Low-light enhancement method and system for cardiac intervention surgery scene based on diffusion model

Abstract

The application discloses a low-light enhancement method and system for a cardiac intervention surgery scene based on a diffusion model, aiming to solve the problem of unclear surgical field in a low-light environment. The application collects multi-angle low-light images through a miniature camera and an external camera system, and performs enhancement processing by using a diffusion model optimized for the cardiac intervention surgery scene. The diffusion model uses two low-light image sources providing different scales and angles as training data sets to capture image details at different levels from micro to macro. By performing the diffusion process at different resolutions, the microstructure and overall tissue layout in the surgical field can be more accurately reconstructed. By introducing prior knowledge related to the cardiac intervention surgery to dynamically adjust the value of noise scheduling, and performing illumination parameter prediction to guide the optimization of diffusion model parameters, the diffusion model can more accurately improve the image quality of the surgical area in real time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical image processing, more particularly, to a low-light enhancement method and system for a cardiac intervention surgery scene based on a diffusion model. BACKGROUND

[0002] Heart valve replacement surgery is a key treatment method in modern cardiac surgery, aiming to replace damaged or diseased heart valves to restore normal heart function. Traditional heart valve replacement surgery is usually performed through a median sternotomy incision. Although this method provides a relatively open view for surgeons, it also increases patient trauma and postoperative recovery time. In recent years, the development of minimally invasive cardiac surgery techniques, such as transcatheter aortic valve replacement (TAVR) and minimally invasive mitral valve repair, has reduced patient trauma but also brought new challenges, especially in terms of surgical field of view.

[0003] In heart valve replacement surgery, accurate visual perception is crucial. Surgeons need to clearly observe the structure of the heart valve, the state of the surrounding tissue, and the placement position of the artificial valve. However, due to the complex anatomy of the heart and the limited operating space required by the surgery, low-light areas often occur, which seriously affect the smooth progress of the surgery.

[0004] The low-light conditions in cardiac intervention surgery mainly come from the following aspects:

[0005] 1. Complexity of heart anatomy: The internal structure of the heart is complex, with multiple chambers and septums that can block light, causing insufficient light in some areas.

[0006] 2. Limitation of surgical incision: To reduce patient trauma, the surgical incision is often small, limiting the placement location and number of light sources.

[0007] 3. Obstruction of surgical instruments: During the surgery, the use of various surgical instruments can block part of the light, causing local shadows.

[0008] 4. Interference of blood and tissue fluid: Bleeding and tissue fluid during the surgery can absorb and scatter light, reducing the lighting effect.

[0009] These low-light conditions have many adverse effects on the surgery:

[0010] 1. Increased surgical difficulty: In the case of insufficient light, doctors have difficulty accurately identifying subtle anatomical structures, increasing the difficulty of surgical operations.

[0011] 2. Extended surgery time: Due to poor visibility, surgeons need to spend more time confirming each step, significantly prolonging the overall surgery time.

[0012] 3. Increased surgery risk: Poor visibility can lead to operational errors, increasing the risk of complications such as bleeding, perivalvular leakage, and conduction block.

[0013] 4. Increased physician fatigue: Long-term work in low-light conditions can exacerbate visual fatigue in physicians, potentially affecting the quality and safety of surgery.

[0014] 5. Impact on surgical teaching: For young surgeons in training, the effectiveness of surgical observation and learning under low-light conditions is greatly reduced.

[0015] Currently, in order to improve the surgical field, the medical community has taken some measures:

[0016] 1. Improve surgical lamp design: Develop brighter and more focused surgical lamps, but this method still cannot solve the problem of deep tissue illumination.

[0017] 2. Use head-mounted lighting devices: These devices can change the direction of light with the movement of the doctor's head, but they often increase the burden on the doctor and have limited light intensity.

[0018] 3. Endoscopic technology: Using endoscopes in some minimally invasive surgeries can provide better visibility, but its application range is limited, and image quality is limited by device performance.

[0019] 4. Fluorescence imaging technology: By injecting fluorescent dyes and using special light sources, some tissue structures can be highlighted, but this method cannot completely improve the overall lighting conditions.

[0020] Although these methods have improved the surgical field to some extent, they still cannot completely solve the problem of low light in heart valve replacement surgery. Especially in deep tissues and narrow spaces, existing technologies still cannot provide ideal lighting effects.

[0021] In addition, existing lighting solutions often only consider the intensity of light, ignoring the importance of light quality. In heart surgery, different tissue structures have different reflection and absorption characteristics of light. Simply increasing the intensity of light may cause some areas to be overexposed, while other areas still lack light. This uneven lighting not only cannot effectively improve the visibility, but also may cause visual interference and affect the doctor's judgment.

[0022] Another issue worth noting is that existing lighting solutions are mostly static and cannot dynamically adjust the light according to different stages and needs of the surgical process. Heart valve replacement surgery is a dynamic process, and the requirements for lighting are not the same at different stages. For example, more focused and bright light may be needed during the suturing stage, while softer and more uniform light may be needed during the observation stage. Lighting systems that lack flexibility cannot meet these dynamic needs.

[0023] Diffusion models were first proposed by Sohl-Dickstein et al. in 2015 and have achieved remarkable results in image generation, super-resolution reconstruction, and image inpainting in recent years. The core idea of this model is to generate or reconstruct images by simulating a process that gradually approaches the data distribution. In the field of image enhancement, diffusion models have shown superior performance to traditional methods, especially in dealing with low light, noise, and blur. However, standard diffusion models still face challenges when directly applied to low-light image enhancement in cardiac intervention surgery scenarios, such as difficulty in noise control, insufficient recovery of local details, and poor adaptability. Therefore, developing a diffusion model-based low-light enhancement method and system suitable for cardiac intervention surgery scenarios remains a technical challenge. SUMMARY

[0024] Due to the problems of the prior art, the present application proposes a diffusion model-based low-light enhancement method and system for cardiac intervention surgery scenarios, aiming to solve the problem of unclear surgical field in low-light environments, which affects the precision and safety of surgical operations. The present application collects multi-angle low-light images through a miniature camera and an external camera system, and uses a diffusion model optimized for cardiac intervention surgery scenarios for enhancement processing.

[0025] To achieve the above-mentioned purpose, in a first aspect, the present application provides a diffusion model-based low-light enhancement method for cardiac intervention surgery scenarios, characterized in that it comprises:

[0026] Step S1, constructing and pre-training a diffusion model for cardiac intervention surgery: collecting image pairs under different lighting conditions during the cardiac intervention surgery process for pre-training of the diffusion model; the diffusion model designs a diffusion process with multiple time steps, each step corresponding to a gradual improvement in image quality; the fine structure inside the heart and the global structure of the surrounding tissue under the motion state of the heart are diffused in high resolution and low resolution, respectively;

[0027] Step S2, obtaining low-light images of the cardiac intervention surgery scenario: using a miniature camera, through a surgical incision or catheter insertion into the heart, to obtain real-time images of the heart interior, while using an external camera system to capture global images of the entire surgical area;

[0028] Step S3, real-time processing of the low-light image in step S2 using the pre-trained diffusion model to obtain an enhanced high-quality image, and online learning and parameter adjustment of the model using image data collected during the surgery: the diffusion model dynamically adjusts the value of the noise schedule according to different stages of the surgery process; predicts the illumination parameters according to the information of the cardiac anatomical structure, the feature information of the surgical instrument and the feature information of different types of valves, and adjusts the parameters in a certain time step in the diffusion model;

[0029] Step S4, feeding back the enhanced high-quality image to the illumination system of the surgical area for dynamic adjustment of the illumination system.

[0030] The diffusion model uses two low-light image sources providing different scales and angles as training data sets to capture image details at different levels from micro to macro. By performing the diffusion process at different resolutions, the microstructure and overall tissue layout in the surgical field can be more accurately reconstructed. By introducing prior knowledge related to cardiac intervention surgery to dynamically adjust the value of the noise schedule and predict the illumination parameters to guide the optimization of the diffusion model parameters, the diffusion model can more accurately improve the image quality of the surgical area in real time.

[0031] Further, in step S1, when pre-training the diffusion model, the parameters including noise level and step size in the diffusion process are optimized.

[0032] Further, in step S3, during the surgery preparation stage, a small noise schedule parameter is used; during the valve replacement process, the noise schedule parameter is increased; and during the suturing stage, the noise schedule parameter is reduced again.

[0033] Further, in step S3, during the illumination parameter prediction, the shadow area caused by the operation of the surgical instrument and the doctor in the surgical area is identified, and the shadow is eliminated by enhancing the image brightness and contrast; the color change of the internal tissue of the heart due to factors including blood flow and illumination angle is identified, and the image is corrected according to the preset standard color model.

[0034] Further, in step S3, a user interaction interface is introduced to allow the doctor to provide real-time feedback on the enhanced image according to his own needs; according to the doctor's feedback, the parameters of the diffusion model are fine-tuned to fine-tune the enhanced image.

[0035] Further, in step S4, the intensity and direction of the surgical lamp are dynamically adjusted according to the brightness distribution in the enhanced image; for the internal area of the heart, controllable micro-LED light sources are used to adjust the internal illumination according to the requirements of the enhanced image.

[0036] In a second aspect, the present application provides a low-light enhancement system for cardiac intervention surgery scene based on diffusion model, characterized by being used to implement the low-light enhancement method for cardiac intervention surgery scene as described above, comprising:

[0037] An image acquisition module that acquires low-light images of the inside of the heart and the entire surgery area in real time from a miniature camera and a global camera;

[0038] A diffusion model processing module based on a GPU-accelerated server, which introduces a multi-scale processing strategy and a lighting parameter prediction module to execute a dynamic image enhancement algorithm;

[0039] An intelligent lighting module that adjusts the parameters of the surgery lighting equipment according to the enhanced images.

[0040] Further, it further comprises a user interaction module to provide an interactive interface for the surgery team; doctors can provide real-time feedback on the enhanced images according to their own needs.

[0041] Further, the diffusion model processing module comprises a dataset, a pre-training module, a model real-time running module, a lighting parameter prediction module, and an image quality assessment module; the model real-time running module obtains high-quality images through low-light image enhancement, and dynamically adjusts the model parameters according to the prediction of the lighting parameter prediction module and the evaluation results of the image quality assessment module.

[0042] Further, the user interaction module allows doctors to mark areas or details that need special attention. The image quality assessment module can receive feedback from the user interaction interface and assess whether the image quality meets the requirements.

[0043] Compared with the prior art, the present application has the following technical effects:

[0044] 1. Effectively improve image quality: enhance the visibility of the valve and surrounding tissues, improve the contrast between surgical instruments and tissues, reduce image noise, and provide more stable visual feedback.

[0045] 2. Customized optimization: integrate prior knowledge of heart anatomy and surgical instruments, accurately enhance different types of valve surgery, and adapt to dynamic scenes such as heart beating and breathing.

[0046] 3. Real-time interaction: real-time interaction with the surgery lighting system, dynamic adjustment of lighting parameters according to enhanced images, and collaborative optimization of the lighting system and image enhancement.

[0047] 4. Improve surgery safety and efficiency: reduce the risk of surgical errors and complications, shorten surgery time, and improve the reaction speed of the surgery team.

[0048] 5. Enhance the effectiveness of surgical teaching and training: It is conducive for doctors to observe and learn complex surgical procedures, and can be used for postoperative analysis and teaching discussion.

[0049] 6. Support remote surgery: High-quality augmented images are suitable for remote transmission, which is conducive to remote experts to provide real-time guidance and improve the precision and safety of robot-assisted surgery.

[0050] 7. Data accumulation and AI-assisted diagnosis: Establish an image database of heart valve diseases to develop automated valve lesion recognition and surgical planning assistance tools. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 The execution flow chart of the low-light enhancement method for the heart intervention surgery scene of an embodiment of the present application.

[0052] Figure 2 The algorithm flow chart of the diffusion model in an embodiment of the present application.

[0053] Figure 3 The example diagram of the low-light image of the heart surgery area of an embodiment of the present application.

[0054] Figure 4 The example diagram of the low-light enhancement effect. Figure 3

[0055] Figure 5 The framework diagram of the low-light enhancement system for the heart intervention surgery scene of an embodiment of the present application.

[0056] Figure 6 The framework diagram of the diffusion model processing module in an embodiment of the present application.

[0057] Figure 7 The framework diagram of the low-light enhancement system for the heart intervention surgery scene of another embodiment of the present application. DETAILED DESCRIPTION

[0058] The present application will be further described below in conjunction with the drawings and specific embodiments, but not as a limitation of the present application.

[0059] In the following detailed description, many specific details are set forth in order to provide a more thorough understanding of the present application. However, it will be apparent to one skilled in the art that well-known algorithms do not show detailed processes in order to avoid obscuring the main idea of the present application.

[0060] In addition, the order of execution of actions, steps, etc. in the devices and methods shown in the claims, specification and drawings can be implemented in any order as long as there is no specific order limitation and the output of the previous process is not used in the subsequent process.

[0061] Embodiment 1​

[0062] This embodiment takes heart valve replacement surgery as an example, see Figure 1 , to provide a low-light enhancement method for heart intervention surgery scene based on diffusion model. The hardware required for the low-light enhancement method is:

[0063] - Miniature camera: inserted into the heart through surgical incision or catheter.

[0064] - GPU accelerated server: used to run diffusion model in real time.

[0065] - Intelligent lighting system: including adjustable surgical lamps and miniature LED light sources.

[0066] - High-resolution display: used to present enhanced images.

[0067] The low-light enhancement method includes the following steps:

[0068] Step S1, construct and pre-train the diffusion model for heart intervention surgery: collect image pairs (low-light images and ideal images) under different lighting conditions during heart intervention surgery for pre-training of the diffusion model. Referring to Figure 2 , the diffusion model designs a diffusion process with multiple time steps, each step corresponding to a gradual improvement in image quality; the fine structure of the heart and the global structure of the surrounding tissue under the motion state of the heart are diffused at high and low resolutions, respectively.

[0069] More specifically, a special diffusion model for heart valve replacement surgery is constructed: when initializing the diffusion model, the obtained low-light image is taken as the starting point. A diffusion process with multiple time steps is designed, each step corresponding to a gradual improvement in image quality. The final output is an enhanced high-quality image that can clearly show the valve structure, surrounding tissue and surgical instruments.

[0070] Optimize the learnable parameters of the diffusion model (train the model):

[0071] Collect a large amount of image data of heart valve replacement surgery, including valve images under different lighting conditions.

[0072] Use these data to train the diffusion model and optimize parameters such as noise level and step size in the diffusion process.

[0073] The core idea of the Diffusion model is to generate or reconstruct images by simulating a process that gradually approaches the data distribution. In the image enhancement task, this principle is used to reconstruct high-quality images from low-quality images.

[0074] 1. Forward diffusion process

[0075] The forward diffusion process is defined as adding Gaussian noise step by step to the image:

[0076]

[0077] where, is the image at the t-th step, is a pre-defined noise schedule.

[0078] 2. The backward diffusion process

[0079] The backward diffusion process aims to recover the original image from the noisy image:

[0080]

[0081] where, and are the mean and variance parameterized by the neural network.

[0082] 3. Training objective

[0083] The training objective of the model is to minimize the following loss function:

[0084]

[0085] where, is the added noise, is the noise predicted by the model.

[0086] To accommodate the unique requirements of heart valve replacement surgery, we introduce a multi-scale processing strategy based on the standard diffusion model to capture image details at different levels from micro to macro. By performing the diffusion process at different resolutions, we can more accurately reconstruct the microstructure and overall tissue layout in the surgical field. At the micro scale: In heart valve replacement surgery, it is crucial to have a clear image of the fine structures of the valve, such as the leaflets and chordae tendineae. The multi-scale model can process images at high resolution, accurately reconstructing these tiny structures, helping doctors assess the severity of valve lesions, choose the appropriate artificial valve, and perform precise suturing operations. At the macro scale: At the same time, doctors also need to observe the movement of the entire heart and the surrounding tissues (such as the atrium, ventricle, and aorta). The multi-scale model can capture these global information at low resolution, helping doctors assess the overall effectiveness of the surgery and identify potential problems (such as bleeding and arrhythmia) early.

[0087] Step S2, acquiring low-light images of the cardiac intervention surgery scene: using a miniature camera, through the surgical incision or catheter insertion into the heart, to acquire real-time images inside the heart, while using an external camera system to capture the global image of the entire surgical area. These two image sources can provide low-light images of different scales and angles, providing rich information for subsequent processing.

[0088] Step S3, real-time processing of the low-light images in step S2 using a pre-trained diffusion model to obtain enhanced high-quality images, and online learning and parameter adjustment of the model using continuously collected image data during the surgery: the diffusion model dynamically adjusts the value of noise scheduling according to different stages of the surgery process; according to the information of cardiac anatomy, the feature information of surgical instruments and the feature information of different types of valves, the light parameter prediction is carried out and the parameter in a certain time step in the diffusion model is adjusted.

[0089] More specifically, the real-time processing flow includes:

[0090] - Input: low-light images obtained from the camera system .

[0091] - Diffusion process: apply a diffusion process of T time steps.

[0092] - Output: enhanced high-quality images .

[0093] To adapt to the special needs of cardiac valve replacement surgery, we introduce prior knowledge related to cardiac valve replacement surgery: integrate information of cardiac anatomy, such as typical morphology of valves, distribution of surrounding blood vessels, etc. Add feature information of commonly used surgical instruments, such as artificial valves, suture needles, etc. to better preserve these key elements during the enhancement process. Consider the characteristics of different types of valves (such as mitral valve, aortic valve), and optimize the image quality of the corresponding area accordingly.

[0094] a. Specific light condition adjustment:

[0095] A light parameter prediction module is added to the model to adjust the brightness and contrast of the image, making the reconstructed image more consistent with the lighting conditions in the operating room. This is achieved by adjusting the parameters in a certain time step in the diffusion model, making the output of the model more suitable for the actual needs of the surgery.

[0096] Shadow removal: surgical instruments and doctors' operations can cause shadows in the surgical area. The light parameter prediction module can identify these shadow areas and eliminate shadows by enhancing image brightness and contrast, improving the clarity of the surgical field.

[0097] Color correction: The color of the internal tissues of the heart can change due to blood flow, light angle, and other factors, which can affect the doctor's judgment of tissue health. The light parameter prediction module can correct the image according to the preset standard color model, presenting a more realistic surgical scene.

[0098] b. Dynamic adjustment of noise scheduling:

[0099] According to different stages of the surgical process (such as the preparation stage, replacement process, and suture stage), dynamically adjust the value of . This is because different stages have different requirements for image clarity, and dynamic adjustment can reduce noise introduction in stages that require high clarity.

[0100] Preparation stage: In the surgical preparation stage, it is necessary to clearly observe the anatomical structure and lesion of the valve, at this time the model can use a smaller noise scheduling parameter to obtain high-definition images.

[0101] Replacement process: During the valve replacement process, attention needs to be paid to the placement position and suture of the artificial valve, at this time the model can appropriately increase the noise scheduling parameter to speed up image processing and ensure real-time performance.

[0102] Suture stage: In the suture stage, it is necessary to clearly observe the direction and tension of the suture line, at this time the model can again reduce the noise scheduling parameter to obtain high-definition images and ensure surgical quality.

[0103] c. Real-time optimization and adaptation:

[0104] Use new image data collected during the operation to perform online learning and adjustment of the model. This ensures that the model can adapt to changing environments and the characteristics of specific operations, such as differences in anatomical structure between different patients.

[0105] Figure 3 and Figure 4 shows the effect of low-light enhancement of the valve area and the entire surgical area during heart surgery. From the figure, it can be seen that the low-light image processed by the diffusion model of the embodiment is enhanced to a high-quality image.

[0106] Step S4, feed back the enhanced high-quality image to the illumination system of the surgical area, and dynamically adjust the illumination system.

[0107] More specifically, an intelligent lighting control system is designed to receive and analyze enhanced image information. According to the brightness distribution in the enhanced image, dynamically adjust the intensity and direction of the surgical lamp. For the internal area of the heart, controllable micro-LED light sources can be used to adjust the internal illumination according to the requirements of the enhanced image.

[0108] Based on the enhanced image, real-time valve recognition and analysis functions can also be developed to help doctors evaluate the valve status.

[0109] From the above method steps, in this embodiment, multi-angle low-light images are collected by a miniature camera and an external camera system, and a diffusion model optimized for a heart intervention surgery scene is used for enhancement processing, which can solve the problem of unclear surgical field in a low-light environment.

[0110] Embodiment 2

[0111] Referring to Figure 5 , the embodiment provides a low-light enhancement system for a heart intervention surgery scene based on a diffusion model, which is used to implement the low-light enhancement method for a heart intervention surgery scene as described in Embodiment 1, and includes:

[0112] The image acquisition module 100 acquires low-light images of the inside of the heart and the entire surgical area in real time from the miniature camera and the global camera;

[0113] The diffusion model processing module 200 based on a GPU-accelerated server introduces a multi-scale processing strategy and a light parameter prediction module, and executes a dynamic image enhancement algorithm.

[0114] The intelligent lighting module 300 adjusts the parameters of the surgical lighting equipment according to the enhanced image.

[0115] Referring to Figure 6 , the diffusion model processing module 200 includes a data set 201, a pre-training module 202, and a model real-time running module 203, a light parameter prediction module 204, and an image quality evaluation module 205; the model real-time running module 203 obtains high-quality images through low-light image enhancement, and dynamically adjusts the model parameters according to the prediction of the light parameter prediction module 204 and the evaluation results of the image quality evaluation module 205.

[0116] This low-light enhancement system for a heart intervention surgery scene based on a diffusion model can effectively improve the quality of the collected images during the surgery and provide more stable visual feedback.

[0117] Embodiment 3

[0118] The embodiment provides a low-light enhancement method and system for a heart intervention surgery scene based on a diffusion model. The steps of the low-light enhancement method are similar to those of Embodiment 1, except that a user interaction interface is also introduced, which allows doctors to provide real-time feedback on the enhanced image according to their own needs; and according to the doctor's feedback, the parameters of the diffusion model are quickly fine-tuned to fine-tune the enhanced image. Referring to Figure 7The low-light enhancement system corresponding thereto is similar to Embodiment 2, but further includes a user interaction module 400 to provide an interactive interface for the surgical team. Different doctors have different preferences for image clarity, contrast, color, etc. The interactive user feedback mechanism allows the doctor to provide feedback on the enhanced image according to his own needs through simple operations (such as a sliding bar, a button), and obtain a visual effect that is more in line with personal habits. The system obtains real-time feedback, and quickly fine-tunes the model parameters to output the image, ensuring that the enhanced image always meets the requirements of surgical precision. For example, the image quality evaluation module receives feedback from the user interaction interface to evaluate whether the image quality meets the requirements. Further, the user interaction module can also allow the doctor to mark areas or details that need special attention, such as circling, highlighting, etc., to facilitate subsequent observation and processing.

[0119] The low-light enhancement method for cardiac intervention surgery described above can be embodied in the form of a computer program product or a software functional unit. If the low-light enhancement method for cardiac intervention surgery described above is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Therefore, the technical solution essentially or the part that contributes to the prior art or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes instructions to make an electronic system (which can be a personal computer, a server, or a network system, etc.) execute all or part of the steps of the method described in each embodiment of the application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0120] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the present embodiments can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0121] In summary, the present application provides a low-light enhancement method and system for cardiac intervention surgery scene based on diffusion model, aiming to solve the problem of unclear surgical field in low-light environment. The present application collects multi-angle low-light images through a miniature camera and an external camera system, and uses a diffusion model optimized for cardiac intervention surgery scene for enhancement processing. The diffusion model uses two low-light image sources providing different scales and angles as training data sets to capture image details at different levels from micro to macro. By performing diffusion process at different resolutions, the microstructure and overall tissue layout in the surgical field can be more accurately reconstructed. By introducing prior knowledge related to cardiac intervention surgery to dynamically adjust the value of noise scheduling, and performing illumination parameter prediction to guide the optimization of diffusion model parameters, the diffusion model can more accurately improve the image quality of the surgical area in real time.

[0122] Those skilled in the art should understand that those skilled in the art can realize variations in combination with the prior art and the above embodiments, which are not described here. Such variations do not affect the essential content of the present application and are not described here.

[0123] The preferred embodiments of the present application are described above. It should be understood that the present application is not limited to the above specific embodiments, and systems and structures not fully described should be understood as implemented in the ordinary way in the art; any person skilled in the art can make many possible changes and modifications to the technical solutions of the present application by using the disclosed methods and technical contents without departing from the scope of the technical solutions of the present application, or modify equivalent embodiments of equivalent changes, which do not affect the essential content of the present application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, without departing from the scope of the technical solutions of the present application, still belongs to the scope of protection of the technical solutions of the present application.

Claims

1. A low-light enhancement method for cardiac interventional surgery scenes based on a diffusion model, characterized by: include: Step S1: Constructing and pre-training a diffusion model for cardiac interventional surgery: collecting image pairs under different lighting conditions during cardiac interventional surgery for pre-training the diffusion model; designing a diffusion process with multiple time steps, each step corresponding to a gradual improvement in image quality; performing the diffusion process at high resolution and low resolution for both the fine structure inside the heart and the global structure of surrounding tissues under cardiac motion; Step S2: Acquire a low-light image of the cardiac interventional surgery scene: Use a miniature camera inserted into the heart through a surgical incision or catheter to acquire real-time images of the heart's interior, while simultaneously using an external camera system to capture a global image of the entire surgical area; Step S3: Using the pre-trained diffusion model to process the low-light image in step S2 in real time to obtain an enhanced high-quality image, and using the image data continuously collected during the operation to perform online learning and parameter adjustment of the model: the diffusion model dynamically adjusts the value of noise scheduling according to different stages of the operation; and predicts illumination parameters based on cardiac anatomical structure information, characteristic information of surgical instruments, and characteristic information of different types of valves, and adjusts the parameters of a certain time step in the diffusion model. Step S4: Feedback the enhanced high-quality image to the lighting system of the surgical area to dynamically adjust the lighting system; In step S3, a small noise scheduling parameter is used during the surgical preparation stage; the noise scheduling parameter is increased during the valve replacement process; and the noise scheduling parameter is reduced again during the suturing stage. When predicting the lighting parameters, the shadow areas in the surgical area caused by the surgical instruments and the doctor's operation are identified, and the shadows are eliminated by enhancing the image brightness and contrast. The changes in the color of the internal heart tissue due to factors including blood flow and lighting angle are identified, and the image is corrected according to a preset standard color model.

2. The low-light enhancement method for cardiac interventional surgery scenes based on a diffusion model according to claim 1, characterized in that: In step S1, when pre-training the diffusion model, the parameters including the noise level and the step size in the diffusion process are optimized.

3. The low-light enhancement method for cardiac interventional surgery scenes based on a diffusion model according to claim 1, characterized in that: In step S3, a user interaction interface is introduced to allow doctors to provide real-time feedback on the enhanced image according to their needs; Based on physician feedback, the enhanced image is fine-tuned by quickly fine-tuning the parameters of the diffusion model.

4. The low-light enhancement method for cardiac interventional surgery scenes based on a diffusion model according to claim 1, characterized in that: In step S4, the intensity and direction of the surgical light are dynamically adjusted according to the brightness distribution in the enhanced image; for the internal area of ​​the heart, a controllable micro-LED light source is used to adjust the internal lighting according to the requirements of the enhanced image.

5. A low-light enhancement system for cardiac interventional surgery scenes based on a diffusion model, characterized by: A method for implementing a low-light enhancement method for a cardiac interventional surgery scene according to any one of claims 1 to 4, comprising: Image acquisition module, which acquires low-light images of the heart interior and the entire surgical area in real time from a micro camera and a global camera; The diffusion model processing module based on the GPU acceleration server introduces a multi-scale processing strategy and an illumination parameter prediction module to execute a dynamic image enhancement algorithm; Intelligent lighting module that adjusts the parameters of surgical lighting equipment based on the enhanced image.

6. The low-light enhancement system for cardiac interventional surgery scenes based on a diffusion model according to claim 5, characterized in that: It also includes a user interaction module that provides an interactive interface for the surgical team; doctors can provide real-time feedback on the enhanced images according to their needs.

7. The low-light enhancement system for cardiac interventional surgery scenes based on a diffusion model according to claim 5 or 6, characterized in that: The diffusion model processing module includes a data set, a pre-training module, a model real-time operation module, an illumination parameter prediction module, and an image quality assessment module; the model real-time operation module obtains high-quality images through low-light image enhancement, and dynamically adjusts model parameters based on the prediction results of the illumination parameter prediction module and the evaluation results of the image quality assessment module.

8. The low-light enhancement system for cardiac interventional surgery scenes based on a diffusion model according to claim 6, characterized in that: The user interaction module allows the doctor to mark areas or details that require special attention.

Images