Bone density measurement method, device, equipment and medium based on deep learning algorithm

Abstract

The present application provides a bone density measurement method, device, equipment and computer-readable storage medium based on a deep learning algorithm. The bone density measurement method based on a deep learning algorithm includes: obtaining a spinal CT image; inputting the spinal CT image into a preset segmentation network model, and outputting the corresponding spinal segmentation result; inputting the spinal segmentation result into a preset regression network model, and outputting the corresponding bone density measurement result. According to the embodiment of the present application, bone density can be accurately measured without the need for special equipment or special personnel.

Description

Technical Field

[0001] The present application belongs to the field of deep learning technology, and in particular, relates to a bone density measurement method, device, equipment and computer-readable storage medium based on a deep learning algorithm. Background Art

[0002] Spinal fractures usually lead to disability and even death in severe cases. Osteoporosis is an important cause of fractures, so prevention of osteoporosis is a key to preventing fractures. Osteoporosis is usually diagnosed by measuring the bone density of the human body to obtain the bone density value.

[0003] There are currently four commonly used methods for measuring bone density on the market.

[0004] The first one is single photon absorption measurement, which uses the principle that the absorption of radioactive substances by bone tissue is proportional to the bone mineral content, and uses radioactive isotopes as light sources to measure the bone mineral content of human limb bones.

[0005] The second method is dual-energy X-ray absorptiometry, which uses an X-ray tube to obtain two types of energy, namely low-energy and high-energy photon peaks, through a certain device. After this photon peak penetrates the body, the scanning system sends the received signal to the computer for data processing to obtain the bone mineral content.

[0006] The third type is quantitative CT, or QCT. QCT can accurately select specific parts of the bone for measurement and bone density, and can evaluate the bone mineral density of cortical bone and cancellous bone separately.

[0007] The fourth method is ultrasonic measurement, which has attracted widespread attention due to its non-radiation and high sensitivity in diagnosing fractures. It uses the speed of sound wave conduction and amplitude attenuation to reflect the amount of bone mineral content, bone structure and bone strength. The instrument used is an ultrasonic bone densitometer.

[0008] However, the above four methods have two disadvantages:

[0009] 1. All of the above methods require special equipment to measure bone density, but some smaller hospitals do not have such equipment.

[0010] 2. The measurement of professional instruments also requires specialized personnel to operate, that is, specialized instrument operators need to be trained.

[0011] Therefore, how to accurately measure bone density without the need for specialized equipment or personnel is a technical problem that technical personnel in this field urgently need to solve. Summary of the invention

[0012] The embodiments of the present application provide a bone density measurement method, device, equipment and computer-readable storage medium based on a deep learning algorithm, which can accurately measure bone density without the need for special equipment or personnel.

[0013] In a first aspect, an embodiment of the present application provides a bone density measurement method based on a deep learning algorithm, comprising:

[0014] Obtain CT images of the spine;

[0015] Input the spinal CT image into the preset segmentation network model and output the corresponding spinal segmentation result;

[0016] The spine segmentation results are input into the preset regression network model, and the corresponding bone density measurement results are output.

[0017] Optionally, after obtaining the spinal CT image, the following is also included:

[0018] Normalize the spine CT images.

[0019] Optionally, the entire network structure of the segmentation network model is performed in parallel using four branches, and each time a new branch is generated, the branches are allowed to fuse information by concatenating;

[0020] Of the four branches output by the final model, only the output of the first branch is used as the final output.

[0021] Optionally, the first branch uses convolutional blocks, each of which consists of 2 convolutional operations, 1 batch normalization, and 1 ReLu activation function;

[0022] The image size remains unchanged throughout the branch, which effectively extracts local detail information of the image.

[0023] Optionally, in branches other than the first branch, the convolution block is replaced with a Transformer Block to effectively extract global information in the image.

[0024] Optionally, the Transformer Block includes a Laplacian operator for detecting second-order derivatives in the image to help locate edges in the image.

[0025] Optional, including:

[0026] Convolve this convolution kernel with the image to get the gradient change;

[0027] Edge detection based on gradient changes.

[0028] In a second aspect, an embodiment of the present application provides a bone density measurement device based on a deep learning algorithm, comprising:

[0029] An image acquisition module, used for acquiring a spinal CT image;

[0030] The spine segmentation module is used to input the spine CT image into a preset segmentation network model and output the corresponding spine segmentation result;

[0031] The bone density measurement module is used to input the spine segmentation results into a preset regression network model and output the corresponding bone density measurement results.

[0032] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device comprising: a processor and a memory storing computer program instructions;

[0033] When the processor executes the computer program instructions, a bone density measurement method based on a deep learning algorithm is implemented.

[0034] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, a bone density measurement method based on a deep learning algorithm is implemented.

[0035] The embodiments of the present application provide a bone density measurement method, device, equipment and computer-readable storage medium based on a deep learning algorithm, which can accurately measure bone density without the need for special equipment or personnel.

[0036] The bone density measurement method based on the deep learning algorithm includes: acquiring a spinal CT image; inputting the spinal CT image into a preset segmentation network model, and outputting a corresponding spinal segmentation result; inputting the spinal segmentation result into a preset regression network model, and outputting a corresponding bone density measurement result. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 It is a flowchart of a bone density measurement method based on a deep learning algorithm provided in one embodiment of the present application;

[0039] Figure 2It is a flowchart of a bone density measurement method based on a deep learning algorithm provided in one embodiment of the present application;

[0040] Figure 3 It is a schematic diagram of the structure of a segmentation network model provided by an embodiment of the present application;

[0041] Figure 4 It is a schematic diagram of the Transformer Block structure provided by an embodiment of the present application;

[0042] Figure 5 It is a structural schematic diagram of a bone density measurement device based on a deep learning algorithm provided by an embodiment of the present application;

[0043] Figure 6 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0044] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0045] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0046] In order to solve the problems of the prior art, the embodiments of the present application provide a bone density measurement method, device, equipment and computer-readable storage medium based on a deep learning algorithm. The following first introduces the bone density measurement method based on a deep learning algorithm provided in the embodiments of the present application.

[0047] Figure 1FIG. 1 is a flow chart of a method for measuring bone density based on a deep learning algorithm provided by an embodiment of the present application. Figure 1 As shown, the bone density measurement method based on deep learning algorithm includes:

[0048] S101, obtaining a spinal CT image;

[0049] In one embodiment, after acquiring the spinal CT image, the method further includes:

[0050] Normalize the spine CT images.

[0051] S102, inputting the spinal CT image into a preset segmentation network model, and outputting the corresponding spinal segmentation result;

[0052] Figure 2 FIG. 1 is a flow chart of a method for measuring bone density based on a deep learning algorithm provided by an embodiment of the present application. Figure 2 As shown in the figure, first, the CT image is segmented into the spine through the segmentation network. Then the L1 to L4 area is extracted through the spine label. The L1 to L4 area is input into the density regression model to predict the bone density.

[0053] In one embodiment, the entire network structure of the segmentation network model is performed in parallel using four branches, and each time a new branch is generated, information is fused between the branches in a concatenated manner;

[0054] Of the four branches output by the final model, only the output of the first branch is used as the final output.

[0055] In one embodiment, the first branch uses convolutional blocks, each of which consists of two convolutional operations, one batch normalization, and one ReLu activation function;

[0056] The image size remains unchanged throughout the branch, which effectively extracts local detail information of the image.

[0057] In one embodiment, in branches other than the first branch, the convolution Block is replaced with the Transformer Block to effectively extract the global information in the image.

[0058] like Figure 3 As shown in the figure, the input image of the entire model is a CT original image, and the output is the corresponding segmentation mask. After normalization, the original image is input into the deep learning model, and after passing through the deep learning network, the corresponding mask image is finally generated.

[0059] The entire network structure uses 4 branches in parallel, and each time a new branch is generated, the branches are concatenated to fuse information. Among the four branches output by the final model, only the head of the first branch is used as the final output.

[0060] In the entire model structure, the first branch uses convolutional blocks. Each convolutional block consists of 2 convolutional operations + 1 batch normalization + 1 ReLu activation function. The image size remains unchanged throughout the branch. In this way, the local detail information of the image can be effectively extracted.

[0061] In the second, third, and fourth branches of the entire model, in order to better extract global information, the convolution block is replaced by the transformer block. Through this module design, the transformer block can effectively extract the global information in the image, thereby achieving the purpose of allowing the convolution block and the transformer block to play their respective advantages.

[0062] In one embodiment, the Transformer Block includes a Laplacian operator for detecting second-order derivatives in an image to help locate edges in the image.

[0063] In one embodiment, it includes:

[0064] Convolve this convolution kernel with the image to get the gradient change;

[0065] Edge detection based on gradient changes.

[0066] Figure 4 is a schematic diagram of the Transformer Block structure provided by an embodiment of the present application, such as Figure 4 As shown, the Laplacian operator is a classic edge detection operator that can process images through convolution operations. The Laplacian operator is usually used to detect the second-order derivatives in an image to help locate the edges in the image. The following is a simple discrete form of the Laplacian operator:

[0067] 0 1 0 1 4 1 0 l 0

[0068] This is a 3x3 convolution kernel, whose central element is -4, the elements of the four neighboring regions are 1, and the other elements are 0. The Laplacian operator performs a convolution operation on the convolution kernel and the image, and calculates the gradient change to achieve edge detection. For a grayscale image I, the convolution operation of the Laplacian operator can be expressed by the following formula:

[0069]

[0070] in, and Represent the second-order derivatives of the image in the horizontal and vertical directions respectively.

[0071] In practical applications, we can use Laplacian convolution kernels of different sizes or perform Gaussian smoothing (L o G operator) to adjust the performance of the operator.

[0072] S103: Input the spine segmentation result into a preset regression network model, and output the corresponding bone density measurement result.

[0073] The present invention proposes a method for measuring bone density using a deep learning algorithm and non-enhanced CT images. CT images are a routine test item for patients during physical examinations. By using non-enhanced CT images, patients can calculate the patient's bone density through a deep learning algorithm without adding additional physical examination items, and then diagnose osteoporosis. On the one hand, it saves patients time and money; on the other hand, it also saves hospital costs. Through the deep learning method, the patient's bone density is regressed based on the non-enhanced CT image.

[0074] Figure 5 : is a schematic diagram of a bone density measurement device based on a deep learning algorithm provided by an embodiment of the present application, the bone density measurement device based on a deep learning algorithm comprises:

[0075] An image acquisition module 501 is used to acquire a spinal CT image;

[0076] The spine segmentation module 502 is used to input the spine CT image into a preset segmentation network model and output the corresponding spine segmentation result;

[0077] The bone density measurement module 503 is used to input the spine segmentation result into a preset regression network model and output the corresponding bone density measurement result.

[0078] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application is shown.

[0079] The electronic device may include a processor 601 and a memory 602 storing computer program instructions.

[0080] Specifically, the processor 601 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0081] The memory 602 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 602 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 602 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 602 may be inside or outside the electronic device. In a particular embodiment, the memory 602 may be a non-volatile solid-state memory.

[0082] In one embodiment, the memory 602 may be a read-only memory (ROM). In one embodiment, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or a flash memory, or a combination of two or more of these.

[0083] The processor 601 reads and executes the computer program instructions stored in the memory 602 to implement any one of the bone density measurement methods based on the deep learning algorithm in the above embodiments.

[0084] In one example, the electronic device may further include a communication interface 603 and a bus 610. Figure 6 As shown, the processor 601, the memory 602, and the communication interface 603 are connected via a bus 610 and communicate with each other.

[0085] The communication interface 603 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0086] Bus 610 includes hardware, software or both, and the parts of electronic equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 610 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.

[0087] In addition, in combination with the bone density measurement method based on deep learning algorithm in the above embodiments, the embodiment of the present application can provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any one of the bone density measurement methods based on deep learning algorithm in the above embodiments is implemented.

[0088] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0089] The functional modules shown in the above-described block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. Programs or code segments can be stored in machine-readable media, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable media" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0090] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0091] The above reference is according to the method of the embodiment of the present application, the flow chart of the device (system) and the computer program product and / or the block diagram described various aspects of the present application.It should be understood that each square box in the flow chart and / or the block diagram and the combination of each square box in the flow chart and / or the block diagram can be realized by computer program instructions.These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the realization of the function / action specified in one or more square boxes of the flow chart and / or the block diagram.Such a processor can be but is not limited to a general-purpose processor, a special-purpose processor, a special application processor or a field programmable logic circuit.It can also be understood that each square box in the block diagram and / or the flow chart and the combination of the square boxes in the block diagram and / or the flow chart can also be realized by the dedicated hardware that performs the specified function or action, or can be realized by the combination of dedicated hardware and computer instructions.

[0092] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A bone density measurement method based on a deep learning algorithm, characterized in that: include: Obtain CT images of the spine; Input the spinal CT image into the preset segmentation network model and output the corresponding spinal segmentation result; The spine segmentation results are input into a preset regression network model, and the corresponding bone density measurement results are output; After obtaining the CT images of the spine, it also includes: Normalize the spinal CT images; The entire network structure of the segmentation network model adopts a parallel approach with four branches, and each time a new branch is generated, the branches are allowed to fuse information through concatenation; Among the four branches output by the final model, only the output of the first branch is used as the final output; The first branch uses convolution blocks, each of which consists of 2 convolutional operations, 1 batchnormalization, and 1 ReLu activation function; The image size remains unchanged throughout the branch, which effectively extracts local detail information of the image; In all branches except the first one, the convolutional block is replaced with the Transformer block to effectively extract the global information in the image. The input image of the entire model is a CT original image, and the output is the corresponding segmentation mask; after normalization, the original image is input into the deep learning model, and after passing through the deep learning network, the corresponding mask image is finally generated; The entire network structure uses 4 branches in parallel, and each time a new branch is generated, the branches are allowed to fuse information through concatenation. Among the four branches output by the final model, only the head of the first branch is used as the final output. In the entire model structure, the first branch uses convolutional blocks. Each convolutional block consists of 2 convolutional operations + 1 batch normalization + 1 ReLu activation function. The image size remains unchanged throughout the branch. In this way, local detail information of the image can be effectively extracted. In the second, third, and fourth branches of the entire model, in order to better extract global information, the convolutional block is replaced by the transformer block. Through this module design, the transformer block can effectively extract global information from the image, thereby achieving the purpose of allowing the convolutional block and the transformer block to play their respective advantages; The Transformer Block includes the Laplacian operator, which is used to detect the second-order derivatives in the image and help locate the edges in the image; Convolve this convolution kernel with the image to get the gradient change; Edge detection based on gradient changes; The Laplacian operator is an edge detection operator that processes images through convolution operations. The Laplacian operator is used to detect the second-order derivatives in the image to help locate the edges in the image. A discrete form of the Laplacian operator: This is a 3x3 convolution kernel, whose central element is 4, the elements of the four upper, lower, left and right neighborhoods are 1, and the other elements are 0; the Laplacian operator performs a convolution operation on the convolution kernel and the image, and calculates the gradient change to achieve edge detection; for a grayscale image I, the convolution operation of the Laplacian operator is expressed by the following formula: in, and Represent the second-order derivatives of the image in the horizontal and vertical directions respectively; In practical applications, the performance of the operator is adjusted by using Laplacian convolution kernels of different sizes or performing Gaussian smoothing operators.

2. A bone density measurement device based on a deep learning algorithm, characterized in that: The device comprises: An image acquisition module, used for acquiring a spinal CT image; The spine segmentation module is used to input the spine CT image into a preset segmentation network model and output the corresponding spine segmentation result; A bone density measurement module is used to input the spine segmentation results into a preset regression network model and output the corresponding bone density measurement results; After obtaining the CT images of the spine, it also includes: Normalize the spinal CT images; The entire network structure of the segmentation network model adopts a parallel approach with four branches, and each time a new branch is generated, the branches are allowed to fuse information through concatenation; Among the four branches output by the final model, only the output of the first branch is used as the final output; The first branch uses convolution blocks, each of which consists of 2 convolutional operations, 1 batchnormalization, and 1 ReLu activation function; The image size remains unchanged throughout the branch, which effectively extracts local detail information of the image; In all branches except the first one, the convolutional block is replaced with the Transformer block to effectively extract the global information in the image. The input image of the entire model is a CT original image, and the output is the corresponding segmentation mask; after normalization, the original image is input into the deep learning model, and after passing through the deep learning network, the corresponding mask image is finally generated; The entire network structure uses 4 branches in parallel, and each time a new branch is generated, the branches are allowed to fuse information through concatenation. Among the four branches output by the final model, only the head of the first branch is used as the final output. In the entire model structure, the first branch uses convolutional blocks. Each convolutional block consists of 2 convolutional operations + 1 batch normalization + 1 ReLu activation function. The image size remains unchanged throughout the branch. In this way, local detail information of the image can be effectively extracted. In the second, third, and fourth branches of the entire model, in order to better extract global information, the convolutional block is replaced by the transformer block. Through this module design, the transformer block can effectively extract global information from the image, thereby achieving the purpose of allowing the convolutional block and the transformer block to play their respective advantages; The Transformer Block includes the Laplacian operator, which is used to detect the second-order derivatives in the image and help locate the edges in the image; Convolve this convolution kernel with the image to get the gradient change; Edge detection based on gradient changes; The Laplacian operator is an edge detection operator that processes images through convolution operations. The Laplacian operator is used to detect the second-order derivatives in the image to help locate the edges in the image. A discrete form of the Laplacian operator: This is a 3x3 convolution kernel, whose central element is 4, the elements of the four upper, lower, left and right neighborhoods are 1, and the other elements are 0; the Laplacian operator performs a convolution operation on the convolution kernel and the image, and calculates the gradient change to achieve edge detection; for a grayscale image I, the convolution operation of the Laplacian operator is expressed by the following formula: in, and Represent the second-order derivatives of the image in the horizontal and vertical directions respectively; In practical applications, the performance of the operator is adjusted by using Laplacian convolution kernels of different sizes or performing Gaussian smoothing operators.

3. An electronic device, characterized in that: The electronic device comprises: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, the bone density measurement method based on the deep learning algorithm as described in claim 1 is implemented.

4. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by the processor, the bone density measurement method based on the deep learning algorithm as claimed in claim 1 is implemented.

Images