Intelligent tracheal intubation surgery system and method for laboratory mice based on mechanical arm linkage

By combining robotic arm linkage and artificial intelligence technology, the automated identification and insertion of endotracheal tubes in laboratory mice has been achieved, solving the problems of high operational risk and time and labor consumption in existing technologies, and improving the accuracy and stability of intubation.

WO2026237032A1PCT designated stage Publication Date: 2026-11-19NORTH CHINA UNIVERSITY OF SCIENCE AND TECHNOLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2025/095294
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-12
Filing Date
2025-05-16
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Existing techniques for tracheal intubation in laboratory mice carry a high risk of failure, are prone to complications, and are time-consuming and labor-intensive, lacking automated and intelligent intubation procedures.

Method used

An intelligent endotracheal intubation system based on robotic arm linkage is adopted. Combining artificial intelligence technology, the system uses image acquisition and image processing modules to perform real-time image analysis, identifies the tracheal position through convolutional neural network, and automatically inserts the endotracheal tube by robotic arm, realizing fully automated operation.

Benefits of technology

It significantly reduces the risk rate and time cost of endotracheal intubation, improves the accuracy and stability of the operation, realizes intelligent identification and automated intubation of narrow cavities, and reduces manpower consumption.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2025095294_19112026_PF_FP_ABST
    Figure CN2025095294_19112026_PF_FP_ABST
Patent Text Reader

Abstract

The present invention provides an intelligent tracheal intubation surgery system and method for laboratory mice based on mechanical arm linkage and relates to the technical field of biomedicine. The system comprises a respiratory anesthesia machine, an anesthesia maintenance box, and a mechanical arm. The anesthesia maintenance box is connected to the respiratory anesthesia machine. A mechanical arm movement track is arranged in the upper interior of the anesthesia maintenance box. A retractable laboratory mouse apparatus plate for securing a laboratory mouse is arranged in the anesthesia maintenance box. An operating port of the mechanical arm is provided with a tracheal catheter release apparatus used for, after prying open the oral cavity of the laboratory mouse, clamping a tracheal catheter, moving deep into the laryngeal cavity, and inserting a tracheal cannula into the trachea of the laboratory mouse. An image acquisition module used for acquiring continuous extracorporeal images and continuous intraoral images in real time is arranged alongside the tracheal catheter. An image processing module is arranged inside the mechanical arm and is used for analyzing the continuous intraoral images, identifying and labeling the trachea, and calculating the time difference between opening and closing of a tracheal duct window. The intelligent tracheal intubation surgery system is based on mechanical arm linkage and enables fully automated tracheal intubation surgery within confined anatomical spaces.
Need to check novelty before this filing date? Find Prior Art

Description

A Robotic Arm-Based Intelligent Endotracheal Intubation System and Method for Laboratory Mice Technical Field

[0001] This invention relates to the field of biomedical technology, and in particular to an intelligent tracheal intubation surgical system and method for laboratory mice based on robotic arm linkage. Background Technology

[0002] Currently, endotracheal intubation is a necessary procedure in animal experiments involving mice and nude rats. Endotracheal intubation involves inserting a specially designed, thin-diameter tube deep into the throat and airway of the mouse or nude rat. This ensures unobstructed airways, adequate oxygen supply, and prevents various complications, providing optimal conditions for the experiment. Meanwhile, the rapid development of artificial intelligence (AI) has driven breakthroughs in various projects across multiple fields. Deep learning CNN technology is widely used in medical imaging, and some AI technologies have already been applied in real-world medical settings, demonstrating high diagnostic accuracy. The use of AI brings new possibilities to experimental procedures and the management of animal injuries and prognoses.

[0003] Artificial endotracheal intubation in the laboratory carries a high risk of failure. Errors or failures in endotracheal intubation can lead to complications, including mucosal bleeding, laryngospasm, bronchospasm, pulmonary aspiration, arrhythmia, laryngeal edema, respiratory inflammation, cardiac arrest, and death, with catastrophic consequences for large / small / nude mice. The success of endotracheal intubation is influenced by various factors, including the choice of instruments, the intubator's experience, and the individual animal's condition. Currently, most endotracheal intubation in large / small / nude mice is performed manually by the experimenter. Errors or lack of experience can lead to tracheal mucosal damage or other complications. Furthermore, the procedure is time-consuming and labor-intensive, particularly unfriendly to inexperienced operators, and can easily cause airway injury or postoperative local infection and complications, affecting data stability and hindering continuous research.

[0004] Human endotracheal intubation technology has been explored by many researchers using artificial intelligence (AI) technology to make the operation simpler and more effective. In the literature Patrini I, Ruperti M, Moccia S, Mattos LS, Frontoni E, De Momi E. Transfer learning for informative-frame selection in laryngoscopic videos through learned features. Med Biol Eng Comput. 2020 Jun; 58(6):1225-1238. doi:10.1007 / s11517-020-02127-7. Epub 2020 Mar 24. PMID:32212052, Patrini et al. developed a method for classifying laryngoscopic video frames based on deep learning. The model can automatically select information-rich laryngoscopic video frames, helping doctors reduce the amount of diagnostic data to be processed. In the literature Wang YY, Hamad AS, Lever TE, Bunyak F. Orthogonal Region Selection Network for Laryngeal Closure Detection in Laryngoscopy Videos. Annu Int Conf IEEE Eng Med Biol Soc. 2020 Jul; 2020:2167-2172. doi:10.1109 / EMBC44109.2020.9176149.PMID:33018436, Wang et al. used convolutional neural networks to detect laryngeal adductor reflex events in laryngoscopy videos. In this context, the application of artificial intelligence (AI) vision-assisted endotracheal intubation technology in large / small / nude mice becomes an effective solution. Currently, there is still a lack of fully automated implementation plans for endotracheal intubation surgery based on the recognition of narrow cavities in laboratory mice. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent tracheal intubation surgical system and method for laboratory mice based on robotic arm linkage. By combining robotic arm linkage with artificial intelligence technology, the system can identify the narrow cavities of laboratory mice and automate the tracheal intubation surgery.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A robotic arm-based intelligent endotracheal intubation surgical system for laboratory mice includes: a respiratory anesthesia machine, an anesthesia maintenance box, and a robotic arm. The anesthesia maintenance box is connected to the respiratory anesthesia machine via a rigid anesthetic gas pipeline. A robotic arm moving track is provided above the anesthesia maintenance box. The anesthesia maintenance box contains a retractable laboratory mouse carrier for placing the laboratory mouse and a laboratory mouse fixation device for fixing the limbs of the laboratory mouse.

[0008] The robotic arm's operating port is equipped with a tracheal tube release device, which is used to pry open the mouse's mouth and then clamp the tracheal tube to move it deeper into the mouse's larynx.

[0009] An image acquisition module is installed next to the endotracheal tube, and an image processing module is installed inside the robotic arm; the image acquisition module is used to acquire continuous images of the experimental mouse in vitro and in the oral cavity in real time, and transmit them to the image processing module.

[0010] The image processing module is used to automatically analyze continuous images inside the oral cavity based on the image data, identify and label the trachea of ​​the experimental mouse, and then automatically calculate the time difference between the opening and closing of the tracheal tube window of the experimental mouse; the robotic arm adjusts the position of the tracheal tube according to the position of the experimental mouse's trachea, and automatically stops moving after determining the optimal tracheal intubation position, and inserts the tracheal tube into the experimental mouse's trachea using the tracheal tube release device.

[0011] Furthermore, the image processing module includes a preprocessing unit, a convolutional neural network unit, and a Python algorithm unit; the Python algorithm unit is used to calculate the time difference between the opening and closing of the tracheal tube window of the experimental mouse; the convolutional neural network unit includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.

[0012] The preprocessing unit preprocesses the continuous images within the oral cavity to obtain an image pixel matrix; the input layer receives three consecutive frames of continuous images and extracts five channels of information for each frame: grayscale, horizontal coordinate gradient, vertical coordinate gradient, x-ray flow, and y-ray flow; the convolutional layer identifies spatial patterns in the image pixel matrix by performing convolution operations on the 3D convolutional kernels of the five channels to obtain features; the pooling layer reduces the size of the matrix to reduce the number of parameters in the entire neural network; the fully connected layer classifies the features, obtains the classification result, and transmits it to the output layer; the output layer outputs the classification result from S4.4 to achieve the identification and labeling of the trachea.

[0013] Furthermore, the robotic arm is a multi-axis robotic arm; the robotic arm's moving track includes a horizontal track and a vertical track; the image acquisition module is a lens, and the lens is equipped with an active lighting device; the endotracheal tube is an airbag endotracheal tube.

[0014] Furthermore, the endotracheal tube release device is a scissor-type endotracheal tube release device, with a small spring placed at the center of the scissor-type endotracheal tube release device. The endotracheal tube is held by the scissor-type endotracheal tube release device so that it is located at the center of the endotracheal tube release device and connected to the spring. The endotracheal tube connected to the scissor-type endotracheal tube release device tightly wraps the air bag endotracheal tube. The endotracheal tube is a symmetrical rigid half-sleeve.

[0015] This invention also provides a method for intelligent endotracheal intubation surgery of laboratory mice based on robotic arm linkage, applied to the aforementioned intelligent endotracheal intubation surgery system for laboratory mice based on robotic arm linkage, comprising the following steps:

[0016] S1. Turn on the respiratory anesthesia machine. The respiratory anesthesia machine continuously circulates isoflurane gas into the anesthesia maintenance chamber to maintain the anesthesia state of the experimental mice until the muscle reflexes of the experimental mice disappear.

[0017] S2, the limbs of the anesthetized experimental mouse are fixed to the retractable experimental mouse carrier plate using the experimental mouse fixation device;

[0018] S3, Start intubation: The image acquisition module acquires continuous images of the experimental mouse in vitro. Based on the continuous images, it is determined whether the endotracheal tube release device has reached the position of the experimental mouse's mouth. The endotracheal tube release device pries open the experimental mouse's mouth to fully expose the larynx. Then, the endotracheal tube is clamped and moved into the experimental mouse's laryngeal cavity.

[0019] S4. After insertion into the oral cavity, the image acquisition module continues to acquire continuous images inside the oral cavity of the experimental mouse. Then, the continuous images inside the oral cavity are preprocessed using the image processing module, and the trachea of ​​the experimental mouse is identified and labeled based on the convolutional neural network. The time difference between the opening and closing of the tracheal tube window of the experimental mouse is calculated.

[0020] S5, the robotic arm adjusts the position of the endotracheal tube according to the position of the mouse's trachea, and automatically stops moving after determining the optimal endotracheal tube position, and uses the endotracheal tube release device to insert the endotracheal tube into the mouse's trachea.

[0021] S6. The robotic arm removes all instruments except the endotracheal tube from the mouse's mouth. The mouse is then manually removed from the anesthesia maintenance box. The respiratory anesthesia machine then automatically stops releasing isoflurane gas, completing the entire endotracheal intubation process for the mouse.

[0022] Furthermore, in step S4, the identification and labeling of the experimental mouse trachea based on a convolutional neural network specifically includes the following steps:

[0023] S4.1 The continuous images inside the oral cavity are preprocessed by the preprocessing unit in the image processing module to obtain an image pixel matrix;

[0024] S4.2 The input layer receives three consecutive frames of continuous images and extracts five channels of information for each frame: grayscale, horizontal coordinate gradient, vertical coordinate gradient, x-axis optical flow, and y-axis optical flow.

[0025] S4.3. The convolutional layer identifies the spatial patterns in the image pixel matrix and performs convolution operations on the 3D convolutional kernels of the five channels to obtain features;

[0026] S4.4. Reduce the size of the image pixel matrix using pooling layers to reduce the number of parameters in the entire neural network; repeat steps S4.3 and S4.4.

[0027] S4.5 The fully connected layer classifies the features, obtains the classification result, and transmits it to the output layer;

[0028] S4.6 Finally, the output layer outputs the classification results from S4.4 to achieve the identification and labeling of the trachea.

[0029] Furthermore, in step S4, calculating the time difference between the opening and closing of the tracheal tube window in the experimental mouse specifically includes:

[0030] After image recognition, the preprocessing unit extracts images of airway window opening and closing and passes them to the Python algorithm unit in the image processing module. The Python algorithm unit automatically calculates the time difference between the opening and closing of the experimental mouse's trachea based on the time difference between the transmitted images.

[0031] Furthermore, the movement principle of the robotic arm is as follows:

[0032] The recorded images are converted to grayscale, the center position of the robotic arm's movement is calculated, and a three-dimensional coordinate axis is established based on the center position. This three-dimensional coordinate axis is used for trajectory planning and motion planning. The coordinates [x,y,z] represent the spatial position points within the robotic arm's movable range, ensuring that the robotic arm's operating ports can reach every spatial position point within the anesthesia maintenance box. A visual vision device is used to ensure the robotic arm's accurate positioning and visual tracking of the experimental mouse.

[0033] According to specific embodiments provided by the present invention, the following technical effects are disclosed: The intelligent endotracheal intubation surgical system and method for laboratory mice based on robotic arm linkage provided by the present invention dynamically identifies and labels the location of the tracheal tube in the larynx of the laboratory mouse and the opening time of the tracheal tube through deep learning CNN (Convolutional Neural Network), while simultaneously linking with a robotic arm to realize the dynamic insertion of the endotracheal tube, achieving fully automated endotracheal intubation. The present invention has the following advantages:

[0034] 1. High accuracy: Artificial intelligence (AI)-assisted endotracheal intubation technology uses deep learning algorithms to visually analyze the location, structure, width, and opening time of the endotracheal tube opening window, which is conducive to the correct insertion of the endotracheal tube.

[0035] 2. Saves time and manpower costs: The fully automated endotracheal intubation process using artificial intelligence (AI) technology can be completed in just 5-7 seconds. It can achieve accurate endotracheal intubation results for a large number of large / small / nude mice by a single person controlling the equipment, which greatly reduces the complexity of endotracheal intubation technology and saves a lot of time and manpower costs.

[0036] 3. Reduced risk rate: Experimental results show that the risk rate of endotracheal intubation experiments performed using this method is only 0.5%-1% compared to manual endotracheal intubation experiments, due to the different methods used. The airways of large / small / nude mice are very narrow. Using deep learning algorithms, combined with the application of vision domain, helps to more accurately locate the intubation channel, resulting in less damage and avoiding complications caused by various errors.

[0037] 4. High stability: This method establishes a fully automated model. The manual endotracheal intubation method has obvious structural errors due to different personnel skill levels, while this method uses artificial intelligence technology to achieve high stability through full automation.

[0038] 5. Intelligent Recognition of Narrow Cavities: The image recognition method of this invention differs from conventional visual domain recognition methods. The tracheas of large / small / nude mice are extremely thin and small, resulting in a narrow field of view. It is not a large-scale visual domain observation where the object can be directly observed from a distance and then gradually magnified upon approach. Visual observation through narrow tubes causes the target to appear rapidly, without a visual buffering process. This technology achieves intelligent recognition of narrow cavities through deep learning algorithms, enabling rapid identification and classification of rapidly appearing objects.

[0039] 6. Automation Technology: This invention utilizes chip-based automated control to move the robotic arm, enabling automated surgery in narrow cavities without the need for human intervention throughout the entire process, thus saving labor costs. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 is a structural diagram of the intelligent endotracheal intubation surgical system for laboratory mice based on robotic arm linkage of the present invention;

[0042] Figure 2 is a flowchart of the intelligent tracheal intubation surgery method for laboratory mice based on robotic arm linkage according to an embodiment of the present invention;

[0043] Figure 3 is a schematic diagram of the intelligent tracheal intubation surgery method for experimental mice based on robotic arm linkage according to an embodiment of the present invention;

[0044] Figure 4 is a schematic diagram of the larynx of a mouse, where A is the oral cavity wall, B is the trachea, C is the epiglottis, D is the esophagus, and E is the endotracheal tube.

[0045] Explanation of reference numerals in the attached diagrams: 1. Anesthesia ventilator; 2. Rigid connecting tubing for anesthetic gas; 3. Rigid connecting tubing for oxygen; 4. Retractable mouse carrier plate; 5. Mouse restraint device; 6. Flexible connecting tubing for oxygen; 7. Lateral track; 8. Longitudinal track; 9. Robotic arm; 10. Robotic arm operating port; 11. Scissor-type endotracheal tube release device; 12. Endotracheal tube; 13. Image acquisition module (lens); 14. Guide wire; 15. Balloon-supported endotracheal tube; 16. Balloon inflation connection tube; 17. Interface between the flexible tubing and various parts of the endotracheal tube; 18. Anesthesia maintenance box. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] The purpose of this invention is to provide an intelligent tracheal intubation surgical system and method for laboratory mice based on robotic arm linkage. By combining robotic arm linkage with artificial intelligence technology, the system can identify the narrow cavities of laboratory mice and automate the tracheal intubation surgery.

[0048] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] Example 1

[0050] As shown in Figure 1, this embodiment of the invention provides an intelligent tracheal intubation surgical system for laboratory mice based on robotic arm linkage, including: an anesthesia ventilator 1, an anesthesia maintenance box 18, and a robotic arm 9;

[0051] The anesthesia maintenance box 18 is used to maintain a certain concentration of anesthetic gas to keep the small animal under anesthesia.

[0052] The anesthesia maintenance box 18 is connected to the respiratory anesthesia machine 1 through a rigid anesthetic gas connecting pipe 2. A robotic arm moving track is provided above the anesthesia maintenance box 18. The robotic arm moving track includes a horizontal track 7 and a vertical track 8. The anesthesia maintenance box 18 is equipped with a pull-out experimental mouse carrier plate 4 for placing experimental mice and an experimental mouse fixation device 5 for fixing the limbs of experimental mice.

[0053] The operating port 10 of the robotic arm 9 is equipped with a tracheal tube release device 11, which is used to pry open the oral cavity of the experimental mouse and then clamp the tracheal tube 12 to move it deeper into the throat cavity of the experimental mouse.

[0054] An image acquisition module 13 is provided next to the tracheal tube 12, and an image processing module is provided inside the robotic arm 9. The image acquisition module is used to acquire continuous images of the experimental mouse in vitro and in the oral cavity in real time, and transmit them to the image processing module. The image processing module is used to automatically analyze the continuous images in the oral cavity based on the image data, identify and label the experimental mouse's trachea, and then automatically calculate the time difference between the opening and closing of the experimental mouse's tracheal tube window. The robotic arm adjusts the position of the tracheal tube 12 according to the position of the experimental mouse's trachea, automatically pauses at a suitable angle, and inserts the tracheal tube 15 into the mouse's trachea using the tracheal tube release device.

[0055] In this embodiment, the robotic arm 9 is connected to the robotic arm movement track via gears, and the controlled port of the robotic arm 9 is connected to the airbag endotracheal cannula, endoscope, and endotracheal cannula release device.

[0056] In this embodiment, the image processing module is a chip, which is built into the robotic arm 9.

[0057] In this embodiment, the image processing module includes a preprocessing unit, a convolutional neural network unit, and a Python algorithm unit; the Python algorithm unit is used to calculate the time difference between the opening and closing of the tracheal tube window of the experimental mouse; the convolutional neural network unit includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.

[0058] The preprocessing unit preprocesses the continuous images within the oral cavity to obtain an image pixel matrix; the input layer receives three consecutive frames of continuous images and extracts five channels of information for each frame: grayscale, horizontal coordinate gradient, vertical coordinate gradient, x-ray flow, and y-ray flow; the convolutional layer identifies spatial patterns in the image pixel matrix to obtain features; the pooling layer reduces the size of the matrix to decrease the number of parameters in the entire neural network; steps S4.3 and S4.4 are repeated; the fully connected layer classifies the features, obtains the classification result, and transmits it to the output layer; the output layer outputs the classification result from S4.4 to achieve tracheal identification and labeling.

[0059] In this embodiment, the robotic arm is a multi-axis robotic arm; the image acquisition module is an endoscope, which is equipped with an active illumination device; the endoscope has a small lens diameter, produces clear images, and the lens does not heat up, and can output images to the robotic arm's artificial intelligence (AI) chipset.

[0060] In this embodiment, the endotracheal tube release device 11 is a scissor-type endotracheal tube release device. A small spring is placed at the center of the scissor-type endotracheal tube release device. The endotracheal tube 12 is held in place by the scissor-type endotracheal tube release device, positioned at the center of the endotracheal tube release device and connected to the spring. The endotracheal tube 12 connected to the scissor-type endotracheal tube release device tightly wraps around the airbag endotracheal tube. The endotracheal tube 12 is a symmetrical rigid half-sleeve. In a further embodiment, the scissor-type endotracheal tube release device can be a scissor-type metal clip.

[0061] In this embodiment, the endotracheal tube 15 is a cuffed endotracheal tube cannula, that is, the front end of the endotracheal tube has a cuff connected to a pressure device, which allows for the inflation and deflation of the cuff. Specifically, the cuff is connected to a rigid oxygen conduit via a balloon inflation connector, and the rigid oxygen conduit is connected to an anesthesia and respiration unit, thereby enabling the inflation and deflation of the cuff.

[0062] In this embodiment, the system also includes an oxygen connection hose 6, which can connect the anesthesia ventilator 1 and the endotracheal tube 15 through the interface 17 between the hose and the various parts of the endotracheal tube, so that the endotracheal tube is filled with anesthetic gas, and after the endotracheal tube 15 is inserted into the mouth of the experimental mouse, it leaves the anesthesia maintenance box 18 for subsequent surgical operations.

[0063] In a further embodiment, the airbag endotracheal cannula contains a guide wire 14 to facilitate insertion of the endotracheal tube into the tracheal window of the mouse.

[0064] The physical properties of the system are as follows:

[0065] Animal range suitable for: all large / small / naked mice;

[0066] Suitable for narrow cavities: all large / small / nude mouse respiratory tracts;

[0067] Anesthesia maintenance box dimensions: 40*40*80cm

[0068] Endoscope resolution: 720P, 24FPS, lens diameter: 2.8mm;

[0069] Robotic arm: multi-axis linkage, arm span 350mm, effective gripping range: radius ≤30cm, area with the central axis as a semicircle, repeatability ±0.5mm;

[0070] Artificial Intelligence (AI): The microprocessor uses a Broadcom BCM2711 64-bit 1.5GHz quad-core + VideoCore VI@500MHz, with an AI computing power of 200GFLOPS. The operating system is Ubuntu mate 20.04LTS + ROS Noetic, and the programming language is Python.

[0071] Example 2

[0072] As shown in Figures 2-3, this embodiment of the invention provides a method for intelligent endotracheal intubation surgery of laboratory mice based on robotic arm linkage, applied to the intelligent endotracheal intubation surgery system for laboratory mice based on robotic arm linkage described in Embodiment 1, and includes the following steps:

[0073] S1. Turn on the respiratory anesthesia machine. The respiratory anesthesia machine continuously circulates isoflurane gas into the anesthesia maintenance chamber to maintain the anesthesia state of the experimental mice until the muscle reflexes of the experimental mice disappear.

[0074] S2. Manually remove the mouse fixation plate from the anesthesia maintenance box. The limbs of the anesthetized mouse are fixed to the retractable mouse carrier plate through the mouse fixation device. Then return the retractable mouse carrier plate to its original position.

[0075] S3. Start intubation: The image acquisition module acquires continuous images of the experimental mice in vitro. Based on the continuous images, it is determined whether the endotracheal tube release device has reached the position of the experimental mouse's mouth. When it reaches the position of the mouth, the endotracheal tube release device pries open the experimental mouse's mouth to fully expose the larynx. Then, the endotracheal tube is clamped and moved into the laryngeal cavity of the large / small / nude mouse. This process is controlled by the robotic arm automatically adjusting the angle and length between each axis to control the insertion of the endotracheal tube into the experimental mouse's mouth.

[0076] S4. After insertion into the oral cavity, the image acquisition module continues to acquire continuous images inside the oral cavity of the experimental mouse. Then, the continuous images inside the oral cavity are preprocessed using the image processing module, and the position of the experimental mouse's trachea is identified and labeled based on the convolutional neural network, and the time difference between the opening and closing of the experimental mouse's tracheal tube window is calculated.

[0077] S5, the robotic arm adjusts the position of the endotracheal tube according to the position of the mouse's trachea, and automatically stops moving after determining the optimal endotracheal tube position, and uses the endotracheal tube release device to insert the endotracheal tube into the mouse's trachea.

[0078] S6. The robotic arm removes all instruments except the endotracheal tube from the mouse's mouth. The mouse is then manually removed from the anesthesia maintenance box. The respiratory anesthesia machine then automatically stops releasing isoflurane gas, completing the entire endotracheal intubation process for the mouse.

[0079] In this embodiment, step S4, which involves identifying and labeling the trachea of ​​experimental mice based on a convolutional neural network, specifically includes the following steps:

[0080] S4.1 The continuous images inside the oral cavity are preprocessed by the preprocessing unit in the image processing module to obtain an image pixel matrix;

[0081] S4.2 The input layer receives three consecutive frames of continuous images and extracts five channels of information for each frame: grayscale, horizontal coordinate gradient, vertical coordinate gradient, x-axis optical flow, and y-axis optical flow.

[0082] S4.3 The convolutional layer identifies the spatial patterns in the image pixel matrix by performing convolution operations on five channels of 3*3*3 3D convolution kernels and extracting information from the image pixel matrix to obtain features; the information in the image pixel matrix includes local features of the image, such as edges, textures, corners, etc.

[0083] S4.4. The pooling layer reduces the size of the image pixel matrix to reduce the number of parameters in the entire neural network; repeat steps S4.3 and S4.4 to increase the number of images;

[0084] S4.5 The fully connected layer classifies the features, obtains the classification result, and transmits it to the output layer;

[0085] S4.6 Finally, the output layer outputs the classification results from S4.4 to achieve the identification and labeling of the trachea;

[0086] The above process uses the 3D-CNN algorithm to capture the three-dimensional features of dynamic images. Because 3D-CNN can fully utilize the convolutional kernels during feature extraction, it greatly reduces the computational and storage requirements of model construction, and lowers the configuration requirements of the construction equipment. In this embodiment, the ResNet-50 residual structure can also be used to address the problems of decreased classification performance, slower network convergence speed, and decreased accuracy caused by increasing the depth of conventional CNN models; the softmax function is applied to construct the model to achieve the task of correct classification and recognition of continuous images.

[0087] In this embodiment, step S4, calculating the time difference between the opening and closing of the tracheal tube window in the experimental mouse, specifically includes:

[0088] After image recognition by the preprocessing unit, the images of airway window opening and closing are extracted and passed to the Python algorithm unit in the image processing module. The Python algorithm unit automatically calculates the time difference of tracheal opening and closing of the experimental mice based on the time difference between the transmitted images, and the experiment is repeated multiple times.

[0089] In this embodiment, the motion principle of the robotic arm is as follows:

[0090] The recorded images were converted to grayscale to calculate the center position of the robotic arm's movement. A three-dimensional coordinate system was established based on this center position for trajectory and motion planning. The coordinates [x, y, z] represent the spatial positions within the robotic arm's movable range, ensuring that the robotic arm's operating ports can reach every spatial position within the anesthesia maintenance chamber. A visual vision device was used to ensure the robotic arm's precise positioning and visual tracking of the experimental mouse. After initial recognition of the model, the model was used to automatically insert an endotracheal tube directly into the experimental mouse.

[0091] This embodiment was selected according to the following criteria to verify the feasibility of the present invention:

[0092] 180 rats, 180 mice, and 180 nude mice were selected (the animals used in the experiment complied with ethical standards, registration number: SQ2024143). Each group of rats was ensured to have different weights, ages, and physical characteristics. Each group was further divided into high-weight, medium-weight, and low-weight groups, with 60 rats in each group (each rat was trained 10 times). The training group consisted of 45 rats, and the validation group consisted of 15 rats.

[0093] Tracheal tubes were manually inserted into some large / small / nude mice in the training group, and dynamic images in the visual lens were recorded. Data processing was performed, and the tissues in the real-time dynamic images were automatically segmented using the conventional Otsu threshold [1]. The tissue regions of each frame image were cut into non-overlapping 250-micrometer square blocks, magnified 25 times, and the size was adjusted to 224×224 pixels. This size was determined based on experience. Experienced clinicians manually excluded non-airway images and manually labeled the obtained images, including secretions and anatomical structures of large and small mice, so that the error rate of the model in automatically identifying the trachea was close to 0. The labeled images were divided into different datasets and saved in PNG format. The data were divided into training set, validation set and test set.

[0094] Errors or mistakes in the endotracheal intubation process were analyzed and modified by the human team until all large / small / nude mice in the training group were trained. Then, 20 large / small / nude mice were randomly selected from the training group for a second endotracheal intubation experiment to analyze the accuracy of the method. Finally, another 15 large / small / nude mice were used for external validation to demonstrate the accuracy and timeliness of the method.

[0095] Example 3

[0096] This invention is a specific example of the intelligent tracheal intubation surgery method for laboratory mice based on robotic arm linkage described in Example 2.

[0097] In this embodiment, mice were randomly selected (the mice selected for the experiment were ethical, and the ethics number of North China University of Technology was SQ2024143), and the experiment was conducted at a room temperature of 24°C, a relative humidity of 55%, with good air circulation and natural light.

[0098] The operator places the anesthetized mouse onto a self-made retractable rat device board and fixes the mouse's limbs.

[0099] Turn on the RWD (Reward) customized respiratory anesthesia machine. The machine continuously injects isoflurane gas into the RWD (Reward) customized anesthesia maintenance chamber to ensure the mice are under anesthesia. Once the mice's muscle reflexes disappear, the experiment begins.

[0100] Using the method described in this invention, upon powering on, the device performs automated operation, smoothly inserting the endotracheal tube into the mouse's laryngeal cavity. (The robotic arm first acquires image data through a 2.8mm custom lens. The built-in STM32 chip uses artificial intelligence (AI) algorithms to identify and calculate the position of the mouse's mouth. The custom DOFBOT robotic arm (motherboard: Raspberry Pi 4B, computing power: 0.2TFLOPS, CPU: Quad-Core Arm Cortex-A72 1.5GHz, GPU: Broadcom VideoCore IV, OLED device) holds and moves the custom endotracheal tube release device to pry open the mouse's mouth, fully exposing the larynx. The robotic arm automatically adjusts the angles and lengths between its axes to adapt to the correct insertion of the endotracheal tube into the rat. After insertion into the mouse's mouth, the lens collects image data from inside the mouse's mouth and uses artificial intelligence (AI) for automatic analysis and preprocessing of continuous dynamic images. CNN (Convolutional Neural Network) is used to perform this preprocessing.) This system (network) collects intraoral images of mice and gathers features for deep learning. The process involves: inputting the image pixel matrix into the neural network via the input layer; identifying spatial patterns in the image using convolutional layers to obtain features; reducing the size of the matrix using pooling layers to decrease the number of parameters in the entire neural network; classifying the features using fully connected layers; and outputting the classification results from the output layer. This achieves the task of tracheal tube identification and annotation. Simultaneously, based on the collected image data, the system automatically calculates the time difference between the opening and closing of the mouse's tracheal tube window. The robotic arm adjusts the position of the tracheal tube according to the tracheal position, automatically pausing at the appropriate angle. A microscopic tracheal release device is used to smoothly insert the tracheal tube into the mouse's trachea, ensuring that the frequency of abdominal undulations matches the frequency of gas inhalation from the ventilator, thus correctly completing the tracheal intubation operation.

[0101] The robotic arm removes all instruments except the endotracheal tube from the animal's mouth, and the mouse is then manually removed from the empty box. The respiratory anesthesia machine automatically stops releasing isoflurane gas, completing the entire endotracheal intubation process.

[0102] The experimental results are as follows:

[0103] The endotracheal intubation was completed in just 5 seconds. After the experiment, the mice showed no abnormalities, and their chest cavity rose and fell rhythmically with the inflow of air from the ventilator. No teeth loosened or fell out, no gastric distension, mucosal bleeding, vocal cord paralysis, laryngospasm, bronchospasm, pulmonary aspiration, laryngeal edema, respiratory inflammation, arrhythmia, cardiac arrest, or death occurred. An experimenter with 5 years of experience in animal endotracheal intubation confirmed that the endotracheal intubation in the mice was successful and without problems or complications.

[0104] Example 4

[0105] This invention is another example of the intelligent tracheal intubation surgery method for laboratory mice based on robotic arm linkage described in Example 2.

[0106] Rats were randomly selected (the animals used in the experiment were ethically selected, registration number: SQ2024143), and the experiment was conducted at a room temperature of 24℃, relative humidity of 55%, with good air circulation and natural light.

[0107] The operator placed the anesthetized rat onto a self-made retractable rat and mouse device board and fixed the rat's limbs.

[0108] Turn on the RWD (Reward) integrated respiratory anesthesia machine. The respiratory anesthesia machine continuously injects isoflurane gas into the RWD (Reward) customized anesthesia maintenance chamber to ensure the rats are under anesthesia. After the rats' muscle reflexes disappear, the experiment begins.

[0109] Using this method, upon powering on, the device operates automatically, smoothly inserting the endotracheal tube into the rat's larynx. (The robotic arm first acquires image data through a 2.8mm custom lens. The built-in STM32 chip uses artificial intelligence (AI) algorithms to identify and calculate the position of the rat's mouth. The robotic arm uses a movable clamp to pry open the rat's mouth, fully exposing the larynx. The DOFBOT robotic arm (motherboard: Raspberry Pi 4B, computing power: 0.2 TFLOPS, CPU: Quad-Core Arm Cortex-A72 1.5GHz, GPU: Broadcom VideoCore IV, OLED device) automatically adjusts the angles and lengths of each axis to ensure correct insertion of the endotracheal tube. After insertion into the rat's mouth, the lens collects intraoral image data and uses AI for automatic analysis and preprocessing of continuous dynamic images. CNN (Convolutional Neural Network) is used.) This system (network) collects intraoral images of rats and gathers features for deep learning. The neural network inputs the pixel matrix of the collected images into an input layer, identifies spatial patterns in the images using convolutional layers to obtain features, reduces the size of the matrix using pooling layers to decrease the number of parameters in the entire neural network, performs feature classification using fully connected layers, and outputs the classification results, thus achieving the task of tracheal tube identification and annotation. Simultaneously, based on the collected image data, the system automatically calculates the time difference between the opening and closing of the rat's tracheal tube window. A robotic arm adjusts the position of the tracheal tube according to the tracheal location, automatically pausing at the appropriate angle. A microscopic tracheal release device is used to smoothly insert the tracheal tube into the rat's trachea, ensuring that the frequency of abdominal undulations matches the frequency of gas inhalation from the ventilator, thus correctly completing the tracheal intubation operation.

[0110] The robotic arm removes all instruments except the endotracheal tube from the animal's mouth, and the rat is then manually removed from the empty box. The respiratory anesthesia machine automatically stops releasing isoflurane gas, completing the entire endotracheal intubation process.

[0111] The experimental results are as follows:

[0112] The endotracheal intubation was completed in just 4.5 seconds. After the experiment, the rats showed no abnormalities; their chest cavity rose and fell rhythmically with the inflow of air from the ventilator. No teeth loosened or fell out, no gastric distension, and no mucosal bleeding, vocal cord paralysis, laryngospasm, bronchospasm, pulmonary aspiration, laryngeal edema, respiratory inflammation, arrhythmia, cardiac arrest, or death occurred. An experimentalist with 5 years of experience in animal endotracheal intubation confirmed that the rat endotracheal intubation was successful and without problems or complications.

[0113] Example 5

[0114] This invention is another example of the intelligent tracheal intubation surgery method for laboratory mice based on robotic arm linkage described in Example 2.

[0115] Nude mice were randomly selected (the animals used in the experiment were ethically selected, registration number: SQ2024143), and the experiment was conducted at a room temperature of 28°C, with a maximum daily temperature difference of no more than 4°C, relative humidity of 50%, in a well-ventilated, sterile environment under natural light.

[0116] 1. The operator places the anesthetized nude mouse onto a self-made retractable rat and mouse device board and fixes the nude mouse's limbs.

[0117] 2. Turn on the RWD (Respiratory Anesthesia) integrated respiratory anesthesia machine. The respiratory anesthesia machine continuously injects isoflurane gas into the RWD (Respiratory Anesthesia) maintenance chamber to ensure the nude mice are anesthetized. The experiment begins after the nude mice lose their muscle reflexes.

[0118] 3. Using this method, upon powering on, the device operates automatically, allowing the airbag endotracheal tube to be smoothly inserted into the larynx of a nude mouse. (The robotic arm first acquires image data through a 2.8mm custom lens. The built-in STM32 chip uses artificial intelligence (AI) algorithms to identify and calculate the position of the mouse's mouth. The robotic arm uses a movable clamp to pry open the mouse's mouth, fully exposing the larynx. The DOFBOT robotic arm (motherboard: Raspberry Pi 4B, computing power: 0.2 TFLOPS, CPU: Quad-Core Arm Cortex-A72 1.5GHz, GPU: Broadcom VideoCore IV, OLED device) automatically adjusts the angles and lengths of each axis to ensure correct insertion of the endotracheal tube into the nude mouse. After insertion into the rat's mouth, the lens collects image data from inside the mouse's mouth and uses artificial intelligence (AI) for automatic analysis and preprocessing of continuous dynamic images. CNN (Convolutional Neural Network) is used.) This system (network) collects intraoral images of nude mice and gathers features for deep learning. The process involves: inputting the image pixel matrix into the neural network via the input layer; identifying spatial patterns in the image using convolutional layers to obtain features; reducing the size of the matrix using pooling layers to decrease the number of parameters in the entire neural network; classifying the features using fully connected layers; and outputting the classification results from the output layer. This achieves the task of tracheal tube identification and annotation. Simultaneously, based on the collected image data, the system automatically calculates the time difference between the opening and closing of the rat's tracheal tube window. A robotic arm adjusts the position of the tracheal tube according to the tracheal location, automatically pausing at the appropriate angle. A microscopic tracheal release device is used to smoothly insert the tracheal tube into the nude mouse's trachea, ensuring that the frequency of abdominal undulations matches the frequency of gas inhalation from the ventilator, thus correctly completing the tracheal intubation operation.

[0119] 4. The robotic arm moves to remove all instruments except the endotracheal tube from the animal's mouth, and the nude mouse is then manually removed from the empty box. The respiratory anesthesia machine automatically stops releasing isoflurane gas, completing the entire endotracheal intubation process.

[0120] The experimental results are as follows:

[0121] The endotracheal intubation was completed in just 4.5 seconds. After the experiment, the nude mice showed no abnormalities; their chest cavity rose and fell rhythmically with the inflow of air from the ventilator. No teeth loosened or fell out, no stomach distension, and no mucosal bleeding, vocal cord paralysis, laryngospasm, bronchospasm, pulmonary aspiration, laryngeal edema, respiratory inflammation, arrhythmia, cardiac arrest, or death occurred. An experimenter with 5 years of experience in animal endotracheal intubation confirmed that the endotracheal intubation in the nude mice was successful and without problems or complications.

[0122] Comparative Example

[0123] Three people with equal skill levels performed manual tracheal intubation on rats. The manual operation and the experiment using this method were carried out simultaneously. Each operator performed tracheal intubation on 50 large / small / nude rats and recorded the time taken and the prognosis.

[0124] Manual operation is fully in line with the requirements of the endotracheal intubation procedure.

[0125] One person used the blind intubation method, one person used the transcervical radiographic intubation method, and one person used the tracheostomy intubation method.

[0126] After weighing the anesthetized experimental mice, they were anesthetized with isoflurane according to their weight.

[0127] Comparative Example 1 – Blind Intubation: The operator fixes the anesthetized mouse in a supine position on a mouse fixation board, pulls the mouse's upper incisors backward to tilt its head back and fix it in place, and uses toothless forceps to grasp the mouse's tongue and pull it upward to open the mouse's larynx as much as possible to expose the airway opening, then performs endotracheal intubation. The tip of the tube is placed against the anterior diaphragm and inserted at a slightly inclined angle. When there is no obvious resistance or abnormal sensation during endotracheal intubation, the guidewire 14 in the tube is removed. The entire procedure needs to be performed carefully and slowly to ensure the accuracy and safety of the intubation operation.

[0128] Comparative Example 2 – Transcervical Transillumination Intubation: The operator fixes the anesthetized mouse in a supine position on a mouse fixation board. A high-intensity cold light source is placed 5 cm from the mouse's neck to irradiate the rat's larynx. Using forceps, the operator gently pulls the mouse's tongue upward to the side of the lower teeth and uses a tongue depressor to press the tongue downward, so that the mouse's mouth is opened to the maximum extent to expose the airway window. The light source illuminates the rat's larynx through the mouse's mouth, and the opening and closing of the mouse's airway can be clearly seen. The operator inserts a tracheal tube into the mouse's larynx at the appropriate time.

[0129] Comparative Example 3 – This comparative example uses the tracheotomy intubation method. The specific operation is as follows: The operator shaves and disinfects the anesthetized experimental mouse, then fixes the experimental mouse and cuts the skin, muscle layer and fascia. The thyroid gland and surrounding tissues are carefully dissected, and the trachea with ring cartilage is fixed after being touched. The trachea is gently opened with a sharp knife until a 1 mm incision is formed, and the catheter is gently inserted downward along the trachea.

[0130] The method described in this invention:

[0131] The automated robot method performs anesthesia using the same methods as manual methods, and then uses technology to automate the operation.

[0132] Record the time taken for the experiment, the extent of experimental damage, and the subsequent prognosis.

[0133] 1. Experiment time

[0134] The average time for the entire endotracheal intubation process for the four individuals was calculated as follows:

[0135] Tracheostomy intubation: 20.15±1.2 min; blind intubation: 6.15±1.8 min; transcervical radiography intubation: 4.27±1.5 min; AI model tracheal intubation process: 0.75±0.2 min.

[0136] 2. Experimental damage

[0137] Transcervical transillumination intubation method results in unclear observation of the tracheal window and a low success rate of intubation.

[0138] The blind insertion method has a low success rate, and the tracheal window is not clearly visible. Only the appearance of the airway window in one direction can be observed, and the complete original appearance of the airway window cannot be seen. Whether the blind insertion method is successful depends on the operator's experience. Judging whether the insertion is successful based on feeling is not friendly to operators with insufficient experience. Moreover, this method is very likely to cause damage to the airway mucosa, airway inflammation, and even suffocation and death of experimental mice.

[0139] Tracheostomy intubation causes stimulation to the mice through the incision, leading to changes in certain data and affecting experimental results. All methods other than automated robotic methods cause varying degrees of damage.

[0140] 3. Subsequent prognosis

[0141] Blind intubation and tracheotomy have a higher proportion of poor prognoses.

[0142] Comparative studies have shown that this method significantly improves the efficiency of tracheal intubation experiments in mice, which is helpful for conducting related experiments.

[0143] Using control experiments as comparative examples, for example, if the reaction temperature is 20-50℃, experimental schemes corresponding to reaction temperatures outside the 20-50℃ range can be provided.

[0144] Table 1 Comparison of time and success rate between this method and other methods ( n=50) Note: Compared with blind intubation, ****P<0.01; compared with transcervical radiography intubation, ****P<0.01; compared with tracheostomy intubation, ****P<0.01.

[0145] The materials required for the above experiments are as follows:

[0146] 1. Blind insertion method:

[0147] Stainless steel endotracheal tube for laboratory mice: RWD;

[0148] Toothless forceps: RWD.

[0149] 2. Transcervical radiography catheterization:

[0150] Tweezers: RWD;

[0151] Pliers: RWD;

[0152] Light source: LED high-intensity light source, RWD.

[0153] 3. Tracheostomy and intubation:

[0154] No. 6 surgical sutures and No. 6 surgical needles: RWD;

[0155] Ophthalmic scissors: RWD;

[0156] 12cm hemostat: RWD;

[0157] Glass dial indicator: RWD;

[0158] 4. This method:

[0159] Endoscope: 2.8mm custom endoscope;

[0160] Anesthesia and ventilator integrated machine: RWD;

[0161] Robotic arm: Custom DOFBOT robotic arm (motherboard: Raspberry Pi 4B, computing power: 0.2 TFLOPS, CPU: Quad-Core Arm Cortex-A7 21.5GHz, GPU: Broadcom VideoCore IV, OLED device);

[0162] AI chip: STM32 chip;

[0163] Anesthesia maintenance chamber: Custom-made RWD 0.128 cubic meter small anesthetic gas release climate chamber;

[0164] Anesthetic gas: Isoflurane gas;

[0165] Endotracheal intubation: RWD custom-made cuffed endotracheal cannula (containing guidewire, cuff, and cannula).

[0166] In summary, the intelligent endotracheal intubation surgical system and method for laboratory mice based on robotic arm linkage provided by this invention utilizes machine vision combined with deep learning CNN (Convolutional Neural Network) to dynamically identify and label the location of the tracheal tube within the laryngeal cavity of large / small / nude mice and the opening time of the tracheal tube, while simultaneously linking with the robotic arm to achieve dynamic insertion of the endotracheal tube, realizing fully automated endotracheal intubation technology. Through artificial intelligence (AI) visual assistance, the success rate of experiments is effectively improved, time is saved, and the risks of endotracheal intubation failure, various complications, and even death in large / small / nude mice due to difficulties in airway exposure caused by the special anatomical characteristics of large / small / nude mice, and operator error are prevented. In conclusion, this invention provides a technology for identifying narrow cavities and assisting in automated endotracheal intubation, offering researchers a more reliable and effective experimental protocol and providing new ideas for the future development and advancement of animal experiments.

[0167] The remaining technical features in this embodiment can be flexibly selected by those skilled in the art to meet different specific practical needs. However, it is obvious to those skilled in the art that these specific details are not necessary to implement the present invention. In other instances, to avoid obscuring the present invention, well-known components, structures, or parts are not specifically described, and all are within the scope of technical protection defined by the claims of the present invention.

[0168] Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of this invention should be within the protection scope of the appended claims. In the above description, numerous specific details have been set forth to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, to avoid obscuring the invention, well-known techniques, such as specific construction details, operating conditions, and other technical conditions, have not been specifically described.

[0169] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A mechanical arm linkage-based intelligent mouse tracheal intubation surgery system, characterized in that, include: The anesthesia machine includes a respiratory anesthesia machine, an anesthesia maintenance box, and a robotic arm. The anesthesia maintenance box is connected to the respiratory anesthesia machine via a rigid anesthetic gas pipeline. A robotic arm moving track is installed above the anesthesia maintenance box. The anesthesia maintenance box also contains a retractable experimental mouse device plate for fixing the limbs of the experimental mouse. The robotic arm's operating port is equipped with a tracheal tube release device, which is used to pry open the mouse's mouth and then clamp the tracheal tube to move it deeper into the mouse's larynx. An image acquisition module is installed next to the endotracheal tube, and an image processing module is installed inside the robotic arm; the image acquisition module is used to acquire continuous images of the experimental mouse in vitro and in the oral cavity in real time, and transmit them to the image processing module. The image processing module is used to automatically analyze continuous images inside the oral cavity based on the image data, identify and label the trachea of ​​the experimental mouse, and then automatically calculate the time difference between the opening and closing of the tracheal tube window of the experimental mouse; the robotic arm adjusts the position of the tracheal tube according to the position of the experimental mouse's trachea, and automatically stops moving after determining the optimal tracheal intubation position, and inserts the tracheal tube into the experimental mouse's trachea using the tracheal tube release device.

2. The intelligent mouse tracheal intubation surgery system based on linkage of mechanical arms according to claim 1, wherein The image processing module includes a preprocessing unit, a convolutional neural network unit, and a Python algorithm unit; the Python algorithm unit is used to calculate the time difference between the opening and closing of the tracheal tube window in the experimental mouse; the convolutional neural network unit includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. The preprocessing unit preprocesses the continuous images within the oral cavity to obtain an image pixel matrix; the input layer receives three consecutive frames of continuous images and extracts five channels of information for each frame: grayscale, horizontal coordinate gradient, vertical coordinate gradient, x-ray flow, and y-ray flow; the convolutional layer identifies spatial patterns in the image pixel matrix by performing convolution operations on the 3D convolutional kernels of the five channels to obtain features; the pooling layer reduces the size of the matrix to reduce the number of parameters in the entire neural network; the fully connected layer classifies the features, obtains the classification result, and transmits it to the output layer; the output layer outputs the classification result from S4.4 to achieve the identification and labeling of the trachea. 3.The intelligent mouse trachea cannula operation system based on linkage of mechanical arm according to claim 1, wherein, The robotic arm is a multi-axis robotic arm; the robotic arm's moving track includes a horizontal track and a vertical track; the image acquisition module is a lens, and the lens is equipped with an active lighting device; the endotracheal tube is an airbag endotracheal tube.

4. The intelligent mouse tracheal intubation surgery system based on linkage of mechanical arms according to claim 1, wherein, The endotracheal tube release device is a scissor-type endotracheal tube release device. A small spring is placed in the center of the scissor-type endotracheal tube release device. The endotracheal tube is held by the scissor-type endotracheal tube release device so that it is located in the center of the endotracheal tube release device and connected to the spring. The endotracheal tube connected to the scissor-type endotracheal tube release device tightly wraps the air bag endotracheal tube. The endotracheal tube is a symmetrical rigid half-sleeve.

5. A method for intelligent tracheal intubation surgery of experimental mice based on linkage of mechanical arms, applied to the intelligent tracheal intubation surgery system of experimental mice based on linkage of mechanical arms in any one of claims 1-4, characterized in that, Includes the following steps: S1. Turn on the respiratory anesthesia machine. The respiratory anesthesia machine continuously circulates isoflurane gas into the anesthesia maintenance chamber to maintain the anesthesia state of the experimental mice until the muscle reflexes of the experimental mice disappear. S2, Fix the limbs of the anesthetized experimental mouse to the retractable experimental mouse device plate; S3, Start intubation: The image acquisition module acquires continuous images of the experimental mouse in vitro. Based on the continuous images, it is determined whether the endotracheal tube release device has reached the position of the experimental mouse's mouth. The endotracheal tube release device pries open the experimental mouse's mouth to fully expose the larynx. Then, the endotracheal tube is clamped and moved into the experimental mouse's laryngeal cavity. S4. After insertion into the oral cavity, the image acquisition module continues to acquire continuous images inside the oral cavity of the experimental mouse. Then, the continuous images inside the oral cavity are preprocessed using the image processing module, and the trachea of ​​the experimental mouse is identified and labeled based on the convolutional neural network. The time difference between the opening and closing of the tracheal tube window of the experimental mouse is calculated. S5, the robotic arm adjusts the position of the endotracheal tube according to the position of the mouse's trachea, and automatically stops moving after determining the optimal endotracheal tube position. Based on the time difference, the robotic arm uses the endotracheal tube release device to insert the endotracheal tube into the mouse's trachea. S6. The robotic arm removes all instruments except the endotracheal tube from the mouse's mouth. The mouse is then manually removed from the anesthesia maintenance box. The respiratory anesthesia machine then automatically stops releasing isoflurane gas, completing the entire endotracheal intubation process for the mouse.

6. The intelligent tracheal intubation surgery method for experimental mice based on linkage of mechanical arms according to claim 5, characterized in that, In step S4, the identification and labeling of the experimental mouse trachea is achieved based on a convolutional neural network, specifically including the following steps: S4.1 The continuous images inside the oral cavity are preprocessed by the preprocessing unit in the image processing module to obtain an image pixel matrix; S4.2 The input layer receives three consecutive frames of continuous images and extracts five channels of information for each frame: grayscale, horizontal coordinate gradient, vertical coordinate gradient, x-axis optical flow, and y-axis optical flow. S4.

3. The convolutional layer identifies the spatial patterns in the image pixel matrix, performs convolution operations on the 3D convolutional kernels of the five channels, and obtains the target features by extracting the image pixel matrix information. S4.

4. Reduce the size of the image pixel matrix using pooling layers to reduce the number of parameters in the entire neural network; repeat steps S4.3 and S4.

4. S4.5 The fully connected layer classifies the features, obtains the classification result, and transmits it to the output layer; S4.6 Finally, the output layer outputs the classification results from S4.4 to achieve the identification and labeling of the trachea.

7. The intelligent tracheal intubation surgery method for experimental mice based on linkage of mechanical arms according to claim 5, characterized in that, In step S4, calculating the time difference between the opening and closing of the tracheal tube window in the experimental mouse specifically includes: After image recognition, the preprocessing unit extracts images of airway window opening and closing and passes them to the Python algorithm unit in the image processing module. The Python algorithm unit automatically calculates the time difference between the opening and closing of the experimental mouse's trachea based on the time difference between the transmitted images.

8. The intelligent mouse tracheal intubation operation method based on linkage of mechanical arms according to claim 5, characterized in that, The movement principle of the robotic arm is as follows: The recorded images are converted to grayscale, the center position of the robotic arm's movement is calculated, and a three-dimensional coordinate axis is established based on the center position. This three-dimensional coordinate axis is used for trajectory planning and motion planning. The coordinates [x,y,z] represent the spatial position points within the robotic arm's movable range, ensuring that the robotic arm's operating ports can reach every spatial position point within the anesthesia maintenance box. A visual vision device is used to ensure the robotic arm's accurate positioning and visual tracking of the experimental mouse.