Spine endoscope image information sensing and automatic enhancement system and method

Abstract

The invention discloses a spine endoscope image information sensing and automatic enhancement system and a spine endoscope image information sensing and automatic enhancement method, and belongs to the technical field of image processing. The high-resolution image generation module uses the trained super-resolution image generation model to process the spinal endoscope video after serialization processing to obtain a high-resolution spinal endoscope image, and the segmentation recognition module carries out segmentation recognition on physiological tissues in the image according to a texture-based threshold segmentation method to obtain a physiological tissue image. And the identification and labeling module is used for fusing the high-resolution spine endoscope image and the physiological tissue image, identifying and labeling the fused high-resolution spine endoscope image by adopting a labeling frame to obtain a target high-resolution spine endoscope image. According to the method, the resolution of the image is improved, the definition of the image is enhanced, the treatment efficiency is improved, the diagnosis and surgical operation efficiency and accuracy of a doctor are improved, and the difficulty and risk of recognition in the surgery are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and in particular relates to a spinal endoscopy image information perception and automatic enhancement system and method. Background Art

[0002] As a major component of the human musculoskeletal system, the spine is responsible for supporting body weight, protecting the spinal cord and nerve roots, and maintaining body balance and mobility. Due to the complexity of its structure and the specificity of its functions, spinal diseases may not only cause severe pain and limited mobility, but may even lead to neurological dysfunction, causing great harm to the patient's physical and mental health. Traditional open surgery is very traumatic to the patient, with slow recovery and high costs. Currently, endoscopic spinal surgery has gradually become an important technology for the treatment of spinal diseases. Endoscopic spinal surgery is a minimally invasive surgery. The basic principle is to use endoscopic technology to send the endoscope and surgical instruments into the location of spinal lesions through a tiny incision on the skin, and perform surgical operations with the assistance of the endoscope to treat spinal diseases. Although this technology has many advantages, it also has some technical problems.

[0003] Due to the optical limitations of endoscopic equipment and the complexity of the spinal structure, the operating space of spinal endoscopes is small, resulting in blurred and low-resolution images, increased treatment time, and time-consuming and labor-intensive treatment, which seriously affects the doctor's diagnosis and surgical operations and increases the risk of surgery. In addition, existing technologies cannot automatically identify physiological tissues in images, such as muscles, bones, and nerves, which further increases the difficulty and risk of surgery.

[0004] Therefore, how to provide an effective technical solution to solve the problems of blurred images, low resolution, and inability to automatically identify physiological tissues in images has become a difficult problem that needs to be solved urgently in the existing technology. Summary of the Invention

[0005] The purpose of the present invention is to provide a spinal endoscopy image information perception and automatic enhancement system and method to solve the problems existing in the prior art such as blurred images, low resolution, and inability to automatically identify physiological tissues in images.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a spinal endoscopy image information perception and automatic enhancement system, comprising an image acquisition module, a high-resolution image generation module, a segmentation and recognition module, and an identification and annotation module; The image acquisition module is used to acquire the spinal endoscopy video acquired by endoscopy, perform serialization processing on the spinal endoscopy video to obtain a spinal endoscopy image sequence, wherein the spinal endoscopy image sequence includes a plurality of time-continuous spinal endoscopy image frames, and upload the plurality of time-continuous spinal endoscopy image frames to the high-resolution image generation module; The high-resolution image generation module is used to process a number of time-continuous spinal endoscopic image frames using the trained super-resolution image generation model to obtain a high-resolution spinal endoscopic image, and upload the high-resolution spinal endoscopic image to the segmentation and recognition module; The segmentation and recognition module is used to segment and recognize physiological tissues in high-resolution spinal endoscopy images according to a texture-based threshold segmentation method to obtain physiological tissue images, and upload the physiological tissue images to the recognition and annotation module; The recognition and annotation module is used to fuse the high-resolution spinal endoscopic image and the physiological tissue image to obtain a fused high-resolution spinal endoscopic image, identify the fused high-resolution spinal endoscopic image to obtain a annotation frame that corresponds one-to-one with the physiological tissue image, and combine the annotation frame with the fused high-resolution spinal endoscopic image to obtain a target high-resolution spinal endoscopic image.

[0007] In one possible design, the image acquisition module is also used to acquire a high-resolution image, perform bicubic downsampling on the high-resolution image to obtain a low-resolution image corresponding to the high-resolution image, and add Gaussian noise to the low-resolution image corresponding to the high-resolution image to obtain a simulated low-resolution image corresponding to the high-resolution image.

[0008] In one possible design, a model training module is further included, which is communicatively connected to the image acquisition module. The model training module is used to input the high-resolution image and the simulated low-resolution image corresponding to the high-resolution image into the super-resolution image generation model for training to obtain a trained super-resolution image generation model.

[0009] In a possible design, after obtaining a sequence of spinal endoscopy images, the following is further included: Performing Gaussian filtering to denoise the spinal endoscopic image sequence to obtain a denoised spinal endoscopic image sequence; Grayscale processing is performed on the denoised spinal endoscopy image sequence to obtain a grayscale image; The grayscale images were standardized and normalized to obtain the preprocessed spinal endoscopy image sequence.

[0010] In one possible design, a trained super-resolution image generation model is used to process a number of temporally consecutive spinal endoscopic image frames to obtain a high-resolution spinal endoscopic image, including: Feature extraction is performed on several temporally continuous spinal endoscopy image frames to obtain a multi-frame feature map, and the multi-frame feature map is temporally and positionally encoded. The multi-frame feature map is matched one-to-one with time to obtain the encoded multi-frame feature map. The feature map at the current moment is used as the current feature map, and the feature map at the previous moment is used as the alignment reference feature map corresponding to the current feature map; when the current moment is the initial moment, the feature map at the initial moment is used as the alignment reference feature map corresponding to the current feature map; Perform frame alignment on the current feature map and the alignment reference feature map to obtain aligned multi-frame feature maps; Perform global feature extraction based on the aligned multi-frame feature maps to obtain a global feature map, perform frame alignment on the global feature map to obtain an aligned global feature map; The aligned multi-frame feature maps and the aligned global feature map are fused to obtain the final feature map, and a high-resolution spinal endoscopic image is reconstructed based on the final feature map.

[0011] In one possible design, the trained super-resolution image generation model is built based on a recurrent neural network and a Transformer network.

[0012] In a possible design, physiological tissues in high-resolution spinal endoscopy images are segmented and identified using a texture-based threshold segmentation method to obtain a physiological tissue image, including: Preprocess the high-resolution spinal endoscopy image to obtain the background image and target image; Use the K-means algorithm to perform preliminary segmentation on the target image to obtain the first image; Using a texture-based threshold segmentation method to segment and identify the target image to obtain a second image; The first image and the second image are fused to obtain a physiological tissue image.

[0013] In one possible design, a texture-based threshold segmentation method is used to segment and identify the target image to obtain a second image, including: Calculate the gray level co-occurrence matrix based on the target image, obtain the gray level co-occurrence matrix of muscle, bone and nerve respectively, and extract the texture features in the gray level co-occurrence matrix of muscle, bone and nerve, wherein the texture features include contrast, energy, entropy, variance, correlation and inverse moment; Calculate the average value of the texture features, set the weights of contrast, energy, entropy, variance, correlation, and inverse moment according to the average value of the texture features, extract the feature value of each pixel in the target image according to the weights of contrast, energy, entropy, variance, correlation, and inverse moment, and obtain the first feature; Extracting muscle tissue in the target image based on the first feature to obtain a muscle tissue image, filling holes in the muscle tissue in the target image to obtain a filled target image; extracting bones and nerves in the filled target image based on the first feature to obtain a bone and nerve image; The muscle tissue image and the bone and nerve images are fused to produce a second image.

[0014] In a second aspect, the present invention provides a method for perceiving and automatically enhancing spinal endoscopic image information, comprising: Acquiring an endoscopic spinal video, and performing serialization processing on the endoscopic spinal video to obtain a spinal endoscopic image sequence, wherein the spinal endoscopic image sequence includes a plurality of temporally continuous spinal endoscopic image frames; Using the trained super-resolution image generation model to process several time-continuous spinal endoscopic image frames, a high-resolution spinal endoscopic image is obtained; The physiological tissue in the high-resolution spinal endoscopy image is segmented and identified using a texture-based threshold segmentation method to obtain a physiological tissue image. The high-resolution spinal endoscopic image and the physiological tissue image are fused to obtain a fused high-resolution spinal endoscopic image, the fused high-resolution spinal endoscopic image is identified to obtain a labeling frame that corresponds one-to-one with the physiological tissue image, and the labeling frame is combined with the fused high-resolution spinal endoscopic image to obtain a target high-resolution spinal endoscopic image.

[0015] In the third aspect, the present invention provides a spinal endoscopic image information perception and automatic enhancement device, comprising a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute a spinal endoscopic image information perception and automatic enhancement method as described in the second invention.

[0016] The beneficial effects of the present invention are as follows: The present invention discloses a spinal endoscopic image information perception and automatic enhancement system and method, comprising an image acquisition module, a high-resolution image generation module, a segmentation and recognition module, and an identification and annotation module. The image acquisition module, after acquiring an endoscopically captured spinal endoscopic video, serializes the endoscopic video to obtain a spinal endoscopic image sequence. The high-resolution image generation module processes the endoscopic image sequence using a trained super-resolution image generation model to obtain a high-resolution spinal endoscopic image. The segmentation and recognition module performs segmentation and recognition on the high-resolution spinal endoscopic image to obtain a physiological tissue image within the high-resolution spinal endoscopic image. The identification and annotation module fuses the physiological tissue image with the high-resolution spinal endoscopic image and annotates the physiological tissue image with an annotation frame to generate a final target high-resolution spinal endoscopic image. The present invention processes the low-resolution endoscopic spinal video captured by the endoscope into a high-resolution spinal endoscopic image with an annotation frame, thereby improving the image resolution and clarity, reducing treatment time, saving efficiency, and enhancing the efficiency and accuracy of doctors' diagnosis and surgical operations. Furthermore, the system identifies physiological tissues in the image and labels them with an annotation frame, reducing the difficulty and risk of identification during surgery and facilitating application and promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A module block diagram of a spinal endoscopy image information perception and automatic enhancement system provided in the first aspect of this embodiment; Figure 2 This is a flowchart of a method for spinal endoscopy image information perception and automatic enhancement provided in the second aspect of this embodiment. DETAILED DESCRIPTION

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the present invention will be briefly introduced below in conjunction with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0019] It should be understood that although the terms first, second, etc. may be used herein to describe various objects, these objects should not be limited by these terms. These terms are merely used to distinguish one object from another. For example, a first object can be referred to as a second object, and similarly, a second object can be referred to as a first object without departing from the scope of the exemplary embodiments of the present invention.

[0020] Example: like Figure 1As shown, the first aspect of this embodiment provides a spinal endoscopy image information perception and automatic enhancement system, including an image acquisition module, a high-resolution image generation module, a segmentation and recognition module, and an identification and annotation module; The image acquisition module is used to acquire the spinal endoscopy video acquired by endoscopy, perform serialization processing on the spinal endoscopy video to obtain a spinal endoscopy image sequence, wherein the spinal endoscopy image sequence includes a plurality of time-continuous spinal endoscopy image frames, and upload the plurality of time-continuous spinal endoscopy image frames to the high-resolution image generation module; The high-resolution image generation module is used to process a number of time-continuous spinal endoscopic image frames using the trained super-resolution image generation model to obtain a high-resolution spinal endoscopic image, and upload the high-resolution spinal endoscopic image to the segmentation and recognition module; The segmentation and recognition module is used to segment and recognize physiological tissues in high-resolution spinal endoscopy images according to a texture-based threshold segmentation method to obtain physiological tissue images, and upload the physiological tissue images to the recognition and annotation module; The recognition and annotation module is used to fuse the high-resolution spinal endoscopic image and the physiological tissue image to obtain a fused high-resolution spinal endoscopic image, identify the fused high-resolution spinal endoscopic image to obtain a annotation frame that corresponds one-to-one with the physiological tissue image, and combine the annotation frame with the fused high-resolution spinal endoscopic image to obtain a target high-resolution spinal endoscopic image.

[0021] Specifically, a target detection algorithm may be used to identify the fused high-resolution spinal endoscopy image. The target detection algorithm is an existing technology and will not be elaborated here.

[0022] In one possible design, the image acquisition module is also used to acquire a high-resolution image, perform bicubic downsampling on the high-resolution image to obtain a low-resolution image corresponding to the high-resolution image, and add Gaussian noise to the low-resolution image corresponding to the high-resolution image to obtain a simulated low-resolution image corresponding to the high-resolution image.

[0023] Specifically, low-resolution images from spinal endoscopy are unclear, unrealistic, poorly illuminated, and unstable. Specifically, low resolution results in blurred details, insufficient contrast, and difficulty distinguishing tissues. Color distortion and lighting issues also hinder image recognition. High-resolution images have a standard pixel density, can distinguish different structures, reproduce the true color of tissues, and have sufficient and uniform brightness. Standard pixel density can be, for example, 2MP or higher. Bicubic downsampling is primarily used to convert high-resolution images into low-resolution images while preserving detail and smoothness. Bicubic downsampling is based on the bicubic interpolation algorithm, a two-dimensional interpolation method that interpolates an image using a cubic polynomial. The bicubic interpolation algorithm calculates the value of the target pixel by fitting a cubic polynomial to the values ​​of the 16 neighboring pixels surrounding the target pixel. After obtaining the low-resolution image through bicubic downsampling, Gaussian noise is added to the low-resolution image to increase its realism, thereby improving the accuracy and robustness of the super-resolution image generation model.

[0024] In one possible design, a model training module is further included, which is communicatively connected to the image acquisition module. The model training module is used to input the high-resolution image and the simulated low-resolution image corresponding to the high-resolution image into the super-resolution image generation model for training to obtain a trained super-resolution image generation model.

[0025] In one possible design, the trained super-resolution image generation model is built based on a recurrent neural network and a Transformer network.

[0026] Furthermore, a recurrent neural network (RNN) is a type of neural network used to process sequential data. It has memory capabilities and can use previous input information to influence current output. The principle of a recurrent neural network is to introduce recurrent connections into the network so that information can be circulated in the network, thereby realizing the storage and processing of sequential data. The Transformer network is a neural network based on the attention mechanism. The principle is to capture information at different positions in the sequence through a multi-head attention mechanism to enhance the learning ability of the model. Each head independently performs self-attention calculations, and then splices or averages the results, thereby improving the robustness and generalization ability of the model.

[0027] In a possible design, after obtaining a sequence of spinal endoscopy images, the following is further included: Performing Gaussian filtering to denoise the spinal endoscopic image sequence to obtain a denoised spinal endoscopic image sequence; Grayscale processing is performed on the denoised spinal endoscopy image sequence to obtain a grayscale image; The grayscale images were standardized and normalized to obtain the preprocessed spinal endoscopy image sequence.

[0028] Specifically, grayscale processing of the denoised spinal endoscopic image sequence can reduce the amount of data processing and improve the operating efficiency of the model. The grayscale image facilitates the subsequent extraction and processing of texture features, avoids interference of other factors on the spinal endoscopic image, and standardizes and normalizes the grayscale image to eliminate the influence between feature dimensions and ensure the fairness of data processing.

[0029] In one possible design, a trained super-resolution image generation model is used to process a number of temporally consecutive spinal endoscopic image frames to obtain a high-resolution spinal endoscopic image, including: Performing feature extraction on a number of temporally continuous spinal endoscopy image frames to obtain a multi-frame feature map, and performing temporal position encoding on the multi-frame feature map to obtain the encoded multi-frame feature map, so that the multi-frame feature map corresponds one-to-one with time, using the feature map at the current moment as the current feature map, and the feature map at the previous moment as the alignment reference feature map corresponding to the current feature map; when the current moment is the initial moment, using the feature map at the initial moment as the alignment reference feature map corresponding to the current feature map; Perform frame alignment on the current feature map and the alignment reference feature map to obtain aligned multi-frame feature maps; Perform global feature extraction based on the aligned multi-frame feature maps to obtain a global feature map, perform frame alignment on the global feature map to obtain an aligned global feature map; The aligned multi-frame feature maps and the aligned global feature map are fused to obtain the final feature map, and a high-resolution spinal endoscopic image is reconstructed based on the final feature map.

[0030] Furthermore, when performing frame alignment processing on the current feature map and the alignment reference feature map, it also includes: The encoded multi-frame feature maps are reversed, and the feature map at time t is used as the current reverse feature map, and the feature map at the previous time is used as the alignment reference reverse feature map corresponding to the current reverse feature map; when the current time is time t, the feature map at time t is used as the alignment reference reverse feature map corresponding to the current reverse feature map; Perform frame alignment on the current reverse feature map and the alignment reference reverse feature map to obtain a reverse aligned multi-frame feature map; The result of frame alignment between the current feature map and the aligned reference feature map is fused with the multi-frame feature maps after reverse alignment.

[0031] In a possible design, physiological tissues in high-resolution spinal endoscopy images are segmented and identified using a texture-based threshold segmentation method to obtain a physiological tissue image, including: Preprocess the high-resolution spinal endoscopy image to obtain the target image; Use the K-means algorithm to perform preliminary segmentation on the target image to obtain the first image; Using a texture-based threshold segmentation method to segment and identify the target image to obtain a second image; The first image and the second image are fused to obtain a physiological tissue image.

[0032] Specifically, the K-means algorithm is a clustering algorithm. The principle of the K-means algorithm is to divide the data set into K clusters through an iterative method so that the sum of the distances between each data point and its assigned cluster center is minimized.

[0033] In one possible design, a texture-based threshold segmentation method is used to segment and identify the target image to obtain a second image, including: Calculate the gray level co-occurrence matrix based on the target image, obtain the gray level co-occurrence matrix of muscle, bone and nerve respectively, and extract the texture features in the gray level co-occurrence matrix of muscle, bone and nerve, wherein the texture features include contrast, energy, entropy, variance, correlation and inverse moment; Calculate the average value of the texture features, normalize the average value of the texture features to obtain the corresponding weight coefficient, set the weights of contrast, energy, entropy, variance, correlation and inverse moment according to the corresponding weight coefficient, extract the characteristic value of each pixel in the target image according to the weights of contrast, energy, entropy, variance, correlation and inverse moment, and obtain the first feature; Extracting muscle tissue in the target image based on the first feature to obtain a muscle tissue image, filling holes in the muscle tissue in the target image to obtain a filled target image; extracting bones and nerves in the filled target image based on the first feature to obtain a bone and nerve image; The muscle tissue image and the bone and nerve images are fused to produce a second image.

[0034] Furthermore, the filled target image is extracted based on the first feature, and the obtained image includes but is not limited to bones and nerves.

[0035] This embodiment provides a spinal endoscopy image information perception and automatic enhancement system, including an image acquisition module, a high-resolution image generation module, a segmentation and recognition module, and an identification and labeling module. The image acquisition module acquires low-resolution spinal endoscopy video acquired by endoscopy, and serializes the low-resolution spinal endoscopy video to obtain a low-resolution spinal endoscopy image. The high-resolution image generation module uses a trained super-resolution image generation model to process the acquired low-resolution spinal endoscopy image to obtain a high-resolution spinal endoscopy image. The segmentation and recognition module segments the high-resolution spinal endoscopy image by a texture-based threshold segmentation method and identifies physiological tissues in the high-resolution spinal endoscopy image. The identification and labeling module labels the identified physiological tissues in the high-resolution spinal endoscopy image to obtain a high-resolution spinal endoscopy image labeled with physiological tissues. By improving the resolution of the spinal endoscopy image, the image clarity is higher, which can improve the efficiency and accuracy of the doctor's diagnosis and surgical operation, and reduce the doctor's identification difficulty and surgical risks.

[0036] like Figure 2 As shown, the second aspect of this embodiment provides a method for perceiving and automatically enhancing spinal endoscopic image information, comprising the following steps: S1. Obtaining an endoscopic spinal endoscopic video, and performing serial processing on the endoscopic spinal video to obtain a spinal endoscopic image sequence, wherein the spinal endoscopic image sequence includes a plurality of temporally continuous spinal endoscopic image frames; S2. Processing a plurality of temporally consecutive spinal endoscopic image frames using the trained super-resolution image generation model to obtain a high-resolution spinal endoscopic image; S3. Segment and identify physiological tissues in high-resolution spinal endoscopy images using a texture-based threshold segmentation method to obtain physiological tissue images; S4. Fuse the high-resolution spinal endoscopic image and the physiological tissue image to obtain a fused high-resolution spinal endoscopic image, identify the fused high-resolution spinal endoscopic image to obtain a labeling frame that corresponds one-to-one with the physiological tissue image, and combine the labeling frame with the fused high-resolution spinal endoscopic image to obtain a target high-resolution spinal endoscopic image.

[0037] This embodiment provides a method for perceiving and automatically enhancing spinal endoscopic image information, including acquiring a spinal endoscopic video acquired by an endoscope, serializing the spinal endoscopic video to obtain a spinal endoscopic image sequence, wherein the spinal endoscopic image sequence includes a number of time-continuous spinal endoscopic image frames; using a trained super-resolution image generation model to process a number of time-continuous spinal endoscopic image frames to obtain a high-resolution spinal endoscopic image; segmenting and identifying physiological tissue in the high-resolution spinal endoscopic image according to a texture-based threshold segmentation method to obtain a physiological tissue image; fusing the high-resolution spinal endoscopic image and the physiological tissue image to obtain a fused high-resolution spinal endoscopic image, identifying the fused high-resolution spinal endoscopic image to obtain a labeling box that corresponds one-to-one to the physiological tissue image, combining the labeling box with the fused high-resolution spinal endoscopic image to obtain a target high-resolution spinal endoscopic image, thereby improving the clarity of the image, reducing the doctor's recognition difficulty during spinal endoscopic surgery, and improving the efficiency and success rate of the surgery.

[0038] The third aspect of this embodiment provides a spinal endoscopic image perception and automatic enhancement device, comprising a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the spinal endoscopic image information perception and automatic enhancement method as described in the second aspect of the embodiment.

[0039] For example, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out memory (FIFO) and / or first-in-last-out memory (FILO), etc.; the processor may include, but is not limited to, a microprocessor of the STM32F105 series, an ARM (Advanced RISC Machines), an X86 or other architecture processor, or a processor with an integrated NPU (neural-network processing units); the transceiver may include, but is not limited to, a WiFi (Wireless Fidelity) wireless transceiver, a Bluetooth wireless transceiver, a General Packet Radio Service (GPRS) wireless transceiver, a ZigBee protocol (a low-power local area network protocol based on the IEEE802.15.4 standard, ZigBee) wireless transceiver, a 3G transceiver, a 4G transceiver and / or a 5G transceiver, etc.

[0040] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

Claims

1. A spinal endoscopy image information perception and automatic enhancement system, characterized in that: It includes image acquisition module, high-resolution image generation module, segmentation and recognition module, and recognition and annotation module; The image acquisition module is used to acquire the spinal endoscopy video acquired by endoscopy, perform serialization processing on the spinal endoscopy video to obtain a spinal endoscopy image sequence, wherein the spinal endoscopy image sequence includes a plurality of time-continuous spinal endoscopy image frames, and upload the plurality of time-continuous spinal endoscopy image frames to the high-resolution image generation module; The high-resolution image generation module is used to process a number of time-continuous spinal endoscopic image frames using the trained super-resolution image generation model to obtain a high-resolution spinal endoscopic image, and upload the high-resolution spinal endoscopic image to the segmentation and recognition module; The segmentation and recognition module is used to segment and recognize physiological tissues in high-resolution spinal endoscopy images according to a texture-based threshold segmentation method to obtain physiological tissue images, and upload the physiological tissue images to the recognition and annotation module; The recognition and annotation module is used to fuse the high-resolution spinal endoscopic image and the physiological tissue image to obtain a fused high-resolution spinal endoscopic image, identify the fused high-resolution spinal endoscopic image to obtain a annotation frame that corresponds one-to-one with the physiological tissue image, and combine the annotation frame with the fused high-resolution spinal endoscopic image to obtain a target high-resolution spinal endoscopic image.

2. A spinal endoscopy image information perception and automatic enhancement system according to claim 1, characterized in that: The image acquisition module is also used to acquire a high-resolution image, perform bicubic downsampling on the high-resolution image to obtain a low-resolution image corresponding to the high-resolution image, and add Gaussian noise to the low-resolution image corresponding to the high-resolution image to obtain a simulated low-resolution image corresponding to the high-resolution image.

3. A spinal endoscopy image information perception and automatic enhancement system according to claim 2, characterized in that: It also includes a model training module, which is communicatively connected to the image acquisition module. The model training module is used to input the high-resolution image and the simulated low-resolution image corresponding to the high-resolution image into the super-resolution image generation model for training to obtain the trained super-resolution image generation model.

4. A spinal endoscopy image information perception and automatic enhancement system according to claim 1, characterized in that: After obtaining the spinal endoscopic image sequence, it also includes: Performing Gaussian filtering to denoise the spinal endoscopic image sequence to obtain a denoised spinal endoscopic image sequence; Grayscale processing is performed on the denoised spinal endoscopy image sequence to obtain a grayscale image; The grayscale images were standardized and normalized to obtain the preprocessed spinal endoscopy image sequence.

5. The spinal endoscopy image information perception and automatic enhancement system according to claim 1, characterized in that: The trained super-resolution image generation model is used to process several time-sequential spinal endoscopic image frames to obtain high-resolution spinal endoscopic images, including: Performing feature extraction on a number of temporally continuous spinal endoscopy image frames to obtain a multi-frame feature map, and performing temporal position encoding on the multi-frame feature map to obtain the encoded multi-frame feature map, so that the multi-frame feature map corresponds one-to-one with time, using the feature map at the current moment as the current feature map, and the feature map at the previous moment as the alignment reference feature map corresponding to the current feature map; when the current moment is the initial moment, using the feature map at the initial moment as the alignment reference feature map corresponding to the current feature map; Perform frame alignment on the current feature map and the alignment reference feature map to obtain aligned multi-frame feature maps; Perform global feature extraction based on the aligned multi-frame feature maps to obtain a global feature map, perform frame alignment on the global feature map to obtain an aligned global feature map; The aligned multi-frame feature maps and the aligned global feature map are fused to obtain the final feature map, and a high-resolution spinal endoscopic image is reconstructed based on the final feature map.

6. A spinal endoscopy image information perception and automatic enhancement system according to claim 1, characterized in that: The trained super-resolution image generation model is built based on recurrent neural networks and Transformer networks.

7. A spinal endoscopy image information perception and automatic enhancement system according to claim 1, characterized in that: The physiological tissue in the high-resolution spinal endoscopy image is segmented and identified using the texture-based threshold segmentation method to obtain the physiological tissue image, including: Preprocess the high-resolution spinal endoscopy image to obtain the target image; Use the K-means algorithm to perform preliminary segmentation on the target image to obtain the first image; Using a texture-based threshold segmentation method to segment and identify the target image to obtain a second image; The first image and the second image are fused to obtain a physiological tissue image.

8. A spinal endoscopy image information perception and automatic enhancement system according to claim 7, characterized in that: The target image is segmented and recognized using a texture-based threshold segmentation method to obtain a second image, including: Calculate the gray level co-occurrence matrix based on the target image, obtain the gray level co-occurrence matrix of muscle, bone and nerve respectively, and extract the texture features in the gray level co-occurrence matrix of muscle, bone and nerve, wherein the texture features include contrast, energy, entropy, variance, correlation and inverse moment; Calculate the average value of the texture features, normalize the average value of the texture features to obtain the corresponding weight coefficient, set the weights of contrast, energy, entropy, variance, correlation and inverse moment according to the corresponding weight coefficient, extract the characteristic value of each pixel in the target image according to the weights of contrast, energy, entropy, variance, correlation and inverse moment, and obtain the first feature; Extracting muscle tissue in the target image based on the first feature to obtain a muscle tissue image, filling holes in the muscle tissue in the target image to obtain a filled target image; extracting bones and nerves in the filled target image based on the first feature to obtain a bone and nerve image; The muscle tissue image and the bone and nerve images are fused to produce a second image.

9. A method for perceiving and automatically enhancing spinal endoscopy image information, characterized in that: include: Acquiring an endoscopic spinal video, and performing serial processing on the endoscopic spinal video to obtain a spinal endoscopic image sequence, wherein the spinal endoscopic image sequence includes a plurality of temporally continuous spinal endoscopic image frames; Using the trained super-resolution image generation model to process several time-continuous spinal endoscopic image frames, a high-resolution spinal endoscopic image is obtained; The physiological tissue in the high-resolution spinal endoscopy image is segmented and identified using a texture-based threshold segmentation method to obtain a physiological tissue image. The high-resolution spinal endoscopic image and the physiological tissue image are fused to obtain a fused high-resolution spinal endoscopic image, the fused high-resolution spinal endoscopic image is identified to obtain a labeling frame that corresponds one-to-one with the physiological tissue image, and the labeling frame is combined with the fused high-resolution spinal endoscopic image to obtain a target high-resolution spinal endoscopic image.

10. A spinal endoscope image information perception and automatic enhancement device, characterized in that: It includes a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the spinal endoscopic image information perception and automatic enhancement method as described in claim 9.

Images