Chest trunk nerve block auxiliary system

Abstract

The invention relates to the technical field of machine vision, and provides a thoracic trunk nerve block auxiliary system, which comprises a wireless ultrasonic probe device used for collecting an ultrasonic image of a thoracic trunk nerve block area of a patient based on a thoracic trunk nerve block scheme of the patient; the thoracic trunk nerve block scheme is generated by the mobile terminal device based on multi-dimensional individual feature information of the patient; the mobile terminal device is used for determining an anatomical structure and a target retardation area based on the ultrasonic image; generating a blocking path based on the thoracic trunk nerve blocking scheme, the anatomical structure and the target blocking area; the target retardation area is an injection area of local anesthetics; and the augmented reality glasses device is used for displaying the ultrasonic image, the anatomical structure, the target retardation area and the retardation path on the real view image of the patient in an overlapping manner. According to the method, individualized retardation scheme recommendation, real-time ultrasonic visual guidance and dynamic retardation path planning are realized, the retardation success rate is improved, and the operation difficulty is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of machine vision, and in particular to a thoracic trunk nerve block auxiliary system. Background Art

[0002] Thoracic trunk nerve blocks are commonly used perioperatively during thoracic surgery and include serratus anterior plane blocks, thoracic paravertebral nerve blocks, and erector spinae plane blocks. Currently available thoracic trunk nerve block adjuncts have numerous shortcomings, including reliance on clinical experience for block protocols, resulting in high subjectivity; hand-eye asynchrony and limited visual field during puncture; time-consuming and specialized ultrasound image interpretation; inaccurate block path planning; and low device integration. These issues result in unstable block effects, complex procedures, and potential for deviation. Summary of the Invention

[0003] The present invention provides a thoracic trunk nerve block auxiliary system to address the defects of the existing technology that rely on experience, resulting in unstable blocking effects, complex operation and easy deviation. It realizes individualized blocking scheme recommendation, real-time ultrasound visualization guidance and dynamic blocking path planning, improves the blocking success rate and reduces the difficulty of operation.

[0004] The present invention provides a thoracic trunk nerve block auxiliary system, comprising a wireless ultrasound probe device, a mobile terminal device and an augmented reality glasses device that are communicatively connected; wherein: The wireless ultrasound probe device is used to acquire an ultrasound image of the patient's thoracic trunk nerve block area based on the patient's thoracic trunk nerve block plan; the thoracic trunk nerve block plan is generated by the mobile terminal device based on the patient's multi-dimensional individual feature information; The mobile terminal device is configured to determine an anatomical structure and a target block area based on the ultrasound image; generate a block path based on the thoracic trunk nerve block protocol, the anatomical structure, and the target block area; the target block area is an injection area for a local anesthetic; The augmented reality glasses device is used to superimpose and display the ultrasound image, the anatomical structure, the target blocking area, and the blocking path on the patient's real view image.

[0005] According to a thoracic trunk nerve block auxiliary system provided by the present invention, the wireless ultrasound probe device is also used to collect ultrasound images of the patient's thoracic trunk nerve block area through a composite imaging mode and an inter-frame interpolation algorithm based on a wireless ultrasound probe matched with the thoracic trunk nerve block scheme.

[0006] According to a thoracic trunk nerve block auxiliary system provided by the present invention, the mobile terminal device includes a block plan formulation module, wherein: The block plan formulation module is used to extract key features from the patient's multi-dimensional individual feature information and generate multiple candidate thoracic trunk nerve block plans based on the key features and an evidence-based rule base; the evidence-based rule base is a rule base constructed based on the principles of evidence-based medicine; The block plan formulation module is further configured to select the optimal candidate thoracic trunk nerve block plan from the multiple candidate thoracic trunk nerve block plans as the patient's thoracic trunk nerve block plan.

[0007] According to the thoracic trunk nerve block auxiliary system provided by the present invention, the block plan formulation module is further configured to, when the recommended dose of local anesthetic in the current optimal candidate thoracic trunk nerve block plan exceeds a threshold, delete the current optimal candidate thoracic trunk nerve block plan, and select the optimal candidate thoracic trunk nerve block plan from the remaining candidate thoracic trunk nerve block plans as the patient's thoracic trunk nerve block plan; The block plan formulation module is further configured to output a warning message when the recommended dose of local anesthetic in all candidate thoracic trunk nerve block plans exceeds a threshold; the warning message is configured to prompt manual intervention.

[0008] According to a thoracic trunk nerve block auxiliary system provided by the present invention, the mobile terminal device includes an ultrasound image preprocessing module and a thoracic nerve block anatomical ultrasound recognition module, wherein: The ultrasound image preprocessing module is used to preprocess the ultrasound image to obtain the preprocessed ultrasound image; The thoracic nerve block anatomical ultrasound recognition module is used to input the pre-processed ultrasound image into the ultrasound image deep learning model to determine the anatomical structure and target block area; Among them, the ultrasound image deep learning model is obtained by using the improved YOLOv8 model as the initial model, using training images marked with anatomical structure labels and target block area labels, and training based on the joint loss function of overlapping area optimization loss and distributed focusing loss; the improved YOLOv8 model is obtained by replacing the CSP-Darknet53 backbone network of the original YOLOv8 model with the CSP-Darknet53++ backbone network.

[0009] According to a thoracic trunk nerve block auxiliary system provided by the present invention, the ultrasound image preprocessing module is also used to denoise the ultrasound image through a wavelet transform-convolutional neural network joint denoising model, and perform multi-scale enhancement on the denoised ultrasound image.

[0010] According to a thoracic trunk nerve block auxiliary system provided by the present invention, the ultrasound image preprocessing module is also used to decompose the ultrasound image through a two-stage cascade segmentation network model to obtain a two-dimensional segmentation result; and reconstruct the two-dimensional segmentation result into a three-dimensional network model through a marching cube algorithm.

[0011] According to a thoracic trunk nerve block auxiliary system provided by the present invention, the mobile terminal device includes a path planning module, wherein: The path planning module is used to generate a puncture starting point and a block target point based on the thoracic trunk nerve block scheme, the anatomical structure and the target block area; based on the puncture starting point and the block target point, a block path is generated; the block target point is the end point of the block path planning.

[0012] According to a thoracic trunk nerve block auxiliary system provided by the present invention, the path planning module is further used to generate candidate skin puncture points based on the block target point; determine whether the target cone interferes with non-interferable tissue, and if so, generate new candidate skin puncture points until the target cone no longer interferes with the non-interferable tissue; The target cone is a cone with a preset taper angle with a line connecting the blocking target point and the candidate skin puncture point as its center line.

[0013] According to a thoracic trunk nerve block auxiliary system provided by the present invention, the path planning module is also used to compare the coordinates of the electromagnetic positioning needle tip with the planned path, and use the Kalman filter algorithm to predict the angle deviation or depth deviation; when the predicted angle deviation or depth deviation meets the deviation threshold trigger condition, the augmented reality glasses device is triggered to superimpose a path correction arrow on the patient's real view image.

[0014] The thoracic trunk nerve block auxiliary system provided by the present invention uses a wireless ultrasound probe device to collect ultrasound images of the patient's thoracic trunk nerve block area, uses a mobile terminal device to generate an individualized nerve block plan based on the patient's multi-dimensional individual feature information, and analyzes the ultrasound image to determine the anatomical structure and target block area, thereby generating a precise block path. Finally, an augmented reality glasses device is used to superimpose and display the ultrasound image, anatomical structure, target block area and block path on the patient's real view, thereby achieving accurate and real-time visual guidance, improving the block success rate and reducing the difficulty of operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 It is a structural schematic diagram of the thoracic trunk nerve block auxiliary system provided by the present invention.

[0017] Figure 2 It is a structural diagram of the artificial intelligence system deployed in the mobile terminal device provided by the present invention.

[0018] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0020] It should be noted that traditional nerve block protocols rely heavily on the clinical experience of healthcare professionals, often using the same type, concentration, and volume of local anesthetic. This approach is highly subjective, lacking objectivity and accuracy, and is easily influenced by individual differences in experience and judgment, leading to inconsistent block effects.

[0021] In addition, in the existing technology, the puncture process usually requires the doctor to intermittently switch between the puncture area and the ultrasound screen. Frequent switching of the field of view will cause the puncture needle to deviate from the original route, which obviously creates greater difficulty in the operation process.

[0022] Safe and successful nerve blocks depend on placing the local anesthetic close enough to the nerve or plexus so that it can reach the nerve fibers, but not so close that the needle contacts the nerve and causes mechanical damage. Finding the "sweet spot" can be somewhat elusive; two-dimensional ultrasound struggles to accurately reflect three-dimensional anatomy, leading to discrepancies in depth and position judgment, and operator experience and insufficient ultrasound skills can lead to block failure. Therefore, accurately identifying the nerve block area and planning the block path are crucial.

[0023] Based on this, the present invention provides a thoracic trunk nerve block auxiliary system, which uses a wireless ultrasound probe device to collect ultrasound images of the patient's thoracic trunk nerve block area, uses a mobile terminal device to generate an individualized nerve block plan based on the patient's multi-dimensional individual feature information, and analyzes the ultrasound image to determine the anatomical structure and target block area, thereby generating a precise block path. Finally, an augmented reality glasses device is used to superimpose and display the ultrasound image, anatomical structure, target block area and block path on the patient's real view, thereby achieving accurate and real-time visual guidance, improving the block success rate and reducing the difficulty of operation.

[0024] The thoracic trunk nerve block auxiliary system of the embodiment of the present invention is as follows: Figure 1 As shown, it includes a wireless ultrasound probe device 100, a mobile terminal device 200 and an augmented reality glasses device 300 that are communicatively connected.

[0025] It should be understood that a wireless ultrasound probe device is used to acquire ultrasound images by transmitting ultrasound signals to a subject, receiving ultrasound echo signals from the subject, and converting the ultrasound echo signals into electrical signals. It can radiate ultrasound signals from the surface of the subject to a target portion within the subject and use the reflected ultrasound signals, i.e., ultrasound echo signals, to non-invasively acquire cross-sectional images of soft tissue or blood flow.

[0026] Specifically, the wireless ultrasound probe device in this embodiment is used to acquire ultrasound images of the patient's thoracic trunk nerve block area based on the patient's thoracic trunk nerve block plan.

[0027] The thoracic nerve block protocol is generated by the mobile terminal based on the patient's multi-dimensional individual characteristics. This protocol includes, but is not limited to, the block type, local anesthetic type, local anesthetic concentration, and volume. The patient's multi-dimensional individual characteristics include, but are not limited to, age, weight, comorbidities, preoperative assessment records, and imaging reports.

[0028] It should be understood that the mobile terminal device has wireless communication capabilities and can exchange data with external devices such as a wireless ultrasound probe device. It has an input unit and a display to facilitate user operation and information viewing.

[0029] In practical applications, a wireless ultrasound probe captures ultrasound images of the patient's thoracic trunk nerve block area. These images can more accurately locate the injection point for the nerve block, ensuring precise delivery of the drug around the target nerve, thereby improving the success rate and safety of the nerve block. The captured ultrasound images are transmitted to a mobile device or other display device via wireless communication technologies (such as Wi-Fi and Bluetooth), allowing users to view and analyze the images in real time.

[0030] Specifically, the mobile terminal device in this embodiment is used to determine the anatomical structure and target blocking area based on the ultrasound image; generate the blocking path based on the thoracic trunk nerve block scheme, anatomical structure and target blocking area; the target blocking area is the injection area of ​​the local anesthetic.

[0031] In one example, reference Figure 2 , Figure 2 FIG. 1 is a schematic diagram of the structure of the artificial intelligence system deployed in the mobile terminal device shown in this embodiment. Figure 2 As shown, the artificial intelligence system includes a user interaction interface, a power management module, an artificial intelligence language module, a thoracic nerve block model feature library module, an AI model training module, a thoracic nerve block anatomical ultrasound recognition module and a path planning module.

[0032] User interface: This includes touchscreen, voice control, gesture recognition, and other features. The touchscreen can display ultrasound images, AI analysis results, and puncture path planning, allowing users to operate and adjust the system. Voice control and gesture recognition further enhance the system's usability, allowing users to control various system functions through voice commands or gestures, reducing interruptions during operation. Furthermore, the user interface should include real-time feedback, allowing users to adjust the display based on their actions to ensure accurate and safe operation.

[0033] Power Management Module: Because the system involves multiple hardware components, the design of the power management module is crucial. This module must be able to provide stable power to the ultrasound equipment, AI computing unit, augmented reality device, sensor module, and more. We use a high-capacity lithium battery pack with fast charging capabilities to ensure continuous system operation during prolonged surgeries. Furthermore, the power management module features overload protection and battery monitoring to ensure system safety and reliability.

[0034] Artificial intelligence language module: Use nerve block related data to train the artificial intelligence language model and build an artificial intelligence language model library. Based on the artificial intelligence language model library, users can interact with patient information such as height, weight, surgical method, etc. in text form before and during the block to formulate the block plan.

[0035] Thoracic Nerve Block Model Feature Library Module: Ultrasound image data for thoracic nerve blocks (including serratus anterior plane block, erector spinae plane block, and thoracic paravertebral block) are collected from clinical settings. The LabelME annotation tool is used to label the various anatomical structures within the ultrasound image data, including the serratus anterior, erector spinae, and paravertebral structures and surrounding structures, such as the serratus anterior, ribs, intercostal muscles, erector spinae, transverse processes, pleura, and paravertebral space. The OpenCV image processing library is used to preprocess the ultrasound images of the blocked area through filtering and resampling to generate high-dimensional image arrays, providing high-quality data for model training. Using Python combined with the FP-Growth algorithm, the module mines ultrasound features of the blocked area using patient information, expert labels, and image arrays. This allows for accurate identification of image texture and other intrinsic features, thereby constructing a thoracic nerve block ultrasound feature library.

[0036] AI model training module: The YOLOv8 model was selected to identify anatomical landmarks in ultrasound images and finely segment anatomical ultrasound planes relevant for thoracic nerve block. Model training was performed using the PyTorch framework. The annotated dataset was split into training and validation sets in an 8:2 ratio. Cross-validation was used to evaluate model performance and optimize hyperparameters. Techniques such as transfer learning and data augmentation were used to improve the model's robustness and generalization across various ultrasound image scenarios.

[0037] Thoracic Nerve Block Anatomy Ultrasound Recognition Module and Path Planning Module: The thoracic nerve block anatomy ultrasound recognition module uses the YOLOv8 model, and the path planning module uses the Mask-RCNN model. The artificial intelligence model is trained using the PyTorch framework. The training data is divided into training and validation sets, and cross-validation is used to evaluate model performance and optimize hyperparameters. Furthermore, transfer learning and data augmentation techniques are used to improve the model's robustness and generalization capabilities in different ultrasound image scenarios.

[0038] In this embodiment, an artificial intelligence system deployed in a mobile terminal device determines the anatomical structure and target blocking area based on ultrasound images; and generates a blocking path based on the thoracic trunk nerve block plan, anatomical structure, and target blocking area.

[0039] Augmented reality glasses are wearable devices primarily used to overlay virtual information onto real-world scenes, providing users with an enhanced visual experience. Specifically, in this embodiment, the augmented reality glasses are used to overlay ultrasound images, anatomical structures, target blockage areas, and blockage pathways onto the patient's real-world view.

[0040] Specifically, augmented reality glasses use a variety of sensors (such as cameras, accelerometers, gyroscopes, and depth sensors) to perceive the user's position and environmental information in real time. This sensor data is fed into simultaneous localization and mapping algorithms to create a digital model of the real scene and track the user's perspective in real time. This enables virtual information to be accurately aligned with the real environment, ensuring that the superimposed image matches the patient's anatomy.

[0041] In practical applications, augmented reality glasses are used to project ultrasound images and AI analysis results into the user's field of view in real time, creating an "in-front-of-eye ultrasound" effect and providing an immersive user experience. The device features a high-resolution display and low-latency image transmission capabilities, ensuring that users can see clear ultrasound images and AI-annotated information in real time. Furthermore, it features head tracking, which adjusts the display based on the user's head movements to ensure image stability and accuracy.

[0042] The thoracic trunk nerve block auxiliary system of this embodiment uses a wireless ultrasound probe device to collect ultrasound images of the patient's thoracic trunk nerve block area, uses a mobile terminal device to generate an individualized nerve block plan based on the patient's multi-dimensional individual characteristic information, and analyzes the ultrasound image to determine the anatomical structure and target block area, thereby generating a precise block path. Finally, an augmented reality glasses device superimposes the ultrasound image, anatomical structure, target block area and block path on the patient's real view, thereby achieving accurate and real-time visual guidance, improving the block success rate and reducing the difficulty of operation.

[0043] It should be noted that each implementation method of the present application can be freely combined, the order can be changed, or it can be executed separately, and does not need to rely on or depend on a fixed execution order.

[0044] In some embodiments, the wireless ultrasound probe device is also used to acquire ultrasound images of the patient's thoracic trunk nerve block area through a composite imaging mode and an inter-frame interpolation algorithm based on a wireless ultrasound probe matched with the thoracic trunk nerve block scheme.

[0045] In this embodiment, a high-frequency linear array probe or a low-frequency convex array probe is selected according to the block depth in the thoracic trunk nerve block scheme, and the composite imaging mode is enabled to reduce beam artifacts. The original frame rate is increased to 60fps through the GPU-accelerated interframe interpolation algorithm to ensure image continuity under respiratory movement; the probe's built-in inertial sensor synchronously records the spatial posture, providing spatiotemporal alignment data for three-dimensional reconstruction, thereby improving the accuracy of nerve positioning and puncture path planning.

[0046] In some embodiments, the mobile terminal device includes a blocking plan formulation module, wherein: The block plan formulation module is used to extract key features from the patient's multi-dimensional individual feature information and generate multiple candidate thoracic trunk nerve block plans based on the key features and an evidence-based rule base; the evidence-based rule base is a rule base constructed based on the principles of evidence-based medicine; The block plan formulation module is further configured to select the optimal candidate thoracic trunk nerve block plan from the multiple candidate thoracic trunk nerve block plans as the patient's thoracic trunk nerve block plan.

[0047] Specifically, the evidence-based rule base is established based on a large amount of clinical research and practical experience, and includes the relationship between different patient characteristics and blockade regimens.

[0048] It should be understood that the model receives structured and unstructured multidimensional individual patient information, such as age, weight, comorbidities, preoperative evaluation records, and imaging reports, and extracts key features through the Transformer model. The model then generates multiple candidate thoracic trunk nerve block protocols based on the extracted key features and an evidence-based rule base.

[0049] After obtaining multiple candidate thoracic trunk nerve block plans, each candidate thoracic trunk nerve block plan is scored according to the preset evaluation criteria, and then the candidate thoracic trunk nerve block plan with the highest score is selected as the patient's thoracic trunk nerve block plan.

[0050] The mobile terminal device of this embodiment generates an individualized nerve block plan by extracting key features from the multi-dimensional individual feature information of the patient and combining it with an evidence-based rule base.

[0051] In some embodiments, the block plan formulation module is further configured to, when the recommended dose of local anesthetic in the current optimal candidate thoracic trunk nerve block plan exceeds a threshold, delete the current optimal candidate thoracic trunk nerve block plan, and select the optimal candidate thoracic trunk nerve block plan from the remaining candidate thoracic trunk nerve block plans as the patient's thoracic trunk nerve block plan; The block plan formulation module is further configured to output a warning message when the recommended dose of local anesthetic in all candidate thoracic trunk nerve block plans exceeds a threshold; the warning message is configured to prompt manual intervention.

[0052] In this embodiment, a threshold is pre-set to measure the effectiveness of the intelligently generated candidate thoracic nerve block plans. If the recommended local anesthetic dose in the currently optimal candidate thoracic nerve block plan exceeds the threshold, the currently optimal candidate thoracic nerve block plan is deleted, and the optimal candidate thoracic nerve block plan is selected from the remaining candidate thoracic nerve block plans as the patient's thoracic nerve block plan.

[0053] Furthermore, if the recommended doses of local anesthetics in all candidate thoracic trunk nerve block schemes exceed the threshold, an early warning message of "manual intervention recommended" is triggered.

[0054] The mobile terminal device of this embodiment further verifies the recommended dosage of local anesthetic in the thoracic trunk nerve block protocol to ensure the validity of the automatically generated thoracic trunk nerve block protocol.

[0055] In some embodiments, the mobile terminal device includes an ultrasound image preprocessing module and a thoracic nerve block anatomical ultrasound recognition module, wherein: The ultrasound image preprocessing module is used to preprocess the ultrasound image to obtain the preprocessed ultrasound image; The thoracic nerve block anatomical ultrasound recognition module is used to input the pre-processed ultrasound image into the ultrasound image deep learning model to determine the anatomical structure and target block area; The ultrasound image deep learning model is based on an improved YOLOv8 model as the initial model, using training images labeled with anatomical structure labels and target block area labels, and is trained based on a joint loss function of overlapping area optimization loss and distributed focusing loss. The improved YOLOv8 model is obtained by replacing the CSP-Darknet53 backbone network of the original YOLOv8 model with the CSP-Darknet53++ backbone network.

[0056] The CSP-Darknet53++ backbone network is an improved version of the YOLOv8-based CSP-Darknet53 backbone network, optimized for detecting small objects (such as nerves and blood vessels) in medical ultrasound images. Its core concept is to improve feature extraction capabilities while maintaining real-time performance through cross-stage partial connections (CSP) and an attention mechanism.

[0057] In this embodiment, the ultrasound image is preprocessed, such as through noise suppression, contrast enhancement, edge sharpening, and normalization, to produce a preprocessed ultrasound image. This preprocessed ultrasound image is then fed into a pretrained ultrasound image deep learning model. This model performs real-time multi-target detection and localization, marking the coordinates and bounding boxes of key tissues such as nerves, blood vessels, and muscles, and generating the 3D spatial coordinates of the local anesthetic injection area (i.e., the target block area).

[0058] Specifically, the CSP-Darknet53++ backbone network is used to extract multi-scale features, and the multi-scale contextual information is fused through the Spatial Pyramid Pooling - Fast (SPPF) module and the Cross Stage Partial Concatenation (CSPC) module to accurately identify the thoracic transverse processes (hyperechoic arc-shaped shadows), intercostal nerves (hypoechoic bundle structures) and blood vessels (pulsating anechoic areas).

[0059] In addition, in view of the small target characteristics of ultrasound images, the ultrasound image deep learning model also introduces a decoupled detection head to separate classification and regression tasks, and adopts a dynamic label allocation strategy to improve the blood vessel-nerve differentiation, accurately mark the coordinates and bounding boxes of key tissues such as nerves, blood vessels, and muscles, and generate the 3D spatial coordinates of the local anesthetic injection area (i.e., the target block area).

[0060] The ultrasound image deep learning model was trained using an improved YOLOv8 model as the initial model, using training images labeled with anatomical structures and target block regions. During training, a joint loss function combining overlap-based CIoU loss and distribution-focused DFL loss was used for iterative optimization. The specific joint loss function is as follows: ; ; in, L CIoU To improve the Complete-IoU loss, we measure the overlap, center distance, and aspect ratio consistency between the predicted box and the real box. ρ is the Euclidean distance, c is the minimum bounding box diagonal length, v Measure aspect ratio similarity, L DFL is the joint distribution of optimized classification confidence and bounding box regression, L angle It is the angle sensitivity loss, which is used to constrain the accuracy of blood vessel / nerve direction prediction.

[0061] The mobile terminal device of this embodiment analyzes the preprocessed ultrasound image based on the ultrasound image deep learning model to accurately and automatically identify the anatomical structure and target blockage area.

[0062] In some embodiments, the ultrasound image preprocessing module is further used to denoise the ultrasound image through a wavelet transform-convolutional neural network joint denoising model, and perform multi-scale enhancement on the denoised ultrasound image.

[0063] It should be noted that direct transmission of images will generate noise, so in this embodiment, a wavelet transform-convolutional neural network (WT-CNN) joint denoising model is used to denoise the ultrasound image.

[0064] Specifically, the image is first decomposed into sub-bands of different frequencies (high-frequency details + low-frequency background) using a wavelet transform. Thresholding is then used to suppress noise (such as speckle noise) while preserving the edges of key tissue structures. A convolutional neural network is then used to perform texture restoration on the low-frequency sub-bands, eliminating speckle noise while preserving the sharpness of neural edges.

[0065] Furthermore, this embodiment also uses multi-scale enhancement, such as a non-uniform illumination correction algorithm based on the Retinex theory, to perform local gamma correction on low-contrast areas (such as deep intercostal nerves). The formula is: ; in, is the original pixel intensity (grayscale value) of the input ultrasound image at coordinate (x, y); For The background light estimation value after Gaussian blurring, γ controls its enhancement intensity.

[0066] In this embodiment, by optimizing the quality of the original ultrasound image, clear and high-contrast input data is provided for subsequent segmentation and modeling.

[0067] In some embodiments, the ultrasound image preprocessing module is further used to decompose the ultrasound image through a two-stage cascade segmentation network model to obtain a two-dimensional segmentation result; and reconstruct the two-dimensional segmentation result into a three-dimensional network model through a marching cube algorithm.

[0068] Here, the two-level cascade segmentation network model includes a first-level YOLOv8 network layer for coarse segmentation of large-scale landmarks such as the thoracic transverse processes and pleural lines, and a second-level attention mechanism network layer (CBAM module) for precise positioning of intercostal nerves and blood vessels.

[0069] Furthermore, in this embodiment, the Dice loss function is used to optimize the small target segmentation in the two-stage cascade segmentation network model, and the formula is: ; in, is the true label, is the predicted probability.

[0070] After decomposing the ultrasound image through a two-stage cascade segmentation network model to obtain a two-dimensional segmentation result, the two-dimensional segmentation result is reconstructed into a three-dimensional network model through a marching cube algorithm.

[0071] In this embodiment, a two-stage cascade segmentation network is used to achieve accurate segmentation of multi-scale targets, and a three-dimensional model is constructed to assist path planning.

[0072] In some embodiments, the mobile terminal device includes a route planning module, wherein: The path planning module is used to generate a puncture starting point and a block target point based on the thoracic trunk nerve block scheme, the anatomical structure and the target block area; based on the puncture starting point and the block target point, a block path is generated; the block target point is the end point of the block path planning.

[0073] In this embodiment, the optimal puncture path is automatically generated through a three-dimensional spatial algorithm based on the anatomical structures (such as nerves and blood vessels) and target blocking areas identified by AI.

[0074] Specifically, the skin needle entry point (puncture starting point) and the center point of drug injection (blocking target point) are first calculated, and then a hierarchical path planning strategy is adopted, such as using The algorithm avoids critical organs and dynamically adjusts by fusing ultrasound and electromagnetic navigation data in real time. It combines multiple constraints to output multiple candidate paths with safety scores (0-100 points), and finally selects the candidate path with the highest safety score as the blocking path.

[0075] In some embodiments, the path planning module is further configured to generate candidate skin puncture points based on the blocked target point; determine whether the target cone interferes with the non-interferable tissue, and if so, generate new candidate skin puncture points until the target cone no longer interferes with the non-interferable tissue; The target cone is a cone with a preset taper angle with a line connecting the blocking target point and the candidate skin puncture point as its center line.

[0076] Here, tissues that must not be interfered with include high-risk anatomical areas such as blood vessels (diameter > 1 mm), pleura, and bony structures.

[0077] In this embodiment, a cone with a preset taper angle (e.g., 15°) is first constructed with the line connecting the blocking target point and the candidate skin puncture point as the center line. A multimodal detection algorithm is used to automatically verify the spatial interference relationship between the cone and dangerous tissues such as blood vessels / pleura. If interference exists, the puncture point position is dynamically optimized until the zero interference condition in the cone space is met.

[0078] In this embodiment, automatic safety verification of the puncture path is achieved through cone space modeling, which improves the efficiency of puncture point planning and reduces the puncture risk.

[0079] In some embodiments, the path planning module is also used to compare the coordinates of the electromagnetic positioning needle tip with the planned path, and use the Kalman filter algorithm to predict the angle deviation or depth deviation; when the predicted angle deviation or depth deviation meets the deviation threshold trigger condition, the augmented reality glasses device is triggered to superimpose a path correction arrow on the patient's real view image.

[0080] In this embodiment, the coordinates of the electromagnetic positioning needle tip are compared with the planned path in real time, and the Kalman filter algorithm is used to predict the deviation trend. If the angle deviation is greater than 5° or the depth error is greater than 2mm, a correction arrow is immediately projected through the augmented reality glasses device. Here, the deviation threshold trigger condition is: or ; in, 、 、 are the deviation values ​​between the real-time coordinates of the electromagnetic needle tip and the planned path in the three-dimensional direction; d th is the maximum allowable position error, is the current angle of the needle predicted by Kalman filtering; The theoretical needle insertion angle for the planned path; θ th is the maximum allowable angular deviation.

[0081] Here, the path correction arrow can be determined based on angle deviation or depth deviation. For example, if the needle tip deviates 2.5mm to the right due to tissue resistance, the augmented reality glasses will immediately project a red arrow pointing leftwards and display "2.5mm right deviation" on the patient's right intercostal space, guiding the user to make adjustments in the opposite direction until the arrow turns green and stabilizes.

[0082] In this embodiment, high-precision dynamic monitoring is achieved by dynamically tracking the coordinates of the electromagnetic positioning needle tip and the planned path, and AR visual intelligent warning is achieved by projecting 3D dynamic arrows through an augmented reality glasses device.

[0083] Figure 3 An example of a physical structure diagram of an electronic device is shown below. Figure 3 As shown, the electronic device includes: a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call the logic instructions in the memory 330 to execute a thoracic trunk nerve block assistance method based on the thoracic trunk nerve block assistance system, the method comprising: Based on the patient's thoracic trunk nerve block plan, collecting an ultrasound image of the patient's thoracic trunk nerve block area; the thoracic trunk nerve block plan is generated by the mobile terminal device based on the patient's multi-dimensional individual feature information; Based on the ultrasound image, determining the anatomical structure and the target block area; generating a block path based on the thoracic trunk nerve block plan, the anatomical structure and the target block area; the target block area is the injection area of ​​the local anesthetic; The ultrasound image, the anatomical structure, the target blocking area, and the blocking path are superimposed and displayed on a real view image of the patient.

[0084] Furthermore, the logic instructions in the aforementioned memory 330 can be implemented as software functional units and stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, is embodied in the form of a software product. This computer software product is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes any medium capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0085] In another aspect, the present invention further provides a computer program product, comprising a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can perform a thoracic trunk nerve block assistance method based on a thoracic trunk nerve block assistance system, the method comprising: Based on the patient's thoracic trunk nerve block plan, collecting an ultrasound image of the patient's thoracic trunk nerve block area; the thoracic trunk nerve block plan is generated by the mobile terminal device based on the patient's multi-dimensional individual feature information; Based on the ultrasound image, determining the anatomical structure and the target block area; generating a block path based on the thoracic trunk nerve block plan, the anatomical structure and the target block area; the target block area is the injection area of ​​the local anesthetic; The ultrasound image, the anatomical structure, the target blocking area, and the blocking path are superimposed and displayed on a real view image of the patient.

[0086] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a thoracic trunk nerve block assistance method based on a thoracic trunk nerve block assistance system is implemented. The method comprises: Based on the patient's thoracic trunk nerve block plan, collecting an ultrasound image of the patient's thoracic trunk nerve block area; the thoracic trunk nerve block plan is generated by the mobile terminal device based on the patient's multi-dimensional individual feature information; Based on the ultrasound image, determining the anatomical structure and the target block area; generating a block path based on the thoracic trunk nerve block plan, the anatomical structure and the target block area; the target block area is the injection area of ​​the local anesthetic; The ultrasound image, the anatomical structure, the target blocking area, and the blocking path are superimposed and displayed on a real view image of the patient.

[0087] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the devices may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0088] Through the description of the above embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or alternatively, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes instructions for enabling a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or portions thereof.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in each of the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A thoracic trunk nerve block auxiliary system, characterized in that include: A wireless ultrasound probe device, a mobile terminal device, and an augmented reality glasses device connected in communication; wherein: The wireless ultrasound probe device is used to acquire an ultrasound image of the patient's thoracic trunk nerve block area based on the patient's thoracic trunk nerve block plan; the thoracic trunk nerve block plan is generated by the mobile terminal device based on the patient's multi-dimensional individual feature information; The mobile terminal device is used to determine the anatomical structure and the target block area based on the ultrasound image; generating a block path based on the thoracic trunk nerve block scheme, the anatomical structure, and the target block area; the target block area is the injection area of ​​the local anesthetic; The augmented reality glasses device is used to superimpose and display the ultrasound image, the anatomical structure, the target blocking area, and the blocking path on the patient's real view image.

2. The thoracic trunk nerve block auxiliary system according to claim 1, characterized in that: The wireless ultrasound probe device is also used to collect ultrasound images of the patient's thoracic trunk nerve block area through a composite imaging mode and an inter-frame interpolation algorithm based on a wireless ultrasound probe matched with the thoracic trunk nerve block scheme.

3. The thoracic trunk nerve block auxiliary system according to claim 1, characterized in that: The mobile terminal device includes a blocking plan formulation module, wherein: The block plan formulation module is used to extract key features from the patient's multi-dimensional individual feature information and generate multiple candidate thoracic trunk nerve block plans based on the key features and an evidence-based rule base; the evidence-based rule base is a rule base constructed based on the principles of evidence-based medicine; The block plan formulation module is further configured to select the optimal candidate thoracic trunk nerve block plan from the multiple candidate thoracic trunk nerve block plans as the patient's thoracic trunk nerve block plan.

4. The thoracic trunk nerve block auxiliary system according to claim 3, characterized in that: The block plan formulation module is further configured to, when the recommended dose of local anesthetic in the current optimal candidate thoracic trunk nerve block plan exceeds a threshold, delete the current optimal candidate thoracic trunk nerve block plan, and select the optimal candidate thoracic trunk nerve block plan from the remaining candidate thoracic trunk nerve block plans as the patient's thoracic trunk nerve block plan; The block plan formulation module is further configured to output a warning message when the recommended dose of local anesthetic in all candidate thoracic trunk nerve block plans exceeds a threshold; the warning message is configured to prompt manual intervention.

5. The thoracic trunk nerve block auxiliary system according to claim 1, characterized in that: The mobile terminal device includes an ultrasound image preprocessing module and a thoracic nerve block anatomical ultrasound recognition module, wherein: The ultrasound image preprocessing module is used to preprocess the ultrasound image to obtain the preprocessed ultrasound image; The thoracic nerve block anatomical ultrasound recognition module is used to input the pre-processed ultrasound image into the ultrasound image deep learning model to determine the anatomical structure and target block area; Among them, the ultrasound image deep learning model is obtained by using the improved YOLOv8 model as the initial model, using training images marked with anatomical structure labels and target block area labels, and training based on the joint loss function of overlapping area optimization loss and distributed focusing loss; the improved YOLOv8 model is obtained by replacing the CSP-Darknet53 backbone network of the original YOLOv8 model with the CSP-Darknet53++ backbone network.

6. The thoracic trunk nerve block auxiliary system according to claim 5, characterized in that: The ultrasonic image preprocessing module is further used to denoise the ultrasonic image through a wavelet transform-convolutional neural network joint denoising model, and perform multi-scale enhancement on the denoised ultrasonic image.

7. The thoracic trunk nerve block auxiliary system according to claim 5, characterized in that: The ultrasound image preprocessing module is further used to decompose the ultrasound image through a two-stage cascade segmentation network model to obtain a two-dimensional segmentation result; and reconstruct the two-dimensional segmentation result into a three-dimensional network model through a marching cube algorithm.

8. The thoracic trunk nerve block auxiliary system according to claim 1, characterized in that: The mobile terminal device includes a path planning module, wherein: The path planning module is used to generate a puncture starting point and a block target point based on the thoracic trunk nerve block scheme, the anatomical structure and the target block area; based on the puncture starting point and the block target point, a block path is generated; the block target point is the end point of the block path planning.

9. The thoracic trunk nerve block auxiliary system according to claim 8, characterized in that: The path planning module is further configured to generate candidate skin puncture points based on the blocked target point; determine whether the target cone interferes with the non-interferable tissue, and if so, generate new candidate skin puncture points until the target cone no longer interferes with the non-interferable tissue; The target cone is a cone with a preset taper angle with a line connecting the blocking target point and the candidate skin puncture point as its center line.

10. The thoracic trunk nerve block auxiliary system according to claim 8, characterized in that: The path planning module is also used to compare the coordinates of the electromagnetic positioning needle tip with the planned path, and use the Kalman filter algorithm to predict the angle deviation or depth deviation; when the predicted angle deviation or depth deviation meets the deviation threshold trigger condition, the augmented reality glasses device is triggered to superimpose a path correction arrow on the patient's real view image.

Images