map;

The surgical incision state importance feature guiding and strengthening expression module is used for carrying out joint constraint expression based on cross-modal feature guiding on the strengthening surgical incision state boundary compensation multi-scale feature map based on the surgical incision state HOG feature vector so as to obtain a texture feature guiding surgical incision state multi-scale fusion feature map;

The surgical incision state abnormality recognition module is used for guiding the surgical incision state multi-scale fusion feature map based on the texture features and determining a recognition result, wherein the recognition result is used for representing whether abnormality exists or not;

the surgical incision state characteristic multi-scale sensing module comprises:

The channel perception enhancement unit is used for carrying out channel perception enhancement processing on the operation incision state boundary compensation multi-scale feature map on different branches so as to obtain a first operation incision state boundary compensation multi-scale channel local activation feature vector, a second operation incision state boundary compensation multi-scale channel compression feature map and a receptive field expansion operation incision state boundary compensation global multi-scale activation feature matrix;

The incision state boundary compensation global multi-scale activation unit is used for multiplying the receptive field expansion operation incision state boundary compensation global multi-scale activation feature matrix and each feature matrix of the second operation incision state boundary compensation multi-scale channel compression feature map along the channel dimension according to position points to obtain a second channel compression operation incision state boundary compensation global multi-scale activation feature map;

The incision state boundary compensation multi-scale channel compression local activation unit is used for taking each position characteristic value in the first operation incision state boundary compensation multi-scale channel local activation characteristic vector as a weighting weight, and weighting each characteristic matrix of the second operation incision state boundary compensation multi-scale channel compression characteristic map along the channel dimension to obtain a second operation incision state boundary compensation multi-scale channel compression local activation characteristic map;

the incision state boundary compensation multi-scale fusion unit is used for carrying out position addition on the second channel compression operation incision state boundary compensation global multi-scale activation characteristic diagram and the second operation incision state boundary compensation multi-scale channel compression local activation characteristic diagram to obtain a second channel compression operation incision state boundary compensation multi-scale fusion activation characteristic diagram;

The reinforced operation incision state boundary compensation multi-scale feature extraction unit is used for carrying out cavity convolution encoding on the second channel compressed operation incision state boundary compensation multi-scale fusion activation feature map so as to obtain a reinforced operation incision state boundary compensation multi-scale feature map;

The channel perception strengthening unit is used for:

Performing point convolution processing on the boundary compensation multi-scale feature map of the surgical incision state in a first branch to obtain a compression feature map of a first surgical incision state boundary compensation multi-scale channel; global averaging is carried out on the first operation incision state boundary compensation multi-scale channel compression feature map so as to obtain a first operation incision state boundary compensation multi-scale channel compression feature vector; performing nonlinear activation processing on the first surgical incision state boundary compensation multi-scale channel compression feature vector to obtain the first surgical incision state boundary compensation multi-scale channel local activation feature vector;

performing point convolution processing on the boundary compensation multi-scale feature map of the surgical incision state in a second branch to obtain a compression feature map of the boundary compensation multi-scale channel of the surgical incision state;

In a third branch, carrying out cavity convolution coding on the boundary compensation multiscale feature map of the surgical incision state to obtain a boundary compensation multiscale receptive field expansion feature map of the surgical incision state; performing point convolution processing on the surgical incision state boundary compensation multiscale receptive field expansion feature map to obtain a receptive field expansion surgical incision state boundary compensation global multiscale feature matrix; and performing nonlinear activation on the receptive field expansion operation incision state boundary compensation global multi-scale feature matrix to obtain the receptive field expansion operation incision state boundary compensation global multi-scale activation feature matrix.

2. The system for identifying the healing state of an incision after operation based on artificial intelligence according to claim 1, wherein the surgical incision state characteristic boundary compensation module is used for: the surgical incision state image is input based on MBCNet including a backbone network and boundary feature extraction branches to obtain the surgical incision state boundary compensation multi-scale feature map.

3. The system for identifying the healing state of an incision after operation based on artificial intelligence according to claim 2, wherein the characteristic of importance of the state of the incision of operation guides the strengthening expression module for: inputting the reinforced surgical incision state boundary compensation multi-scale feature map and the surgical incision state HOG feature vector into a MetaNet-based cross-modal joint constraint encoder to obtain the texture feature guided surgical incision state multi-scale fusion feature map.

4. A post-operative incision healing state identification system based on artificial intelligence according to claim 3, wherein the operative incision state abnormality identification module is configured to: and inputting the multi-scale fusion characteristic map of the surgical incision state guided by the texture characteristics into a healing state recognition module based on a classifier to obtain a recognition result, wherein the recognition result is used for indicating whether an abnormality exists.

5. The artificial intelligence based post-operative incision healing state identification system of claim 4, wherein the operative incision state anomaly identification module comprises:

The unfolding unit is used for unfolding the multi-scale fusion feature map of the texture feature guided operation incision state into classification feature vectors based on row vectors or column vectors;

the full-connection coding unit is used for carrying out full-connection coding on the classification characteristic vectors by using a plurality of full-connection layers of the classifier so as to obtain coded classification characteristic vectors;

and the classification result generation unit is used for passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the identification result.