Anesthesia puncture auxiliary positioning method

Abstract

The application discloses an anesthetic puncture auxiliary positioning method and relates to the technical field of medical image processing. The method comprises the following steps: acquiring multi-modal image data of an anesthetic region, performing image synchronization and fusion on the multi-modal image data; dividing the fused multi-modal comprehensive image into a plurality of image blocks according to a preset number, and calculating the gray dynamic change degree of each image block; removing noise and enhancing features of each image block based on the gray dynamic change degree of each image block; generating an accurate three-dimensional model of a spinal canal region by using the processed multi-modal image data through volume rendering, and displaying the three-dimensional model in real time through an augmented reality device; labeling important anatomical structures on the generated three-dimensional model, and calculating an optimal puncture path that avoids key structures in the three-dimensional space by using an A* algorithm. The application improves the accuracy and operability of anesthetic puncture auxiliary positioning through multi-modal image fusion, three-dimensional reconstruction and path planning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical image processing, in particular to an anesthetic puncture auxiliary positioning method. BACKGROUND

[0002] Anesthetic puncture technology is an indispensable important link in modern medicine, and its accuracy and safety are directly related to the success rate of surgery and the rehabilitation of patients. The traditional anesthetic puncture positioning method mainly relies on the experience of doctors and single modal image guidance, such as ultrasound or X-ray guidance. However, these methods have certain limitations in identifying complex anatomical structures and accurate positioning. In recent years, multi-modal image fusion technology can present anatomical structure information more comprehensively by integrating ultrasound, magnetic resonance, CT and other multi-modal image data, and improve the accuracy and safety of puncture, but there are still some problems to be solved.

[0003] Firstly, the synchronization and fusion of multi-modal image data still face technical challenges, and the spatio-temporal registration accuracy between different modalities needs to be improved. Secondly, the existing image processing methods are not ideal in noise removal and feature enhancement, and it is difficult to fully extract and utilize the key information in multi-modal images. Thirdly, the application of three-dimensional visualization technology in anesthetic puncture auxiliary positioning is not mature enough, and it is difficult to realize accurate three-dimensional reconstruction and real-time display of anatomical structures. Finally, the existing puncture path planning algorithm often only considers two-dimensional planes, and lacks comprehensive consideration of complex anatomical structures in three-dimensional space, resulting in suboptimal selection of puncture paths. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an anesthetic puncture auxiliary positioning method to solve the problems of multi-modal medical image fusion, three-dimensional visualization reconstruction and real-time display, and intelligent puncture path planning in anesthetic puncture auxiliary positioning.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present application provides an anesthetic puncture auxiliary positioning method, which comprises: acquiring multi-modal image data of an anesthetic region; performing image synchronization and fusion on the multi-modal image data to generate multi-modal comprehensive images; dividing the fused multi-modal comprehensive images into a plurality of image blocks according to a preset number, and calculating the gray dynamic change degree of each image block; performing noise removal and feature enhancement processing on each image block based on the gray dynamic change degree of each image block; generating an accurate three-dimensional model of a spinal canal region by volume rendering method using the processed multi-modal image data, and displaying the three-dimensional model in real time through an augmented reality device; and labeling important anatomical structures on the generated three-dimensional model, and calculating an optimal puncture path that avoids key structures in the three-dimensional space by using an A* algorithm.

[0008] As a preferred scheme of the anesthetic puncture auxiliary positioning method, the acquisition of the multi-modal image data of the anesthetic region comprises the following steps:

[0009] The ultrasound probe, the near-infrared light sensor and the elastography device are used to collect ultrasound images, near-infrared light images and elastography data of a target puncture region of the waist of the patient.

[0010] As a preferred scheme of the anesthetic puncture auxiliary positioning method, the image synchronization and fusion on the multi-modal image data to generate multi-modal comprehensive images comprises the following steps:

[0011] A unified time reference is used to generate a timestamp, and an accurate timestamp is automatically embedded when the image data of each modality is collected;

[0012] An image processing software is used to extract timestamp information from the metadata of each modality image;

[0013] The time difference between different modal images is calculated, a fixed delay is identified and recorded, and the time sequence of each modality image is adjusted by interpolation and cropping methods;

[0014] The SIFT algorithm is applied to the ultrasound images, the NIR images and the elastography images to detect key feature points, and the coordinates and descriptors of each feature point are recorded;

[0015] The FLANN matcher is initialized, the index parameters and search parameters are set, the feature point matching is performed on the NIR-ultrasound and elastography-ultrasound respectively, the potential matching pairs are generated, and the ratio test is used to filter the false matches;

[0016] The coordinate set of the matched point pair is extracted, the affine transformation matrix of the NIR-ultrasound and the elastic imaging-ultrasound is calculated respectively, the affine transformation matrix obtained is used for spatial transformation of the NIR image and the elastic imaging image, and the NIR image and the elastic imaging image which are spatially aligned with the ultrasound image are generated;

[0017] The aligned ultrasound image, NIR image and elastic imaging data are normalized to unify the gray scale range, and the normalized multi-modal image data is input into the pre-trained fusion U-Net model for forward propagation to generate a fused multi-modal comprehensive image.

[0018] As a preferred scheme of the anesthetic puncture auxiliary positioning method, the fused multi-modal comprehensive image is evenly divided into a plurality of image blocks according to a preset number, and the gray dynamic change degree of each image block is calculated, including the following steps,

[0019] The number and size of the image blocks are set, a fixed grid method is used, starting from the top left corner of the image, the image is divided row by row, and whether each image block covers different anatomical regions is verified;

[0020] Before the puncture operation starts, at least two ultrasound images are collected as an initial image group, and the baseline gray scale distribution is recorded;

[0021] During the actual puncture process, ultrasound images are collected in real time at fixed time intervals, and new images are sequentially added to the image group to form a time series image data set;

[0022] For each image block, the gray scale distribution sequence is extracted from the time series image data set, and the mean gray scale sequence in the baseline stage is calculated;

[0023] The DTW distance between the gray scale distribution sequence of each image block at each time and the mean sequence is calculated by using the dtaidistance library in Python to realize the DTW algorithm;

[0024] The gray dynamic change degree of the image block is calculated in combination with the gray scale fluctuation index and the DTW distance.

[0025] As a preferred scheme of the anesthetic puncture auxiliary positioning method, the gray dynamic change degree of each image block is calculated, and the noise removal and feature enhancement of each image block are performed based on the gray dynamic change degree of each image block, including the following steps,

[0026] For each pair of image blocks, the difference in the gray dynamic change degree is calculated;

[0027] The similarity weight between the image blocks is calculated, and a similarity weight matrix between the image blocks is constructed;

[0028] The size of the neighborhood range is defined, for each pixel point, the image block in which it is located is positioned, and the image block with high similarity is selected in the similarity weight matrix;

[0029] The weights of all neighborhood image blocks are accumulated to obtain a normalization factor, the pixel value after denoising is calculated, and the denoising processing of the whole image is completed to obtain the filtered denoised multi-modal image;

[0030] The denoised multi-modal image is normalized, and the normalized multi-modal image is input into the pre-trained fusion U-Net model for forward propagation to generate the enhanced multi-modal image.

[0031] As a preferred scheme of the anesthesia puncture auxiliary positioning method, wherein: the denoised multi-modal image data is used to generate an accurate three-dimensional model of the spinal canal region by volume rendering method, and the three-dimensional model is displayed in real time by the augmented reality device, including the following steps,

[0032] According to the depth of the puncture region and the organ structure, the three-dimensional resolution of the voxel is set, and the linear interpolation method is used to fill the blank area between the two-dimensional images;

[0033] The voxel data is input into the Marching Cubes algorithm, the threshold is set, and it is judged whether the vertex of each small cube in the voxel data is above or below the threshold;

[0034] According to the state of the cube, the table is looked up to determine the triangle combination to be generated, the accurate position of the polygon vertex is calculated by linear interpolation, and the grid structure is formed;

[0035] The Quadratic Decimation method is used to optimize the grid structure by reducing the number of polygons, and the Laplacian smoothing algorithm is applied to reduce the grid noise and smooth the surface curve;

[0036] Microsoft HoloLens 2 is selected as the AR device, optical tracking technology is used, multiple LED markers are installed on the patient's waist and the augmented reality device, and a high frame rate camera is configured for real-time marker recognition and position calculation;

[0037] The optical tracking system is calibrated initially, and the spatial relationship between the AR device and the actual anatomical structure is corrected;

[0038] The optimized three-dimensional surface grid model is imported into the visualization software of the AR device, and the position and direction of the three-dimensional model are adjusted according to the spatial transformation matrix obtained by calibration;

[0039] The data of the optical tracking system is continuously received, and the position and direction of the three-dimensional model are adjusted in real time.

[0040] As a preferred scheme of the anesthesia puncture auxiliary positioning method, wherein: on the generated three-dimensional model, important anatomical structures are marked, and an A* algorithm is used to calculate an optimal puncture path in three-dimensional space to avoid key structures, including the following steps,

[0041] The fused multi-modal image data is processed using an image segmentation algorithm to automatically identify and segment the key anatomical structures, and the automatic segmentation result is manually corrected.

[0042] The three-dimensional space of the puncture area is discretized into grid nodes, and each node represents a point in space.

[0043] A 26-neighbor connection method is used to define the connection relationship between nodes.

[0044] The Euclidean distance is selected as the heuristic function to estimate the straight-line distance from the current node to the target node.

[0045] The actual cost and heuristic cost are combined to calculate the total cost of each node in the A* algorithm.

[0046] The starting node is added to the open list, and the node with the smallest total cost is selected from the open list , removed from the open list and added to the closed list.

[0047] All neighbor nodes of the node are traversed.

[0048] If the node is already in the closed list, it is skipped; if the node is not in the open list, it is added to the open list and its parent node is recorded as ; if the node is already in the open list but the new actual cost is smaller, its total cost and parent node are updated.

[0049] Repeat step until the target node is selected from the open list and the path search is complete; if the open list is empty, an effective path cannot be found.

[0050] Starting from the target node, the parent nodes are traced back step by step until the starting node is reached, forming the entire path, and the sequence of nodes obtained by backtracking is reversed to obtain the complete puncture path from the starting point to the target point.

[0051] In a second aspect, the embodiments of the present application provide a computer device comprising a memory and a processor, the memory storing a computer program, wherein: the computer program is executed by the processor to implement any step of the anesthesia puncture auxiliary positioning method according to the first aspect of the present application.

[0052] In a third aspect, an embodiment of the present application provides a computer readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements any step of the method for assisting positioning of anesthetic puncture according to the first aspect of the present application.

[0053] The present application has the following beneficial effects: the multi-modal image data of the region to be anesthetized is acquired, and the multi-modal image data is synchronized and fused to generate a multi-modal comprehensive image, thereby improving the comprehensiveness and accuracy of the image data; the fused multi-modal comprehensive image is evenly divided into a plurality of image blocks, and the degree of gray dynamic change of each image block is calculated, thereby effectively capturing the feature changes of different anatomical regions; the image blocks are subjected to targeted noise removal and feature enhancement processing, thereby effectively retaining the key anatomical structure information and enhancing the image readability and interpretability; an accurate three-dimensional model is generated by a volume rendering method, and is displayed in real time in combination with augmented reality technology, thereby realizing intuitive presentation of complex anatomical structures, important anatomical structures are labeled on the three-dimensional model, and an A* algorithm is used to calculate an optimal puncture path, thereby improving the accuracy and safety of anesthetic puncture. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0055] Figure 1 The flowchart of the method for assisting positioning of anesthetic puncture in embodiment 1.

[0056] Figure 2 The flowchart of the A* algorithm in embodiment 1. DETAILED DESCRIPTION

[0057] In order to make the above-mentioned objects, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0058] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0059] Second, the "one embodiment" or "an embodiment" appearing in the specification herein refers to a specific feature, structure, or characteristic in at least one implementation of the application. The "in one embodiment" appearing in different places in the specification herein does not refer to the same embodiment, nor is it mutually exclusive with other embodiments.

[0060] Embodiment 1, Reference Figure 1 and Figure 2 As a first embodiment of the present application, the embodiment provides an auxiliary positioning method for anesthetic puncture, comprising the following steps:

[0061] S1, acquiring multi-modal image data of an anesthetic region;

[0062] Specifically, a high-resolution real-time ultrasound probe with a model of XYZ-5000 is selected, and the working frequency is set to 5 MHz to ensure sufficient penetration depth (about 5 cm) and high spatial resolution (less than 0.5 mm), with real-time imaging capability and high frame rate (≥60 frames / s), supporting multi-planar imaging, which is convenient for real-time observation of the dynamic position of the puncture needle; an NIR-800 near-infrared light reflection sensor is selected, with a working wavelength range of 700-900 nm, high spectral resolution and low noise characteristics, which can enhance the visibility of subcutaneous blood vessels and tissues, and provide rich vascular structure information in combination with ultrasound images; an ElasticPro-300 elastic imaging module is used, with a compression force range set between 20-50 grams, and a resolution of 0.2 mm, which is suitable for high-precision acquisition of soft tissue elastic characteristics, and can display the elastic distribution of the tissue in real time, assisting in identifying the hardness change of the tissue in the puncture region;

[0063] All modules (ultrasound probe, near-infrared light sensor, and elastic imaging module) are interconnected through a USB 3.0 interface to ensure high-speed data transmission, and a unified control software platform (such as MedFusion Control Suite) is used to realize unified control and synchronous operation of the various modal devices;

[0064] The main controller (such as a high-performance computer or an embedded system) integrates various modal devices, sets a unified time reference, ensures time synchronization of different modal image data, uses hardware trigger signals (such as TTL signals) to synchronize image acquisition of each device, and avoids data misplacement caused by time deviation;

[0065] S1、Calibration and Synchronization of the Integrated Device, including the following steps,

[0066] S1.1、Select the L4-L5 intervertebral space in the patient's lumbar region as the target puncture area, as it has clear anatomical structure and a relatively safe needle movement path. Use ultrasound guidance to preliminarily confirm the accurate position of the puncture point and avoid serious positions of blood vessels and nerves.

[0067] S1.2、Use special fixing devices such as fixing plates and clamps to securely fix the integrated device at the predetermined position on the patient's lumbar region. Adjust the position and angle of the fixing device to ensure that the ultrasound probe, NIR sensor, and elastography module have a field of view covering the target puncture area. Avoid device displacement during the puncture process. Check the stability of the device fixation by slightly moving the patient's body position to verify the reliability of the fixing device and ensure the stability of the device during the entire puncture process.

[0068] S1.3、Start all modules of the integrated device, including the ultrasound probe, NIR sensor, and elastography module. Confirm that each module is in good working condition. Check the real-time display of image data through the control software. Select the "synchronous acquisition" mode in the control software to ensure that image data from different modalities is acquired synchronously according to a unified time reference. Start acquisition and record the timestamp of each image to ensure accurate synchronization of data in time and space.

[0069] S2、Image Synchronization and Fusion of Multimodal Image Data, including the following steps,

[0070] S2.1、Use a unified time reference to generate timestamps to ensure that image data from each modality in the multimodal device configuration is automatically embedded with precise timestamps during acquisition. Use image processing software such as OpenCV or MATLAB to read the metadata fields of image files from each modality image. Extract timestamps and convert them into a comparable time series format for subsequent synchronization processing. Use the YYYY-MM-DD HH:MM:SS:MS format to ensure millisecond-level accuracy to meet real-time synchronization requirements.

[0071] Calculate the time difference between different modal images, identify and record the fixed delay (such as trigger signal delay), adjust the time sequence of each modal image by interpolation and cropping method, so that the image data is strictly aligned in time; use visualization tools (such as Matplotlib or VTK) to display the images of the same time point of different modalities by superimposition, check the alignment effect of different modal images in time, and confirm that the key anatomical structures and puncture needle positions are consistent in each modal image.

[0072] S2.2, SIFT (Scale Invariant Feature Transform) algorithm is used for feature point detection, because of its invariance in scale and rotation, it is suitable for complex registration requirements of multi-modal images;

[0073] SIFT algorithm is used to detect key feature points for ultrasound images and NIR images, and cv2.SIFT_create() function in OpenCV is used to initialize SIFT detector, detectAndCompute method is called for each modal image to extract feature points and descriptors, and the coordinates and descriptors of each feature point are recorded for subsequent matching;

[0074] FLANN (Fast Library for Approximate Nearest Neighbors) matching algorithm is used because of its high efficiency and accuracy in handling a large number of feature points; initialize the FLANN matcher, set the index parameters and search parameters, and perform feature point matching for NIR-ultrasound and elastography-ultrasound respectively, generate potential matching pairs, use Lowe's ratio test to filter false matches, and keep reliable feature point pairs;

[0075] Extract the coordinate set of the matching point pairs, calculate the affine transformation matrix of NIR-ultrasound and elastography-ultrasound respectively, use the cv2.estimateAffine2D() function of OpenCV to calculate the best affine transformation matrix, and use the calculated affine transformation matrix to perform spatial transformation on the NIR image and elastography image, so that they are aligned with the ultrasound image.

[0076] S2.3, a pre-trained fusion U-Net model is used, which is based on the classic U-Net architecture and increases the multi-modal data input channel, suitable for the fusion of ultrasound, near-infrared light and elastography data;

[0077] The input layer is a three-channel input corresponding to ultrasound, NIR and elastography data. Each layer includes convolution, batch normalization (BN), activation function (ReLU) and pooling operation. High-resolution features and low-resolution features are fused by using up-sampling deconvolution and skip connections to generate a high-quality integrated image. A sigmoid activation function is used to ensure that the gray value of the output image is within the range of 0 and 1.

[0078] The aligned ultrasound image, NIR image and elastography data are normalized to unify the gray scale range. The normalized multi-modal image data is input into the pre-trained fusion U-Net model for forward propagation to generate a fused multi-modal integrated image.

[0079] S3, the fused multi-modal integrated image is evenly divided into a plurality of image blocks according to a preset number, and the gray scale dynamic change degree of each image block is calculated, including the following steps,

[0080] S3.1, set the number of image blocks to 50 to ensure that each image block covers approximately the same anatomical region. The image block size is set to 100x100 pixels. The image resolution and clinical requirements are adjusted. A fixed grid method is used to start from the top left corner of the image and divide it row by row to ensure that each image block does not overlap and completely covers the target puncture area.

[0081] Verify whether each image block covers different anatomical regions to avoid excessive or insufficient coverage of a certain area and ensure the comprehensiveness of dynamic change analysis. Visualize the divided image blocks using tools such as Matplotlib to confirm the reasonableness of each coverage area. Calculate the proportion of tissue types covered by each block to ensure uniform distribution of multi-modal image information among the image blocks.

[0082] S3.2, before the puncture operation starts, collect at least two ultrasound images as an initial image group to record the baseline gray scale distribution for subsequent dynamic change analysis. First, ensure that the device is working properly and the sensor data is stable. Second, collect two ultrasound images at fixed time intervals (e.g., every 1 second) and store the image data in an image database with associated timestamps and image block numbers. Finally, preprocess the collected baseline images (e.g., denoising, enhancement) to ensure the accuracy of the gray scale distribution.

[0083] During the actual puncture process, the ultrasound images are collected in real time at fixed time intervals, and the new images are sequentially added to the image group to form a time sequence image dataset. Specifically, the puncture operation starts, the real-time image acquisition module is started, the ultrasound probe and the puncture needle are kept synchronous movement to avoid image blur, a timer (such as the time module in Python) is used, the image acquisition is triggered every 0.5 seconds, the collected image blocks are sequentially added to the corresponding image block group to form a time sequence image dataset.

[0084] S3.3, to accurately identify the image changes caused by the puncture needle, a dynamic time warping (DTW) algorithm is used to measure the deviation degree between the gray distribution sequence and the mean sequence, so as to calculate the gray dynamic change degree of each image block; for each image block , the gray distribution sequence is extracted from the time sequence image dataset , wherein =1, 2, …, , is the number of time sequence images, the mean gray sequence of the baseline stage is calculated , the expression is

[0085]

[0086] wherein, is the number of baseline time sequence images (such as 2 images);

[0087] The DTW algorithm is used to measure the similarity between the image block gray sequence and the mean sequence , the best matching path is solved by minimizing the Euclidean distance, and the DTW distance is calculated, the expression is

[0088]

[0089] wherein, is the gray distribution sequence of the current image block, is the reference gray distribution sequence (mean), represents the gray value of the th position in the sequence , is the sequence length, that is, the total number of time sequence images;

[0090] The DTW algorithm is realized by using the dtaidistance library in Python, and the DTW distance between the gray distribution sequence of the current image block and the mean sequence at time is calculated;

[0091] Combining the gray scale fluctuation index and the DTW distance , the gray scale dynamic change degree of the image block is calculated, and the gray scale dynamic change degree of the image block is calculated , the expression is,

[0092]

[0093] wherein, is the gray scale dynamic change degree of the image block , and is the gray scale fluctuation index (such as the standard deviation of the gray scale value) of the image block at the time

[0094] The gray scale fluctuation index is used to measure the gray scale change degree of the image block at the time , and the standard deviation or variance of the gray scale value can be used as an index;

[0095] For all image blocks , the gray scale dynamic change degree calculated is sorted from high to low, and the image blocks with the highest change degree are selected to infer the possible area of the puncture needle, and a threshold value or the top 5 image blocks with the highest change degree are selected as the key area of the puncture needle positioning.

[0096] S4, based on the gray scale dynamic change degree of each image block, noise removal and feature enhancement are performed on each image block, including the following steps,

[0097] S4.1, according to the calculated gray scale dynamic change degree of each image block , the adjustment parameter =0.001 is set as the basis for similarity weight calculation to control the weight decay rate and ensure the sensitivity and accuracy of the weight calculation;

[0098] For each pair of image blocks, the difference of the gray scale dynamic change degree is calculated, and the similarity weight between the image blocks is calculated, and the expression is,

[0099]

[0100] wherein, is the similarity weight of the image block and the image block , and the greater the weight value indicates the higher the similarity between the image block pairs, and the smaller the weight value indicates the lower the similarity,​​​​ , For image blocks and image blocks The degree of dynamic change in grayscale is measured; a similarity weight matrix W is generated between image patches for subsequent filtering.

[0101] S4.2 By combining dynamic similarity weights and nonlocal mean filtering, efficient noise removal is achieved while preserving key anatomical structures and puncture needle features;

[0102] Specifically, define the size of the neighborhood range, such as 20×20 pixels, to cover related similar image patches, for each pixel. Locate the image patch containing it, and select the image patch with high similarity to it from the similarity weight matrix W. ;

[0103] The weights of all neighboring image patches are summed to obtain a normalization factor, and the denoised pixel values ​​are then calculated. The expression is,

[0104]

[0105] in, These are the pixel values ​​after noise reduction. This is a normalization factor to ensure that the filtered pixel values ​​are within a reasonable range. For image blocks With neighboring image patches Similarity weights, For pixels The non-local neighborhood range, For neighboring pixels The original pixel values;

[0106] After denoising the entire image, a filtered and denoised multimodal image is obtained. ;

[0107] Furthermore, efficient matrix operation libraries (such as NumPy and OpenCL) are used to accelerate the filtering process, ensuring real-time processing requirements. Parallel computing techniques are employed to perform parallel filtering on image patches, significantly improving processing speed. During the filtering process, gradient weights or edge detection algorithms (such as the Sobel operator) are used to protect image edges, prevent edge blurring, and preserve the sharp edges of critical structures such as puncture needles. Based on the actual image noise level, similarity weights and neighborhood range parameters are adjusted to optimize the filtering effect. The optimal parameter combination (e.g., ...) is determined experimentally. =0.001, neighborhood range=20×20 pixels), to balance noise removal and structure preservation.

[0108] S4.3, a pre-trained fusion U-Net (Fusion U-Net) model is adopted, which is based on the classic U-Net architecture and has the ability to fuse multi-modal data, suitable for feature enhancement of ultrasound, near-infrared light and elastography data; the input layer is a 3-channel input corresponding to the denoised ultrasound, NIR light and elastography data respectively, each layer contains convolution, batch normalization (Batch Normalization, BN) and ReLU activation function, high-resolution features and low-resolution features are fused by using up-sampling deconvolution and skip connections, and Sigmoid activation function is used to ensure that the output image gray value is in the range of 0, 1;

[0109] The denoised multi-modal image is normalized, and the normalized multi-modal image is input into the pre-trained fusion U-Net model for forward propagation to generate an enhanced multi-modal image.

[0110] Through the skip connection mechanism of the U-Net network, high-resolution features are retained, the edge and contrast of the puncture needle are enhanced, and the detailed features of the puncture needle and key anatomical structures are extracted through the convolution layer to ensure that they are clearly visible in the enhanced image. The enhanced image is de-normalized to restore the gray value range, and morphological processing (such as dilation or erosion) is used to further enhance the features of the puncture needle and improve its contrast and visibility in the image.

[0111] S5, using the denoised multi-modal image data, a precise three-dimensional model of the spinal canal region is generated by volume rendering method, and the three-dimensional model is displayed in real time through an augmented reality device, including the following steps,

[0112] S5.1, the registered two-dimensional image sequence is converted into three-dimensional voxel data (Voxel Data), realizing the spatial information expansion from two-dimensional to three-dimensional;

[0113] Specifically, according to the depth of the puncture region and the organ structure, the three-dimensional resolution of the voxel is set, such as 1x1x1mm, to ensure the details and accuracy of the three-dimensional model, and the linear interpolation method is used to fill the blank area between the two-dimensional images to ensure the coherence and integrity of the voxel data.

[0114] In order to ensure that the three-dimensional data after voxelization processing has no obvious discontinuity and void, and to maintain the coherence and integrity of the three-dimensional model, a three-dimensional visualization tool (such as VTK or Mayavi) can be used to visualize the voxel data, check the coherence and accuracy of the model, and identify and repair the voids and discontinuities in the voxel data through threshold segmentation and connected component analysis.

[0115] S5.2. According to the clinical needs and the use of the three-dimensional model, the Marching Cubes algorithm is selected for surface reconstruction because it can effectively extract the surface grid model in the voxel data and is suitable for accurate modeling of anatomical structures. Marching Cubes is a classic three-dimensional surface reconstruction algorithm that traverses each small cube in the voxel data, determines the position and shape of the surface polygon according to the threshold value combination of the voxel value inside, and generates a grid model.

[0116] By detecting whether the vertex of each small cube in the voxel data is above or below the threshold value, the position and shape of the surface polygon are determined according to the vertex state to construct a three-dimensional surface grid. The specific process is as follows: the cube state is defined as,

[0117]

[0118] wherein, =1 if the vertex of the voxel value Vi is greater than or equal to the threshold value, otherwise =0.

[0119] According to the cube state, the triangle combination to be generated is determined by looking up the table, the accurate position of the polygon vertex is calculated by linear interpolation, and the grid structure is formed.

[0120] The Quadratic Decimation method is used to optimize the grid structure by reducing the number of polygons, so that the rendering and calculation efficiency is improved while maintaining the original shape. The Quadratic Decimation method is used to simplify the triangular mesh, which reduces the number of polygons while trying to maintain the details of the geometric shape. This method uses a quadratic simplification technique to reduce the complexity of the mesh, which is commonly used to optimize the calculation and rendering performance.

[0121] The Laplacian smoothing algorithm is applied to reduce grid noise and smooth surface curves.

[0122] S5.3. To realize real-time superimposed display of the three-dimensional model in the doctor's field of view, a suitable augmented reality device is selected and accurate positioning and tracking configuration is performed.

[0123] Further, Microsoft HoloLens 2 is selected as the AR device because it has high-resolution display, accurate head tracking and gesture recognition function, and is suitable for complex medical environment.

[0124] Optical tracking technology (such as the Motion Capture system based on LED markers) is used to achieve high-precision synchronization between the device and the patient's position. Multiple LED markers are installed on the patient's waist and the augmented reality device for recognition by the optical tracking camera. A high frame rate camera (such as 60 FPS) is configured to perform real-time marker recognition and position calculation.

[0125] Initial calibration of the optical tracking system is performed to correct the spatial relationship between the AR device and the actual anatomical structure, ensuring accurate superposition of the 3D model with the patient's position. Specifically, multiple known location points are marked on the patient's waist, and the corresponding points in the AR device are recorded to achieve precise spatial alignment. The position and orientation of the 3D model are adjusted using a spatial transformation matrix to achieve high-precision superposition.

[0126] The optimized 3D surface mesh model is imported into the visualization software of the AR device, such as Unity 3D or Unreal Engine. Based on the spatial transformation matrix obtained from the calibration, the position and orientation of the 3D model are adjusted to precisely align it with the patient's actual anatomical position.

[0127] It continuously receives and analyzes data from the optical tracking system, obtains the spatial transformation matrix at the current time point, dynamically updates the spatial transformation of the 3D model in the AR device, and adjusts the position and orientation of the 3D model in real time to ensure its synchronous movement with the patient's anatomical structure.

[0128] S6. On the generated 3D model, mark important anatomical structures, and use the A* algorithm to calculate the optimal puncture path in 3D space that avoids key structures, including the following steps.

[0129] S6.1 In order to generate a 3D model with key structure labels and provide obstacle information for path planning, image segmentation algorithms such as U-Net based on deep learning are used to process the fused multimodal image data, automatically identify and segment key anatomical structures (such as nerves and blood vessels), and manually correct the automatic segmentation results to ensure the accuracy of the labeling.

[0130] S6.2 Determine the three-dimensional boundary of the puncture area, set the mesh resolution (e.g., 1 mm³ per voxel), and use a uniform mesh generation method to divide the puncture area into several mesh nodes, each node representing a point in space.

[0131] Assign position coordinates to each node ( , , The marker nodes are marked as being located inside key anatomical structures, serving as obstacle nodes and providing spatial location and obstacle information for path planning.

[0132] Define the connection relationships between nodes, using a 26-neighborhood connection method, meaning each node can be connected to at most 26 neighboring nodes (including all adjacent nodes in the top, bottom, left, right, and diagonal directions). For each node ( , , ), identify all its possible neighboring nodes ( , , )satisfy Eliminate all obstructing nodes and ensure that only feasible neighboring nodes are connected.

[0133] S6.3. Choose Euclidean distance as the heuristic function to estimate the straight-line distance from the current node n to the target node. The expression is:

[0134]

[0135] in,( , , ) represents the coordinates of the target node, ( , , ) is the current node The coordinates;

[0136] Extract the current node coordinates ( , , ) and its parent node coordinates ( , , ), calculate the Euclidean distance between the current node and its parent node, expressed as,

[0137]

[0138] The actual cost of the current node The actual cost of updating to the parent node Add the distance between nodes:

[0139]

[0140] The actual cost of the starting node 0;

[0141] By calculating the total cost of each node Combined with actual costs and heuristic costs The optimal path is determined by the expression,

[0142]

[0143] wherein, is the actual cost from the start node to the current node , usually the Euclidean or Manhattan distance between nodes, is the heuristic estimated cost (e.g. Euclidean distance) from the current node to the goal node;

[0144] The A* algorithm is used to search for the optimal path from the start node to the goal node in a 3D grid. The start node is added to the Open List, and its total cost is calculated and recorded, and the search process of the A* algorithm is started;

[0145] The node with the smallest total cost is selected from the Open List , removed from the Open List and added to the Closed List. All neighbor nodes of the current node are traversed , all neighbor nodes located in critical anatomical structures are skipped, the actual cost from the start node to the neighbor node is calculated , , and the heuristic cost and the total cost are calculated;

[0146] If the node is already in the Closed List, it is skipped. If the node is not in the Open List, it is added to the Open List, and its parent node is recorded as . If the node is already in the Open List, but the new is smaller, its total cost and parent node are updated;

[0147] The process is repeated until the goal node is selected from the Open List, and the path search is completed. If the Open List is empty, a valid path cannot be found;

[0148] Starting from the goal node, the parent nodes are traced back step by step until the start node is reached, forming the entire path. The sequence of nodes obtained by backtracking is reversed to obtain the complete puncture path from the start node to the goal node.

[0149] The embodiment also provides a computer device suitable for the case of the anesthesia puncture auxiliary positioning method, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the anesthesia puncture auxiliary positioning method proposed in the above embodiment.

[0150] The computer device can be a terminal, which includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved by WIFI, a carrier network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, a trackball or a touchpad arranged on the shell of the computer device, or an external keyboard, a touchpad or a mouse, etc.

[0151] The embodiment also provides a storage medium having a computer program stored thereon, the program being executed by a processor to implement the method for assisting positioning of anesthetic puncture proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0152] In summary, the present application realizes the complementation of the advantages of different imaging technologies by acquiring multi-modal image data of the area to be anesthetized and synchronizing and fusing the same to generate multi-modal comprehensive images, significantly improving the comprehensiveness and accuracy of the image data; the multi-modal comprehensive images after fusion are evenly divided into a plurality of image blocks, and the degree of gray dynamic change of each image block is calculated, realizing the localized analysis of the image data and effectively capturing the characteristic changes of different anatomical regions; based on the degree of gray dynamic change of each image block, targeted noise removal and feature enhancement processing are performed on each image block, realizing the significant improvement of the image quality, effectively retaining the key anatomical structure information while suppressing the interference factors, providing high-quality data input for subsequent three-dimensional reconstruction, and enhancing the image readability and interpretability; using the processed multi-modal image data, an accurate three-dimensional model is generated through volume rendering, and real-time display is realized in combination with augmented reality technology, realizing the intuitive presentation of complex anatomical structures, further, important anatomical structures are labeled on the three-dimensional model, and an optimal puncture path is calculated using an A* algorithm, realizing safe and efficient puncture navigation, and significantly improving the accuracy and safety of the anesthetic puncture.

[0153] Example 2, referring to Table 1, is a second embodiment of the present application, and gives experimental simulation data of the anesthetic puncture auxiliary positioning method to further verify the technical solution of the present application.

[0154] This embodiment is aimed at assisting positioning for lumbar anesthetic puncture. The experimental preparation and implementation process are as follows:

[0155] Firstly, 50 patients who need to receive lumbar anesthesia are selected, with an age range of 25-65 years old and a body mass index (BMI) between 18.5 and 30. The patients are randomly divided into two groups: an experimental group (25 people) and a control group (25 people). The experimental group uses the anesthetic puncture auxiliary positioning method of the present application, and the control group uses the traditional ultrasound-guided method.

[0156] For the experimental group, Philips EPIQ Elite ultrasound system equipped with L12-3 linear array probe is used for ultrasound imaging, FLIR One Pro near-infrared light sensor is used to acquire near-infrared light images, and Siemens ACUSON S2000 HELX Evolution ultrasound system Virtual Touch elastography function is used to collect elastography data. All devices are synchronized through a high-precision time synchronization system to ensure the time consistency of multi-modal data.

[0157] The image processing and fusion stage uses a pre-trained U-Net model for image fusion, which is trained on a data set containing 5000 multi-modal medical images. The image block division uses an 8x8 grid, and each image block is 64x64 pixels in size.

[0158] To generate an accurate three-dimensional model, a high-precision lumbar spine model was created using a Voxel8 3D printer for system calibration and accuracy verification. Microsoft HoloLens 2 was used as the AR device, combined with the OptiTrack Prime13W optical tracking system for real-time positioning.

[0159] During the actual puncture process, the following parameters were recorded: puncture success rate, puncture attempt number, puncture time, patient pain score (VAS score), and complication incidence. Meanwhile, three experienced anesthesiologists were invited to score the difficulty level of the puncture process (1-10 points, with 10 points representing the easiest).

[0160] The details are shown in Table 1 below:

[0161] Table 1 Experimental data table

[0162]

[0163] By comparing the data of the experimental group and the control group, it can be clearly seen that the method of the present invention is significantly superior to the traditional ultrasound-guided method in many aspects.

[0164] The puncture success rate of the experimental group reached 96%, which was 14.29% higher than the 84% of the control group. This indicates that the method of the present invention can provide more accurate positioning information, greatly improving the accuracy of puncture. For the average number of puncture attempts, the experimental group only needed 1.2 attempts to successfully puncture, while the control group needed 2.3 attempts. The method of the present invention reduced the number of puncture attempts by 47.83%, not only improving efficiency but also reducing trauma to patients. The average puncture time of the experimental group was 3.5 minutes, which was 51.39% less than the 7.2 minutes of the control group. This significant time saving not only improves surgical efficiency but also reduces patient discomfort time on the operating table. In terms of average patient pain score, the VAS score of the experimental group was 2.1, which improved by 51.16% compared to the 4.3 of the control group. This indicates that the method of the present invention greatly reduces the pain of patients during the puncture process, improving patient comfort. The complication incidence of the experimental group was only 2%, while that of the control group was 8%, a reduction of 75%. This improvement indicates that the method of the present invention greatly improves surgical safety and reduces potential medical risks. In the anesthesiologist score, the experimental group scored 8.7, which was 40.32% higher than the 6.2 of the control group. This shows that the method of the present invention not only has excellent objective indicators, but also has the recognition of professional physicians in subjective operation experience.

[0165] In summary, the present application has achieved remarkable effects in improving puncture success rate, reducing operation time, reducing patient pain and complication risk, etc. This not only improves medical efficiency, but also greatly improves patient experience.

[0166] It should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A method of assisting positioning of an anaesthetic puncture, characterized by: The application relates to a method for generating a precise three-dimensional model of a spinal canal region and a puncture path for a spinal anesthesia needle. The method comprises the following steps: a plurality of modal image data of a region to be anesthetized is acquired, the plurality of modal image data is image-synchronized and fused, and a plurality of modal comprehensive image is generated; the plurality of modal comprehensive image after fusion is evenly divided into a plurality of image blocks according to a preset number, and a gray dynamic change degree of each image block is calculated; based on the gray dynamic change degree of each image block, noise removal and feature enhancement processing are performed on each image block, including the following steps: for each pair of image blocks, a difference in the gray dynamic change degree is calculated, a similarity weight between the image blocks is calculated, a similarity weight matrix between the image blocks is constructed, a size of a neighborhood range is defined, for each pixel point, a corresponding image block is located, and a high-similarity image block is selected in the similarity weight matrix, a weight of all neighborhood image blocks is accumulated to obtain a normalization factor, a pixel value after denoising is calculated, denoising processing of a full image is completed, and a plurality of modal image after filtering and denoising is obtained; the plurality of modal image after denoising is normalized, the plurality of modal image after normalization is input into a pre-trained fusion U-Net model, forward propagation is performed, and a plurality of modal image after enhancement is generated; precise three-dimensional model of the spinal canal region is generated by using the plurality of modal image data after processing through volume rendering method, and the three-dimensional model is displayed in real time through an augmented reality device; on the generated three-dimensional model, important anatomical structures are labeled, and an optimal puncture path avoiding key structures is calculated in a three-dimensional space by using an A* algorithm, including the following steps: The three-dimensional space of the puncture region is discretized into grid nodes, each node representing a point in space, and each node is assigned a position coordinate (x, y, z) , , ), and whether the node is located inside a key anatomical structure is marked as an obstacle node; The connection relationship between nodes is defined by using 26-neighbor connection mode, and all possible adjacent nodes of each node are identified , , ) meet , all obstacle nodes are excluded; The Euclidean distance is chosen as the heuristic function, estimating the straight-line distance from the current node to the goal node. an image segmentation algorithm is used to process the plurality of modal image data after fusion, and key anatomical structures are automatically recognized and segmented, and the automatic segmentation result is manually corrected; adding the start node to the open list, selecting the node with the lowest total cost from the open list removing it from the open list and adding it to the closed list traversing the nodes all the neighboring nodes of the node ; If the node is already in the closed list, skip; if the node is not in the open list, add it to the open list and record its parent as ; if the node is already in the open list but the new actual cost is smaller, update its total cost and parent. a total cost of each node in the A* algorithm is calculated by combining an actual cost and a heuristic cost; the steps are repeated until a target node is selected from an open list, and path searching is completed; if the open list is empty, an effective path cannot be found; 2. The anesthetic puncture assisting positioning method according to claim 1, wherein: starting from the target node, the path is backtracked along parent nodes step by step until a starting node is reached, the entire path is formed, the node sequence obtained by backtracking is reversed, and a complete puncture path from the starting node to the target node is obtained. The plurality of modal image data of the region to be anesthetized is acquired, including the following steps:

3. The method of claim 2, wherein: ultrasonic image, near-infrared light image and elastic imaging data of a target puncture region of a patient are collected through an ultrasonic probe, a near-infrared light sensor and an elastic imaging device. The plurality of modal image data is image-synchronized and fused to generate the plurality of modal comprehensive image, including the following steps: a unified time reference is used to generate a timestamp, and accurate timestamps are automatically embedded when image data of each mode is collected; timestamp information is extracted from metadata of each modal image by using an image processing software; time differences between different modal images are calculated, a fixed delay is identified and recorded, and the time sequence of each modal image is adjusted by using an interpolation and clipping method; SIFT algorithms are respectively applied to ultrasonic images, NIR images and elastic imaging images to detect key feature points, and coordinates and descriptors of each feature point are recorded; Initialize the FLANN matcher, set the index parameters and search parameters, and perform feature point matching on NIR-ultrasound and elastography-ultrasound respectively to generate potential matching pairs, and filter false matches using the ratio test; Extract the coordinate set of the matching point pairs, calculate the affine transformation matrix of NIR-ultrasound and elastography-ultrasound respectively, and use the calculated affine transformation matrix to perform spatial transformation on the NIR image and elastography image, generating NIR image and elastography image that are spatially aligned with the ultrasound image; Normalize the aligned ultrasound image, NIR image and elastography data, unify the gray scale range, input the normalized multi-modal image data into the pre-trained fusion U-Net model, perform forward propagation, and generate the fused multi-modal comprehensive image.

4. The method of claim 3, wherein: Divide the fused multi-modal comprehensive image into several image blocks according to the preset number, calculate the gray dynamic change degree of each image block, including the following steps, Set the number of image blocks and the size of image blocks, use the fixed grid method, start from the top left corner of the image, divide row by row, and verify whether each image block covers different anatomical regions; Before the start of the puncture operation, collect at least two ultrasound images as the initial image group, and record the baseline gray scale distribution; During the actual puncture process, real-time collection of ultrasound images at fixed time intervals, and the new images are sequentially added to the image group to form a time series image dataset; For each image block, extract the gray scale distribution sequence from the time series image dataset, and calculate the mean gray scale sequence in the baseline stage; Use the dtaidistance library in Python to implement the DTW algorithm, calculate the DTW distance between the gray scale distribution sequence of each image block at each time and the mean sequence; Combine the gray scale fluctuation index and the DTW distance to calculate the gray dynamic change degree of the image block.

5. The method of claim 4, wherein: Using the denoised multi-modal image data, generate an accurate three-dimensional model of the spinal canal region by volume rendering method, and display the three-dimensional model in real time through augmented reality equipment, including the following steps, According to the depth of the puncture area and the organ structure, set the three-dimensional resolution of the voxel, and use linear interpolation method to fill the blank area between two-dimensional images; Input the voxel data into the Marching Cubes algorithm, set the threshold, and determine whether the vertices of each small cube in the detected voxel data are above or below the threshold; According to the state of the cube, look up the table to determine the triangle combination to be generated, calculate the accurate position of the polygon vertex through linear interpolation, and form a grid structure; Use the Quadratic Decimation method to optimize the grid structure by reducing the number of polygons, apply the Laplacian smoothing algorithm to reduce grid noise and smooth the surface curve; Select Microsoft HoloLens 2 as the AR device, use optical tracking technology, install multiple LED markers on the patient's waist and augmented reality equipment, and configure a high-frame-rate camera for real-time marker recognition and position calculation; Perform initial calibration of the optical tracking system to correct the spatial relationship between the AR device and the actual anatomical structure; The optimized three-dimensional surface mesh model is imported into the visualization software of the AR device, and the position and direction of the three-dimensional model are adjusted according to the space transformation matrix obtained through calibration. The data of the optical tracking system are continuously received, and the position and direction of the three-dimensional model are adjusted in real time.

6. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: The processor executes the computer program to realize the steps of the anesthetic puncture auxiliary positioning method in any one of claims 1-5.

7. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the anesthetic puncture auxiliary positioning method in any one of claims 1-5.

Images