Intelligent cold cathode imaging device

Abstract

The present invention relates to the field of medical device technology, and provides an intelligent cold cathode imaging device comprising: a C-arm equipped with a beam splitter and a digital imaging plate; a drive motor for driving the C-arm to rotate about its central axis; a steering mechanism, the output end of which is connected to the outer wall of the C-arm, the steering mechanism driving the C-arm to rotate along a path around the center point of the C-arm; and a main controller equipped with a deep learning module comprising a neural network feature extraction mechanism, an attention mechanism, and a three-dimensional convolution mechanism for generating three-dimensional and two-dimensional images of annotated target bones. The present invention achieves multi-angle X-ray imaging with a simple and reliable structure, greatly assisting doctors in observing and diagnosing injured parts and enabling more efficient patient treatment in field hospitals with limited medical staff.

Description

Technical Field

[0001] The present invention relates to the technical field of medical devices, and in particular to an intelligent cold cathode imaging device. Background Art

[0002] The C-arm X-ray machine is an intelligent full-pulse fluoroscopy instrument widely used in interventional departments, orthopedics, surgery, orthopedic surgery, urology, spinal surgery, abdominal surgery, pain management, cardiology, gastroenterology, gynecology and operating rooms.

[0003] For example, the utility model patent with application number 201820193397.4 only discloses a C-arm equipped with a beam spotter and a digital imaging plate connected to a lifting frame and a mounting frame. The lifting frame moves the C-arm up and down, and the mounting frame moves the C-arm horizontally, which cannot achieve multi-angle imaging of the C-arm.

[0004] Therefore, in order to ensure stable imaging quality, the structure of traditional mobile C-arm X-ray machines is difficult to adapt to the tense situation in field hospitals, and cannot better assist doctors in field hospitals to quickly determine bone injuries through multi-angle imaging. Summary of the Invention

[0005] To this end, the present invention provides an intelligent cold cathode imaging device, which realizes multi-angle X-ray imaging with a simple and reliable structure by rotating a C-arm equipped with a beam splitter and a digital imaging plate around its central axis and flipping around its center point.

[0006] The present invention also processes multi-angle imaging through deep learning based on the attention mechanism to form three-dimensional and two-dimensional images with labeled medical features, which greatly assists doctors in observing and diagnosing injured parts, and enables more efficient treatment of patients when medical staff in field hospitals are limited.

[0007] To achieve the above objectives, the present invention provides an intelligent cold cathode imaging device, comprising:

[0008] The C-arm has a beam emitter at one end of its inner wall that emits cold cathode X-rays, and a digital imaging plate that receives cold cathode X-rays at the other end, which is used to form bone images at any angle when X-rays scan the corresponding bones of the human body;

[0009] a drive motor, connected to the C-arm for driving the C-arm to rotate circumferentially around its central axis;

[0010] A steering mechanism, an output end of which is connected to the outer peripheral wall of the C-arm, wherein the steering mechanism drives the C-arm to flip along a flipping path surrounding the center point of the C-arm;

[0011] a main controller, communicatively connected to the beam splitter, the digital imaging board, and the steering mechanism, the main controller receiving multi-angle imaging of the digital imaging board and imaging angles of the steering mechanism;

[0012] The main controller is equipped with a deep learning module, which includes a neural network feature extraction mechanism, an attention mechanism, and a three-dimensional convolution mechanism;

[0013] The neural network feature extraction mechanism is used to extract features of the target bone imaged from multiple angles and generate a two-dimensional image and a three-dimensional image that reasonably connect the target bone;

[0014] The attention mechanism is used to enhance the correlation features between the multiple features;

[0015] The three-dimensional convolution mechanism is used to splice the associated features and generate a two-dimensional annotated image and a three-dimensional annotated image of the target bone for intelligent cold cathode imaging.

[0016] Furthermore, the crawler includes a back crawler and a lateral crawler, the back crawler is provided on the outer peripheral wall of the C-arm, and the lateral crawler is provided on at least one side wall of the C-arm excluding the outer peripheral wall;

[0017] The outer peripheral wall of the C-shaped arm is fixed with a guide member, and the rear crawler and the side crawler are separated from the corresponding outer peripheral wall and side wall, and then pass through the guide member and are wound around the driving wheel;

[0018] The steering mechanism includes a steering motor and a steering member;

[0019] The steering motor is arranged in the installation box, and the steering member is fixed to the guide member;

[0020] The steering

[0021] The output shaft of the motor is connected to the steering member through the opening of the mounting box, so as to drive the guide member and the C-arm to rotate around the center point of the C-arm.

[0022] Furthermore, the intelligent cold cathode imaging device further comprises a lifting platform and a base, wherein the bottom end of the lifting platform is mounted on the base, and the top end is detachably connected to the mounting box.

[0023] Furthermore, a horizontally arranged electric guide rail is provided at the top of the lifting platform, and a slider is fixedly provided at the bottom of the installation box. The slider is detachably connected to the electric guide rail to realize the detachable connection between the lifting platform and the installation box;

[0024] When the electric guide rail pushes the slider to move, it drives the C-arm to move horizontally.

[0025] Furthermore, the multi-angle imaging at least includes: when the steering mechanism outputs a first flip angle, the drive motor outputs a first rotation angle and the digital imaging plate acquires a first image, the drive motor outputs the first rotation angle and the digital imaging plate acquires a second image, and when the steering mechanism outputs a second flip angle, the drive motor outputs the first rotation angle and the digital imaging plate acquires a third image, the drive motor outputs the second rotation angle and the digital imaging plate acquires a fourth image;

[0026] The main controller uses the neural network feature extraction mechanism to process the first to fourth images to generate feature vectors 1 to 4 of the target bone respectively;

[0027] The main controller processes the feature vectors 1 to 4 through the attention mechanism to enhance the associated features of the target skeleton;

[0028] The main controller uses a three-dimensional convolution mechanism to deeply fuse feature vectors one to four according to the associated features, generates a two-dimensional image and a three-dimensional image that reasonably connect the target bones in the first imaging to the fourth imaging, and annotates the two-dimensional image and the three-dimensional image according to the associated features to generate the two-dimensional annotated image and the three-dimensional annotated image.

[0029] Furthermore, the attention mechanism includes a compression network and an incentive network;

[0030] The main controller receives sample data of damaged bones of the human body and trains the neural network feature extraction mechanism, the excitation network and the three-dimensional convolution mechanism according to the sample data.

[0031] Furthermore, multiple groups of compression networks and excitation networks constitute multiple channels of the attention mechanism;

[0032] The main controller processes the eigenvectors one to four respectively through each group of compression network and excitation network to enhance different associated features of the target bone.

[0033] Furthermore, in the three-dimensional convolution mechanism, the main controller performs feature splicing on the feature vectors one to four processed by the attention mechanism to generate splicing features and the two-dimensional image, and performs three-dimensional convolution on the splicing features to generate multi-view features and the three-dimensional image.

[0034] Furthermore, the deep learning module also includes a feature annotation mechanism, which includes at least a pooling layer, a fully connected layer and an activation function;

[0035] The main controller sequentially passes the multi-view features into the pooling layer, the fully connected layer and the activation function to generate a two-dimensional annotated image and a three-dimensional annotated image with the target skeleton annotated.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] 1. The C-arm equipped with a beam splitter and a digital imaging plate rotates around its central axis and flips around its center point, thereby achieving multi-angle X-ray imaging with a simple and reliable structure.

[0038] 2. The flexible movement of the C-arm is achieved through tracks, lifting platforms and electric guide rails. The intelligent cold cathode imaging device is easy to assemble and maintain, and can adapt to the environment of field hospitals.

[0039] 3. Through deep learning based on the attention mechanism, multi-angle imaging is processed to form three-dimensional images and two-dimensional images with annotated medical features for intelligent cold cathode imaging. This greatly assists doctors in observing and diagnosing injured parts, and enables more efficient treatment of patients when medical staff in field hospitals are limited.

[0040] 4. Through the multi-channel attention mechanism, compared with using only neural networks, the consistency of the associated features of the extracted target human skeleton is further guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Schematic diagram of the structure of an intelligent cold cathode imaging device according to an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of a specific model of a deep learning module according to an embodiment of the present invention;

[0043] Figure 3 Schematic diagram of a specific model of the attention mechanism and three-dimensional convolution mechanism of an embodiment of the present invention.

[0044] In the figure: 1. C-arm; 11. Back track; 12. Side track; 13. Angle detection port; 14. Beam detector; 15. Digital imaging board; 21. Guide box; 22. Steering member; 3. Mounting box; 31. Steering motor; 32. Drive motor; 33. Electric guide rail; 41. Lifting platform; 42. Housing; 43. Frame; 5. Base; 51. Universal wheel; 52. Mounting plate. DETAILED DESCRIPTION

[0045] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0046] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0047] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0048] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0049] like Figures 1 to 3 As shown, the present invention provides an intelligent cold cathode imaging device, which realizes multi-angle X-ray imaging with a simple and reliable structure by rotating a C-arm equipped with a beam splitter and a digital imaging plate around its central axis and flipping around its center point.

[0050] The present invention also processes multi-angle imaging through deep learning based on the attention mechanism to form three-dimensional and two-dimensional images with labeled medical features, which greatly assists doctors in observing and diagnosing injured parts, and enables more efficient treatment of patients when medical staff in field hospitals are limited.

[0051] in, Figure 1 Schematic diagram of the structure of an intelligent cold cathode imaging device according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a specific model of a deep learning module according to an embodiment of the present invention; Figure 3 Schematic diagram of a specific model of the attention mechanism and three-dimensional convolution mechanism of an embodiment of the present invention;

[0052] It includes a C-arm 1, a dorsal track 11, a lateral track 12, an angle detection port 13, a beam detector 14, a digital imaging board 15, a guide box 21, a steering part 22, a mounting box 3, a steering motor 31, a drive motor 32, an electric guide rail 33, a lifting platform 41, a shell 42, a frame 43, a base 5, a universal wheel 51, and a mounting plate 52.

[0053] Example 1

[0054] like Figure 1 As shown, this embodiment proposes an intelligent cold cathode imaging device, comprising: a C-arm 1, one end of the inner wall of which is provided with a beam emitter 14 for emitting cold cathode X-rays, and the other end is provided with a digital imaging plate 15 for receiving cold cathode X-rays, so as to form a bone image of any angle when the X-ray scans the corresponding bones of the human body; a drive motor 32, which is in transmission connection with the C-arm 1 and is used to drive the C-arm 1 to rotate circumferentially around its central axis; a steering mechanism, whose output end is connected to the outer wall of the C-arm 1, and the steering mechanism drives the C-arm 1 to flip along a flipping path around the center point of the C-arm 1; a main controller, which is connected to the beam emitter 14, the digital imaging plate 15 and the steering mechanism The communication connection is used, and the main controller receives the multi-angle imaging of the digital imaging plate 15 and the imaging angle of the steering mechanism; the main controller is equipped with a deep learning module, and the deep learning module includes a neural network feature extraction mechanism, an attention mechanism and a three-dimensional convolution mechanism; the neural network feature extraction mechanism is used to extract the features of the target bones of the multi-angle imaging, and generate a two-dimensional image and a three-dimensional image that reasonably connect the target bones; the attention mechanism is used to enhance the correlation features between multiple features; the three-dimensional convolution mechanism is used to splice the correlation features and generate a two-dimensional annotated image and a three-dimensional annotated image of the target bones for intelligent cold cathode imaging.

[0055] It should be noted that the target bone has bone features similar to those of samples of damaged human bones, and the samples of damaged human bones include characteristic images of fractures, arthritis, osteoporosis and bone tumors.

[0056] Furthermore, a crawler is fixedly provided on the side wall of the C-arm 1; a mounting box 3 is provided on one side of the C-arm 1, and a driving wheel wrapped around the crawler and the driving motor 32 are provided inside the mounting box 3, and the output shaft of the driving motor 32 is connected to the driving wheel to form a transmission connection between the driving motor 32 and the C-arm 1 and drive the C-arm machine to rotate around its central axis.

[0057] Furthermore, the crawler includes a dorsal crawler 11 and a lateral crawler 12, the dorsal crawler 11 is arranged on the outer peripheral wall of the C-arm 1, and the lateral crawler 12 is arranged on at least one side wall of the C-arm 1 except the outer peripheral wall; the outer peripheral wall of the C-arm 1 is fixed with a guide member, and the dorsal crawler 11 and the lateral crawler 12 are separated from the corresponding outer peripheral wall and side wall, and then wound around the driving wheel through the guide member; the steering mechanism includes a steering motor 31 and a steering member 22; the steering motor 31 is arranged in the installation box 3, and the steering member 22 is fixed to the guide member; the output shaft of the steering motor 31 is connected to the steering member 22 through the opening of the installation box 3, so as to drive the guide member and the C-arm 1 to flip around the center point of the C-arm 1.

[0058] Specifically, the guide member is preferably a rotating drum, the steering member 22 is a reduction mechanism of the motor, preferably a reduction gear, and the output shaft of the steering motor 31 is transmission-connected to the steering member 22 , thereby achieving stable transmission of the steering motor 31 .

[0059] It can be understood that in some other possible embodiments, the rear crawler track 11 and the lateral crawler track 12 can be replaced by a rear guide rail and a lateral guide rail.

[0060] Furthermore, the intelligent cold cathode imaging device further includes a lifting platform 41 and a base 5 . The bottom end of the lifting platform 41 is mounted on the base 5 , and the top end is detachably connected to the installation box 3 .

[0061] Furthermore, a horizontally arranged electric guide rail 33 is provided at the top of the lifting platform 41, and a slider is fixedly provided at the bottom of the installation box 3. The slider is detachably connected to the electric guide rail 33 to realize the detachable connection between the lifting platform 41 and the installation box 3; when the electric guide rail 33 pushes the slider to move, it drives the C-arm 1 to move horizontally.

[0062] like Figure 1 As shown, the specific structure of the intelligent cold cathode imaging device described in this embodiment is as follows: Figure 1In the coordinate system, the central axis of the C-arm 1 in its initial position is parallel to the x-axis and arranged horizontally. The center point of the C-arm 1 is the center of the arc on which the C-arm is located. A beam splitter 14 is located at the top of the inner wall of the C-arm 1, and a digital imaging plate 15 is located at the bottom of the inner wall of the C-arm 1. The beam splitter 14 and the digital imaging plate 15 are arranged symmetrically along the z-axis, resulting in symmetry between the upper and lower halves of the C-arm 1. An angle detection port 13 is provided on the inner wall at the symmetric location. An angle detection port 13 houses a laser level, making it easier for medical personnel to accurately position the C-arm 1. A guide box 21 is fixedly provided on the outer peripheral wall of the C-arm 1 corresponding to the horizontal direction of the angle detection port 13. The guide box 21 is provided with the guide member. The back track 11 covering the outer peripheral wall of the C-arm 1 and the two lateral tracks 12 covering the front and rear sides of the C-arm 1 extend into the interior of the guide box 21, pass through one or more guide members, realize steering, and are respectively wound around a driving wheel of the mounting box 3. A steering member 22 is provided on one side of the guide box 21, and a mounting box 3 is provided on one side of the steering member 22. A steering motor 31 is provided in the mounting box 3 at a position close to the C-arm 1, and a drive motor 32 is provided at a position away from the C-arm 1, which meets the need for greater torque when the C-arm 1 is flipped. The lifting platform 41 is preferably an electric lifting platform, the outside of which is surrounded by a shell 42. The base 5 includes universal wheels 51 and a mounting plate 52. The mounting plate 52 is detachably connected to the lifting platform 41, the outer shell 42 and the frame 43. The part of the mounting plate 52 close to the C-arm 1 is connected to the universal wheels 51, and the part away from the C-arm 1 is provided with the frame 43. The frame 43 not only ensures the stability of the lifting platform 41, but also facilitates medical staff to push the intelligent cold cathode imaging device.

[0063] like Figure 1 As shown, the assembly process of the intelligent cold cathode imaging device described in this embodiment is: the shell 42, the electric guide rail 33 and the lifting platform 41 are integrated and fixed on the mounting plate 52, the frame 43 is installed, and the electric guide rail 33 at the top of the lifting platform 41 is installed with the mounting plate 52, and the mounting box 52 is integrated with the steering member 22, the guide member box 21, and the C-arm 1.

[0064] In this embodiment, the C-arm 1 equipped with a beam splitter 14 and a digital imaging plate 15 rotates around its central axis and flips around its center point, thereby achieving multi-angle X-ray imaging with a simple and reliable structure; the flexible movement of the C-arm 1 is achieved through the tracks, lifting platform 41 and electric guide rails 33, and the intelligent cold cathode imaging device is easy to assemble and maintain, and is suitable for the environment of a field hospital.

[0065] Example 2

[0066] like Figures 1 to 3As shown, this embodiment, based on the first embodiment, proposes an intelligent cold cathode imaging device, which has a deep learning module based on the attention mechanism, and the deep learning module can generate a two-dimensional annotated image and a three-dimensional annotated image of the target bone.

[0067] Specifically, such as Figure 2 As shown, the multi-angle imaging at least includes: when the steering mechanism outputs a first flip angle, the drive motor 32 outputs a first rotation angle and the digital imaging board 15 acquires a first image, the drive motor 32 outputs the first rotation angle and the digital imaging board 15 acquires a second image, and when the steering mechanism outputs a second flip angle, the drive motor 32 outputs the first rotation angle and the digital imaging board 15 acquires a third image, the drive motor 32 outputs the second rotation angle and the digital imaging board 15 acquires a fourth image;

[0068] The main controller uses the neural network feature extraction mechanism to process the first to fourth images to generate feature vectors 1 to 4 of the target bone respectively;

[0069] The main controller processes the feature vectors 1 to 4 through the attention mechanism to enhance the associated features of the target skeleton;

[0070] The main controller uses a three-dimensional convolution mechanism to deeply fuse feature vectors one to four according to the associated features, generates a two-dimensional image and a three-dimensional image that reasonably connect the target bones in the first imaging to the fourth imaging, and annotates the two-dimensional image and the three-dimensional image according to the associated features to generate the two-dimensional annotated image and the three-dimensional annotated image.

[0071] Furthermore, the attention mechanism includes a compression network and an excitation network that process eigenvectors one to four in sequence; the main controller receives sample data of damaged human bones and trains the neural network feature extraction mechanism, the excitation network and the three-dimensional convolution mechanism based on the sample data.

[0072] Furthermore, multiple groups of compression networks and excitation networks constitute multiple channels of the attention mechanism; the main controller processes the feature vectors one to four respectively through each group of compression networks and excitation networks to enhance different associated features of the target skeleton.

[0073] Specifically, such as Figure 3 As shown, the multiple channels of the attention mechanism are parallel channels.

[0074] Furthermore, in the three-dimensional convolution mechanism, the main controller performs feature splicing on the feature vectors one to four processed by the attention mechanism to generate splicing features and the two-dimensional image, and performs three-dimensional convolution on the splicing features to generate multi-view features and the three-dimensional image.

[0075] Furthermore, the deep learning module also includes a feature labeling mechanism, which includes at least a pooling layer, a fully connected layer and an activation function; the main controller passes the multi-view features and the features into the pooling layer, the fully connected layer and the activation function in sequence to generate a two-dimensional labeled image and a three-dimensional labeled image of the labeled target skeleton.

[0076] Specifically, the activation function is a Sigmoid (threshold function) layer.

[0077] like Figure 2 and 3 As shown in the figure, the attention mechanism of the deep learning module described in this embodiment includes compression and excitation. The excitation part calculates a weight value for each channel in the corresponding feature map based on feature vectors one to four obtained during the compression process, normalizes the weight value through the Sigmoid function to obtain a weight vector, and multiplies all elements in the weight vector by the corresponding channel of the input feature.

[0078] Specifically, the attention mechanism and the three-dimensional convolution mechanism adopt an early fusion strategy to deeply fuse visual features from different perspectives at an earlier stage, and can maintain semantic consistency between visual features from different perspectives, which is more conducive to downstream tasks.

[0079] The attention mechanism is CAM (Channel Attention Module), which uses channel attention to enhance key information between multi-view features and suppress information with low task relevance. The three-dimensional convolution mechanism is the 3D-CM module, which uses three-dimensional convolution to deeply fuse multi-view features. Three-dimensional convolution can establish connections between visual features from different viewpoints, thereby learning the spatial structure and detail information contained in multi-view images. It is far better at extracting spatial structural features than two-dimensional convolution.

[0080] Specifically, in Figure 2In the method, the neural network feature extraction mechanism is a CNN (Convolutional Neural Network, channel attention) network, which extracts local visual features of all multi-view images through the CNN network, and then inputs these features into the multi-view feature fusion module for deep fusion to obtain multi-view features. The multi-view features contain complete and semantically consistent visual information, which can be used for processing downstream tasks. In order to verify the information extraction capability of the model, the downstream task here is classification and annotation. After the average pooling layer, the multi-view features are compressed into a one-dimensional feature vector of length C, and then the feature vector is sent to the fully connected layer and the Sigmoid layer to realize classification and annotation. Finally, the results of the classification and annotation task are used to generate two-dimensional annotated images and three-dimensional annotated images.

[0081] It should be noted that this embodiment preferably uses CNN as the backbone network of the deep learning module, that is, CNN is preferably used as the neural network feature extraction mechanism, and other types of neural networks, such as RNN neural network, may also be used.

[0082] In this embodiment, multi-angle imaging is processed through deep learning based on the attention mechanism to form three-dimensional images and two-dimensional images with labeled medical features, which greatly assists doctors in observing and diagnosing injured parts, and achieves more efficient treatment of patients when medical staff in field hospitals are limited; through the multi-channel attention mechanism, compared with using only neural networks, the consistency of the extracted target human skeleton features is further guaranteed.

[0083] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0084] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. An intelligent cold cathode imaging device, characterized in that: include: A C-arm (1) has an inner wall at one end provided with a beam emitter (14) for emitting cold cathode X-rays, and at the other end provided with a digital imaging plate (15) for receiving cold cathode X-rays, for enabling X-rays to scan corresponding bones of a human body to form bone images at any angle; a drive motor (32) in transmission connection with the C-arm (1) for driving the C-arm (1) to rotate circumferentially around its central axis; A steering mechanism, the output end of which is connected to the outer peripheral wall of the C-arm (1), wherein the steering mechanism drives the C-arm (1) to flip along a flipping path around the center point of the C-arm (1); a main controller, communicatively connected to the beam splitter (14), the digital imaging plate (15) and the steering mechanism, the main controller receiving multi-angle imaging of the digital imaging plate (15) and the imaging angle of the steering mechanism; The main controller is equipped with a deep learning module, which includes a neural network feature extraction mechanism, an attention mechanism, and a three-dimensional convolution mechanism; The neural network feature extraction mechanism is used to extract features of the target bone imaged from multiple angles and generate a two-dimensional image and a three-dimensional image that reasonably connect the target bone; The attention mechanism is used to enhance the correlation features between the multiple features; The three-dimensional convolution mechanism is used to splice the associated features and generate a two-dimensional annotated image and a three-dimensional annotated image of the target bone for intelligent cold cathode imaging.

2. The intelligent cold cathode imaging device according to claim 1, characterized in that: A crawler track is fixedly provided on the side wall of the C-arm (1); A mounting box (3) is provided on one side of the C-arm (1), and a driving wheel wound around the crawler and the driving motor (32) are provided inside the mounting box (3). The output shaft of the driving motor (32) is connected to the driving wheel, so that the driving motor (32) forms a transmission connection with the C-arm (1) and drives the C-arm (1) to rotate around its central axis.

3. The intelligent cold cathode imaging device according to claim 2, characterized in that: The crawler comprises a back crawler (11) and a lateral crawler (12), wherein the back crawler (11) is arranged on the outer peripheral wall of the C-arm (1), and the lateral crawler (12) is arranged on at least one side wall of the C-arm (1) excluding the outer peripheral wall; The outer peripheral wall of the C-shaped arm (1) is fixed with a guide member, and the back crawler (11) and the side crawler (12) are separated from the corresponding outer peripheral wall and side wall, and then passed through the guide member and wound around the driving wheel; The steering mechanism comprises a steering motor (31) and a steering member (22); The steering motor (31) is arranged in the installation box (3), and the steering member (22) is fixed to the guide member; The output shaft of the steering motor (31) is connected to the steering member (22) through the opening of the mounting box (3) to drive the guide member and the C-arm (1) to rotate around the center point of the C-arm (1).

4. The intelligent cold cathode imaging device according to claim 2, characterized in that: It also includes a lifting platform (41) and a base (5), wherein the bottom end of the lifting platform (41) is mounted on the base (5) and the top end is detachably connected to the installation box (3).

5. The intelligent cold cathode imaging device according to claim 4, characterized in that: A horizontally arranged electric guide rail (33) is provided at the top of the lifting platform (41), and a slider is fixedly provided at the bottom of the installation box (3). The slider is detachably connected to the electric guide rail (33) to achieve detachable connection between the lifting platform (41) and the installation box (3); When the electric guide rail (33) pushes the slider to move, it drives the C-arm (1) to move horizontally.

6. The intelligent cold cathode imaging device according to any one of claims 1 to 5, characterized in that: The multi-angle imaging at least includes: when the steering mechanism outputs a first flip angle, the drive motor (32) outputs a first rotation angle and the digital imaging plate (15) acquires a first image, the drive motor (32) outputs the first rotation angle and the digital imaging plate (15) acquires a second image, and when the steering mechanism outputs a second flip angle, the drive motor (32) outputs the first rotation angle and the digital imaging plate (15) acquires a third image, the drive motor (32) outputs the second rotation angle and the digital imaging plate (15) acquires a fourth image; The main controller uses the neural network feature extraction mechanism to process the first to fourth images to generate feature vectors 1 to 4 of the target bone respectively; The main controller processes the feature vectors one to four through the attention mechanism to enhance the associated features of the target skeleton; The main controller uses a three-dimensional convolution mechanism to deeply fuse feature vectors one to four according to the associated features, generates a two-dimensional image and a three-dimensional image that reasonably connect the target bones in the first imaging to the fourth imaging, and annotates the two-dimensional image and the three-dimensional image according to the associated features to generate the two-dimensional annotated image and the three-dimensional annotated image.

7. The intelligent cold cathode imaging device according to claim 6, characterized in that: The attention mechanism includes a compression network and an excitation network; The main controller receives sample data of human bone injuries and trains the neural network feature extraction mechanism, the excitation network and the three-dimensional convolution mechanism according to the sample data.

8. The intelligent cold cathode imaging device according to claim 7, characterized in that: Multiple groups of compression networks and excitation networks constitute multiple channels of the attention mechanism; The main controller processes the eigenvectors one to four respectively through each group of compression network and excitation network to enhance different associated features of the target bone.

9. The intelligent cold cathode imaging device according to claim 6, characterized in that: In the three-dimensional convolution mechanism, the main controller performs feature splicing on the feature vectors one to four processed by the attention mechanism to generate splicing features and the two-dimensional image, and performs three-dimensional convolution on the splicing features to generate multi-view features and the three-dimensional image.

10. The intelligent cold cathode imaging device according to claim 9, characterized in that: The deep learning module also includes a feature annotation mechanism, which includes at least a pooling layer, a fully connected layer and an activation function; The main controller sequentially passes the multi-view features into the pooling layer, the fully connected layer and the activation function to generate a two-dimensional annotated image and a three-dimensional annotated image with the target skeleton annotated.

Images