Image Segmentation Method, Device, Medium and Product Applicable to Bone Imaging Data

Abstract

The present invention discloses an image segmentation method, device, medium and product applicable to skeletal image data. The method includes: determining an initial point set corresponding to an initial active contour set in a skeletal CT image, where all bones in the skeletal CT image are within the initial active contour; determining a weight matrix of the skeletal CT image based on a weight function, the weight matrix including the weighted sum of the gray weight and the gradient weight of each pixel in the skeletal CT image, and the weight function including a gray term and a gradient term that are both monotonically decreasing functions; solving an iterative equation based on the initial point set, the weight matrix and a predetermined iteration step length to obtain a target active contour that coincides with the bone boundary, the iterative equation being the gradient flow equation for minimizing an energy functional, and the energy functional optimizing the active contour based on the weighted sum of the weighted perimeter and the weighted area of the active contour; and determining a bone segmentation result corresponding to the target active contour in the skeletal CT image. Embodiments of the present invention can improve the accuracy of the bone segmentation result.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to an image segmentation method, device, medium and product applicable to bone image data. Background Art

[0002] Existing technologies usually perform bone segmentation on bone CT images based on boundary detection methods, region detection methods, and bone segmentation models. However, the accuracy of the bone segmentation results determined based on these methods is relatively low. For example, non-bone with a relatively high gray value is easily regarded as bone, or the accuracy of the determined bone boundary is relatively low because the cortical bone and cancellous bone cannot be accurately identified.

[0003] Therefore, there is a need to provide an image segmentation method applicable to bone image data to improve the accuracy of bone segmentation results. Summary of the Invention

[0004] The present invention provides an image segmentation method, device, medium and product applicable to bone image data to solve the technical problem that the accuracy of existing bone segmentation methods is relatively low.

[0005] According to one aspect of the present invention, there is provided an image segmentation method applicable to bone image data, including:

[0006] Determine an initial point set corresponding to an initial active contour set in a bone CT image, where all bones in the bone CT image are within the initial active contour;

[0007] Based on a weight function, determine a weight matrix of the bone CT image, where the weight matrix includes a weighted sum of the gray weight and the gradient weight of each pixel in the bone CT image, the weight function includes a gray term and a gradient term that are both monotonically decreasing functions, the gray term is used to determine the gray weight of pixels that do not meet the soft tissue gray condition, and the gradient term is used to determine the gradient weight of pixels that do not meet the soft tissue gradient condition;

[0008] Based on the initial point set, the weight matrix, and a predetermined iteration step, solve an iteration equation to obtain a target active contour that coincides with the bone boundary. The iteration equation is a gradient flow equation for minimizing an energy functional, and the energy functional optimizes the active contour based on a weighted sum of the weighted perimeter and the weighted area of the active contour. The weights used in the determination of the weighted perimeter and the weighted area come from the weight matrix.

[0009] According to another aspect of the present invention, there is provided an electronic device, where the electronic device includes:

[0010] At least one processor; and

[0011] A memory communicatively connected to the at least one processor; wherein,

[0012] The memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor is enabled to execute the image segmentation method applicable to skeletal image data according to any embodiment of the present invention.

[0013] According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for implementing, when executed by a processor, the image segmentation method applicable to skeletal image data according to any embodiment of the present invention.

[0014] According to another aspect of the present invention, there is provided a computer program product, characterized in that the computer program product includes a computer program which, when executed by a processor, implements the image segmentation method applicable to skeletal image data according to any embodiment.

[0015] In a skeletal CT image, the pixel gray value of the skeletal region is higher than that of the surrounding soft tissue region, and the pixel gradient change value at the skeletal boundary is higher than that of other regions. In the technical solution provided by the embodiments of the present invention, the weight function includes a gray term and a gradient term that are both monotonically decreasing functions. The gray term is used to determine the gray weight of pixels that do not meet the soft tissue gray condition, and the gradient term is used to determine the gradient weight of pixels that do not meet the soft tissue gradient condition; based on this feature, the weights of soft tissue pixels and skeletal pixels can be differentiated, and the contribution of soft tissue to the weighted area and weighted perimeter determined by the energy functional can be weakened, thereby reducing the influence of soft tissue on the accuracy of the weighted area and weighted perimeter of the active contour, and in the process of minimizing the energy functional, according to the characteristics of the weighted sum determined by the weight function, the active contour can be controlled to accurately approach the skeletal boundary, obtaining an accurate target active contour, and thus based on the target active contour, a skeletal segmentation result with high accuracy can be determined.

[0016] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 is a flowchart of an image segmentation method applicable to bone imaging data according to an embodiment of the present invention;

[0019] Figure 2 is a schematic diagram of an initial active contour according to an embodiment of the present invention;

[0020] Figure 3 is a schematic diagram of a weight matrix according to an embodiment of the present invention;

[0021] Figure 4 is a schematic diagram of a target active contour according to an embodiment of the present invention;

[0022] Figure 5 is another flowchart of an image segmentation method applicable to bone imaging data according to an embodiment of the present invention;

[0023] Figure 6 is a schematic diagram of an image segmentation method applicable to bone imaging data according to an embodiment of the present invention;

[0024] Figure 7 is a comparison diagram of bone segmentation results according to an embodiment of the present invention;

[0025] Figure 8 is a schematic diagram of active contours of different algorithms at the same number of iteration steps according to an embodiment of the present invention;

[0026] Figure 9 is a schematic diagram of bone segmentation results of bone CT images of different parts determined based on different algorithms according to an embodiment of the present invention;

[0027] Figure 10 is a schematic diagram of the structure of an image segmentation device applicable to bone imaging data according to an embodiment of the present invention;

[0028] Figure 11 is another schematic diagram of the structure of an image segmentation device applicable to bone imaging data according to an embodiment of the present invention;

[0029] Figure 12 is another schematic diagram of the structure of an image segmentation device applicable to bone imaging data according to an embodiment of the present invention;

[0030] Figure 13 is another schematic diagram of the structure of an image segmentation device applicable to bone imaging data according to an embodiment of the present invention;

[0031] Figure 14 is a schematic diagram of the structure of an electronic device for implementing the image segmentation method applicable to bone imaging data according to an embodiment of the present invention. Detailed implementation manners

[0032] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0034] Figure 1 The figure is a flowchart of an image segmentation method applicable to bone image data according to an embodiment of the present invention. This embodiment is applicable to the case of determining the bone segmentation result of a bone CT image based on an active contour. This method can be executed by an image segmentation device applicable to bone image data. The image segmentation device applicable to bone image data can be implemented in the form of hardware and / or software, and the image segmentation device applicable to bone image data can be configured in the processor of an electronic device. As Figure 1 shown, the method includes:

[0035] S110. Determine an initial point set corresponding to an initial active contour set in a bone CT image, and all bones in the bone CT image are within the initial active contour.

[0036] Among them, the bone CT image is a CT image including bones, such as a foot CT image, a leg CT image, a hip CT image, etc.

[0037] The initial active contour can be a closed curve of any shape, such as a circle, a rectangle (see the red rectangle in Figure 2 ), etc., as long as all bones in the bone CT image are within the initial active contour.

[0038] In one embodiment, in response to a contour drawing operation in a skeletal CT image in the contour drawing mode, a corresponding initial active contour is generated. Since this embodiment is mainly targeted at doctors who have good image reading ability, they can add accurate initial active contours to the skeletal CT image. Therefore, this embodiment can quickly determine accurate initial active contours.

[0039] In one embodiment, the bone distribution range in the skeletal CT image is recognized based on a pre-trained bone recognition model, and then the bone distribution range is expanded by a set size to obtain an initial active contour; or a closed curve including the bone distribution range is used as the initial active contour. Since the pre-trained bone recognition model may not accurately recognize the bone boundary, but it can recognize the bone distribution range, the initial active contour determined based on the bone distribution range has high accuracy, achieving the technical effect of automatically adding an accurate initial active contour to the skeletal CT image.

[0040] After the initial active contour is determined, all pixels included in the initial active contour are used as an initial point set. In this embodiment, the bone image is represented as I, where I: Ω → R, Ω is the image domain of the bone image I, and R is the real number space. The active contour is represented as φ, which is a level set function on the image domain Ω, and its zero level set is {x ∈ Ω|φ(x) = 0}, representing the considered active contour. It is agreed that the region where φ(x) < 0 is the inside of the active contour, which is the target region; the region where φ(x) > 0 is the outside of the active contour, which is the background region.

[0041] S120. Determine the weight matrix of the skeletal CT image based on a weight function. The weight matrix includes the weighted sum of the gray weight and the gradient weight of each pixel in the skeletal CT image. The weight function includes a gray term and a gradient term that are both monotonically decreasing functions. The gray term is used to determine the gray weight of pixels that do not meet the soft tissue gray condition, and the gradient term is used to determine the gradient weight of pixels that do not meet the soft tissue gradient condition.

[0042] The weight matrix includes the weights of each pixel in the bone image (see Figure 3 ), and the weight of each pixel is the weighted sum of the gray weight and the gradient weight of each pixel. Figure 3 Among them, the weights of most soft tissue pixels are relatively large, and the weights of pixels at the bone boundary are much smaller than the weights of surrounding pixels.

[0043] In the skeletal CT image, the gray value of bone pixels is higher than that of surrounding soft tissue pixels, and pixels distributed on the bone boundary have a high gradient. Therefore, the weighted sum of the gray weight and the gradient weight is used as the weight of each pixel to improve the accuracy of determining the weights of each pixel.

[0044] A monotonically decreasing function is a function in which the output data decreases as the independent variable increases.

[0045] The soft tissue gray level condition is used to remove some or all of the soft tissue pixels in the bone CT image, and the gray level term is used to determine the gray level weights of the remaining pixels in the bone CT image after removing some or all of the soft tissue; the soft tissue gradient condition is also used to remove some or all of the soft tissue pixels in the bone CT image, and the gradient term is used to determine the gradient weights of the remaining pixels in the bone CT image after removing some or all of the soft tissue. It can be understood that the pixel combination corresponding to the soft tissue gray level condition and the pixel combination corresponding to the soft tissue gradient condition may be the same or different. The gray level weight of the soft tissue pixels corresponding to the soft tissue gray level condition is 1, and the gradient weight of the soft tissue pixels corresponding to the soft tissue gradient condition is also 1, realizing the contribution of the gray level weight of the corresponding pixels removed by the soft tissue gray level condition to the weighted sum and the contribution of the gradient weight of the corresponding pixels removed by the soft tissue gradient condition to the weighted sum.

[0046] In one embodiment, the soft tissue gray level condition is a gray level threshold; the soft tissue gradient condition is a gradient threshold; the gray level threshold is greater than or equal to a first gray level value and less than or equal to a second gray level value; the first gray level value is the maximum value of the gray level values of the soft tissue pixels used in the browsing of the bone CT image; the second gray level value is the minimum value of the gray level values of the bone pixels used in the browsing of the bone CT image; the gradient threshold is greater than or equal to zero and less than or equal to a third gray level value, and the third gray level value is equal to the difference between the second gray level value and the first gray level value.

[0047] Based on this, the gray level term is used to make the gray level weight of the pixel with a higher gray level value lower for the pixel with a gray level value higher than the gray level threshold; the gradient term is used to make the gradient weight of the pixel with a higher gradient value lower for the pixel with a gradient value higher than the gradient threshold. Since the gray level value and the gradient value of the pixels at the bone boundary are much larger than those of the surrounding pixels, the weights of the pixels on the bone boundary are significantly smaller than those of the pixels in other regions, so that in the process of minimizing the energy functional, the active contour can be accurately controlled to approach the bone boundary.

[0048] Specifically, in clinical orthopedic diagnosis, doctors usually need to browse the bone CT images of patients under the bone window range. In the field of mathematics, browsing under the bone window range refers to the process of truncating and mapping the CT value into the image gray level range according to the window width and window position of the bone window, which can be expressed by the following formula:

[0049]

[0050] Due to the significant difference in density between bones and surrounding soft tissues, there is also a significant difference in their CT values. When the window width and window level are given, the gray value difference between bone pixels and soft tissue pixels can be obtained through the above gray mapping calculation.

[0051] Generally speaking, when different doctors view CT images of a specific part, although they will not adjust the window width and window level to a certain fixed value, they can empirically adjust them to a certain range. In this embodiment, by using the range of bone window width and window level and combining with the CT value range of bones and soft tissues, the upper bound of the gray value range of soft tissue pixels and the lower bound of the gray value range of bone pixels are calculated, and based on this upper bound and lower bound, the gray threshold and gradient threshold corresponding to the weight function are determined, so that the weight function can accurately differentiate the weights of each pixel and reduce the influence of other tissues and organs on the contraction of the active contour.

[0052] Set the range of window width and window level as: W ω ∈[ω1,ω2], W t ∈[l1,l2], and assume that the CT value of soft tissue is less than or equal to S, and the CT value of bone is greater than or equal to B. By solving the following two optimization problems, the upper bound θ1 and lower bound θ2 of the corresponding gray value are calculated, and θ1 is called the first gray value, and θ2 is called the second gray value:

[0053] θ1=maxm(z;W ω , W l ) s.t. z≤S, ω1≤W ω ≤ω2, l1≤W l ≤l2,

[0054] θ2=minm(z;W ω , W l ) s.t. z≥B, ω1≤W ω ≤ω2, l1≤W l ≤l2. It can be seen that θ1≤θ2, and θ2 - θ1 is called the third gray value.

[0055] Considering the characteristics of high gray value and high gradient value at the bone boundary, both gray information and gradient information are embedded into the weight function. The weight function can be selected as:

[0056]

[0057] where γ>0, which is used to adjust the ratio of the gray term to the gradient term, that is, the weighting coefficient between the two. f1 is the function corresponding to the soft tissue gray condition, and f2 is the function corresponding to the soft tissue gradient condition. Hereinafter, the weight function included in the above formula will be briefly denoted as

[0058] In one embodiment, Here z+ = max{z, 0}, where δ1, δ2 > 0, and ε1 ∈ [θ1, θ2], ε2 ∈ [0, θ2 - θ1] are selected. It should be noted that the selection of f1 and f2 includes but is not limited to this form. In another embodiment, the selection method is f1(z) = exp((z - ε1) + ), f2(z) = exp((z - ε2) + ), where exp(·) represents the natural exponential operation, and any function that is monotonically increasing within the range of positive real numbers can achieve this function.

[0059] In the determination process of the weighted sum of the gray weight and the gradient weight, the above weight function filters out the influence of some soft tissue information on the weighted sum, realizing the weight differentiation between soft tissue pixels and bone pixels; at the same time, it takes into account the gray information and gradient information of the pixels. And since both the gray weight and the gradient term are monotonically decreasing functions, the weights of the pixels at the bone boundaries in the bone CT image are significantly smaller than the weights of the pixels at other positions. Therefore, based on the weight matrix of the bone CT image determined by the weight function, during the process of minimizing the energy functional, it better guides the active contour to evolve towards the bone boundary, avoiding the influence of other tissues and organs on the accuracy of the evolution result.

[0060] S130. Solve the iteration equation based on the initial point set, the weight matrix, and the predetermined iteration step size to obtain the target active contour that coincides with the bone boundary. The iteration equation is the gradient flow equation for minimizing the energy functional, and the energy functional optimizes the active contour based on the weighted sum of the weighted perimeter and the weighted area of the active contour. The weights used in the determination process of the weighted perimeter and the weighted area come from the weight matrix.

[0061] In one embodiment, the energy functional is:

[0062]

[0063] where H(φ(x)) is zero when φ(x) is less than or equal to zero and is 1 when φ(x) is greater than zero. That is to say, for the pixels outside the active contour, H(φ(x)) is 1, and in other cases, H(φ(x)) is 0; is the weight function, which is used to determine the weight matrix of the bone CT image. α is the ratio between the weighted perimeter term and the weighted area term, and this value is an empirical value, and its value range can be selected as (0, 2). Among them, the in this energy functional is the weighted perimeter term, which is used to determine the weighted perimeter of the active contour; the in this energy functional is the weighted area term, which is used to determine the weighted area inside the active contour.

[0064] It should be noted that the weighted perimeter term and the weighted area term in the energy functional can also be implemented in other ways, as long as the perimeter and area of the active contour can be accurately determined. This embodiment will not elaborate here.

[0065] After the energy functional is determined, the gradient flow equation for minimizing the energy functional is determined, and then the gradient flow equation is discretized to obtain an iterative equation; then, based on the initial point set, the initial weight matrix, and a predetermined iteration step size, the iterative equation is iteratively calculated until the set stopping condition is satisfied to obtain the target active contour. As Figure 4 shown, the red closed curve is the target active contour, and this target active contour coincides with the bone boundary in the bone CT image.

[0066] The gradient flow equation for minimizing the energy functional is:

[0067]

[0068] After the gradient flow equation is discretized, an iterative equation is obtained, and the iterative equation is:

[0069]

[0070] where h is the time step, φ n is the active contour in the previous iteration round, and φ n+1 is the active contour in the current iteration round, is the gradient operator, is the divergence operator.

[0071] S140. In the bone CT image, determine the bone segmentation result corresponding to the target active contour.

[0072] Since the target active contour coincides with the bone boundary in the bone CT image, the bone CT image can be segmented based on this target active contour to obtain the bone segmentation result.

[0073] Specifically, the target active contour is used as the bone boundary, the area inside the target active contour is used as the bone, and the area outside the target active contour is used as the background area. Therefore, based on the target active contour, the target active contour and the area inside the target active contour can be segmented from the bone CT image as the bone.

[0074] Since the gray information and gradient information of each pixel are fully considered in the process of determining the target active contour, the target active contour and the bone boundary have a high degree of coincidence. Therefore, the bone area can be completely segmented from the bone CT image based on the target active contour, and it has a high generalization ability for different bone parts.

[0075] Due to the physiological structures of different bone parts, some bones in the bone segmentation result may include non-bone holes or slits. In view of this, in one embodiment, a hole removal operation and / or a slit removal operation are performed on the bone part in the bone segmentation result to obtain an updated bone segmentation result; the updated bone segmentation result is binarized to obtain a mask for the target bone. Among them, holes can be obtained by expanding an internal active contour, and slits can be obtained by a region generation algorithm; of course, other methods can also be used to detect the existence of holes and / or slits, and when holes and / or slits exist, operations such as hole and / or slit filling and dilation are performed to remove the holes and / or slits in the bone segmentation result to obtain an updated bone segmentation result; then the updated bone segmentation result is binarized, for example, the pixels in the target bone region in the updated bone segmentation result are assigned 1, and the non-bone regions are assigned 0, etc., to obtain a mask for the target bone.

[0076] It can be understood that since the target bone in the target bone segmentation result or the mask for the target bone both include the three-dimensional information of the target bone in the bone image, the three-dimensional spatial structure of the target bone can be determined based on the bone segmentation result or the mask.

[0077] The embodiments of the present invention can be used for the segmentation of real bones and can also be used for the segmentation of implant objects similar to bones.

[0078] In a bone CT image, the gray value of bone pixels is higher than that of surrounding soft tissue pixels, and the gradient change value of pixels at the bone boundary is higher than that of pixels in other regions; in the technical solution provided by the embodiments of the present invention, the weight function includes a gray term and a gradient term that are both monotonically decreasing functions. The gray term is used to determine the gray weight of pixels that do not meet the soft tissue gray condition, and the gradient term is used to determine the gradient weight of pixels that do not meet the soft tissue gradient condition; this feature makes the weights of soft tissue pixels significantly different from those of bone pixels, weakens the contribution of soft tissues to the weighted area and weighted perimeter determined by the energy functional, thereby reducing the influence of soft tissues on the accuracy of the weighted area and weighted perimeter of the active contour, and in the process of minimizing the energy functional, controlling the active contour to accurately approximate the bone boundary to obtain an accurate target active contour, and thus determining a bone segmentation result with high accuracy based on the target active contour.

[0079] Figure 5 This is a flowchart of an image segmentation method applicable to bone image data provided by the embodiments of the present invention. In this embodiment, the predetermined iteration step size is refined on the basis of the foregoing embodiments. As Figure 5 shown, the method includes:

[0080] S220. Determine an initial point set corresponding to an initial active contour set in the bone CT image, where all bones in the bone CT image are within the initial active contour.

[0081] S220. Determine a weight matrix of the bone CT image based on a weight function. The weight matrix includes a weighted sum of the gray weight and the gradient weight of each pixel in the bone CT image. The weight function includes a gray term and a gradient term, both of which are monotonically decreasing functions. The gray term is used to determine the gray weight of pixels that do not meet the soft tissue gray condition, and the gradient term is used to determine the gradient weight of pixels that do not meet the soft tissue gradient condition.

[0082] S230. Solve an iterative equation based on the initial point set, the weight matrix, and a predetermined iteration step size to obtain a target active contour that coincides with the bone boundary. The iterative equation is a gradient flow equation for minimizing an energy functional. The energy functional optimizes the active contour based on a weighted sum of the weighted perimeter and the weighted area of the active contour. The weights used in the determination of the weighted perimeter and the weighted area are from the weight matrix. The predetermined iteration step size includes a time step size and an adaptive step size based on the boundary distance. The adaptive step size is determined based on the distance between each pixel in the active contour and the bone boundary. The distance between each pixel in the active contour and the bone boundary is extracted from a first boundary distance set. The first boundary distance set includes the normalized distance between each pixel in the bone CT image and the bone boundary.

[0083] The time step size is greater than 0 and less than 0.5. In this embodiment, the time step size is preferably greater than 0 and less than 0.2.

[0084] The adaptive step size can be understood as a spatial step size. Specifically, based on the distance between each pixel in the previous active contour and the bone boundary, the spatial step size in the current iteration round is adaptively determined, and the active contour in the current iteration round is determined according to the previous active contour (the active contour in the previous iteration round) and the spatial step size. Moreover, as the distance between the active contour and the bone boundary shortens, the spatial step size also shortens, and the iterative evolution speed decreases.

[0085] Since the first boundary distance set includes the normalized distance between each pixel in the bone CT image and the bone boundary, the distance between each pixel in the current active contour and the bone boundary can be determined according to the first boundary distance set. In this way, when it is detected that the distance between each pixel in the current active contour and the bone boundary is less than or equal to a set threshold, the stop iterative evolution operation is triggered, and the current active contour is used as the target active contour.

[0086] In one embodiment, the first boundary distance set can be implemented through the following steps:

[0087] Step a1. Determine a bone boundary set corresponding to the bone CT image.

[0088] To determine the distance between each pixel in a bone CT image and the bone boundary, it is necessary to first determine the bone boundary in the bone CT image. In one embodiment, the bone CT image is analyzed by a bone boundary extraction algorithm (such as the Canny operator, etc.), and all the pixels distributed on the bone boundary in the bone CT image are obtained. The set of all the pixels distributed on the bone boundary is used as the bone boundary set (E bone ). Specifically:

[0089] The gray values near the bone boundary are usually higher, while the gray values at non-bone positions are lower. Therefore, this feature can be used to filter out non-bone boundaries and obtain the bone boundary. When implementing the algorithm, its feature function can be expressed as:

[0090]

[0091] where {φ0≤0} is the set of internal elements of the initial active contour, K is the convolution kernel used to filter out non-bone boundaries, η is a threshold, is the set of pixels on all the boundaries of the bone CT image.

[0092] In another embodiment, the bone CT image is input into a pre-trained bone boundary extraction model to obtain the bone boundary set corresponding to the bone CT image.

[0093] It can be understood that for a fracture scenario, the fracture boundary corresponding to the fracture gap can also be regarded as the bone boundary. Therefore, the bone boundary set in this embodiment also includes the fracture boundary set.

[0094] When there is a fracture in the bone CT image, but the boundary at the fracture is not obvious or even there is no boundary, it will cause the active contour to not stay at the fracture. The present invention incorporates the artificial interaction information at the fracture into the boundary distance, so that the active contour can stay at the fracture, so that the bone segmentation result can be used in the fracture recovery scenario.

[0095] Specifically, in the fracture edge drawing mode, the doctor uses a drawing tool to outline the fracture in the bone CT image. The processor responds to the confirmation operation for the fracture edge drawing result in the fracture edge drawing mode and determines all the pixels in the bone CT image corresponding to the fracture edge drawing result; all the pixels in the bone CT image corresponding to the fracture edge drawing result are used as the fracture boundary set. Figure 6 The red rectangle in the left bone CT image is the initialized active contour (i.e., the initial active contour), the yellow curve in the bone CT image is the fracture prompt, and the set of pixels corresponding to the fracture prompt is the fracture boundary set.

[0096] Since doctors have good fracture recognition ability, the fracture boundary set (E prompt ) determined based on the fracture curves drawn by doctors in skeletal CT images has high accuracy. Moreover, the skeletal boundary set as a fracture prompt allows the active contour to stop at the fracture, improving the accuracy of iterative stopping, as well as the interactivity and flexibility of the determination of the target active contour.

[0097] After the fracture boundary set is determined, the union of the fracture boundary set and the skeletal boundary set is used as the updated skeletal boundary set

[0098] Step a2: Determine the distance between each pixel in the skeletal CT image and the corresponding skeletal boundary of the skeletal boundary set, and obtain the distance between each pixel in the skeletal CT image and the skeletal boundary.

[0099] Step a3: Take the summary result of the distances between all pixels and the skeletal boundary as the second boundary distance set, and normalize each distance in the second boundary distance set to obtain the first boundary distance set.

[0100] Specifically, the second boundary distance set can be expressed as: d(x, E bone ), where x ∈ Ω.

[0101] After the second boundary distance set is determined, the following formula is used to complete the normalization of each distance in the second boundary distance set to obtain the first boundary distance set:

[0102]

[0103] For the skeletal boundary set including the fracture boundary set, the corresponding second boundary distance set is: where x ∈ Ω. The first boundary distance set corresponding to the second boundary distance set is determined by the following formula:

[0104]

[0105] When

[0106] If this function is used as the adaptive step size, when the active contour is far from the skeletal boundary, the evolution speed of the active contour is fast. As the distance decreases, the evolution speed gradually slows down until the active contour reaches the skeletal boundary, at which point the evolution speed is zero and the evolution stops, thus achieving the purpose of a stable algorithm. Figure 6 Shows the adaptive step size incorporating the boundary distance, that is, the adaptive step size determined based on the first skeletal boundary distance set, and the width of each color distribution region corresponds to an adaptive step size.

[0107] The gradient flow function including an adaptive step size can be expressed as:

[0108]

[0109] After discretization, an iterative equation is obtained, and the iterative equation is:

[0110]

[0111] where h is the time step size.

[0112] S240. In the skeletal CT image, determine the skeletal segmentation result corresponding to the target active contour.

[0113] In this embodiment, the skeletal boundary set used to determine the second skeletal boundary distance set corresponds to the skeletal boundary in the skeletal CT image, and the target active contour also corresponds to the skeletal boundary in the skeletal CT image; however, the former has lower accuracy and cannot distinguish between the inside and outside of the bone, while the latter is the opposite. That is to say, in this embodiment, a target active contour with higher accuracy is determined based on the combination of the energy functional and the skeletal boundary set with lower accuracy.

[0114] To improve the accuracy of the skeletal segmentation result, as Figure 6 shown, after the skeletal segmentation result is determined, post-processing is performed on the skeletal segmentation result, such as hole removal operation and / or slit removal operation, etc., to update the skeletal segmentation result.

[0115] The technical solution provided by the embodiment of the present invention, since the adaptive step size is the spatial step size, during the evolution process of the energy functional, the specific value of the adaptive step size can be adjusted according to the distance between the active contour and the skeletal boundary, so as to achieve the purpose of controlling the evolution speed; moreover, it can be foreseen that when the distance between the active contour and the skeletal boundary is less than the set threshold or zero, the iterative evolution can be directly stopped, improving the accuracy of the stop timing of the iterative evolution and the accuracy of the target active contour.

[0116] Figure 7 This is a comparison diagram of the skeletal segmentation results provided by the embodiment of the present invention, specifically the comparison result between the skeletal segmentation result provided by the embodiment of the present invention and the skeletal segmentation result determined based on the Chan-Vese algorithm.

[0117] The skeletal CT images in the embodiments of the present invention may be pelvic CT images or ankle CT images. The window width range and window level range based on which the gray values of each pixel in the skeletal CT images are determined are [1000 HU, 1500 HU] and [250 HU, 350 HU] respectively; and it is considered that the CT value of soft tissue does not exceed 100 HU, and the CT value of bone does not fall below 300 HU. By solving the optimization problem, it is obtained that the gray value of the pixel of soft tissue does not exceed 102, and the gray value of the pixel of bone does not fall below 115. Appropriate parameters ε1 ∈ [102, 115] and ε2 ∈ [0, 13] are selected. δ1 = 1, δ2 = 2 and γ = 1 are fixedly selected. When constructing the adaptive step size, the Canny operator is selected for edge extraction, and the convolution kernel for filtering non-bone boundaries adopts a 3×3 uniformly distributed convolution kernel, and the threshold η = 70.

[0118] As Figure 7 shown, since the Chan-Vese algorithm approximates the foreground and background of an image using two constants, it cannot well adapt to the characteristic that there are large differences in the density and texture between the cortical bone and cancellous bone in the bone structure, and thus cannot segment the complete bone region. In the embodiments of the present invention, the cortical bone and cancellous bone are accurately segmented simultaneously.

[0119] In the pelvic and ankle CT images, numerical experiments and comparisons are also carried out on the method described in the embodiments of the present invention and the original geodesic active contour method, and the results are as Figure 8 shown. Figure 8 shows the initial active contours of the two methods and the iterative results at 1000 steps, 2000 steps, 3000 steps, 6000 steps and 7000 steps. It can be seen that in the original geodesic active contour method, the active contour is easily affected by the boundaries of other tissue organs, and the active contour has become unstable at 7000 steps of iteration. In the embodiments of the present invention, the active contour is not affected by the boundaries of other tissue organs, and the shape of the active contour is stably maintained during the iteration process. At the same time, the active contour also stays well at these positions.

[0120] Figure 9Schematic diagrams of bone segmentation results of CT images of different parts of bones determined based on different algorithms are provided. Specifically, they are the comparison results of the method described in the embodiments of the present invention, the SAM method, the continuous maximum flow algorithm, the Chan-Vese algorithm, the U-Net method and the ground truth label. The continuous maximum flow algorithm is the same as the Chan-Vese algorithm and cannot segment out the complete bone region. SAM is a large model algorithm, and its results are from the official website. The white square is the prompt contour, and the blue contour is the bone contour found by the algorithm. It can be seen that SAM can only give the approximate contour of the segmentation target and still cannot give accurate results in terms of segmentation details, especially at the fracture site. The U-Net used in the experiment was trained for 20 rounds on the pelvic CT image data of 90 patients, and the test results were slightly better on pelvic images but could not be generalized to the ankle. The generalization ability has always been a key issue for deep learning algorithms. In contrast, the algorithm proposed in the present invention belongs to the traditional mathematical model algorithm and has an advantage in generalization ability, and better results are obtained for both the pelvis and the ankle.

[0121] The present invention simultaneously considers the following four numerical indicators to measure the quality of the segmentation results. It is agreed that G represents the ground truth label and S represents the segmentation result. ∂ represents the boundary of the set, and |·| represents the measure. In this embodiment, the Dice coefficient and the Jaccard coefficient are used to evaluate the prediction accuracy of the segmentation result at the pixel level. The results obtained by both are in the range of [0%, 100%], and the larger the value, the better the segmentation result.

[0122] Among them,

[0123] In this embodiment, the Hausdorff Distance (HD) and the Average Symmetric Surface Difference (ASSD) are used to evaluate the closeness between the bone boundary in the bone segmentation result and the bone boundary in the ground truth label. The smaller the value, the better the bone segmentation. Among them,

[0124]

[0125] Based on the Dice coefficient, the Jaccard coefficient, the Hausdorff distance and the average symmetric surface distance, the comparison results between the bone segmentation results determined in the embodiments of the present invention and the ground truth label, as well as the comparison results between the bone segmentation results determined by other algorithms and the ground truth label are determined. Specifically, as shown in Table 1. It can be seen from Table 1 that the index data of the bone segmentation results determined in the embodiments of the present invention are generally the best.

[0126] Table 1 Index data display table of bone segmentation results corresponding to different methods

[0127]

[0128] In summary, the accuracy of the bone segmentation result determined in the embodiment of the present invention is significantly higher than that of the bone segmentation results determined by other existing bone segmentation methods.

[0129] Figure 10 It is a schematic structural diagram of an image segmentation device applicable to bone image data provided by an embodiment of the present invention. As Figure 10 shown, the device includes:

[0130] An initial point set module 310, configured to determine an initial point set corresponding to an initial active contour set in a bone CT image, and all bones in the bone CT image are within the initial active contour;

[0131] A weight matrix module 320, configured to determine a weight matrix of the bone CT image based on a weight function, where the weight matrix includes a weighted sum of the gray weight and the gradient weight of each pixel in the bone CT image, the weight function includes a gray term and a gradient term that are both monotonically decreasing functions, the gray term is used to determine the gray weight of pixels that do not meet the soft tissue gray condition, and the gradient term is used to determine the gradient weight of pixels that do not meet the soft tissue gradient condition;

[0132] An iterative solution module 330, configured to solve an iterative equation based on the initial point set, the weight matrix, and a predetermined number of iterative steps to obtain a target active contour that coincides with the bone boundary. The iterative equation is a gradient flow equation for minimizing an energy functional, and the energy functional optimizes the active contour based on a weighted sum of the weighted perimeter and the weighted area of the active contour. The weights used in the determination of the weighted perimeter and the weighted area come from the weight matrix;

[0133] A segmentation module 340, configured to determine a bone segmentation result corresponding to the target active contour in the bone CT image.

[0134] In one embodiment, the soft tissue gray condition is a gray threshold;

[0135] The soft tissue gradient condition is a gradient threshold;

[0136] The gray threshold is greater than or equal to a first gray value and less than or equal to a second gray value;

[0137] The first gray value is the maximum value of the soft tissue pixel gray values used in the browsing of the bone CT image;

[0138] The second gray value is the minimum value of the bone pixel gray values used in the browsing of the bone CT image;

[0139] The gradient threshold is greater than or equal to zero and less than or equal to a third grayscale value, and the third grayscale value is equal to the difference between the second grayscale value and the first grayscale value.

[0140] In one embodiment, the predetermined iteration step size includes a time step size and an adaptive step size based on a boundary distance, the adaptive step size is determined based on the distance between each pixel in the active contour and a bone boundary, and the distance between each pixel in the active contour and a bone boundary is extracted from a first boundary distance set;

[0141] The first boundary distance set includes the normalized distance between each pixel in the bone CT image and the bone boundary.

[0142] In one embodiment, the boundary distance module 350 (see Figure 11 ) determines a first boundary distance set, the boundary distance module comprising:

[0143] A bone boundary unit, used to determine a bone boundary set corresponding to the bone CT image;

[0144] A distance unit, used to determine the distance between each pixel in the bone CT image and the bone boundary corresponding to the bone boundary set, so as to obtain the distance between each pixel in the bone CT image and the bone boundary;

[0145] The normalization unit is used to summarize the distances between all the pixels and the bone boundaries as a second boundary distance set, and normalize each distance in the second boundary distance set to obtain a first boundary distance set.

[0146] In one embodiment, the boundary distance module further includes a fracture boundary unit, the fracture boundary unit including:

[0147] A fracture boundary subunit, used for acquiring a pixel set at a fracture location in the bone CT image, and using the pixel set at the fracture location as a fracture boundary set;

[0148] An updating subunit is used to add the fracture boundary set to the bone boundary set to update the bone boundary set.

[0149] In one embodiment, the fracture boundary subunit is specifically used to:

[0150] In response to a confirmation operation on a fracture edge delineation result in a fracture edge drawing mode, determining all pixels in the bone CT image corresponding to the fracture edge delineation result;

[0151] All pixels in the bone CT image corresponding to the fracture edge delineation result are taken as the fracture boundary set.

[0152] In one embodiment, as Figure 12 shown, the apparatus further includes a post-processing module 360, and the post-processing module 360 is configured to:

[0153] perform a hole removal operation and / or a slit removal operation on the bone part in the bone segmentation result to obtain an updated bone segmentation result;

[0154] perform binarization on the updated bone segmentation result to obtain a mask for the target bone.

[0155] In one embodiment, as Figure 13 shown, the apparatus further includes an application module 370, and the application module 370 is configured to:

[0156] determine tool parameters based on the bone segmentation result, where the tool parameters are parameters of a bone fixation tool for the fracture site.

[0157] In a bone CT image, the gray value of a bone pixel is higher than that of a surrounding soft tissue pixel, and the gradient change value of a pixel at the bone boundary is higher than that of a pixel in other regions; in the technical solution provided by the embodiment of the present invention, the weight function includes a gray term and a gradient term that are both monotonically decreasing functions. The gray term is used to determine the gray weight of a pixel that does not meet the soft tissue gray condition, and the gradient term is used to determine the gradient weight of a pixel that does not meet the soft tissue gradient condition; this feature makes the weight of a soft tissue pixel significantly different from that of a bone pixel, weakens the contribution of the soft tissue to the weighted area and weighted perimeter determined by the energy functional, thereby reducing the influence of the soft tissue on the accuracy of the weighted area and weighted perimeter of the active contour, and during the process of minimizing the energy functional, controlling the active contour to accurately approach the bone boundary to obtain an accurate target active contour, so as to determine a bone segmentation result with high accuracy based on the target active contour.

[0158] The image segmentation apparatus for bone image data provided by the embodiment of the present invention can execute the image segmentation method for bone image data provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.

[0159] Figure 14FIG. 0 shows a schematic structural diagram of an electronic device 10 that can be used to implement embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0160] As Figure 14 shown, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. The memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. The input / output (I / O) interface 15 is also connected to the bus 14.

[0161] A plurality of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0162] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as an image segmentation method applicable to skeletal image data.

[0163] In some embodiments, an image segmentation method applicable to bone imaging data can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the image segmentation method applicable to bone imaging data described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the image segmentation method applicable to bone imaging data by any other suitable means (e.g., by means of firmware).

[0164] The various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuitry, integrated circuit systems, field-programmable gate arrays (FPGA), application-specific integrated circuits (ASIC), application-specific standard products (ASSP), systems-on-chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special or general programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0165] The computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer programs are executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer programs can be executed entirely on the machine, partially on the machine, as a stand-alone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0166] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0167] To provide for interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide for interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0168] The systems and techniques described herein can be implemented in a computing system that includes backend components (such as, for example, a data server), or a computing system that includes middleware components (such as, for example, an application server), or a computing system that includes frontend components (such as, for example, a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (such as, for example, a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0169] A computing system may include a client and a server. The client and the server are generally far from each other and usually interact via a communication network. The relationship between the client and the server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, and solves the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0170] An embodiment of the present invention also provides a computer program product, including a computer program which, when executed by a processor, implements the image segmentation method applicable to skeletal image data provided in any embodiment of the present application.

[0171] In the process of implementing the computer program product, computer program code for performing the operations of the present invention can be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network - including a local area network (LAN) or a wide area network (WAN) - or can be connected to an external computer (for example, by using an Internet service provider to connect through the Internet).

[0172] It should be understood that various forms of the processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is made herein.

[0173] The above specific embodiments do not constitute a limitation to the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An image segmentation method suitable for bone image data, characterized in that: include: Determine an initial point set corresponding to an initial active contour set in the bone CT image, wherein all bones in the bone CT image are in the initial active contour; Determining a weight matrix of the bone CT image based on a weight function, the weight matrix comprising a weighted sum of a grayscale weight and a gradient weight of each pixel in the bone CT image, the weight function comprising a grayscale term and a gradient term both being monotonically decreasing functions, the grayscale term being used to determine a grayscale weight of a pixel that does not meet a soft tissue grayscale condition, and the gradient term being used to determine a gradient weight of a pixel that does not meet a soft tissue gradient condition; Solving an iterative equation based on the initial point set, the weight matrix and a predetermined iteration step size to obtain a target active contour that coincides with the bone boundary, wherein the iterative equation is a gradient flow equation that minimizes an energy functional, wherein the energy functional optimizes the active contour based on a weighted sum of a weighted perimeter and a weighted area of ​​the active contour, and the weights used in determining the weighted perimeter and the weighted area come from the weight matrix; In the bone CT image, a bone segmentation result corresponding to the target active contour is determined.

2. The method according to claim 1, characterized in that The soft tissue grayscale condition is a grayscale threshold; The soft tissue gradient condition is a gradient threshold; The grayscale threshold is greater than or equal to the first grayscale value and less than or equal to the second grayscale value; The first grayscale value is the maximum value of the soft tissue pixel grayscale values ​​used in browsing the bone CT image; The second grayscale value is the minimum value of the bone pixel grayscale values ​​used in browsing the bone CT image; The gradient threshold is greater than or equal to zero and less than or equal to a third grayscale value, and the third grayscale value is equal to the difference between the second grayscale value and the first grayscale value.

3. The method according to claim 1, characterized in that The predetermined iteration step size includes a time step size and an adaptive step size based on a boundary distance, wherein the adaptive step size is determined based on a distance between each pixel in the active contour and a bone boundary, and the distance between each pixel in the active contour and a bone boundary is extracted from a first boundary distance set; The first boundary distance set includes the normalized distance between each pixel in the bone CT image and the bone boundary.

4. The method according to claim 3, characterized in that The first boundary distance set is determined by the following steps: Determine a bone boundary set corresponding to the bone CT image; Determine the distance between each pixel in the bone CT image and the bone boundary corresponding to the bone boundary set to obtain the distance between each pixel in the bone CT image and the bone boundary; The summary result of the distances between all the pixels and the bone boundary is taken as a second boundary distance set, and each distance in the second boundary distance set is normalized to obtain a first boundary distance set.

5. The method according to claim 4, characterized in that Before determining the distance between each pixel in the bone CT image and the bone boundary corresponding to the bone boundary set, the method further includes: Acquire a pixel set at a fracture location in the bone CT image, and use the pixel set at the fracture location as a fracture boundary set; The fracture boundary set is added to the bone boundary set to update the bone boundary set.

6. The method according to claim 5, characterized in that The step of acquiring a pixel set at a fracture location in the bone CT image and using the pixel set at the fracture location as a fracture boundary set includes: In response to a confirmation operation on a fracture edge delineation result in a fracture edge drawing mode, determining all pixels in the bone CT image corresponding to the fracture edge delineation result; All pixels in the bone CT image corresponding to the fracture edge delineation result are taken as the fracture boundary set.

7. The method according to claim 1, characterized in that After determining the bone segmentation result corresponding to the target active contour in the bone CT image, the method further includes: Performing a hole removal operation and / or a slit removal operation on the bone part in the bone segmentation result to obtain an updated bone segmentation result; The updated bone segmentation result is binarized to obtain a mask for the target bone.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the image segmentation method applicable to bone image data described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the image segmentation method applicable to bone image data as described in any one of claims 1-7 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the computer program implements the image segmentation method applicable to bone image data according to any one of claims 1 to 7.

Images