Percutaneous pedicle screw placement method and device based on deep learning

Abstract

The application provides a kind of based on deep learning pedicle screw channel planning method and device, the method comprises: extraction procedure, extract the pedicle CT image in the spine CT image to be extracted;Layering procedure, the pedicle CT image is sliced and layered along the pedicle cross section, confirm the pedicle boundary, and obtain a plurality of pedicle two-dimensional images in turn;Planning procedure, the center point of the pedicle two-dimensional image is obtained in turn, the center point is fitted, and the pedicle screw channel is obtained.The pedicle screw channel planning method and device based on deep learning provided in the embodiment of the application, by sequentially executing extraction, slicing, layering and confirming the boundary to the spine CT image, the pedicle two-dimensional image is obtained, the center point of the two-dimensional image is obtained in turn and the obtained center point is fitted to obtain the final pedicle screw channel, the rapid acquisition of the pedicle screw channel is realized, and the operation efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of medical imaging technology, and in particular to a method and apparatus for planning pedicle screw implantation channels based on deep learning. Background Technology

[0002] Currently, pedicle screw fixation plays a crucial role in human spinal surgery and is considered one of the most stable methods of spinal fixation. This technique requires the pedicle screw to be positioned within the pedicle; if it perforates the pedicle, it can cause serious damage, such as injury to the spinal cord or vertebral blood vessels, and even endanger the patient's life. Therefore, ensuring accurate placement of pedicle screws has become a key research objective.

[0003] Traditional pedicle screw implantation surgery relies on the surgeon's expertise, adjusting the surgical path based on C-arm X-ray images taken in the operating room. Due to varying surgeon skill levels, this often leads to different degrees of complications. While computer-aided navigation-based pedicle screw implantation surgery uses preoperative or intraoperative medical images (CT scans, X-rays) and bony landmarks and position tracking sensors to establish relationships between the surgical space, image space, and robotic space, it then translates the surgical instruments and robotic end effector into the image space, creating a virtual reality to assist the surgeon in the pedicle screw implantation procedure. However, this method is complex, magnifies errors, and the positional relationships seen on the navigation images are often misleading, failing to accurately and promptly reflect the relationship between the tools and the pedicle, thus lacking true real-time monitoring. Summary of the Invention

[0004] To address the problems in the prior art, this invention provides a deep learning-based method and apparatus for pedicle screw implantation channel planning.

[0005] This invention provides a deep learning-based method for pedicle screw implantation channel planning, comprising:

[0006] Extraction process: Extract pedicle CT images from the spinal CT images to be extracted;

[0007] The layering process involves slicing the pedicle CT image along the cross-section of the pedicle to confirm the pedicle boundary, thereby obtaining multi-layered two-dimensional images of the pedicle.

[0008] The planning process involves sequentially acquiring the center point of the pedicle two-dimensional image, fitting the center point, and obtaining the pedicle screw implantation channel.

[0009] According to a deep learning-based pedicle screw implantation channel planning method provided by the present invention, the method further includes: extending the pedicle screw implantation channel, taking the intersection with the zygomatic spine as the starting point of the pedicle screw implantation channel, taking the intersection with the vertebral body as the ending point of the pedicle screw implantation channel, and obtaining the length of the pedicle screw implantation channel based on the starting point and the ending point.

[0010] According to the deep learning-based pedicle screw implantation channel planning method provided by the present invention, before the extraction process, the method further includes: a preprocessing process, which normalizes the three-dimensional inter-slice spacing in the spinal CT image to be processed, and centers the normalized data to 0 values ​​to obtain the spinal CT image to be extracted.

[0011] According to the present invention, a deep learning-based method for pedicle screw implantation channel planning, wherein confirming the pedicle boundary and sequentially obtaining multi-layer pedicle two-dimensional images specifically includes: performing opening and closing operations on the sliced ​​and layered pedicle CT images, counting the number of connected components in the sliced ​​and layered pedicle CT images, determining whether the number of connected components in the sliced ​​and layered pedicle CT images is greater than a preset threshold, and if the condition is met, retaining the connected component as a target connected component; if the condition is not met, discarding the connected component, and sequentially obtaining multi-layer pedicle two-dimensional images based on the target connected components.

[0012] According to a deep learning-based pedicle screw implantation channel planning method provided by the present invention, the center point of the two-dimensional pedicle image is obtained using the gray-scale centroid method, the formula of which is:

[0013]

[0014]

[0015] Where (u,v) represents the coordinates of a pixel, f(u,v) represents the gray value of pixel (u,v), and Ω represents the set of pixels in the pedicle 2D image. The x-coordinate of the center point The ordinate represents the center point.

[0016] According to a deep learning-based pedicle screw implantation channel planning method provided by the present invention, the step of fitting the center point to obtain the pedicle screw implantation channel specifically includes: obtaining a fitted straight line equation based on a fitted straight line function and a minimum sum of squared errors function based on the center point, and obtaining the pedicle screw implantation channel based on the fitted straight line equation; wherein, the fitted straight line function is expressed as: z = ax + by + c, and the minimum sum of squared errors function is expressed as: a, b, and c represent the parameters to be determined for the fitted linear function, n represents the order of the center points, and N represents the number of center points.

[0017] According to a deep learning-based pedicle screw implantation channel planning method provided by the present invention, the spinal CT image data to be processed is divided into a training dataset, a validation dataset, and a test dataset; wherein, the training dataset is used to train a neural network, the validation dataset is used to adjust the hyperparameters of the neural network, and the test dataset is used to verify the accuracy of the neural network.

[0018] The present invention also provides a deep learning-based pedicle screw implantation channel planning device, comprising:

[0019] The extraction module is used to extract pedicle CT images from the spinal CT images to be extracted.

[0020] The layering module is used to: perform layering processing on the pedicle CT image along the cross-section of the pedicle to obtain multi-layer pedicle two-dimensional images in sequence;

[0021] The planning module is used to: sequentially acquire the center point of the two-dimensional image of the pedicle, fit the center point, and acquire the pedicle screw implantation channel.

[0022] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the deep learning-based pedicle screw implantation channel planning method described above.

[0023] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the deep learning-based pedicle screw implantation channel planning method as described above.

[0024] The present invention provides a deep learning-based method and apparatus for pedicle screw implantation channel planning. This method extracts pedicle CT images from spinal CT images, then performs layered processing and boundary verification to obtain two-dimensional pedicle images. The center points of these two-dimensional images are then sequentially acquired, and multiple center points are fitted to obtain the final pedicle screw implantation channel. The entire acquisition of this pedicle screw implantation channel relies on automatic calculation using image processing technology, eliminating the need for human assistance from doctors. This achieves automatic planning of the implantation channel and standardizes and accelerates the process of acquiring the implantation channel. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is one of the flowcharts of the deep learning-based pedicle screw implantation channel planning method provided by the present invention;

[0027] Figure 2 This is the second flowchart of the deep learning-based pedicle screw implantation channel planning method provided by the present invention;

[0028] Figure 3 This is a schematic diagram of pedicle CT images and annotations provided by the present invention;

[0029] Figure 4 This invention provides a two-dimensional image of the pedicle with a center point;

[0030] Figure 5 This is a schematic diagram of the pedicle screw implantation channel provided by the present invention;

[0031] Figure 6 This is a schematic diagram of the deep learning-based pedicle screw implantation channel planning device provided by the present invention.

[0032] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0034] Figure 1 This is one of the flowcharts illustrating the deep learning-based pedicle screw implantation channel planning method provided by the present invention, such as... Figure 1 As shown, the method includes:

[0035] S110, Extraction process, extracting pedicle CT images from the spinal CT images to be extracted;

[0036] S120, Layered Process: The pedicle CT image is sliced ​​and layered along the cross-section of the pedicle to confirm the pedicle boundary and obtain multi-layered pedicle two-dimensional images in sequence.

[0037] S130, planning process: sequentially obtain the center point of the pedicle two-dimensional image, fit the center point, and obtain the pedicle screw implantation channel.

[0038] It should be noted that the original spinal CT images are preprocessed to obtain the spinal CT images to be extracted. Compared with the original spinal CT images, the interslice spacing between the spinal CT images to be extracted is uniformized. This is beneficial for the subsequent training of the network model using this data, thus accelerating the training speed of the network model.

[0039] The spinal CT image data to be extracted is input into a 3D segmentation network model. The pedicle CT image is obtained through the extraction processing of the 3D segmentation network model. Then, the pedicle CT image is sliced ​​and layered along the cross-section of the pedicle through the 3D segmentation network model to obtain multi-layer slice images. The pedicle boundaries on the multi-layer slice images are confirmed in turn to obtain multi-layer pedicle two-dimensional images.

[0040] The extraction and layering of spinal CT images were completed using a 3D segmentation network model. Then, by confirming the pedicle boundaries, a series of orderly distributed two-dimensional images of the pedicles were obtained. The center points of the two-dimensional images of the pedicles were calculated, and the obtained center points were fitted with a fitted straight line equation. The fitted straight line was used as the pedicle screw implantation channel.

[0041] The deep learning-based pedicle screw implantation channel planning method provided by this invention extracts pedicle CT images from spinal CT images, then performs layer processing and boundary confirmation to obtain two-dimensional pedicle images. The center points of the two-dimensional images are obtained sequentially, and multiple center points are fitted to obtain the final pedicle screw implantation channel. The acquisition of this pedicle screw implantation channel relies entirely on automatic calculation using image processing technology, without the need for human assistance from doctors. This achieves automatic planning of the implantation channel and standardizes and accelerates the process of acquiring the implantation channel.

[0042] According to a deep learning-based pedicle screw implantation channel planning method provided by the present invention, the method further includes: extending the pedicle screw implantation channel, using the intersection with the vertebral column as the starting point of the pedicle screw implantation channel, and the intersection with the vertebral body as the ending point of the pedicle screw implantation channel, and obtaining the length of the pedicle screw implantation channel based on the starting point and the ending point.

[0043] It should be noted that in the process of obtaining the pedicle screw implantation channel using the 3D segmentation network model, the output pedicle screw implantation channel is limited to the length inside the pedicle and cannot be directly applied to the actual surgical process. Based on this, the above-mentioned pedicle screw implantation channel is extended, and the intersection of the extended implantation channel with the zigzag ridge is taken as the starting point of the pedicle screw implantation channel, and the intersection with the vertebral body is taken as the ending point of the pedicle screw implantation channel. The distance between the above-mentioned starting point and ending point is taken as the final length of the pedicle screw implantation channel.

[0044] The deep learning-based pedicle screw implantation channel planning method provided by this invention uses the intersection of the extension line of the pedicle screw implantation channel obtained by the 3D segmentation network model with the zigzag ridge as the starting point of the pedicle screw implantation channel and the intersection with the vertebral body as the ending point. Based on this, the specific starting and ending points of the pedicle screw implantation channel are clearly defined, and the length of the pedicle screw implantation channel is obtained at the same time. Based on this, the pedicle screw implantation channel can be planned more accurately, which is beneficial to its application in actual surgery and ensures the accuracy of the actual screw implantation process.

[0045] According to the deep learning-based pedicle screw implantation channel planning method provided by the present invention, before the extraction process, the method further includes: a preprocessing process, which normalizes the three-dimensional inter-slice spacing in the spinal CT image to be processed, and centers the normalized data to 0 values ​​to obtain the spinal CT image to be extracted.

[0046] The window width and level of the original spinal CT images were adjusted to remove interference other than bone structures. However, the original spinal CT image data had varying interslice spacing. To address this, resampling was used to change the interslice spacing in all three dimensions of the CT image data to 1 mm. Furthermore, to facilitate data processing and accelerate program convergence, the data underwent normalization and zero-value centering to obtain the spinal CT images to be extracted. Normalization involves subtracting the mean from the variable and then dividing by the standard deviation. Zero-value centering involves subtracting the mean from the variable, essentially a translation process. After translation, the center of all data is (0,0). Through centering and standardization, data with a mean of 0 and a standard deviation of 1, following a standard normal distribution, was obtained. This eliminates errors caused by different units, inherent variations, or large numerical differences.

[0047] The deep learning-based pedicle screw implantation channel planning method provided by this invention achieves image data standardization by performing normalization and zero-value centering on the spinal CT image to be processed sequentially. Based on this, the convergence speed of the network model can be accelerated when applying the standard image data. At the same time, when the data is input into the network model, the processing speed of the network model can also be improved.

[0048] According to the present invention, a deep learning-based method for pedicle screw implantation channel planning is provided. In this invention, the step of confirming the pedicle boundary and sequentially obtaining multi-layer pedicle two-dimensional images specifically includes: performing opening and closing operations on the sliced ​​and layered pedicle CT images, counting the number of connected components in the sliced ​​and layered pedicle CT images, determining whether the number of connected components in the sliced ​​and layered pedicle CT images is greater than a preset threshold. If the condition is met, the connected component is retained as a target connected component; if the condition is not met, the connected component is discarded, and multi-layer pedicle two-dimensional images are sequentially obtained based on the target connected components.

[0049] In image processing, opening and closing operations are common techniques, specifically divided into opening and closing operations. Opening involves erosion followed by dilation, while closing involves dilation followed by erosion. Erosion eliminates boundary points, causing the boundary to shrink inwards; it can be used to eliminate small and meaningless objects. Dilation merges all background points in contact with an object into that object, causing the boundary to expand outwards; it can be used to fill holes in an object. Opening operations can remove small grain noise and break the adhesion between target objects, while closing operations can connect neighboring objects in an image, smoothing their boundaries without significantly changing their area.

[0050] A connected component refers to an image region consisting of foreground pixels with the same pixel value and adjacent positions. Images with fewer than a preset threshold of connected components are considered interference images and are discarded. Connected components that meet the preset threshold are retained as target connected components, and optimized pedicle 2D images are obtained based on the target connected components.

[0051] It should be noted that the preset threshold can be set in advance according to actual needs. In this invention, the preset threshold is set to 2.

[0052] The deep learning-based pedicle screw implantation channel planning method provided by this invention performs opening and closing operations on the pedicle CT images after slice layering and performs image filtering based on the number of connected components. Based on this, irrelevant images are removed, and the images are optimized. This helps to avoid interference from irrelevant images in the pedicle screw implantation channel planning process and achieve accurate planning.

[0053] According to a deep learning-based pedicle screw implantation channel planning method provided by the present invention, the center point of the two-dimensional pedicle image is obtained using the gray-scale centroid method, and the formula for the gray-scale centroid method is as follows:

[0054]

[0055]

[0056] Where (u,v) represents the coordinates of a pixel, f(u,v) represents the gray value of pixel (u,v), and Ω represents the set of pixels in the pedicle 2D image. The x-coordinate of the center point The ordinate represents the center point.

[0057] For targets with uneven brightness (such as light spots or streaks), the gray-scale centroid method can calculate the coordinates of the weighted centroids based on the target's light intensity distribution, which are then used as tracking points; this is also called the density centroid algorithm. This invention fully considers the difference in brightness between the pedicle boundary and its interior in a two-dimensional pedicle image, and based on this, accurately determines the center point of the two-dimensional pedicle image.

[0058] The deep learning-based pedicle screw implantation channel planning method provided by this invention achieves accurate acquisition of the center point of the pedicle in a two-dimensional image through the gray-scale centroid method, which is beneficial for subsequent accurate planning of the pedicle screw implantation channel based on the center point.

[0059] According to a deep learning-based pedicle screw implantation channel planning method provided by the present invention, the step of fitting the center point to obtain the pedicle screw implantation channel specifically includes: obtaining a fitted straight line equation based on a fitted straight line function and a minimum error sum of squares function of the center point, and obtaining the pedicle screw implantation channel based on the fitted straight line equation; wherein, the fitted straight line function is expressed as: z = ax + by + c, and the minimum error sum of squares function is expressed as: a, b, and c represent the parameters to be determined for the fitted linear function, n represents the order of the center points, and N represents the number of center points.

[0060] The center points are obtained by fitting using the least squares method. The basic principle of least squares is to minimize the variance. Specifically, after obtaining the center points of the two-dimensional pedicle images, the fitting line function is set as z = ax + by + c, and the center points are specifically represented as (x... i ,y i ,z i Meanwhile, the minimum sum of squared errors function is used as the basis for solving the parameters a, b, and c in the fitted straight line function. Specifically, the obtained center points are substituted into the minimum sum of squared errors function in turn. In the process, after obtaining the specific values ​​of parameters a, b, and c, they are substituted into the fitting linear function to obtain the fitting linear equation. This fitting linear equation represents the specific path of the pedicle screw implantation channel. Here, a, b, and c represent the parameters to be determined in the fitting linear function, n represents the order of the center points, N represents the number of center points, and i takes the value [1, n].

[0061] The deep learning-based pedicle screw implantation channel planning method provided by this invention sets a fitting linear function and obtains the parameters in the fitting linear function by using the minimum error sum of squares function, and finally obtains the fitting linear equation, realizing the mathematical expression of the pedicle screw implantation channel and ensuring the visual and quantitative description of the pedicle screw implantation channel.

[0062] According to the deep learning-based pedicle screw implantation channel planning method provided by the present invention, the spinal CT image data to be processed is divided into a training dataset, a validation dataset, and a test dataset; wherein, the training dataset is used to train the neural network, the validation dataset is used to adjust the hyperparameters of the neural network, and the test dataset is used to verify the accuracy of the neural network.

[0063] It should be noted that in this invention, the spinal CT image data to be processed is split into a training dataset, a validation dataset, and a test dataset. The ratio of the data volume between the training dataset and the test dataset is 7:3, while the data volume of the validation dataset is 20% of that of the training dataset.

[0064] The deep learning-based pedicle screw implantation channel planning method provided by this invention splits the spinal CT image data to be processed into training datasets, validation datasets, and test datasets, and reasonably limits the proportional relationship between each dataset. Based on this, the quality of the network model trained by the above datasets can be improved, ensuring that the network model has good prediction accuracy.

[0065] Figure 2 This is the second flowchart of the deep learning-based pedicle screw implantation channel planning method provided by the present invention, as shown below. Figure 2 As shown, the method includes:

[0066] Step 1: Obtain spinal CT data;

[0067] Step 2, 3D pedicle segmentation: The pedicle CT data is obtained by extracting spinal CT data using a 3D segmentation network model.

[0068] Step 3, Mask cross-sectional layering: Using a 3D segmentation network model, slice the pedicle CT data along the cross-section of the pedicle to obtain image slices of the pedicle in each layer.

[0069] Step 4: Calculate the pedicle boundary of the cross section, and obtain the target connected region by using opening and closing operations and connected component filtering process to determine the boundary of the pedicle 2D image.

[0070] Step 5: Calculate the center point of the section mask boundary. The gray-scale centroid method is used to calculate the center point of the section mask boundary, which is the center point of the pedicle in the two-dimensional image of each layer.

[0071] Step 6: Fit the pedicle anatomical axis. Use the least squares method to fit the center point obtained in Step 5, and obtain the pedicle anatomical axis.

[0072] Step 7: Calculate the start and end points and length. Extend the pedicle anatomical axis obtained in Step 6, and take the intersection with the zigzag ridge as the start point and the intersection with the vertebral body as the end point. The distance between the start and end points is taken as the final length.

[0073] Step 8, End.

[0074] The deep learning-based pedicle screw implantation channel planning method provided by this invention extracts pedicle CT images from spinal CT images, then performs layer processing and boundary confirmation to obtain a two-dimensional pedicle image. The center points of the two-dimensional image are then sequentially acquired, and multiple center points are fitted to obtain the final pedicle screw implantation channel. The entire acquisition of this pedicle screw implantation channel relies on automatic calculation using image processing technology, requiring no human assistance from the surgeon. This achieves automatic planning of the implantation channel and standardizes and accelerates the acquisition process. Furthermore, the calculation of the start and end points clarifies the length of the pedicle screw implantation channel, further improving the accuracy of surgical path planning and ensuring the smooth progress of the surgery.

[0075] Figure 3 This is a schematic diagram of pedicle CT images and annotations provided by the present invention, such as... Figure 3 As shown, 3-1 is a spinal CT image, and 3-2 is the pedicle region. In practical applications, different colors are used to display the pedicle region in 3-2 to achieve the annotation effect.

[0076] Figure 4 This invention provides a two-dimensional image of the pedicle with a center point, such as... Figure 4 As shown, the white area represents the two-dimensional image of the pedicle, the adjacent black and white areas represent the boundary of the two-dimensional image of the pedicle, and the black dots represent the center point of the two-dimensional image of the pedicle.

[0077] Figure 5 This is a schematic diagram of the pedicle screw implantation channel provided by the present invention, as shown below. Figure 5 As shown, the pedicle screw implantation channel is extended, and the intersection of the extended line of the implantation channel with the zigzag ridge is taken as the starting point of the pedicle screw implantation channel, and the intersection with the vertebral body is taken as the ending point of the pedicle screw implantation channel. The distance between the starting point and the ending point is taken as the final length of the pedicle screw implantation channel.

[0078] The deep learning-based pedicle screw implantation channel planning method provided by this invention uses the intersection of the extension line of the pedicle screw implantation channel with the zigzag ridge as the starting point of the pedicle screw implantation channel and the intersection with the vertebral body as the ending point. Based on this, the specific starting and ending points of the pedicle screw implantation channel are clearly defined, and the length of the pedicle screw implantation channel is obtained at the same time. Based on this, the pedicle screw implantation channel can be planned more accurately, which is beneficial to its application in actual surgery and ensures the accuracy of the actual screw implantation process.

[0079] Figure 6 This is a schematic diagram of the deep learning-based pedicle screw implantation channel planning device provided by the present invention, as shown below. Figure 6 As shown, the device includes: an extraction module 610, a layering module 620, and a planning module 630, wherein,

[0080] Extraction module 610 is used to: extract pedicle CT images from the spinal CT images to be extracted;

[0081] The layering module 620 is used to: slice and layer the pedicle CT image along the cross-section of the pedicle, confirm the pedicle boundary, and obtain multi-layer pedicle two-dimensional images in sequence;

[0082] The planning module 630 is used to: sequentially acquire the center point of the two-dimensional image of the pedicle, fit the center point, and acquire the pedicle screw implantation channel.

[0083] The deep learning-based pedicle screw implantation channel planning device provided by this invention extracts pedicle CT images from spinal CT images, then performs layer processing and boundary confirmation to obtain two-dimensional pedicle images. The center points of the two-dimensional images are obtained sequentially, and multiple center points are fitted to obtain the final pedicle screw implantation channel. The acquisition of this pedicle screw implantation channel relies entirely on automatic calculation using image processing technology, without the need for human assistance from doctors. This achieves automatic planning of the implantation channel and standardizes and accelerates the process of acquiring the implantation channel.

[0084] According to the present invention, a deep learning-based pedicle screw implantation channel planning device is provided, the device further comprising an intersection module, the intersection module being used to: extend the pedicle screw implantation channel, with the intersection point with the zygomatic ridge as the starting point of the pedicle screw implantation channel, and the intersection point with the vertebral body as the ending point of the pedicle screw implantation channel, and to obtain the length of the pedicle screw implantation channel based on the starting point and the ending point.

[0085] The deep learning-based pedicle screw implantation channel planning device provided by this invention uses the intersection of the extension line of the pedicle screw implantation channel obtained by the 3D segmentation network model with the zigzag ridge as the starting point of the pedicle screw implantation channel and the intersection with the vertebral body as the ending point. Based on this, the specific starting and ending points of the pedicle screw implantation channel are clearly defined, and the length of the pedicle screw implantation channel is obtained at the same time. Based on this, the pedicle screw implantation channel can be planned more accurately, which is beneficial to its application in actual surgery and ensures the accuracy of the actual screw implantation process.

[0086] According to the present invention, a deep learning-based pedicle screw implantation channel planning device is provided, the device further includes a preprocessing module, which is used to: normalize the three-dimensional inter-slice spacing in the spinal CT image to be processed, and center the normalized data to 0 values ​​to obtain the spinal CT image to be extracted.

[0087] The deep learning-based pedicle screw implantation channel planning device provided by this invention achieves image data standardization by performing normalization and zero-value centering on the spinal CT image to be processed in sequence. Based on this, the convergence speed of the network model can be accelerated when applying the standard image data processing network model, and the processing speed of the network model can also be improved when the data is input into the network model.

[0088] According to the deep learning-based pedicle screw implantation channel planning device provided by the present invention, the layering module 620, when used to confirm the pedicle boundary and sequentially obtain multi-layer pedicle two-dimensional images, specifically performs the following operations: performs opening and closing operations on the pedicle CT images after slice layering, counts the number of connected components in the slice layering CT images, determines whether the number of connected components in the slice layering CT images is greater than a preset threshold, and if the condition is met, retains the connected component as a target connected component; if the condition is not met, discards the connected component, and sequentially obtains multi-layer pedicle two-dimensional images based on the target connected component.

[0089] The deep learning-based pedicle screw implantation channel planning device provided by this invention performs opening and closing operations on the pedicle CT images after slice layering and performs image filtering based on the number of connected components. Based on this, irrelevant images are removed, and the images are optimized. This helps to avoid interference from irrelevant images in the pedicle screw implantation channel planning process and achieves accurate planning.

[0090] According to the present invention, a deep learning-based pedicle screw implantation channel planning device is provided. The device uses a gray-scale centroid method to obtain the center point of the two-dimensional pedicle image. The formula for the gray-scale centroid method is:

[0091]

[0092]

[0093] Where (u,v) represents the coordinates of a pixel, f(u,v) represents the gray value of pixel (u,v), and Ω represents the set of pixels in the pedicle 2D image. The x-coordinate of the center point The ordinate represents the center point.

[0094] The deep learning-based pedicle screw implantation channel planning device provided by this invention achieves accurate acquisition of the center point of the pedicle in a two-dimensional image through the gray-scale centroid method, which is beneficial for subsequent accurate planning of the pedicle screw implantation channel based on the center point.

[0095] According to a deep learning-based pedicle screw implantation channel planning device provided by the present invention, the planning module 630, when fitting the center point to obtain the pedicle screw implantation channel, is specifically used to obtain a fitted straight line equation based on the fitted straight line function and the minimum sum of squared errors function of the center point, and obtain the pedicle screw implantation channel based on the fitted straight line equation; wherein, the fitted straight line function is expressed as: z=ax+by+c, and the minimum sum of squared errors function is expressed as: a, b, and c represent the parameters to be determined for the fitted linear function, n represents the order of the center points, and N represents the number of center points.

[0096] The deep learning-based pedicle screw implantation channel planning device provided by this invention sets a fitting straight line function and obtains the parameters in the fitting straight line function by using the minimum error sum of squares function, and finally obtains the fitting straight line equation, realizing the mathematical expression of the pedicle screw implantation channel and ensuring the visual and quantitative description of the pedicle screw implantation channel.

[0097] According to the present invention, a deep learning-based pedicle screw implantation channel planning device further includes a splitting module, which is used to split the spinal CT image data to be processed into a training dataset, a validation dataset, and a test dataset; wherein the training dataset is used to train a neural network, the validation dataset is used to adjust the hyperparameters of the neural network, and the test dataset is used to verify the accuracy of the neural network.

[0098] The deep learning-based pedicle screw implantation channel planning device provided by this invention splits the spinal CT image data to be processed into training datasets, validation datasets, and test datasets, and reasonably limits the proportional relationship between each dataset. Based on this, the quality of the network model trained by the above datasets can be improved, ensuring that the network model has good prediction accuracy.

[0099] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7 As shown, the electronic device may include a processor 710, a communication interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other through the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute a deep learning-based pedicle screw implantation channel planning method. This method includes: an extraction process, extracting pedicle CT images from the spinal CT images to be extracted; a layering process, slicing the pedicle CT images along the cross-section of the pedicle to confirm the pedicle boundaries, and sequentially obtaining multi-layer pedicle two-dimensional images; and a planning process, sequentially obtaining the center points of the pedicle two-dimensional images, fitting the center points, and obtaining the pedicle screw implantation channel.

[0100] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0101] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the deep learning-based pedicle screw implantation channel planning method provided by the above methods, the method comprising: an extraction process, extracting pedicle CT images from a spinal CT image to be extracted; a layering process, performing slicing and layering processing on the pedicle CT images along the cross-section of the pedicle, confirming the pedicle boundaries, and sequentially obtaining multi-layer pedicle two-dimensional images; and a planning process, sequentially obtaining the center points of the pedicle two-dimensional images, fitting the center points, and obtaining the pedicle screw implantation channel.

[0102] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned deep learning-based pedicle screw implantation channel planning methods. The method includes: an extraction process, extracting pedicle CT images from a spinal CT image to be extracted; a layering process, slicing the pedicle CT images along the pedicle cross-section to layer them, confirming the pedicle boundaries, and sequentially obtaining multiple layers of pedicle two-dimensional images; and a planning process, sequentially obtaining the center points of the pedicle two-dimensional images, fitting the center points, and obtaining the pedicle screw implantation channel.

[0103] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0104] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part 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, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A deep learning-based method for pedicle screw implantation channel planning, characterized in that, include: The extraction process involves extracting pedicle CT images from the spinal CT images to be extracted. Specifically, the spinal CT image data to be extracted is input into a 3D segmentation network model, and the pedicle CT images are obtained through the extraction processing of the 3D segmentation network model. The process involves slicing the pedicle CT image along its cross-section to identify the pedicle boundaries, thereby obtaining multiple layers of pedicle 2D images. Specifically, the 3D segmentation network model is used to slice the pedicle CT image along its cross-section to obtain multiple slice images. The pedicle boundaries on each slice image are then identified to obtain multiple layers of pedicle 2D images. Specifically, this includes: performing opening and closing operations on the pedicle CT image after slice-layer processing; counting the number of connected components in the slice-layer pedicle CT image; determining whether the number of connected components in the slice-layer pedicle CT image is greater than a preset threshold; if the condition is met, the connected component is retained as a target connected component; if the condition is not met, the connected component is discarded; and the multi-layer pedicle two-dimensional image is obtained sequentially based on the target connected component. Here, the connected component refers to an image region composed of foreground pixels with the same pixel value and adjacent positions. The planning process involves sequentially acquiring the center point of the pedicle two-dimensional image, fitting the center point, and obtaining the pedicle screw implantation channel.

2. The deep learning-based pedicle screw implantation channel planning method according to claim 1, characterized in that, The method further includes: The pedicle screw implantation channel is extended, with the intersection with the vertebral ridge as the starting point of the pedicle screw implantation channel and the intersection with the vertebral body as the ending point of the pedicle screw implantation channel. The length of the pedicle screw implantation channel is obtained based on the starting point and the ending point.

3. The deep learning-based pedicle screw implantation channel planning method according to claim 1, characterized in that, Prior to the extraction process, the method further includes: The preprocessing procedure involves normalizing the three-dimensional interslice spacing in the spinal CT image to be processed, and then centering the normalized data to zero to obtain the spinal CT image to be extracted.

4. The deep learning-based pedicle screw implantation channel planning method according to claim 1, characterized in that, The center point of the pedicle 2D image is obtained using the gray-scale centroid method, the formula of which is: Where (u,v) represents the coordinates of a pixel, f(u,v) represents the gray value of pixel (u,v), and Ω represents the set of pixels in the pedicle 2D image. The x-coordinate of the center point The ordinate represents the center point.

5. The deep learning-based pedicle screw implantation channel planning method according to claim 1, characterized in that, The process of fitting the center point to obtain the pedicle screw implantation channel specifically includes: The fitted line equation is obtained based on the fitted line function and the minimum sum of squared errors function of the center point, and the pedicle screw implantation channel is obtained based on the fitted line equation; wherein, the fitted line function is expressed as: z = ax + by + c, and the minimum sum of squared errors function is expressed as: a, b, and c represent the parameters to be determined for the fitted linear function, n represents the order of the center points, and N represents the number of center points.

6. The deep learning-based pedicle screw implantation channel planning method according to claim 3, characterized in that, The spinal CT image data to be processed is split into a training dataset, a validation dataset, and a test dataset; wherein, the training dataset is used to train the neural network, the validation dataset is used to adjust the hyperparameters of the neural network, and the test dataset is used to verify the accuracy of the neural network.

7. A deep learning-based pedicle screw implantation channel planning device, characterized in that, include: The extraction module is used to: extract pedicle CT images from the spinal CT images to be extracted, wherein the spinal CT image data to be extracted is input into a 3D segmentation network model, and the pedicle CT images are obtained through the extraction processing of the 3D segmentation network model; The layering module is used to: slice and layer the pedicle CT image along the cross-section of the pedicle, confirm the pedicle boundary, and obtain multi-layer pedicle two-dimensional images in sequence; wherein, the pedicle CT image is sliced ​​and layered along the cross-section of the pedicle via the 3D segmentation network model to obtain multi-layer slice images in sequence; and the pedicle boundary on the multi-layer slice images is confirmed in sequence to obtain multi-layer pedicle two-dimensional images. Specifically, this is used to: perform opening and closing operations on the pedicle CT image after slice-layer processing, count the number of connected components in the pedicle CT image after slice-layer processing, determine whether the number of connected components in the pedicle CT image after slice-layer processing is greater than a preset threshold, and if the condition is met, retain the connected component as a target connected component; if the condition is not met, discard the connected component, and obtain the multi-layer pedicle two-dimensional image sequentially based on the target connected component, wherein the connected component refers to an image region composed of foreground pixels with the same pixel value and adjacent positions in the image; The planning module is used to: sequentially acquire the center point of the two-dimensional image of the pedicle, fit the center point, and acquire the pedicle screw implantation channel.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the deep learning-based pedicle screw implantation channel planning method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the deep learning-based pedicle screw implantation channel planning method as described in any one of claims 1 to 6.

Images