An Adaptive Segmentation Method and System for Spinal Images

Abstract

The present application provides an adaptive segmentation method and system for spinal images. By extracting the local texture features of each pixel point in the spinal image, multiple feature pairs are constructed based on all the local texture features and the gray gradient of each pixel point; multiple joint probability distributions corresponding to the spinal image and a segmentation threshold interval during spinal image segmentation are determined according to all the feature pairs, and the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold in the segmentation threshold interval are determined; based on all the joint probability distributions, the intra-class features and inter-class features of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold are respectively determined, and the segmentation confidence of each pre-segmentation threshold is calculated according to all the intra-class features and inter-class features; the target spinal region is segmented from the spinal image according to the optimal segmentation threshold extracted based on all the segmentation confidences. It can achieve the adaptive segmentation of spinal images under the condition that other tissue structures around the target spine are complex and variable.

Description

Technical Field

[0001] This application relates to the technical field of medical image processing. More specifically, this application relates to an adaptive segmentation method and system for spinal images. Background Art

[0002] Medical image processing is an important technology in the cross - field of modern medicine and artificial intelligence. With the continuous development of medical imaging technologies, such as X - rays, ultrasounds, etc., a large amount of medical images have been generated. These images contain rich patient health information. However, traditional medical image processing mainly relies on manual viewing, which has limitations such as low efficiency, strong subjectivity, and easy fatigue. Therefore, processing medical images through computer vision and deep learning algorithms can effectively analyze and identify medical images, thus providing an important basis for surgical navigation and treatment planning.

[0003] In existing medical image processing, medical image processing mainly digitally analyzes and optimizes medical images through computer algorithms. First, pre - process the medical image to improve the image quality; then, separate the region of interest (such as a diseased tissue) from the medical image; then, extract key features from the separated region of interest; finally, perform intelligent diagnosis assistance based on the extracted key features. However, in spinal image processing, due to the complex and variable other tissue structures around the target spine in spinal region images, fixed - threshold segmentation is difficult to adapt to the variation differences under different spinal states, easily resulting in fuzzy segmentation (i.e., over - segmentation or under - segmentation) of the target spine region in spinal images. Therefore, how to achieve adaptive segmentation of spinal images under the condition that other tissue structures around the target spine are complex and variable has become a problem faced by the industry. Summary of the Invention

[0004] This application provides an adaptive segmentation method and system for spinal images, which can achieve adaptive segmentation of spinal images under the condition that other tissue structures around the target spine are complex and variable.

[0005] In a first aspect, this application provides an adaptive segmentation method for spinal images, including the following steps:

[0006] Collect spinal images of a target patient;

[0007] Based on a preset image sliding window, perform local feature extraction on each pixel point in the spinal image to obtain the local texture feature of each pixel point, and construct a feature binary group for each pixel point through all the local texture features and the gray - level gradient of each pixel point;

[0008] Determine multiple joint probability distributions corresponding to the spinal image and the segmentation threshold interval during spinal image segmentation according to all the feature pairs, and divide all the pixels in the spinal image according to each pre-segmentation threshold in the segmentation threshold interval to obtain the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold;

[0009] Based on all the joint probability distributions, respectively determine the intra-class features and inter-class features of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold, and then calculate the segmentation confidence of each pre-segmentation threshold according to all the intra-class features and inter-class features;

[0010] Extract the optimal segmentation threshold of the spinal image from all the pre-segmentation thresholds according to each segmentation confidence, and segment the target spinal region from the spinal image according to the optimal segmentation threshold.

[0011] In some embodiments, local feature extraction is performed on each pixel point in the spinal image based on a preset image sliding window, and obtaining the local texture feature of each pixel point specifically includes:

[0012] Obtain a preset image sliding window and set the sliding step length of the image sliding window;

[0013] Take each pixel point as the center of the image sliding window, and slide the image sliding window on the spinal image according to the sliding step length until each pixel point in the spinal image is traversed, so as to obtain the local pixel block corresponding to each pixel point;

[0014] Determine the local texture feature of each pixel point according to the local pixel block corresponding to each pixel point.

[0015] In some embodiments, constructing the feature pair of each pixel point through all the local texture features and the gray level gradient of each pixel point specifically includes:

[0016] Obtain the local texture feature and gray level gradient corresponding to each pixel point;

[0017] Determine the texture influence coefficient and gray level gradient influence coefficient of each pixel point;

[0018] Based on the texture influence coefficient and gray level gradient influence coefficient of each pixel point, weight the local texture feature and gray level gradient corresponding to each pixel point to obtain the weighted local texture feature and weighted gray level gradient value of each pixel point;

[0019] Concatenate the weighted local texture feature and weighted gray level gradient value of each pixel point, and then obtain the feature pair of each pixel point.

[0020] In some embodiments, determining the multiple joint probability distributions corresponding to the spinal image and the segmentation threshold interval in spinal image segmentation according to all feature pairs specifically includes:

[0021] Obtain the total number of pixels in the spinal image;

[0022] Extract each different feature pair from all the feature pairs, and calculate the number of times each different feature pair appears in the spinal image;

[0023] Determine the multiple joint probability distributions corresponding to the spinal image according to the total number of pixels and the number of times each different feature pair appears in the spinal image;

[0024] Extract the feature pair with the most occurrences and the feature pair with the least occurrences in the spinal image;

[0025] Determine the upper limit of the segmentation threshold in spinal image segmentation according to the set of pixel points corresponding to the feature pair with the most occurrences;

[0026] Determine the lower limit of the segmentation threshold in spinal image segmentation through the set of pixel points corresponding to the feature pair with the least occurrences;

[0027] Construct the segmentation threshold interval in spinal image segmentation from the upper limit of the segmentation threshold and the lower limit of the segmentation threshold.

[0028] In some embodiments, dividing all the pixels in the spinal image according to each pre-segmentation threshold in the segmentation threshold interval to obtain the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold specifically includes:

[0029] Select one pre-segmentation threshold in the segmentation threshold interval as the selected pre-segmentation threshold;

[0030] Compare the selected pre-segmentation threshold with the gray value of each pixel point in the spinal image, divide the pixel points with gray values greater than or equal to the selected pre-segmentation threshold into the target clustering cluster, and divide the pixel points with gray values less than the selected pre-segmentation threshold into the background clustering cluster, so as to obtain the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold;

[0031] Continue to determine the background clustering cluster and the target clustering cluster corresponding to the remaining pre-segmentation thresholds.

[0032] In some embodiments, calculating the segmentation confidence of each pre-segmentation threshold according to all intra-class features and inter-class features specifically includes:

[0033] Obtain the inter-class features and intra-class features corresponding to each pre-segmentation threshold;

[0034] For each inter-class feature and intra-class feature corresponding to a pre-segmentation threshold, perform segmentation score verification to obtain the segmentation confidence of each pre-segmentation threshold.

[0035] In some embodiments, a spinal image of a target patient is acquired by computed tomography.

[0036] In a second aspect, the present application provides an adaptive segmentation system for spinal images, including:

[0037] An acquisition module for acquiring a spinal image of a target patient;

[0038] A processing module for performing local feature extraction on each pixel point in the spinal image based on a preset image sliding window to obtain the local texture feature of each pixel point, and constructing a feature binary group of each pixel point through all the local texture features and the gray gradient of each pixel point;

[0039] The processing module is further configured to determine a plurality of joint probability distributions corresponding to the spinal image and a segmentation threshold interval during spinal image segmentation according to all the feature binary groups, divide all the pixels in the spinal image according to each pre-segmentation threshold in the segmentation threshold interval, and obtain a background clustering cluster and a target clustering cluster corresponding to each pre-segmentation threshold;

[0040] The processing module is further configured to respectively determine the intra-class feature and the inter-class feature of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold based on all the joint probability distributions, and further calculate the segmentation confidence of each pre-segmentation threshold according to all the intra-class features and inter-class features;

[0041] An execution module for extracting the optimal segmentation threshold of the spinal image from all the pre-segmentation thresholds according to each segmentation confidence, and segmenting a target spinal region from the spinal image according to the optimal segmentation threshold.

[0042] In a third aspect, the present application provides a computer device, which includes a memory and a processor. The memory stores code, and the processor is configured to obtain the code and execute the above-mentioned adaptive segmentation method for spinal images.

[0043] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned adaptive segmentation method for spinal images is implemented.

[0044] The technical solutions provided by the disclosed embodiments of the present application have the following beneficial effects:

[0045] In the adaptive segmentation method and system for spinal images provided by this application, first, spinal images of a target patient are collected; second, local feature extraction is performed on each pixel point in the spinal images based on a preset image sliding window to obtain the local texture features of each pixel point, and a feature pair of each pixel point is constructed through all the local texture features and the gray level gradient of each pixel point; further, a plurality of joint probability distributions corresponding to the spinal images and a segmentation threshold interval during spinal image segmentation are determined according to all the feature pairs, and all the pixels in the spinal images are divided according to each pre-segmentation threshold in the segmentation threshold interval to obtain a background clustering cluster and a target clustering cluster corresponding to each pre-segmentation threshold; then, the intra-class features and inter-class features of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold are respectively determined based on all the joint probability distributions, and further, the segmentation confidence of each pre-segmentation threshold is calculated according to all the intra-class features and inter-class features; finally, the optimal segmentation threshold of the spinal images is extracted from all the pre-segmentation thresholds according to each segmentation confidence, and a target spinal region is segmented from the spinal images according to the optimal segmentation threshold.

[0046] It can be seen that the present application can achieve adaptive segmentation of spinal images under the complex and variable other tissue structures around the target spine. First, based on a preset image sliding window, local texture features of each pixel point in the spinal image are extracted, and a feature binary group of each pixel point is constructed through all the local texture features and the gray gradient of each pixel point to provide a more comprehensive feature description, making the segmentation algorithm more accurate when identifying the spinal region. Second, according to all the feature binary groups, multiple joint probability distributions corresponding to the spinal image and a segmentation threshold interval during spinal image segmentation are determined. By constructing the joint probability distribution, the distribution of all pixel points can be statistically obtained to estimate the main demarcation point between the spinal region and the background region. By determining the segmentation threshold interval, the optimal segmentation threshold in the spinal image can be effectively extracted. Further, all pixels in the spinal image are divided according to each pre-segmentation threshold in the segmentation threshold interval, and the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold are obtained, which can more finely distinguish the spinal region and the background region to better adapt to different brightness or contrast of the image, ensure accurate extraction of the spinal region, and further avoid fuzzy segmentation caused by the complex and variable other tissue structures around the target spine. Still further, based on all the joint probability distributions, the intra-class features and inter-class features of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold are respectively determined to effectively evaluate the discrimination degree between the target clustering cluster and the background clustering cluster under different thresholds, and then find the threshold that can best distinguish the target from the background to optimize the segmentation result. Then, based on all the intra-class features and inter-class features, the optimal segmentation threshold of the spinal image is obtained to adaptively obtain the best segmentation threshold. Finally, the target spinal region is segmented from the spinal image according to the optimal segmentation threshold to obtain the most accurate segmentation effect, reduce the cases of mis-segmentation and missed-segmentation, and further improve the accuracy and reliability of spinal image processing. In summary, the technical solution provided by the present application can achieve adaptive segmentation of spinal images under the complex and variable other tissue structures around the target spine. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is an exemplary flowchart of an adaptive segmentation method of a spinal image according to some embodiments of the present application;

[0048] Figure 2 is an exemplary flowchart of determining local texture features according to some embodiments of the present application;

[0049] Figure 3 is an exemplary flowchart of determining segmentation confidence according to some embodiments of the present application;

[0050] Figure 4 is a schematic structural diagram of an adaptive segmentation system of a spinal image according to some embodiments of the present application;

[0051] Figure 5 It is a schematic structural diagram of a computer device for implementing an adaptive segmentation method of spinal images as shown in some embodiments of the present application. Detailed implementation manners

[0052] To better understand the technical solution of the present application, the technical solution of the present application will be described in detail below in conjunction with the accompanying drawings of the specification and specific implementation manners.

[0053] Referring to Figure 1 , this figure is an exemplary flowchart of an adaptive segmentation method of spinal images as shown in some embodiments of the present application. The adaptive segmentation method 100 of the spinal images mainly includes the following steps:

[0054] In step 101, a spinal image of a target patient is acquired.

[0055] In specific implementation, a spinal image of a target patient can be acquired by computed tomography (CT). The spinal image represents an image for displaying spinal structure and pathological information, and can be specifically obtained through medical imaging technology. In addition, in other embodiments, other acquisition devices or acquisition sensors can also be used to acquire the spinal image of the target patient. For example, magnetic resonance imaging (MRI) scanning, ultrasonic scanning, which are not limited here.

[0056] In step 102, local feature extraction is performed on each pixel point in the spinal image based on a preset image sliding window to obtain the local texture feature of each pixel point, and a feature binary group of each pixel point is constructed through all the local texture features and the gray gradient of each pixel point.

[0057] In some embodiments, referring to Figure 2 shown in this figure, which is an exemplary flowchart of determining local texture features as shown in some embodiments of the present application. In this embodiment, local feature extraction is performed on each pixel point in the spinal image based on a preset image sliding window, and the local texture feature of each pixel point can be implemented by the following steps:

[0058] First, in step 1021, a preset image sliding window is obtained, and the sliding step length of the image sliding window is set;

[0059] Then, in step 1022, each pixel point is used as the center of the image sliding window, and the image sliding window slides on the spinal image according to the sliding step length until each pixel point in the spinal image is traversed, so as to obtain the local pixel block corresponding to each pixel point;

[0060] Finally, in step 1023, the local texture feature of each pixel is determined according to the local pixel block corresponding to each pixel.

[0061] In specific implementation, a preset image sliding window can be obtained through the scoliosis correction database. In this application, the image sliding window is set to a matrix size of 3×3. In addition, in other embodiments, it can also be set to a matrix size of 5×5 or 7×7, which is not limited here. In addition, in this application, the sliding step of the image sliding window is set to 1 pixel.

[0062] It should be noted that in this embodiment, the scoliosis correction database represents a database storing data related to scoliosis correction, including recognition data of spinal images and data such as scoliosis samples; in this application, the image sliding window represents a fixed window for extracting local features of an image.

[0063] In specific implementation, each pixel is used as the center of the image sliding window, and the image sliding window slides on the spinal image in the order from left to right and from top to bottom according to the sliding step until each pixel in the spinal image is traversed, so as to obtain the local pixel block corresponding to each pixel; it should be noted that in this embodiment, each pixel corresponds to a local pixel block, and the local pixel block is constructed with the corresponding pixel as the center point. The local pixel block represents the local pixel area in the spinal image. By determining the local pixel area, the local features corresponding to each pixel can be effectively extracted.

[0064] Among them, in some embodiments, determining the local texture feature of each pixel according to the local pixel block corresponding to each pixel can be implemented by the following steps, that is:

[0065] Select a pixel as the selected pixel, and obtain all the gray values in the local pixel block corresponding to the selected pixel;

[0066] Take the mean value of the maximum gray value and the minimum gray value among all the gray values as the local texture feature of the selected pixel;

[0067] Continue to determine the local texture features of the remaining pixels.

[0068] It should be noted that in this application, the local texture feature represents a feature value for measuring the texture information in the local neighborhood of a pixel. By determining the local texture feature, the spinal shape and spinal overview of the target patient can be effectively analyzed.

[0069] In some embodiments, constructing the feature binary group of each pixel through all the local texture features and the gray gradient of each pixel can be implemented by the following steps, that is:

[0070] Obtain the local texture feature and grayscale gradient corresponding to each pixel point;

[0071] Determine the texture influence coefficient and grayscale gradient influence coefficient of each pixel point;

[0072] Based on the texture influence coefficient and grayscale gradient influence coefficient of each pixel point, weight the local texture feature and grayscale gradient corresponding to each pixel point to obtain the weighted local texture feature and weighted grayscale gradient value of each pixel point;

[0073] Concatenate the weighted local texture feature and weighted grayscale gradient value of each pixel point to obtain the feature binary group of each pixel point.

[0074] When specifically implemented, first, obtain the local texture feature and grayscale gradient corresponding to each pixel point; second, determine the texture influence coefficient and grayscale gradient influence coefficient of each pixel point, that is: obtain the local pixel block corresponding to each pixel point, for each pixel point, calculate the information entropy of all normalized pixel values in the local pixel block corresponding to the pixel point, and use the calculated information entropy as the texture influence coefficient of the pixel point, calculate the standard deviation of all normalized grayscale values in the local pixel block corresponding to the pixel point, and use the calculated standard deviation as the grayscale gradient influence coefficient of the pixel point, so as to obtain the texture influence coefficient and grayscale gradient influence coefficient of each pixel point. In addition, in other embodiments, other methods may also be used to determine the texture influence coefficient and grayscale gradient influence coefficient of each pixel point, which is not limited here; then, based on the texture influence coefficient and grayscale gradient influence coefficient of each pixel point, weight the local texture feature and grayscale gradient corresponding to each pixel point to obtain the weighted local texture feature and weighted grayscale gradient value of each pixel point, that is: for each pixel point, multiply the local texture feature and grayscale gradient of the pixel point by the corresponding texture influence coefficient and grayscale gradient influence coefficient respectively, so as to obtain the weighted local texture feature and weighted grayscale gradient value of each pixel point; finally, concatenate the weighted local texture feature and weighted grayscale gradient value of each pixel point to obtain the feature binary group of each pixel point, that is: combine the weighted local texture feature and weighted grayscale gradient value of each pixel point to obtain the feature binary group of each pixel point, and the feature binary group includes the weighted local texture feature and weighted grayscale gradient value.

[0075] It should be noted that in this embodiment, the texture influence coefficient represents a coefficient for measuring the weight of the local texture feature of a pixel point in the construction of the feature binary tuple. By determining the texture influence coefficient, it can ensure that the importance of texture features can be reasonably allocated in different types of regions. In this embodiment, the gray-scale gradient influence coefficient represents a coefficient for measuring the weight of the gray-scale gradient information of a pixel point in the construction of the feature binary tuple, and is used to adjust the contribution degree of gradient features in different regions. In this embodiment, the weighted local texture feature represents the value obtained by weighted adjustment of the local texture feature. In this embodiment, the weighted gray-scale gradient value represents the value obtained by weighted adjustment of the gray-scale gradient. In this application, the feature binary tuple represents a two-dimensional feature vector constructed by the local texture feature and the gray-scale gradient. The feature binary tuple combines the local texture and gradient features, and can provide a more comprehensive feature description, making the segmentation algorithm more accurate when identifying the spine region.

[0076] In step 103, according to all the feature binary tuples, determine multiple joint probability distributions corresponding to the spine image and the segmentation threshold interval during spine image segmentation. Divide all the pixels in the spine image according to each pre-segmentation threshold in the segmentation threshold interval to obtain the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold.

[0077] In some embodiments, determining multiple joint probability distributions corresponding to the spine image and the segmentation threshold interval during spine image segmentation according to all the feature binary tuples can adopt the following steps, that is:

[0078] Obtain the total number of pixels in the spine image;

[0079] Extract each different feature binary tuple from all the feature binary tuples, and calculate the number of times each different feature binary tuple appears in the spine image;

[0080] Determine multiple joint probability distributions corresponding to the spine image according to the total number of pixels and the number of times each different feature binary tuple appears in the spine image;

[0081] Extract the feature binary tuple with the most occurrences and the feature binary tuple with the fewest occurrences in the spine image;

[0082] Determine the upper limit of the segmentation threshold during spine image segmentation according to the set of pixel points corresponding to the feature binary tuple with the most occurrences;

[0083] Determine the lower limit of the segmentation threshold during spine image segmentation through the set of pixel points corresponding to the feature binary tuple with the fewest occurrences;

[0084] Construct the segmentation threshold interval during spine image segmentation from the upper limit of the segmentation threshold and the lower limit of the segmentation threshold.

[0085] In specific implementation, the total number of pixels of the spinal image can be obtained through the image processing tool OpenCV, and various different feature pairs can be extracted from all the feature pairs through the data processing tool Python, and the number of occurrences of each different feature pair in the spinal image can be calculated.

[0086] In specific implementation, multiple joint probability distributions corresponding to the spinal image are determined according to the total number of pixels and the number of occurrences of each different feature pair in the spinal image, that is: for each different feature pair, the quotient of the number of occurrences of the different feature pair in the spinal image and the total number of pixels is used as the joint probability distribution corresponding to the different feature pair, so as to obtain multiple joint probability distributions corresponding to the spinal image. The multiple joint probability distributions include the joint probability distributions corresponding to each different feature pair. In addition, in other embodiments, other calculation methods can also be used to calculate the multiple joint probability distributions corresponding to the spinal image, which are not limited here.

[0087] It should be noted that the joint probability distribution in this application represents the probability distribution of the co-occurrence of multiple random variables. Specifically, the joint probability distribution in this application is the joint occurrence probability of the weighted local texture feature and the weighted gray gradient value in the spinal image. In spinal image processing, there are usually significant differences in texture and gradient characteristics between the target area (i.e., the spine itself) and the background area in the spinal image. For example, the spine area usually has relatively consistent texture features (such as smooth bone edges and high density), while the background area (soft tissue, blank part) may show larger gradient changes or more complex texture distributions. By constructing the joint probability distribution, the distribution of all pixel points can be statistically analyzed to estimate the main demarcation point between the spine area and the background area. The joint probability distribution plays a core role in spinal image processing. It can not only be used to statistically analyze the distribution of pixel point features, but also be used to extract the optimal segmentation threshold interval, thereby improving the distinguishability between the spine area and the background area.

[0088] In specific implementation, the upper limit of the segmentation threshold for spine image segmentation is determined according to the set of pixel points corresponding to the feature binary group with the most occurrences, that is: the set of pixel points corresponding to the feature binary group with the most occurrences is obtained through a data processing tool, the gray mean value of all pixel points in the set of pixel points is calculated, and the gray mean value is used as the upper limit of the segmentation threshold for spine image segmentation, where the data processing tool can be Python. In addition, in other embodiments, other calculation methods can also be used to calculate the upper limit of the segmentation threshold for spine image segmentation, which is not limited here. It should be noted that the set of pixel points corresponding to the feature binary group with the most occurrences in this embodiment represents the set formed by combining the pixel points corresponding to the feature binary group with the most occurrences. The upper limit of the segmentation threshold in this embodiment represents the highest segmentation threshold for distinguishing the spine region and the background region.

[0089] In specific implementation, the lower limit of the segmentation threshold for spine image segmentation is determined according to the set of pixel points corresponding to the feature binary group with the least occurrences, that is: the set of pixel points corresponding to the feature binary group with the least occurrences is obtained through a data processing tool, the gray mean value of all pixel points in the set of pixel points is calculated, and the gray mean value is used as the upper limit of the segmentation threshold for spine image segmentation, where the data processing tool can be Python. In addition, in other embodiments, other calculation methods can also be used to calculate the upper limit of the segmentation threshold for spine image segmentation, which is not limited here. It should be noted that the set of pixel points corresponding to the feature binary group with the least occurrences in this embodiment represents the set formed by combining the pixel points corresponding to the feature binary group with the least occurrences. The lower limit of the segmentation threshold in this embodiment represents the lowest segmentation threshold for distinguishing the spine region and the background region.

[0090] In specific implementation, the segmentation threshold interval for spine image segmentation is constructed from the upper limit of the segmentation threshold and the lower limit of the segmentation threshold, that is: the upper limit of the segmentation threshold and the segmentation threshold are combined to obtain the segmentation threshold interval for spine image segmentation.

[0091] It should be noted that the segmentation threshold interval in this application represents a numerical range for distinguishing the spine region and the background region, and this range is jointly determined by the upper limit of the segmentation threshold and the lower limit of the segmentation threshold. In spine image processing, since the gray levels of tissue structures in the spine region image vary greatly, fixed-threshold segmentation is difficult to adapt to the individual differences of different patients, and it is easy to cause fuzzy segmentation of the target spine region in the spine image. Therefore, by determining the segmentation threshold interval, the optimal segmentation threshold in the spine image can be effectively extracted, and then the target spine region in the spine image can be further segmented.

[0092] In some embodiments, dividing all pixels in the spinal image according to each pre-segmentation threshold in the segmentation threshold interval to obtain the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold can be achieved by the following steps, namely:

[0093] Select a pre-segmentation threshold in the segmentation threshold interval as the selected pre-segmentation threshold;

[0094] Compare the selected pre-segmentation threshold with the gray value of each pixel point in the spinal image. Divide the pixel points with gray values greater than or equal to the selected pre-segmentation threshold into the target clustering cluster, and divide the pixel points with gray values less than the selected pre-segmentation threshold into the background clustering cluster, thereby obtaining the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold;

[0095] Continue to determine the background clustering cluster and the target clustering cluster corresponding to the remaining pre-segmentation thresholds.

[0096] Specifically, first, select a pre-segmentation threshold in the segmentation threshold interval as the selected pre-segmentation threshold, and the pre-segmentation threshold is obtained by uniform sampling between the lower limit and the upper limit of the segmentation threshold; then, compare the selected pre-segmentation threshold with the gray value of each pixel point in the spinal image. Divide the pixel points with gray values greater than or equal to the selected pre-segmentation threshold into the target clustering cluster, and divide the pixel points with gray values less than the selected pre-segmentation threshold into the background clustering cluster, thereby obtaining the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold; finally, continue to determine the background clustering cluster and the target clustering cluster corresponding to the remaining pre-segmentation thresholds according to the determination method of "compare the selected pre-segmentation threshold with the gray value of each pixel point in the spinal image. Divide the pixel points with gray values greater than or equal to the selected pre-segmentation threshold into the target clustering cluster, and divide the pixel points with gray values less than the selected pre-segmentation threshold into the background clustering cluster, thereby obtaining the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold".

[0097] It should be noted that in this application, the background clustering cluster represents a set of pixels divided into the background area. The background clustering cluster contains pixels with relatively low gray values, which do not belong to the target spine area but represent the surrounding non-spine parts, such as soft tissues, air, or other background information; the target clustering cluster in this application represents a set of pixels divided into the target spine area. The target clustering cluster contains pixels with relatively high gray values, and these pixels represent the areas of concern in the spine image, that is, the spine or spine-related structures. By determining the target clustering cluster and the background clustering cluster under different pre-segmentation thresholds, the spine area and the background area can be more precisely distinguished. By adjusting different thresholds, it is possible to better adapt to different brightness or contrast of the image and ensure the accurate extraction of the spine area. For example, for the background and target areas in the spine image with relatively close gray levels, the most suitable segmentation effect can be found by trying different thresholds multiple times. In addition, the spine image may contain noise, artifacts, or irregular shapes. By dividing the background and target clustering clusters under different pre-segmentation thresholds, these interference factors can be more flexibly dealt with. For example, some noise may affect the segmentation effect under a specific threshold, but through multi-threshold analysis, the noise can be effectively removed and the target spine area can be correctly calibrated.

[0098] In step 104, based on all the joint probability distributions, the within-class features and between-class features of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold are respectively determined, and then the segmentation confidence of each pre-segmentation threshold is calculated based on all the within-class features and between-class features.

[0099] In some embodiments, determining the within-class features and between-class features of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold based on all the joint probability distributions can be implemented by the following steps, that is:

[0100] Select a pre-segmentation threshold as the selected pre-segmentation threshold;

[0101] Determine the background feature quantity of the selected pre-segmentation threshold according to the corresponding set of joint probability distributions in the background clustering cluster corresponding to the selected pre-segmentation threshold;

[0102] Determine the target feature quantity of the selected pre-segmentation threshold through the corresponding set of joint probability distributions in the target clustering cluster corresponding to the selected pre-segmentation threshold;

[0103] Obtain the set of pixel points in the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold;

[0104] Determine the within-class features and between-class features of the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold based on the background feature quantity, the target feature quantity, and the set of pixel points in the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold;

[0105] Continue to determine the intra-class features and inter-class features of the background clustering cluster and the target clustering cluster corresponding to the remaining pre-segmentation thresholds.

[0106] In specific implementation, determine the background feature quantity of the selected pre-segmentation threshold according to the corresponding joint probability distribution set in the background clustering cluster corresponding to the selected pre-segmentation threshold, that is: obtain the corresponding joint probability distribution set in the background clustering cluster corresponding to the selected pre-segmentation threshold, and the joint probability distribution set is composed of the joint probability distributions corresponding to each pixel point in the background clustering cluster corresponding to the pre-segmentation threshold. Calculate the mean value of all the joint probability distributions in the joint probability distribution set, and use the mean calculation result as the background feature quantity of the selected pre-segmentation threshold. In addition, in other embodiments, other calculation methods can also be used to calculate the background feature quantity of the selected pre-segmentation threshold, which is not limited here. It should be noted that in this embodiment, the background feature quantity represents the statistical feature describing the joint probability distribution in the background clustering cluster, and the background feature quantity is usually used for quantitative analysis of the pixel distribution in the background region, so as to perform effective classification and threshold setting in subsequent image segmentation.

[0107] In specific implementation, determine the target feature quantity of the selected pre-segmentation threshold through the corresponding joint probability distribution set in the target clustering cluster corresponding to the selected pre-segmentation threshold, that is: obtain the corresponding joint probability distribution set in the target clustering cluster corresponding to the selected pre-segmentation threshold, and the joint probability distribution set is composed of the joint probability distributions corresponding to each pixel point in the target clustering cluster corresponding to the pre-segmentation threshold. Calculate the mean value of all the joint probability distributions in the joint probability distribution set, and use the mean calculation result as the target feature quantity of the selected pre-segmentation threshold. In addition, in other embodiments, other calculation methods can also be used to calculate the target feature quantity of the selected pre-segmentation threshold, which is not limited here. It should be noted that in this embodiment, the target feature quantity represents the statistical feature describing the joint probability distribution in the target clustering cluster, and the target feature quantity is usually used for quantitative analysis of the pixel distribution in the target region, so as to perform effective classification and threshold setting in subsequent image segmentation.

[0108] In specific implementation, use the data processing tool Python to obtain the pixel point sets in the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold. In addition, in other embodiments, other data processing tools can also be used to obtain the pixel point sets in the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold, which is not limited here.

[0109] Among them, in some embodiments, determining the intra-class features and inter-class features of the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold according to the background feature quantity, the target feature quantity, and the pixel point sets in the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold can be implemented by the following steps, that is:

[0110] Calculate the gray variance of the set of pixel points of the background clustering cluster corresponding to the selected pre-segmentation threshold, and use the calculated gray variance as the within-class gray feature of the background clustering cluster corresponding to the selected pre-segmentation threshold;

[0111] Calculate the gray variance of the set of pixel points of the target clustering cluster corresponding to the selected pre-segmentation threshold, and use the calculated gray variance as the within-class gray feature of the target clustering cluster corresponding to the selected pre-segmentation threshold;

[0112] Determine the within-class features of the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold according to the within-class gray feature of the background clustering cluster, the within-class gray feature of the target clustering cluster, the background feature quantity, and the target feature quantity;

[0113] Calculate the gray mean value of the set of pixel points of the background clustering cluster corresponding to the selected pre-segmentation threshold;

[0114] Calculate the gray mean value of the set of pixel points of the target clustering cluster corresponding to the selected pre-segmentation threshold;

[0115] Determine the between-class features of the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold according to the gray mean value of the set of pixel points of the background clustering cluster, the gray mean value of the set of pixel points of the target clustering cluster, the background feature quantity, and the target feature quantity.

[0116] In specific implementation, first, calculate the gray variance of the set of pixel points of the background clustering cluster corresponding to the selected pre-segmentation threshold, and use the calculated gray variance as the intra-class gray feature of the background clustering cluster corresponding to the selected pre-segmentation threshold; second, calculate the gray variance of the set of pixel points of the target clustering cluster corresponding to the selected pre-segmentation threshold, and use the calculated gray variance as the intra-class gray feature of the target clustering cluster corresponding to the selected pre-segmentation threshold; further, determine the intra-class feature between the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold according to the intra-class gray feature of the background clustering cluster, the intra-class gray feature of the target clustering cluster, the background feature quantity and the target feature quantity, that is: sum the product result of the intra-class gray feature of the background clustering cluster and the background feature quantity and the product result of the intra-class gray feature of the target clustering cluster and the target feature quantity, and use the sum result as the intra-class feature between the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold; furthermore, calculate the gray mean value of the set of pixel points of the background clustering cluster corresponding to the selected pre-segmentation threshold; then, calculate the gray mean value of the set of pixel points of the target clustering cluster corresponding to the selected pre-segmentation threshold; finally, determine the inter-class feature between the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold according to the gray mean value of the set of pixel points of the background clustering cluster, the gray mean value of the set of pixel points of the target clustering cluster, the background feature quantity and the target feature quantity, that is: calculate the variance of the product result of the gray mean value of the set of pixel points of the background clustering cluster and the background feature quantity and the product result of the gray mean value of the set of pixel points of the target clustering cluster and the target feature quantity, and use the variance calculation result as the inter-class feature between the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold.

[0117] It should be noted that the intra-class gray feature of the background clustering cluster in this embodiment represents the descriptive feature of the gray values of the pixel points inside the background clustering cluster; the intra-class gray feature of the target clustering cluster in this embodiment represents the descriptive feature of the gray values of the pixel points inside the target clustering cluster; the intra-class feature in this application represents the fusion descriptive feature of the gray values of the pixel points in the background clustering cluster and the target clustering cluster; the inter-class feature in this application represents the differential descriptive feature of the gray values of the pixel points between the background clustering cluster and the target clustering cluster. By determining the intra-class feature and the inter-class feature between the background clustering cluster and the target clustering cluster, the distinguishability between the target clustering cluster and the background clustering cluster under different thresholds can be effectively evaluated. Specifically, the intra-class feature reflects the concentration degree of the pixel values in the same class, while the inter-class feature measures the separation degree between different classes. By comparing the performances of these features under different thresholds, the threshold that can best distinguish the target from the background can be found to optimize the segmentation result.

[0118] In some embodiments, refer to Figure 3As shown, the figure is an exemplary flowchart for determining the segmentation confidence according to some embodiments of the present application. In this embodiment, the segmentation confidence of each pre-segmentation threshold calculated based on all intra-class features and inter-class features can be implemented by the following steps:

[0119] First, in step 1041, obtain the inter-class features and intra-class features corresponding to each pre-segmentation threshold;

[0120] Then, in step 1042, perform segmentation score verification on the inter-class features and intra-class features corresponding to each pre-segmentation threshold to obtain the segmentation confidence of each pre-segmentation threshold.

[0121] Specifically, in implementation, first, obtain the inter-class features and intra-class features corresponding to each pre-segmentation threshold; then, perform segmentation score verification on the inter-class features and intra-class features corresponding to each pre-segmentation threshold to obtain the segmentation confidence of each pre-segmentation threshold, that is: for each pre-segmentation threshold, take the quotient of the inter-class features corresponding to the pre-segmentation threshold and the intra-class features as the segmentation confidence of the pre-segmentation threshold, and thus obtain the segmentation confidence of each pre-segmentation threshold.

[0122] It should be noted that in the present application, the segmentation confidence represents the degree of trust of the pre-segmentation threshold in the correctness of the segmentation result. The segmentation confidence reflects the reliability of the selected segmentation threshold. A high segmentation confidence indicates that there is a clear distinction between the target spine region and the background region of the spine image, and the pre-segmentation threshold is the best under the current spine image conditions. By selecting the segmentation result with a high confidence, the accuracy of image segmentation can be improved, and the situations of mis-segmentation, missed segmentation, and the segmentation ambiguity caused by a fixed threshold can be avoided.

[0123] In step 105, extract the optimal segmentation threshold of the spine image from all the pre-segmentation thresholds according to each segmentation confidence, and segment the target spine region from the spine image according to the optimal segmentation threshold.

[0124] In some embodiments, the step of extracting the optimal segmentation threshold of the spine image from all the pre-segmentation thresholds according to each segmentation confidence can be as follows, that is:

[0125] Obtain the pre-segmentation thresholds corresponding to each segmentation confidence;

[0126] Extract the pre-segmentation threshold corresponding to the maximum segmentation confidence from all the pre-segmentation thresholds as the optimal segmentation threshold of the spine image.

[0127] It should be noted that in this application, the optimal segmentation threshold represents the best threshold that can distinguish the target spine region from the background region. By determining the optimal segmentation threshold, the most accurate segmentation effect can be obtained, reducing the cases of mis-segmentation and missed-segmentation, and thus improving the accuracy and reliability of spine image processing.

[0128] In some embodiments, the target spine region can be segmented from the spine image according to the optimal segmentation threshold by the following steps, that is:

[0129] Compare the optimal segmentation threshold with the gray values of each pixel in the spine image, set the pixels with gray values greater than or equal to the optimal segmentation threshold as the target spine region, and set the pixels less than the optimal segmentation threshold as the background region, so as to obtain the target spine region segmented from the spine image.

[0130] It should be noted that in this application, the target spine region represents the detailed region where the spine itself is located in the spine image. By determining the target spine region, the auxiliary effectiveness of the scoliosis correction assistance system can be effectively and intelligently improved.

[0131] In addition, on the other hand of this application, in some embodiments, this application provides an adaptive segmentation system for spine images. Refer to Figure 4 , this figure is a schematic structural diagram of the adaptive segmentation system for spine images according to some embodiments of this application. The adaptive segmentation system 200 for spine images includes: an acquisition module 201, a processing module 202, and an execution module 203, which are described as follows:

[0132] The acquisition module 201 is mainly used in this application to acquire the spine image of the target patient;

[0133] The processing module 202 is mainly used in this application to perform local feature extraction on each pixel point in the spine image based on a preset image sliding window to obtain the local texture feature of each pixel point, and construct a feature binary group for each pixel point through all the local texture features and the gray gradient of each pixel point;

[0134] The processing module 202 is further used to determine multiple joint probability distributions corresponding to the spine image and the segmentation threshold interval during spine image segmentation according to all the feature binary groups, and divide all the pixels in the spine image according to each pre-segmentation threshold in the segmentation threshold interval to obtain the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold;

[0135] In addition, the processing module 202 is further configured to respectively determine the intra-class features and inter-class features of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold based on all the joint probability distributions, and then calculate the segmentation confidence of each pre-segmentation threshold according to all the intra-class features and inter-class features;

[0136] The execution module 203. In this application, the execution module 203 is mainly configured to extract the optimal segmentation threshold of the spine image from all the pre-segmentation thresholds according to each segmentation confidence, and segment the target spine region from the spine image according to the optimal segmentation threshold.

[0137] In addition, this application also provides a computer device, which includes a memory and a processor. The memory stores code, and the processor is configured to obtain the code and execute the above-mentioned adaptive segmentation method for spine images.

[0138] In some embodiments, refer to Figure 5 ., this figure is a schematic structural diagram of a computer device for implementing the adaptive segmentation method of spine images according to some embodiments of this application. The adaptive segmentation method of spine images in the above embodiments can be implemented by Figure 5 The computer device shown. The computer device 300 includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0139] The processor 301 can be a general-purpose central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more for controlling the execution of the adaptive segmentation method of spine images in this application.

[0140] The communication bus 302 can be used to transmit information between the above components.

[0141] The memory 303 can be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM), or other types of dynamic storage devices that can store information and instructions. It can also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 303 can exist independently and be connected to the processor 301 through the communication bus 302. The memory 303 can also be integrated with the processor 301.

[0142] Among them, the memory 303 is used to store the program code for executing the solution of this application, and is controlled by the processor 301 to execute. The processor 301 is used to execute the program code stored in the memory 303. The program code can include one or more software modules. The determination of the adaptive segmentation method for spinal images in the above embodiments can be implemented by one or more software modules in the processor 301 and the program code in the memory 303.

[0143] The communication interface 304 uses any device such as a transceiver to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0144] In a specific implementation, as an embodiment, the computer device can include multiple processors, and each of these processors can be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, the processor can refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0145] The computer device described above may be a general-purpose computer device or a special-purpose computer device. In a specific implementation, the computer device may be a desktop computer, a laptop computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. The embodiments of the present application do not limit the type of the computer device.

[0146] In addition, the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the adaptive segmentation method of the spinal image described above is implemented.

[0147] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concept. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0148] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. An adaptive segmentation method for spinal images, characterized in that It includes the following steps: Collect the spinal image of the target patient; Based on a preset image sliding window, perform local feature extraction on each pixel point in the spinal image to obtain the local texture feature of each pixel point, and construct a feature binary group for each pixel point through all the local texture features and the gray gradient of each pixel point; Determine multiple joint probability distributions corresponding to the spinal image and the segmentation threshold interval during spinal image segmentation according to all the feature binary groups, and divide all the pixels in the spinal image according to each pre-segmentation threshold in the segmentation threshold interval to obtain the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold; Based on all the joint probability distributions, respectively determine the intra-class features and inter-class features of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold, and then calculate the segmentation confidence of each pre-segmentation threshold according to all the intra-class features and inter-class features; Extract the optimal segmentation threshold of the spinal image from all the pre-segmentation thresholds according to each segmentation confidence, and segment the target spinal region from the spinal image according to the optimal segmentation threshold.

2. The method according to claim 1, wherein Based on a preset image sliding window, performing local feature extraction on each pixel point in the spinal image to obtain the local texture feature of each pixel point specifically includes: Obtain a preset image sliding window and set the sliding step of the image sliding window; Take each pixel point as the center of the image sliding window, and slide the image sliding window on the spinal image according to the sliding step until each pixel point in the spinal image is traversed, so as to obtain the local pixel block corresponding to each pixel point; Determine the local texture feature of each pixel point according to the local pixel block corresponding to each pixel point.

3. The method according to claim 1, characterized in that Constructing the feature binary group for each pixel point through all the local texture features and the gray gradient of each pixel point specifically includes: Obtain the local texture feature and gray gradient corresponding to each pixel point; Determine the texture influence coefficient and gray gradient influence coefficient of each pixel point; Based on the texture influence coefficient and gray gradient influence coefficient of each pixel point, weight the local texture feature and gray gradient corresponding to each pixel point to obtain the weighted local texture feature and weighted gray gradient value of each pixel point; Concatenate the weighted local texture feature and weighted gray gradient value of each pixel point to obtain the feature binary group of each pixel point.

4. The method according to claim 1, wherein Determining multiple joint probability distributions corresponding to the spinal image and the segmentation threshold interval during spinal image segmentation according to all the feature binary groups specifically includes: Obtain the total number of pixels in the spinal image; Extract each different feature binary group from all the feature binary groups and calculate the number of times each different feature binary group appears in the spinal image; Determine multiple joint probability distributions corresponding to the spinal image according to the total number of pixels and the number of times each different feature binary group appears in the spinal image; Extract the feature binary group with the most occurrences and the feature binary group with the fewest occurrences in the spinal image; Determine the upper limit of the segmentation threshold during spinal image segmentation according to the set of pixel points corresponding to the most frequently occurring feature binary group; Determine the lower limit of the segmentation threshold during spinal image segmentation through the set of pixel points corresponding to the least frequently occurring feature binary group; Construct a segmentation threshold interval for spinal image segmentation from the upper limit and the lower limit of the segmentation threshold.

5. The method according to claim 1, characterized in that, Dividing all pixels in the spinal image according to each pre-segmentation threshold in the segmentation threshold interval to obtain the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold specifically includes: Select a pre-segmentation threshold in the segmentation threshold interval as the selected pre-segmentation threshold; Compare the selected pre-segmentation threshold with the gray value of each pixel point in the spinal image, divide the pixel points with gray values greater than or equal to the selected pre-segmentation threshold into the target clustering cluster, and divide the pixel points with gray values less than the selected pre-segmentation threshold into the background clustering cluster, thereby obtaining the background clustering cluster and the target clustering cluster corresponding to the selected pre-segmentation threshold; Continue to determine the background clustering cluster and the target clustering cluster corresponding to the remaining pre-segmentation thresholds.

6. The method according to claim 1, characterized in that Calculating the segmentation confidence of each pre-segmentation threshold based on all intra-class features and inter-class features specifically includes: Obtain the inter-class features and intra-class features corresponding to each pre-segmentation threshold; Perform segmentation score verification on the inter-class features and intra-class features corresponding to each pre-segmentation threshold to obtain the segmentation confidence of each pre-segmentation threshold.

7. The method according to claim 1, characterized in that Collect the spinal image of the target patient through computed tomography.

8. An adaptive segmentation system for spinal images, characterized in that, Including: A collection module for collecting the spinal image of the target patient; A processing module for extracting local features of each pixel point in the spinal image based on a preset image sliding window to obtain the local texture features of each pixel point, and constructing the feature binary group of each pixel point through all the local texture features and the gray gradient of each pixel point; The processing module is further configured to determine multiple joint probability distributions corresponding to the spinal image and a segmentation threshold interval for spinal image segmentation according to all the feature binary groups, divide all pixels in the spinal image according to each pre-segmentation threshold in the segmentation threshold interval to obtain the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold; The processing module is further configured to respectively determine the intra-class features and inter-class features of the background clustering cluster and the target clustering cluster corresponding to each pre-segmentation threshold based on all the joint probability distributions, and then calculate the segmentation confidence of each pre-segmentation threshold according to all the intra-class features and inter-class features; An execution module for extracting the optimal segmentation threshold of the spinal image from all the pre-segmentation thresholds according to each segmentation confidence, and segmenting the target spinal region from the spinal image according to the optimal segmentation threshold.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory stores code, and the processor is configured to obtain the code and execute the adaptive segmentation method of the spinal image according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the adaptive segmentation method for spinal images as described in any one of claims 1 to 7.

Images