WO2021114622A1

WO2021114622A1 - Spinal-column curvature measurement method, apparatus, computer device, and storage medium

Info

Publication number: WO2021114622A1
Application number: PCT/CN2020/099252
Authority: WO
Inventors: 吴海萍; 陶蓉; 徐尚良; 毋戈; 吕传峰
Original assignee: 平安科技（深圳）有限公司
Priority date: 2020-06-08
Filing date: 2020-06-30
Publication date: 2021-06-17
Also published as: CN111671454A; CN111671454B

Abstract

Provided are a spinal-column curvature measurement method, apparatus, computer device, and storage medium, relating to the field of artificial intelligence, and used in the field of smart medical care. The method comprises: inputting a spinal column image to be measured into a preset neural network model, and extracting a candidate frame image having the highest inclination probability from the spinal column image to be measured; extracting the key points associated with the vertebrae from the candidate frame image in the model by means of preset key points; coding the key points of each vertebra by means of a key point correction model and obtaining a coding result; according to the coding result, correcting the position of the key points associated with the vertebrae in the candidate frame image, and according to the position correction result, correcting the candidate frame image; according to the key points in the corrected candidate frame image, determining the Cobb angle used to evaluate the curvature of the spinal column. The invention also relates to blockchain technology, said Cobb angle being stored on the blockchain. The invention is used for saving the time cost and labor cost of measuring the Cobb angle.

Description

脊柱弯曲角度测量方法、装置、计算机设备及存储介质Spinal bending angle measurement method, device, computer equipment and storage medium

本申请要求于2020年6月8日提交中国专利局、申请号为202010514042.2，发明名称为“脊柱弯曲角度测量方法、装置、计算机设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。This application claims the priority of the Chinese patent application filed with the Chinese Patent Office on June 8, 2020, with the application number 202010514042.2, and the invention titled "Spine bending angle measurement method, device, computer equipment and storage medium". The entire content of the Chinese patent application is approved. The reference is incorporated in this application.

技术领域Technical field

本申请涉及人工智能的图像处理领域，尤其涉及一种脊柱弯曲角度测量方法、装置、计算机设备及存储介质。This application relates to the field of artificial intelligence image processing, and in particular to a method, device, computer equipment, and storage medium for measuring the angle of spine curvature.

背景技术Background technique

脊柱侧弯是一种常见脊柱畸形病变，在青少年和中老年人群中都有一定的发病率。Scoliosis is a common spinal deformity, which has a certain incidence in adolescents and middle-aged and elderly people.

发明人意识到，目前在脊柱侧弯筛查的过程中，最大的难点就是脊柱侧弯的测量，通过传统图像学方法的边缘检测去定位椎骨的边缘和角等位置，这种方法存在精度低和受图像质量影响大的问题，并且该方法对于病变区域的边缘和点的识别成功率会大幅下降，因此，临床上依赖人工测量脊柱侧弯，该方法存在时间成本和人力成本高的问题，且测量精度不高。The inventor realized that in the current scoliosis screening process, the biggest difficulty is the measurement of scoliosis. The edge detection of traditional imaging methods is used to locate the edges and corners of the vertebrae. This method has low accuracy. And the problem that is greatly affected by the image quality, and the success rate of the method for identifying the edges and points of the diseased area will be greatly reduced. Therefore, clinically relying on manual measurement of scoliosis, this method has the problems of high time cost and labor cost. And the measurement accuracy is not high.

因此，本领域技术人员亟需寻找新的一种技术方案来解决上述提到的测量精度不高等问题。Therefore, those skilled in the art urgently need to find a new technical solution to solve the aforementioned problem of low measurement accuracy.

发明内容Summary of the invention

基于此，有必要针对上述技术问题，提供一种脊柱弯曲角度测量方法、装置、计算机设备及存储介质，用于节省测量Cobb角的时间成本和人力成本，并提升测量精度。Based on this, it is necessary to provide a method, device, computer device, and storage medium for measuring the angle of spine curvature in response to the above technical problems, so as to save the time and labor costs of measuring the Cobb angle, and to improve the measurement accuracy.

一种脊柱弯曲角度测量方法，包括：A method for measuring the bending angle of the spine includes:

自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位；所述脊柱冠状位影像数据集中包含多张所述待测量脊柱图像；Acquire the to-be-measured spine images containing the preset region of interest from the spine coronal image data set, and adjust all the acquired spine images to be measured to a uniform graphic size and window width; the spine coronal image data set Containing multiple images of the spine to be measured;

将调整后的所述待测量脊柱图像输入至预设神经网络模型，分别自所述待测量脊柱图像的预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，得到多张候选框图像；每张所述候选框图像中已定位出位于脊柱中的椎骨，且一张所述候选框图像只对应一块所述椎骨；Input the adjusted spine image to be measured into a preset neural network model, and extract the candidate frame images with the highest tilt probability from the preset regions of interest of the spine image to be measured to obtain multiple candidate frame images Vertebrae located in the spine have been located in each candidate frame image, and one candidate frame image corresponds to only one of the vertebrae;

将所述候选框图像输入至预设关键点提取模型中，自所述候选框图像中提取与已定位的所述椎骨关联的多分辨率关键点；一块椎骨关联四个关键点；Input the candidate frame image into a preset key point extraction model, and extract the multi-resolution key points associated with the positioned vertebra from the candidate frame image; one vertebra is associated with four key points;

通过关键点矫正模型对各所述椎骨的所述关键点的相关性进行编码并获取编码结果，根据所述编码结果对所述候选框图像中与所述椎骨关联的所述关键点进行位置矫正，并根据位置矫正结果对所述候选框图像进行修正；Encode the correlation of the key points of each vertebra through the key point correction model and obtain the encoding result, and perform position correction on the key points associated with the vertebra in the candidate frame image according to the encoding result , And correct the candidate frame image according to the position correction result;

根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角。According to the key points in the corrected candidate frame image, the Cobb angle for evaluating the curvature of the spine is determined.

一种脊柱弯曲角度测量装置，包括：A device for measuring the bending angle of the spine includes:

调整模块，用于自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位；所述脊柱冠状位影像数据集中包含多张所述待测量脊柱图像；The adjustment module is used to obtain the to-be-measured spine image containing the preset region of interest from the spine coronal image data set, and adjust all the obtained to-be-measured spine images to a uniform graphic size and window width and window level; the spine The coronal image data set contains multiple images of the spine to be measured;

提取模块，用于将调整后的所述待测量脊柱图像输入至预设神经网络模型，分别自所述待测量脊柱图像的预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，得到多张候选框图像；每张所述候选框图像中已定位出位于脊柱中的椎骨，且一张所述候选框图像只对应一块所述椎骨；The extraction module is used to input the adjusted spine image to be measured into a preset neural network model, and extract the candidate frame images with the highest tilt probability from the preset regions of interest of the spine image to be measured, respectively, to obtain Multiple candidate frame images; each of the candidate frame images has located vertebrae located in the spine, and one candidate frame image corresponds to only one of the vertebrae;

修正模块，用于通过关键点矫正模型对各所述椎骨的所述关键点的相关性进行编码并获取编码结果，根据所述编码结果对所述候选框图像中与所述椎骨关联的所述关键点进行位置矫正，并根据位置矫正结果对所述候选框图像进行修正；The correction module is used to encode the correlation of the key points of each of the vertebrae through the key point correction model and obtain the encoding result, and according to the encoding result, perform correction on the candidate frame image associated with the vertebrae. Performing position correction on the key points, and correcting the candidate frame image according to the position correction result;

确定模块，用于根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角。The determining module is configured to determine the Cobb angle for evaluating the curvature of the spine according to the key points in the corrected candidate frame image.

一种计算机设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现上述脊柱弯曲角度测量方法，例如，实现以下步骤：A computer device includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and the processor implements the above-mentioned method for measuring spinal curvature angle when the processor executes the computer program, for example, Implement the following steps:

一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现上述脊柱弯曲角度测量方法，例如，实现以下步骤：A computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, realizes the above-mentioned spinal curvature angle measurement method, for example, the following steps are realized:

上述脊柱弯曲角度测量方法、装置、计算机设备及存储介质，可节省时间成本和人力成本，并存在测量精度高的优点。The above-mentioned method, device, computer equipment and storage medium for measuring the angle of spine curvature can save time and labor costs, and have the advantages of high measurement accuracy.

附图说明Description of the drawings

图1是本申请一实施例中脊柱弯曲角度测量方法的一应用环境示意图；FIG. 1 is a schematic diagram of an application environment of the method for measuring the bending angle of the spine in an embodiment of the present application;

图2是本申请一实施例中脊柱弯曲角度测量方法的一流程图；FIG. 2 is a flow chart of a method for measuring the bending angle of the spine in an embodiment of the present application;

图3是本申请一实施例中脊柱弯曲角度测量装置的结构示意图；FIG. 3 is a schematic diagram of the structure of the spine bending angle measuring device in an embodiment of the present application;

图4是本申请一实施例中计算机设备的一示意图。Fig. 4 is a schematic diagram of a computer device in an embodiment of the present application.

具体实施方式Detailed ways

下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行描述。The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

本申请的技术方案可应用于人工智能、智慧城市和/或数字医疗技术领域，可提升脊柱弯曲角度测量精度，以进行健康管理，实现智慧医疗。The technical solution of the present application can be applied to the fields of artificial intelligence, smart city and/or digital medical technology, and can improve the accuracy of spine bending angle measurement for health management and smart medical treatment.

本申请提供的脊柱弯曲角度测量方法，可应用在如图1的应用环境中，其中，客户端通过网络与服务器进行通信。其中，客户端可以但不限于各种个人计算机、笔记本电脑、智能手机、平板电脑和便携式可穿戴设备。服务器可以用独立的服务器或者是多个服务器组成的服务器集群来实现。The method for measuring the bending angle of the spine provided in this application can be applied to the application environment as shown in FIG. 1, where the client communicates with the server through the network. Among them, the client can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices. The server can be implemented as an independent server or a server cluster composed of multiple servers.

在一实施例中，如图2所示，提供一种脊柱弯曲角度测量方法，以该方法应用在图1中的服务器为例进行说明，包括如下步骤：In an embodiment, as shown in FIG. 2, a method for measuring the angle of spine curvature is provided. Taking the method applied to the server in FIG. 1 as an example for description, the method includes the following steps:

S10，自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位；所述脊柱冠状位影像数据集中包含多张所述待测量脊柱图像；S10: Obtain a spine image to be measured that includes a preset region of interest from the spine coronal image data set, and adjust all the acquired spine images to be measured to a uniform graphic size and window width; the spine coronal image The data set contains multiple images of the spine to be measured;

可理解地，脊柱冠状位影像数据集可来源于医院存放待测量脊柱图像的数据库，因此脊柱冠状位影像数据集中可包含多人的待测量脊柱图像，且该待测量脊柱图像可通过医院中的专门的拍片设备(如computed tomography，简称为CT，其是一种功能齐全的病情探测仪器)进行获取，本实施例可从脊柱冠状位影像数据集中选取出包含预设感兴趣区域的待测量脊柱图像(该待测量脊柱图像与每一个测量者的身份信息互相关联，在本实施例每一次对待测量脊柱图像进行处理的过程中，该处理过程应当是针对同一个人的待测量脊柱图像)，其中，待测量脊柱图像应当为多模态多尺寸，因此本实施例可通过预处理(裁剪和图像插值方法)将所有待测量脊柱图像调整为统一的图形尺寸(分辨率为512*256)和窗宽窗位(窗宽是CT图像上显示的CT值范围，在此CT值范围内的组织和病变均以不同的模拟灰度显示；窗位是窗的中心位置，同样的窗宽，由于窗位不同，其所包括CT值范围的CT值也有差异)。本实施例通过调整手段可将脊柱冠状位影像数据集中不同类型的待测量脊柱图像转换为统一的形式的待测量脊柱图像，以便于后续的处理方法能按照统一标准处理待测量脊柱图像。Understandably, the spine coronal image data set can be derived from the hospital storing the spine images to be measured. Therefore, the spine coronal image data set can contain multiple people's spine images to be measured, and the spine images to be measured can be passed through the hospital Special filming equipment (such as computed tomography, abbreviated as CT, which is a full-featured disease detection instrument) can be obtained. In this embodiment, the spine to be measured containing the preset region of interest can be selected from the coronal image data set of the spine Image (the spine image to be measured is correlated with the identity information of each measurer. In this embodiment, each time the spine image to be measured is processed, the processing procedure should be for the spine image to be measured of the same person), wherein , The spine image to be measured should be multi-modal and multi-size. Therefore, in this embodiment, all spine images to be measured can be adjusted to a uniform graphic size (resolution of 512*256) and window through preprocessing (cropping and image interpolation methods). Wide window level (window width is the range of CT values displayed on the CT image, the tissues and lesions within this range of CT values are displayed in different simulated gray scales; the window level is the center of the window, the same window width, due to the window For different bits, the CT value of the CT value range included is also different). In this embodiment, different types of spine images to be measured in the coronal image data set of the spine can be converted into uniform forms of spine images to be measured through adjustment means, so that subsequent processing methods can process the spine images to be measured in accordance with uniform standards.

S20，将调整后的所述待测量脊柱图像输入至预设神经网络模型，分别自所述待测量脊柱图像的预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，得到多张候选框图像；每张所述候选框图像中已定位出位于脊柱中的椎骨，且一张所述候选框图像只对应一块所述椎骨；S20. Input the adjusted spine image to be measured into a preset neural network model, and extract the candidate frame images with the highest tilt probability from the preset regions of interest of the spine image to be measured, respectively, to obtain multiple candidates Frame images; each of the candidate frame images has located vertebrae located in the spine, and one candidate frame image corresponds to only one of the vertebrae;

可理解地，预设神经网络模型包括但不限于RPN(是能把一个任意尺度的图片作为输入，并输出一系列的矩形object proposals的框架)、FPN(是一种利用常规CNN模型来高效提取图片中各维度特征的框架)、Yolo(是基于一个单独的end-to-end网络，完成从原始图像的输入到物***置和类别的输出的框架)和Maskrcnn(是可以完成目标分类、目标检测、语义分割、实例分割、人体姿势识别等多种任务的框架)等，该预设神经网络模型主要是用于生成候选框图像，其中，候选框图像中应当包含脊柱(脊柱中包含了17块椎骨)中的一整块的椎骨，一张椎骨应当对应一张候选框图像，具体地，通过本实施例中的预设神经网络模型可从椎骨的形状中提前确定出该候选框图像中的椎骨是否发生了倾斜(比如脊柱侧弯的患者椎骨带有倾斜角度，特别是脊柱中的端椎，该端椎相比正常的端椎，形状会存在差异)，也可定位出各椎骨在脊柱中的具***置，以提取出发生倾斜概率最高的候选框图像。本实施例主要通过具有深度网络的预设神经网络模型精准定位出位于脊柱中的各个椎骨(可通过该模型回归各个椎骨的中心点、宽度和高度)，并生成出包含各个椎骨的候选框图像。Understandably, the preset neural network model includes, but is not limited to, RPN (a frame that can take a picture of any scale as input and output a series of rectangular object proposals), FPN (a kind of efficient extraction using a conventional CNN model) The frame of each dimensional feature in the picture), Yolo (based on a single end-to-end network, which completes the frame from the input of the original image to the output of the object position and category) and Maskrcnn (the target classification and target detection can be completed) , Semantic segmentation, instance segmentation, human pose recognition and other tasks. The preset neural network model is mainly used to generate candidate frame images. Among them, the candidate frame image should contain the spine (the spine contains 17 blocks). One piece of vertebrae in the vertebrae), one vertebrae should correspond to one candidate frame image. Specifically, the preset neural network model in this embodiment can determine the candidate frame image in advance from the shape of the vertebrae. Whether the vertebrae are tilted (for example, the vertebrae in patients with scoliosis have tilt angles, especially the end vertebrae in the spine, which have different shapes compared to the normal end vertebrae), and the vertebrae can also be located in the spine In order to extract the candidate frame image with the highest probability of tilting. This embodiment mainly uses a preset neural network model with a deep network to accurately locate each vertebra in the spine (the center point, width, and height of each vertebra can be returned through the model), and generate a candidate frame image containing each vertebra .

S30，将所述候选框图像输入至预设关键点提取模型中，自所述候选框图像中提取与已定位的所述椎骨关联的多分辨率关键点；一块椎骨关联四个关键点；S30. Input the candidate frame image into a preset key point extraction model, and extract the multi-resolution key points associated with the positioned vertebra from the candidate frame image; one vertebra is associated with four key points;

可理解地，预设关键点提取模型包括CPN(CascadedPyramid Network，Cascaded指的是级联的意思，代表了网络级联了2个类似的模块GolbalNet和RefineNet，Pyramid指的是类似于FPN的金字塔网络结构，可用于识别人体部位中的关键点)，该预设关键点提取模型主要用于从候选框图像中提取出关键点，其中，由于一张候选框图像代表一块椎骨，而从候选框图像中提取的关键点只能为一块椎骨的四个顶点，因此每块椎骨都会关联到四个关键点，若从待测量脊柱图像能提取到一块脊柱中17块椎骨所对应的候选框图像时，本实施例则可自候选框图像中提取到17*4＝68个关键点，需要说明的是，预设关键点提取模型是一种级联金字塔结构，本实施例借用该预设关键点提取模型能够同时兼顾人体关节点的局部信息以及全局信息，因此通过该预设关键点提取模型可在深度网络的多个分辨率上做关键点的提取，进而可提取到具有多分辨率特点的关键点。本实施例主要通过具有深度网络的预设关键点提取模型精准识别出包含椎骨的候选框图像中的多分辨率关键点，且由于候选框图像中的椎骨已被精准定位，因此对于本实施例来说，也可提高提取关键点的工作效率。Understandably, the preset key point extraction model includes CPN (CascadedPyramid Network, Cascaded refers to the meaning of cascade, which means that the network cascades two similar modules, GolbalNet and RefineNet, and Pyramid refers to a pyramid network similar to FPN. Structure, which can be used to identify key points in human body parts). The preset key point extraction model is mainly used to extract key points from candidate frame images. Among them, since a candidate frame image represents a vertebra, and the candidate frame image The key points extracted can only be the four vertices of a vertebra, so each vertebra will be associated with four key points. If the candidate frame image corresponding to 17 vertebrae in a spine can be extracted from the spine image to be measured, In this embodiment, 17*4=68 key points can be extracted from the candidate frame image. It should be noted that the preset key point extraction model is a cascaded pyramid structure, and this embodiment borrows the preset key point extraction The model can take into account the local and global information of the human joint points at the same time, so the model can be extracted from multiple resolutions of the deep network through the preset key point extraction model, and then the key with multi-resolution characteristics can be extracted point. This embodiment mainly uses a preset key point extraction model with a deep network to accurately identify multi-resolution key points in a candidate frame image containing vertebrae, and because the vertebrae in the candidate frame image have been accurately positioned, for this embodiment In other words, the work efficiency of extracting key points can also be improved.

S40，通过关键点矫正模型对各所述椎骨的所述关键点的相关性进行编码并获取编码结果，根据所述编码结果对所述候选框图像中与所述椎骨关联的所述关键点进行位置矫正，并根据位置矫正结果对所述候选框图像进行修正；S40: Encode the correlation of the key points of each of the vertebrae through a key point correction model and obtain an encoding result, and perform processing on the key points associated with the vertebrae in the candidate frame image according to the encoding result. Position correction, and correct the candidate frame image according to the position correction result;

可理解地，基于RNN(Recurrent Neural Network，循环神经网络，其中，神经网络是一种节点定向连接成环的人工神经网络)的关键点矫正模型是用于对关键点进行矫正，由于脊柱中椎骨关联的关键点有着较强分布规律(关键点与关键点之间的相关性可决定出分布规律，具体地，该相关性可由基于RNN的关键点矫正模型中的LSTM模块进行识别，该基于RNN的关键点矫正模型在训练过程中，可从步骤S20定位出的候选框图像提取出关于椎块的特征向量，并从步骤S30提取出候选框图像中关键点的位置信息，最后将特征向量和位置信息输入至基于RNN的关键点矫正模型中的LSTM模块进行训练)，比如，每块椎骨的四个关键点会平均分布在脊柱的两侧，一旦分布规律出现异常，则需对各椎骨关联的关键点进行位置矫正(正常分布规律的关键点可为

异常分布规律的关键点可为

其中，1代表关键点，而0代表不存在关键点，可见，从分布规律也可看出关键点的位置信息)，具体地，本实施例通过各椎骨的关键点的相关性来对关键点进行编码(将自注意力机制作为基于RNN的关键点矫正模型中的一层网络结构，则可使用自注意力进行编码，编码的作用是为了对不定长的序列转换成一个定长的向量)，得到了各块椎骨的编码结果，并通过解码器对编码结果进行解码，根据解码结果中各种位置向量(可确定出关键点具体坐标的位置信息)对异常分布规律的关键点进行位置矫正，并可根据位置矫正结果去代替掉候选框图像中的原先异常分布规律的关键点。本实施例主要通过自注意力机制对关键点的相关性进行编码，以提升关键点位置信息的识别效率，进而提高位置矫正效率。 Understandably, the key point correction model based on RNN (Recurrent Neural Network, cyclic neural network, where the neural network is an artificial neural network in which the nodes are oriented and connected into a ring) is used to correct the key points, because the vertebrae in the spine The key points of the association have a strong distribution law (the correlation between key points and key points can determine the distribution law. Specifically, the correlation can be identified by the LSTM module in the key point correction model based on RNN, which is based on RNN In the training process of the key point correction model, the feature vector about the vertebral block can be extracted from the candidate frame image located in step S20, and the position information of the key point in the candidate frame image can be extracted from step S30, and finally the feature vector and The position information is input to the LSTM module in the RNN-based key point correction model for training). For example, the four key points of each vertebra will be evenly distributed on both sides of the spine. Once the distribution pattern is abnormal, each vertebra needs to be associated Position correction of the key points (the key points of the normal distribution law can be

The key point of the abnormal distribution law can be

Among them, 1 represents a key point, and 0 represents that there is no key point. It can be seen that the position information of the key point can also be seen from the distribution law). Specifically, this embodiment uses the correlation of the key points of each vertebra to compare the key points. Encoding (using the self-attention mechanism as a layer of network structure in the key point correction model based on RNN, self-attention can be used for encoding, and the function of encoding is to convert a variable-length sequence into a fixed-length vector) , Get the coding result of each vertebra, and decode the coding result through the decoder, according to the various position vectors in the decoding result (the position information of the specific coordinates of the key points can be determined), the position correction of the key points of the abnormal distribution law is carried out , And can replace the key points of the original abnormal distribution law in the candidate frame image according to the position correction result. In this embodiment, the relevance of key points is mainly encoded by the self-attention mechanism, so as to improve the efficiency of identifying the position information of the key points, thereby improving the efficiency of position correction.

S50，根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角。S50: Determine the Cobb angle for evaluating the curvature of the spine according to the key points in the corrected candidate frame image.

可理解地，脊柱侧弯的诊断标准主要是通过Cobb角，也即该Cobb角是可用于评估脊柱弯曲角度，其中，Cobb角可以是指脊柱中的头侧端椎上缘的垂线与脊柱中的尾侧端椎下缘垂线的交角。Understandably, the diagnostic criteria for scoliosis is mainly based on the Cobb angle, that is, the Cobb angle can be used to evaluate the angle of spine curvature, where the Cobb angle can refer to the vertical line of the upper edge of the cranial end vertebra in the spine and the spine The angle of intersection of the perpendicular to the lower edge of the caudal end vertebra.

进一步地，所述自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位，包括：Further, the acquisition of the to-be-measured spine images including the preset region of interest from the coronal image data set of the spine, and the adjustment of all the acquired spine images to be measured to a uniform graphic size and window width and window level, include:

通过预设方向投影定位出所述脊柱冠状位影像数据集中包含预设感兴趣区域的所述待测量脊柱图像；Locating the to-be-measured spine image that includes a preset region of interest in the coronal image data set of the spine through a preset direction projection;

对所述待测量脊柱图像按照预设比例表进行裁剪，通过图像插值方法将裁剪后的所述待测量脊柱图像调整为统一图形尺寸，并通过CT成像技术将统一图形尺寸的所述待测量脊柱图像调整为统一的窗宽窗位。The spine image to be measured is cropped according to a preset scale table, the cropped spine image to be measured is adjusted to a uniform graphic size by an image interpolation method, and the spine to be measured with a uniform graphic size is converted by CT imaging technology The image is adjusted to a uniform window width and level.

可理解地，预设方向包括水平方向和垂直方向，投影可通过专门的投影设备，可用于定位出包含预设感兴趣区域的待测量脊柱图像；图像插值可为三线性插值法，通过该三线性插值法可进一步地为裁剪过后的待测量脊柱图像进行调整(避免过度裁剪而使待测量脊柱图像失去太多图像内容)；统一的窗宽窗位可为骨窗。本实施例主要是将待测量脊柱图像调整为统一的图形尺寸和窗宽窗位，且本实施例中的图像插值法可从低分辨率图像生成高分辨率图像。Understandably, the preset direction includes the horizontal direction and the vertical direction. The projection can be through a special projection device, which can be used to locate the spine image to be measured that contains the preset region of interest; the image interpolation can be a trilinear interpolation method, and the three The linear interpolation method can further adjust the cropped spine image to be measured (to avoid excessive cropping and the spine image to be measured loses too much image content); the uniform window width and window level can be the bone window. This embodiment mainly adjusts the spine image to be measured to a uniform graphic size and window width and window level, and the image interpolation method in this embodiment can generate a high-resolution image from a low-resolution image.

进一步地，所述预设神经网络模型包括RPN；所述自所述待测量脊柱图像的所述预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，包括：Further, the preset neural network model includes RPN; the extraction of the candidate frame image with the highest tilt probability from the preset region of interest of the spine image to be measured includes:

利用所述RPN中的卷积层从包含所述预设感兴趣区域的所述待测量脊柱图像中获取特征图；Using the convolutional layer in the RPN to obtain a feature map from the to-be-measured spine image containing the preset region of interest;

将所述特征图各点的多维特征与所述RPN中的全连接层连接后，获取所述全连接层输出的预测结果；所述预测结果表征出所述全连接层输出的achor box包含目标的目标概率；After connecting the multi-dimensional features of each point of the feature map with the fully connected layer in the RPN, the prediction result output by the fully connected layer is obtained; the prediction result indicates that the achor box output by the fully connected layer contains the target Target probability;

提取出所述目标概率大于预设概率的所述预测结果，将所述预测结果中包含目标的所述achor box进行平移和缩放后，生成预设数量的候选框图像，并利用非极大值筛选出发生倾斜概率最高的所述候选框图像。Extract the prediction result whose target probability is greater than the preset probability, translate and zoom the achor box containing the target in the prediction result, generate a preset number of candidate frame images, and use non-maximum values The candidate frame image with the highest probability of occurrence of tilt is screened out.

可理解地，卷积层可为RPN中的VGG16等卷积神经网络的卷积层，第一特征图也可通过滑动窗口对待测量脊柱图像进行卷积；特征图中的各点的多维特征可与两个全连接层进行连接，其中，每个全连接层的都可输出achor box，每个achor box两个值分别表示包含目标的目标概率和不包含目标的目标概率；候选框图像可通过回归值对achor box进行平移和缩放而产生；非极大值可抑制筛选出预测分数较高的(也即发生倾斜概率最高)候选框图像。Understandably, the convolutional layer can be a convolutional layer of a convolutional neural network such as VGG16 in RPN. The first feature map can also be convolved through a sliding window to be measured on the spine image; the multi-dimensional features of each point in the feature map can be Connect with two fully connected layers, where each fully connected layer can output an achor box, and each achor box has two values representing the target probability that contains the target and the target probability that does not contain the target; the candidate frame image can pass The regression value is generated by panning and zooming the achor box; the non-maximum value can inhibit the screening of candidate frame images with higher prediction scores (that is, the highest probability of tilting).

进一步地，所述预设关键点提取模型包括CPN；所述自所述候选框图像中提取与已定位的所述脊柱中的所述椎骨关联的多分辨率关键点，包括：Further, the preset key point extraction model includes CPN; the extracting from the candidate frame image the multi-resolution key points associated with the vertebrae in the spine that has been positioned includes:

通过所述CPN中的GlobalNe对各所述候选框图像的第一关键点进行检测；所述第一关键点为属于第一分辨率范围的可见的关键点；Detect the first key point of each candidate frame image through GlobalNe in the CPN; the first key point is a visible key point belonging to a first resolution range;

通过所述CPN中的RefineNe对各所述候选框图像的第二关键点进行检测；所述第二关键点为属于第二分辨率范围的不可见的关键点；Detect the second key points of each candidate frame image through RefineNe in the CPN; the second key points are invisible key points belonging to a second resolution range;

提取检测成功的所述第一关键点和所述第二关键点，并将所述第一关键点和所述第二关键点作为所述脊柱中所有椎骨的多分辨率关键点。The first key point and the second key point that are successfully detected are extracted, and the first key point and the second key point are used as multi-resolution key points of all vertebrae in the spine.

可理解地，GlobalNe和RefineNe是CPN中的两个模块；第一关键点的可见或第二关键点的不可见，可反映出两类关键点在候选框图像中的分辨率(第一分辨率范围和第二分辨率范围，且两个分辨率范围可根据需求进行自行设置)，可见的第一关键点是可通过预设关键点提取模型直接提取到，不可见的第二关键点需通过增大视野才能确定出关键点的位置信息，进而才能通过关键点的位置信息检测到第二关键点。Understandably, GlobalNe and RefineNe are two modules in CPN; the visibility of the first key point or the invisible of the second key point can reflect the resolution of the two types of key points in the candidate frame image (the first resolution Range and the second resolution range, and the two resolution ranges can be set according to requirements), the first visible key point can be directly extracted through the preset key point extraction model, and the invisible second key point needs to pass Only by increasing the field of view can the position information of the key point be determined, and then the second key point can be detected through the position information of the key point.

进一步地，所述Cobb角存储于区块链中，所述根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角，包括：Further, the Cobb angle is stored in a blockchain, and the determination of the Cobb angle for evaluating the curvature of the spine based on the key points in the corrected candidate frame image includes:

以每一张所述候选框图像中的所述椎骨关联的四个所述关键点组装成一个包含第一边线和第二边线的预设形状，并将每一张所述候选框图中的所述第一边线的中心点和第二边线的中心点之间的连线，记录为所述候选框图像对应的所述椎骨的计算线段；所述第一边线和第二边线分别由四个所述关键点中两个不同关键点连接而成；Assemble the four key points associated with the vertebrae in each candidate frame image into a preset shape containing a first edge line and a second edge line, and combine the four key points in each candidate frame image The line between the center point of the first sideline and the center point of the second sideline is recorded as the calculated line segment of the vertebra corresponding to the candidate frame image; the first sideline and the second sideline are respectively determined by Two different key points among the four key points are connected;

计算所有所述计算线段中的任意两条所述计算线段之间所形成的夹角，得到一个包含计算的所有所述夹角的矩阵，从所述矩阵筛选出其中两条所述计算线段之间形成的目标夹角，并将所述目标夹角作为Cobb角。Calculate the included angle formed between any two of the calculated line segments to obtain a matrix containing all the calculated included angles, and filter out one of the two calculated line segments from the matrix The target included angle formed between the two, and the target included angle is taken as the Cobb angle.

可理解地，本实施例中存在17张包含椎骨的矩阵的候选框图像，每一张候选框图像可提取到一条计算线段，每一条计算线段与所有计算线段中的任意一条计算线段之间的夹角可形成一个17*17的矩阵；目标夹角指倾斜角度最大的两条计算线段之间所形成的夹角；目标夹角指PT(Proximal Thoracic，近胸的)的夹角，而MT(Main Thoracic，主胸的)和TL(Thoraco lumbar/Lumbar，腰部的)的夹角通过形成目标夹角的计算线段和其他计算线段(位于主胸和腰部位置的计算线段)之间的夹角所形成。需要说明的是，通过本实施例的算法也是可计算出各个椎骨相对于水平位置发生的倾斜角度(以水平位置作为一条水平线段，可求出水平线段与计算线段之间的夹角)。Understandably, there are 17 candidate frame images containing a matrix of vertebrae in this embodiment. Each candidate frame image can be extracted from a calculation line segment, and each calculation line segment is between each calculation line segment and all calculation line segments. The included angle can form a 17*17 matrix; the target included angle refers to the included angle formed between the two calculated line segments with the largest inclination angle; the target included angle refers to the included angle of PT (Proximal Thoracic), and MT The angle between (Main Thoracic, main chest) and TL (Thoraco lumbar/Lumbar, waist) is the angle between the calculation line segment that forms the target angle and other calculation line segments (the calculation line segments located at the main chest and waist positions) Formed. It should be noted that the algorithm of this embodiment can also calculate the inclination angle of each vertebra relative to the horizontal position (taking the horizontal position as a horizontal line segment, and the angle between the horizontal line segment and the calculated line segment can be calculated).

另外需要强调的是，为进一步保证上述Cobb角的私密和安全性，上述Cobb角还可以存储于一区块链的节点中。其中，本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain)，本质上是一个去中心化的数据库，是一串使用密码学方法相关联产生的数据块，每一个数据块中包含了一批次网络交易的信息，用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。区块链提供的去中心化的完全分布式DNS服务通过网络中各个节点之间的点对点数据传输服务就能实现域名的查询和解析，可用于确保某个重要的基础设施的操作***和固件没有被篡改，可以监控软件的状态和完整性，发现不良的篡改，并确保所传输的数据没用经过篡改，将所述无监督领域自适应网络模型存储在区块链中，能够确保无监督领域自适应网络模型的私密和安全性。In addition, it should be emphasized that, in order to further ensure the privacy and security of the aforementioned Cobb angle, the aforementioned Cobb angle may also be stored in a node of a blockchain. Among them, the blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer. The decentralized and fully distributed DNS service provided by the blockchain can realize the query and resolution of domain names through the point-to-point data transmission service between various nodes in the network, which can be used to ensure that the operating system and firmware of an important infrastructure are not available. If it is tampered with, it can monitor the status and integrity of the software, find bad tampering, and ensure that the transmitted data has not been tampered with. Store the unsupervised domain adaptive network model in the blockchain, which can ensure the unsupervised domain The privacy and security of the adaptive network model.

进一步地，所述根据修正后的所述候选框图像中的所述关键点，确定用于评估脊柱弯曲角度的Cobb角之后，还包括：根据所述Cobb角评估出所述脊柱弯曲角度。本实施例评估的柱弯曲角度可以表征测量者的脊柱侧弯的问题。Further, after determining the Cobb angle for evaluating the curvature of the spine based on the key points in the corrected candidate frame image, the method further includes: evaluating the curvature of the spine based on the Cobb angle. The column bending angle evaluated in this embodiment can characterize the scoliosis problem of the examiner.

综上所述，上述提供了一种脊柱弯曲角度测量方法，使用多种模型的结合去识别出关键点，并基于关键点计算出Cobb角，相比现有的边缘检测方法，并不存在识别椎骨的边缘、角和点等位置精度低的问题(以候选框图像作为识别图像，该候选框图像只代表脊柱的一块椎骨，并可通过预设关键点提取模型中可从候选框图像中高精准提取出关于脊柱各块椎骨的多分辨率关键点)，也不受待测量脊柱图像的图像质量影响(将获取的待测量脊柱图像调整为统一的图形尺寸和窗宽窗位，并可通过预设关键点提取模型可识别出多种分辨率下的候选框图像)，而相比现有的人工手动测量方法，无需担忧时间成本和人力成本的问题，可见，本方法计算Cobb角的方法相比于现有方法可节省时间成本和人力成本，并存在测量精度高的优点。本方法可应用于智慧医疗领域中，从而推动智慧城市的建设。In summary, the above provides a method for measuring the curvature of the spine, which uses a combination of multiple models to identify key points, and calculates the Cobb angle based on the key points. Compared with the existing edge detection methods, there is no recognition The problem of low position accuracy of the edges, corners and points of the vertebrae (the candidate frame image is used as the recognition image, the candidate frame image only represents a vertebra of the spine, and the model can be extracted from the candidate frame image with high precision through the preset key points Extract the multi-resolution key points about each vertebra of the spine), and it is not affected by the image quality of the spine image to be measured (adjust the acquired spine image to be measured to a uniform graphic size and window width and window level, and can be pre-defined Assuming that the key point extraction model can identify candidate frame images at multiple resolutions), compared with the existing manual manual measurement method, there is no need to worry about the time cost and labor cost. It can be seen that the method of calculating the Cobb angle in this method is similar Compared with the existing method, it can save time and labor cost, and has the advantage of high measurement accuracy. This method can be applied in the field of smart medical care to promote the construction of smart cities.

应理解，上述实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。It should be understood that the size of the sequence number of each step in the foregoing embodiment does not mean the order of execution, and the execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiment of the present application.

在一实施例中，提供一种脊柱弯曲角度测量装置，该脊柱弯曲角度测量装置与上述实施例中脊柱弯曲角度测量方法一一对应。如图3所示，该脊柱弯曲角度测量装置包括调整模块11、第一提取模块12、第二提取模块13、修正模块14和确定模块15。各功能模块详细说明如下：In one embodiment, a device for measuring the bending angle of the spine is provided, and the device for measuring the bending angle of the spine corresponds to the method for measuring the bending angle of the spine in the above-mentioned embodiment. As shown in FIG. 3, the spine bending angle measurement device includes an adjustment module 11, a first extraction module 12, a second extraction module 13, a correction module 14 and a determination module 15. The detailed description of each functional module is as follows:

调整模块11，用于自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位；所述脊柱冠状位影像数据集中包含多张所述待测量脊柱图像；The adjustment module 11 is configured to obtain a to-be-measured spine image including a preset region of interest from the spine coronal image data set, and adjust all the obtained to-be-measured spine images to a uniform graphic size and window width and window level; The spine coronal image data set includes a plurality of the spine images to be measured;

第一提取模块12，用于将调整后的所述待测量脊柱图像输入至预设神经网络模型，分别自所述待测量脊柱图像的预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，得到多张候选框图像；每张所述候选框图像中已定位出位于脊柱中的椎骨，且一张所述候选框图像只对应一块所述椎骨；The first extraction module 12 is configured to input the adjusted spine image to be measured into a preset neural network model, and extract the candidate frame with the highest tilt probability from the preset interest area of the spine image to be measured, respectively Image to obtain multiple candidate frame images; each of the candidate frame images has located a vertebra in the spine, and one candidate frame image corresponds to only one of the vertebrae;

第二提取模块13，用于将所述候选框图像输入至预设关键点提取模型中，自所述候选框图像中提取与已定位的所述椎骨关联的多分辨率关键点；一块椎骨关联四个关键点；The second extraction module 13 is configured to input the candidate frame image into a preset key point extraction model, and extract the multi-resolution key points associated with the positioned vertebra from the candidate frame image; a piece of vertebrae is associated Four key points;

修正模块14，用于通过关键点矫正模型对各所述椎骨的所述关键点的相关性进行编码并获取编码结果，根据所述编码结果对所述候选框图像中与所述椎骨关联的所述关键点进行位置矫正，并根据位置矫正结果对所述候选框图像进行修正；The correction module 14 is used to encode the correlation of the key points of each of the vertebrae through a key point correction model and obtain the encoding result. According to the encoding result, all of the candidate frame images that are associated with the vertebrae are encoded. Performing position correction on the key points, and correcting the candidate frame image according to the position correction result;

确定模块15，用于根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角。The determining module 15 is configured to determine the Cobb angle for evaluating the curvature of the spine according to the key points in the corrected candidate frame image.

进一步地，所述调整模块包括：Further, the adjustment module includes:

定位子模块，用于通过预设方向投影定位出所述脊柱冠状位影像数据集中包含预设感兴趣区域的所述待测量脊柱图像；A positioning sub-module, configured to locate the to-be-measured spine image in the coronal image data set of the spine that includes a preset region of interest through a preset direction projection;

调整子模块，用于对所述待测量脊柱图像按照预设比例表进行裁剪，通过图像插值方法将裁剪后的所述待测量脊柱图像调整为统一图形尺寸，并通过CT成像技术将统一图形尺寸的所述待测量脊柱图像调整为统一的窗宽窗位。The adjustment sub-module is used to crop the spine image to be measured according to a preset scale table, adjust the cropped spine image to be measured to a uniform graphic size through an image interpolation method, and use CT imaging technology to adjust the uniform graphic size The spine image to be measured is adjusted to a uniform window width and window level.

进一步地，所述第一提取模块包括：Further, the first extraction module includes:

第一获取子模块，用于利用RPN中的卷积层从包含所述预设感兴趣区域的所述待测量脊柱图像中获取特征图；The first acquisition sub-module is configured to use the convolutional layer in the RPN to acquire a feature map from the to-be-measured spine image containing the preset region of interest;

第二获取子模块，用于将所述特征图各点的多维特征与所述RPN中的全连接层连接后，获取所述全连接层输出的预测结果；所述预测结果表征出所述全连接层输出的achor box包含目标的目标概率；The second acquisition sub-module is used to connect the multi-dimensional features of each point of the feature map with the fully connected layer in the RPN, and then obtain the prediction result output by the fully connected layer; the prediction result represents the fully connected layer; The achor box output by the connection layer contains the target probability of the target;

第一筛选子模块，用于提取出所述目标概率大于预设概率的所述预测结果，将所述预测结果中包含目标的所述achor box进行平移和缩放后，生成预设数量的候选框图像，并利用非极大值筛选出发生倾斜概率最高的所述候选框图像。The first screening sub-module is used to extract the prediction result whose target probability is greater than the preset probability, and after panning and zooming the achor box containing the target in the prediction result, to generate a preset number of candidate frames And use non-maximum values to filter out the candidate frame image with the highest tilt probability.

进一步地，所述第二提取模块包括：Further, the second extraction module includes:

第一检测子模块，用于通过所述CPN中的GlobalNe对各所述候选框图像的第一关键点进行检测；所述第一关键点为属于第一分辨率范围的可见的关键点；The first detection sub-module is configured to detect the first key point of each candidate frame image through GlobalNe in the CPN; the first key point is a visible key point belonging to a first resolution range;

第二检测子模块，用于通过所述CPN中的RefineNe对各所述候选框图像的第二关键点进行检测；所述第二关键点为属于第二分辨率范围的不可见的关键点；The second detection sub-module is configured to detect the second key points of each candidate frame image through RefineNe in the CPN; the second key points are invisible key points belonging to a second resolution range;

提取子模块，用于提取检测成功的所述第一关键点和所述第二关键点，并将所述第一关键点和所述第二关键点作为所述脊柱中所有椎骨的多分辨率关键点。The extraction sub-module is used to extract the first key point and the second key point that are successfully detected, and use the first key point and the second key point as the multi-resolution of all vertebrae in the spine key point.

进一步地，所述确定模块包括：Further, the determining module includes:

记录子模块，用于以每一张所述候选框图像中的所述椎骨关联的四个所述关键点组装成一个包含第一边线和第二边线的预设形状，并将每一张所述候选框图中的所述第一边线的中心点和第二边线的中心点之间的连线，记录为所述候选框图像对应的所述椎骨的计算线段；所述第一边线和第二边线分别由四个所述关键点中两个不同关键点连接而成；The recording sub-module is used to assemble the four key points associated with the vertebrae in each candidate frame image into a preset shape including a first edge line and a second edge line, and combine each The line between the center point of the first sideline and the center point of the second sideline in the candidate block diagram is recorded as the calculated line segment of the vertebra corresponding to the candidate frame image; the first sideline And the second sideline are respectively formed by connecting two different key points among the four key points;

第二筛选子模块，用于计算所有所述计算线段中的任意两条所述计算线段之间所形成的夹角，得到一个包含计算的所有所述夹角的矩阵，从所述矩阵筛选出其中两条所述计算线段之间形成的目标夹角，并将所述目标夹角作为Cobb角。The second screening sub-module is used to calculate the angle formed between any two of the calculation line segments in all the calculation line segments, to obtain a matrix containing all the calculated angles, and to filter out from the matrix The target included angle formed between the two calculated line segments is taken as the Cobb angle.

关于脊柱弯曲角度测量装置的具体限定可以参见上文中对于脊柱弯曲角度测量方法的限定，在此不再赘述。上述脊柱弯曲角度测量装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。For the specific definition of the spine curvature angle measuring device, please refer to the above definition of the spine curvature angle measuring method, which will not be repeated here. Each module in the above-mentioned spine bending angle measurement device can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

在一个实施例中，提供了一种计算机设备，该计算机设备可以是服务器，其内部结构图可以如图4所示。该计算机设备包括通过***总线连接的处理器、存储器、网络接口和数据库。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作***、计算机程序和数据库。该内存储器为非易失性存储介质中的操作***和计算机程序的运行提供环境。该计算机设备的数据库用于存储脊柱弯曲角度测量方法中涉及到的数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种脊柱弯曲角度测量方法。In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. 4. The computer equipment includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide calculation and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer equipment is used to store the data involved in the method of measuring the spine bending angle. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program is executed by the processor to realize a method for measuring the bending angle of the spine.

在一个实施例中，提供了一种计算机设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，处理器执行计算机程序时实现上述实施例中脊柱弯曲角度测量方法的步骤，例如图2所示的步骤S10至步骤S40。或者，处理器执行计算机程序时实现上述实施例中脊柱弯曲角度测量装置的各模块/单元的功能，例如图3所示模块11至模块14的功能。为避免重复，这里不再赘述。In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor. The processor executes the computer program to implement the method for measuring the spine bending angle in the above-mentioned embodiment. , For example, step S10 to step S40 shown in FIG. 2. Or, when the processor executes the computer program, the function of each module/unit of the spine curvature angle measurement device in the above-mentioned embodiment is realized, for example, the function of the module 11 to the module 14 shown in FIG. 3. To avoid repetition, I won’t repeat them here.

在一个实施例中，提供了一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现上述实施例中脊柱弯曲角度测量方法的步骤，例如图2所示的步骤S10至步骤S40。或者，计算机程序被处理器执行时实现上述实施例中脊柱弯曲角度测量装置的各模块/单元的功能，例如图3所示模块11至模块14的功能。为避免重复，这里不再赘述。可选的，本申请涉及的存储介质如计算机可读存储介质可以是非易失性的，也可以是易失性的。In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for measuring the spine bending angle in the above-mentioned embodiment are implemented, for example, the steps shown in FIG. 2 S10 to step S40. Or, when the computer program is executed by the processor, the function of each module/unit of the spine bending angle measurement device in the above-mentioned embodiment is realized, for example, the functions of the modules 11 to 14 shown in FIG. 3. To avoid repetition, I won’t repeat them here. Optionally, the storage medium involved in this application, such as a computer-readable storage medium, may be non-volatile or volatile.

本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage. In the medium, when the computer program is executed, it may include the processes of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and/or volatile memory. Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not a limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain)，本质上是一个去中心化的数据库，是一串使用密码学方法相关联产生的数据块，每一个数据块中包含了一批次网络交易的信息，用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。The blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，仅以上述各功能单元、模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能单元、模块完成，即将所述装置的内部结构划分成不同的功能单元或模块，以完成以上描述的全部或者部分功能。Those skilled in the art can clearly understand that, for the convenience and conciseness of description, only the division of the above functional units and modules is used as an example. In practical applications, the above functions can be allocated to different functional units and modules as needed. Module completion, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

以上所述实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围，均应包含在本申请的保护范围之内。The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, a person of ordinary skill in the art should understand that it can still implement the foregoing The technical solutions recorded in the examples are modified, or some of the technical features are equivalently replaced; these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the application, and should be included in Within the scope of protection of this application.

Claims

一种脊柱弯曲角度测量方法，其中，包括：A method for measuring the bending angle of the spine, which includes:

自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位；所述脊柱冠状位影像数据集中包含多张所述待测量脊柱图像；Acquire the to-be-measured spine images containing the preset region of interest from the spine coronal image data set, and adjust all the acquired spine images to be measured to a uniform graphic size and window width; the spine coronal image data set Containing multiple images of the spine to be measured;

将调整后的所述待测量脊柱图像输入至预设神经网络模型，分别自所述待测量脊柱图像的预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，得到多张候选框图像；每张所述候选框图像中已定位出位于脊柱中的椎骨，且一张所述候选框图像只对应一块所述椎骨；Input the adjusted spine image to be measured into a preset neural network model, and extract the candidate frame images with the highest tilt probability from the preset regions of interest of the spine image to be measured to obtain multiple candidate frame images Vertebrae located in the spine have been located in each candidate frame image, and one candidate frame image corresponds to only one of the vertebrae;

将所述候选框图像输入至预设关键点提取模型中，自所述候选框图像中提取与已定位的所述椎骨关联的多分辨率关键点；一块椎骨关联四个关键点；Input the candidate frame image into a preset key point extraction model, and extract the multi-resolution key points associated with the positioned vertebra from the candidate frame image; one vertebra is associated with four key points;

通过关键点矫正模型对各所述椎骨的所述关键点的相关性进行编码并获取编码结果，根据所述编码结果对所述候选框图像中与所述椎骨关联的所述关键点进行位置矫正，并根据位置矫正结果对所述候选框图像进行修正；Encode the correlation of the key points of each vertebra through the key point correction model and obtain the encoding result, and perform position correction on the key points associated with the vertebra in the candidate frame image according to the encoding result , And correct the candidate frame image according to the position correction result;

根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角。According to the key points in the corrected candidate frame image, the Cobb angle for evaluating the curvature of the spine is determined.
根据权利要求1所述的脊柱弯曲角度测量方法，其中，所述自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位，包括：The method for measuring the angle of spine curvature according to claim 1, wherein the spine image to be measured containing a preset region of interest is acquired from the coronal image data set of the spine, and all the acquired spine images to be measured are adjusted to be unified The graphic size and window width and level of the window, including:

通过预设方向投影定位出所述脊柱冠状位影像数据集中包含预设感兴趣区域的所述待测量脊柱图像；Locating the to-be-measured spine image that includes a preset region of interest in the coronal image data set of the spine through a preset direction projection;

对所述待测量脊柱图像按照预设比例表进行裁剪，通过图像插值方法将裁剪后的所述待测量脊柱图像调整为统一图形尺寸，并通过CT成像技术将统一图形尺寸的所述待测量脊柱图像调整为统一的窗宽窗位。The spine image to be measured is cropped according to a preset scale table, the cropped spine image to be measured is adjusted to a uniform graphic size by an image interpolation method, and the spine to be measured with a uniform graphic size is converted by CT imaging technology The image is adjusted to a uniform window width and level.
根据权利要求1所述的脊柱弯曲角度测量方法，其中，所述预设神经网络模型包括RPN；所述自所述待测量脊柱图像的所述预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，包括：The method for measuring the angle of spine curvature according to claim 1, wherein the preset neural network model includes RPN; and the one with the highest tilt probability is extracted from the preset region of interest of the spine image to be measured Candidate frame images, including:

利用所述RPN中的卷积层从包含所述预设感兴趣区域的所述待测量脊柱图像中获取特征图；Using the convolutional layer in the RPN to obtain a feature map from the to-be-measured spine image containing the preset region of interest;

将所述特征图各点的多维特征与所述RPN中的全连接层连接后，获取所述全连接层输出的预测结果；所述预测结果表征出所述全连接层输出的achor box包含目标的目标概率；After connecting the multi-dimensional features of each point of the feature map with the fully connected layer in the RPN, the prediction result output by the fully connected layer is obtained; the prediction result indicates that the achor box output by the fully connected layer contains the target Target probability;

提取出所述目标概率大于预设概率的所述预测结果，将所述预测结果中包含目标的所述achor box进行平移和缩放后，生成预设数量的候选框图像，并利用非极大值筛选出发生倾斜概率最高的所述候选框图像。Extract the prediction result whose target probability is greater than the preset probability, translate and zoom the achor box containing the target in the prediction result, generate a preset number of candidate frame images, and use non-maximum values The candidate frame image with the highest probability of occurrence of tilt is screened out.
根据权利要求1所述的脊柱弯曲角度测量方法，其中，所述预设关键点提取模型包括CPN；所述自所述候选框图像中提取与已定位的所述脊柱中的所述椎骨关联的多分辨率关键点，包括：The method for measuring the curvature angle of the spine according to claim 1, wherein the preset key point extraction model includes CPN; and the extraction from the candidate frame image is associated with the positioned vertebrae in the spine Key points of multi-resolution, including:

通过所述CPN中的GlobalNe对各所述候选框图像的第一关键点进行检测；所述第一关键点为属于第一分辨率范围的可见的关键点；Detect the first key point of each candidate frame image through GlobalNe in the CPN; the first key point is a visible key point belonging to a first resolution range;

通过所述CPN中的RefineNe对各所述候选框图像的第二关键点进行检测；所述第二关键点为属于第二分辨率范围的不可见的关键点；Detect the second key points of each candidate frame image through RefineNe in the CPN; the second key points are invisible key points belonging to a second resolution range;

提取检测成功的所述第一关键点和所述第二关键点，并将所述第一关键点和所述第二关键点作为所述脊柱中所有椎骨的多分辨率关键点。The first key point and the second key point that are successfully detected are extracted, and the first key point and the second key point are used as multi-resolution key points of all vertebrae in the spine.
根据权利要求1所述的脊柱弯曲角度测量方法，其中，所述Cobb角存储于区块链中，所述根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角，包括：The method for measuring the spine bending angle according to claim 1, wherein the Cobb angle is stored in a blockchain, and the key points in the candidate frame image after the correction are determined to be used for evaluating the spine Cobb angle of bending angle, including:

以每一张所述候选框图像中的所述椎骨关联的四个所述关键点组装成一个包含第一边线和第二边线的预设形状，并将每一张所述候选框图中的所述第一边线的中心点和第二边线的中心点之间的连线，记录为所述候选框图像对应的所述椎骨的计算线段；所述第一边线和第二边线分别由四个所述关键点中两个不同关键点连接而成；Assemble the four key points associated with the vertebrae in each candidate frame image into a preset shape containing a first edge line and a second edge line, and combine the four key points in each candidate frame image The line between the center point of the first sideline and the center point of the second sideline is recorded as the calculated line segment of the vertebra corresponding to the candidate frame image; the first sideline and the second sideline are respectively determined by Two different key points among the four key points are connected;

计算所有所述计算线段中的任意两条所述计算线段之间所形成的夹角，得到一个包含计算的所有所述夹角的矩阵，从所述矩阵筛选出其中两条所述计算线段之间形成的目标夹角，并将所述目标夹角作为Cobb角。Calculate the included angle formed between any two of the calculated line segments to obtain a matrix containing all the calculated included angles, and filter out one of the two calculated line segments from the matrix The target included angle formed between the two, and the target included angle is taken as the Cobb angle.
根据权利要求1-5任一项所述的脊柱弯曲角度测量方法，其中，所述方法还包括：The method for measuring the angle of curvature of the spine according to any one of claims 1 to 5, wherein the method further comprises:

根据所述Cobb角评估出所述脊柱弯曲角度。The angle of curvature of the spine is evaluated based on the Cobb angle.
一种脊柱弯曲角度测量方法装置，其中，包括：A method and device for measuring the bending angle of the spine, which comprises:

调整模块，用于自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位；所述脊柱冠状位影像数据集中包含多张所述待测量脊柱图像；The adjustment module is used to obtain the to-be-measured spine image containing the preset region of interest from the spine coronal image data set, and adjust all the obtained to-be-measured spine images to a uniform graphic size and window width and window level; the spine The coronal image data set contains multiple images of the spine to be measured;

第一提取模块，用于将调整后的所述待测量脊柱图像输入至预设神经网络模型，分别自所述待测量脊柱图像的预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，得到多张候选框图像；每张所述候选框图像中已定位出位于脊柱中的椎骨，且一张所述候选框图像只对应一块所述椎骨；The first extraction module is configured to input the adjusted spine image to be measured into a preset neural network model, and extract the candidate frame image with the highest tilt probability from the preset regions of interest of the spine image to be measured, respectively , Obtain multiple candidate frame images; each of the candidate frame images has located a vertebra in the spine, and one candidate frame image corresponds to only one of the vertebrae;

第二提取模块，用于将所述候选框图像输入至预设关键点提取模型中，自所述候选框图像中提取与已定位的所述椎骨关联的多分辨率关键点；一块椎骨关联四个关键点；The second extraction module is configured to input the candidate frame image into a preset key point extraction model, and extract the multi-resolution key points associated with the positioned vertebra from the candidate frame image; one vertebra is associated with four Key point

修正模块，用于通过关键点矫正模型对各所述椎骨的所述关键点的相关性进行编码并获取编码结果，根据所述编码结果对所述候选框图像中与所述椎骨关联的所述关键点进行位置矫正，并根据位置矫正结果对所述候选框图像进行修正；The correction module is used to encode the correlation of the key points of each of the vertebrae through the key point correction model and obtain the encoding result, and according to the encoding result, perform correction on the candidate frame image associated with the vertebrae. Performing position correction on the key points, and correcting the candidate frame image according to the position correction result;

确定模块，用于根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角。The determining module is configured to determine the Cobb angle for evaluating the curvature of the spine according to the key points in the corrected candidate frame image.
根据权利要求7所述的脊柱弯曲角度测量方法装置，其中，所述调整模块包括：The method and device for measuring the angle of spinal curvature according to claim 7, wherein the adjustment module comprises:

定位子模块，用于通过预设方向投影定位出所述脊柱冠状位影像数据集中包含预设感兴趣区域的所述待测量脊柱图像；A positioning sub-module, configured to locate the to-be-measured spine image in the coronal image data set of the spine that includes a preset region of interest through a preset direction projection;

调整子模块，用于对所述待测量脊柱图像按照预设比例表进行裁剪，通过图像插值方法将裁剪后的所述待测量脊柱图像调整为统一图形尺寸，并通过CT成像技术将统一图形尺寸的所述待测量脊柱图像调整为统一的窗宽窗位。The adjustment sub-module is used to crop the spine image to be measured according to a preset scale table, adjust the cropped spine image to be measured to a uniform graphic size through an image interpolation method, and use CT imaging technology to adjust the uniform graphic size The spine image to be measured is adjusted to a uniform window width and window level.
一种计算机设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，其中，所述处理器执行所述计算机程序时实现以下步骤：A computer device includes a memory, a processor, and a computer program that is stored in the memory and can run on the processor, wherein the processor implements the following steps when the processor executes the computer program:

自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位；所述脊柱冠状位影像数据集中包含多张所述待测量脊柱图像；Acquire the to-be-measured spine images containing the preset region of interest from the spine coronal image data set, and adjust all the acquired spine images to be measured to a uniform graphic size and window width; the spine coronal image data set Containing multiple images of the spine to be measured;

将调整后的所述待测量脊柱图像输入至预设神经网络模型，分别自所述待测量脊柱图像的预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，得到多张候选框图像；每张所述候选框图像中已定位出位于脊柱中的椎骨，且一张所述候选框图像只对应一块所述椎骨；Input the adjusted spine image to be measured into a preset neural network model, and extract the candidate frame images with the highest tilt probability from the preset regions of interest of the spine image to be measured to obtain multiple candidate frame images Vertebrae located in the spine have been located in each candidate frame image, and one candidate frame image corresponds to only one of the vertebrae;

将所述候选框图像输入至预设关键点提取模型中，自所述候选框图像中提取与已定位的所述椎骨关联的多分辨率关键点；一块椎骨关联四个关键点；Input the candidate frame image into a preset key point extraction model, and extract the multi-resolution key points associated with the positioned vertebra from the candidate frame image; one vertebra is associated with four key points;

通过关键点矫正模型对各所述椎骨的所述关键点的相关性进行编码并获取编码结果，根据所述编码结果对所述候选框图像中与所述椎骨关联的所述关键点进行位置矫正，并根据位置矫正结果对所述候选框图像进行修正；Encode the correlation of the key points of each vertebra through the key point correction model and obtain the encoding result, and perform position correction on the key point associated with the vertebra in the candidate frame image according to the encoding result , And correct the candidate frame image according to the position correction result;

根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角。According to the key points in the corrected candidate frame image, the Cobb angle for evaluating the curvature of the spine is determined.
根据权利要求9所述的计算机设备，其中，所述自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位时，具体实现以下步骤：9. The computer device according to claim 9, wherein the spine image to be measured containing a preset region of interest is obtained from the coronal image data set of the spine, and all the acquired spine images to be measured are adjusted to a uniform graphic size In the case of window width and window level, the following steps are specifically implemented:

通过预设方向投影定位出所述脊柱冠状位影像数据集中包含预设感兴趣区域的所述待测量脊柱图像；Locating the to-be-measured spine image that includes a preset region of interest in the coronal image data set of the spine through a preset direction projection;

对所述待测量脊柱图像按照预设比例表进行裁剪，通过图像插值方法将裁剪后的所述待测量脊柱图像调整为统一图形尺寸，并通过CT成像技术将统一图形尺寸的所述待测量脊柱图像调整为统一的窗宽窗位。The spine image to be measured is cropped according to a preset scale table, the cropped spine image to be measured is adjusted to a uniform graphic size by an image interpolation method, and the spine to be measured with a uniform graphic size is converted by CT imaging technology The image is adjusted to a uniform window width and level.
根据权利要求9所述的计算机设备，其中，所述预设神经网络模型包括RPN；所述自所述待测量脊柱图像的所述预设感兴趣区域中提取出发生倾斜概率最高的候选框图像时，具体实现以下步骤：9. The computer device according to claim 9, wherein the preset neural network model comprises RPN; and the candidate frame image with the highest tilt probability is extracted from the preset region of interest of the spine image to be measured When, specifically implement the following steps:

利用所述RPN中的卷积层从包含所述预设感兴趣区域的所述待测量脊柱图像中获取特征图；Using the convolutional layer in the RPN to obtain a feature map from the to-be-measured spine image containing the preset region of interest;

将所述特征图各点的多维特征与所述RPN中的全连接层连接后，获取所述全连接层输出的预测结果；所述预测结果表征出所述全连接层输出的achor box包含目标的目标概率；After connecting the multi-dimensional features of each point of the feature map with the fully connected layer in the RPN, the prediction result output by the fully connected layer is obtained; the prediction result indicates that the achor box output by the fully connected layer contains the target Target probability;

提取出所述目标概率大于预设概率的所述预测结果，将所述预测结果中包含目标的所述achor box进行平移和缩放后，生成预设数量的候选框图像，并利用非极大值筛选出发生倾斜概率最高的所述候选框图像。Extract the prediction result whose target probability is greater than the preset probability, translate and zoom the achor box containing the target in the prediction result, generate a preset number of candidate frame images, and use non-maximum values The candidate frame image with the highest probability of occurrence of tilt is screened out.
根据权利要求9所述的计算机设备，其中，所述预设关键点提取模型包括CPN；所述自所述候选框图像中提取与已定位的所述脊柱中的所述椎骨关联的多分辨率关键点时，具体实现以下步骤：8. The computer device according to claim 9, wherein the preset key point extraction model comprises CPN; and the extraction of the multi-resolution associated with the vertebrae in the spine that has been positioned from the candidate frame image At the key point, the following steps are specifically implemented:

通过所述CPN中的GlobalNe对各所述候选框图像的第一关键点进行检测；所述第一关键点为属于第一分辨率范围的可见的关键点；Detect the first key point of each candidate frame image through GlobalNe in the CPN; the first key point is a visible key point belonging to a first resolution range;

通过所述CPN中的RefineNe对各所述候选框图像的第二关键点进行检测；所述第二关键点为属于第二分辨率范围的不可见的关键点；Detect the second key points of each candidate frame image through RefineNe in the CPN; the second key points are invisible key points belonging to a second resolution range;

提取检测成功的所述第一关键点和所述第二关键点，并将所述第一关键点和所述第二关键点作为所述脊柱中所有椎骨的多分辨率关键点。The first key point and the second key point that are successfully detected are extracted, and the first key point and the second key point are used as multi-resolution key points of all vertebrae in the spine.
根据权利要求9所述的计算机设备，其中，所述Cobb角存储于区块链中，所述根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角时，具体实现以下步骤：9. The computer device according to claim 9, wherein the Cobb angle is stored in a blockchain, and the key points in the candidate frame image after correction are used to determine the angle of curvature of the spine. In the case of Cobb angle, the specific steps are as follows:

以每一张所述候选框图像中的所述椎骨关联的四个所述关键点组装成一个包含第一边线和第二边线的预设形状，并将每一张所述候选框图中的所述第一边线的中心点和第二边线的中心点之间的连线，记录为所述候选框图像对应的所述椎骨的计算线段；所述第一边线和第二边线分别由四个所述关键点中两个不同关键点连接而成；Assemble the four key points associated with the vertebrae in each candidate frame image into a preset shape containing a first edge line and a second edge line, and combine the four key points in each candidate frame image The line between the center point of the first sideline and the center point of the second sideline is recorded as the calculated line segment of the vertebra corresponding to the candidate frame image; the first sideline and the second sideline are respectively determined by Two different key points among the four key points are connected;

计算所有所述计算线段中的任意两条所述计算线段之间所形成的夹角，得到一个包含计算的所有所述夹角的矩阵，从所述矩阵筛选出其中两条所述计算线段之间形成的目标夹角，并将所述目标夹角作为Cobb角。Calculate the included angle formed between any two of the calculated line segments to obtain a matrix containing all the calculated included angles, and filter out one of the two calculated line segments from the matrix The target included angle formed between the two, and the target included angle is taken as the Cobb angle.
根据权利要求9-13任一项所述的计算机设备，其中，所述处理器执行所述计算机程序时还实现以下步骤：The computer device according to any one of claims 9-13, wherein the processor further implements the following steps when executing the computer program:

根据所述Cobb角评估出所述脊柱弯曲角度。The angle of curvature of the spine is evaluated based on the Cobb angle.
一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，其中，所述计算机程序被处理器执行时实现以下步骤：A computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the following steps:

自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位；所述脊柱冠状位影像数据集中包含多张所述待测量脊柱图像；Acquire the to-be-measured spine images containing the preset region of interest from the spine coronal image data set, and adjust all the acquired spine images to be measured to a uniform graphic size and window width; the spine coronal image data set Containing multiple images of the spine to be measured;

将调整后的所述待测量脊柱图像输入至预设神经网络模型，分别自所述待测量脊柱图像的预设感兴趣区域中提取出发生倾斜概率最高的候选框图像，得到多张候选框图像；每张所述候选框图像中已定位出位于脊柱中的椎骨，且一张所述候选框图像只对应一块所述椎骨；Input the adjusted spine image to be measured into a preset neural network model, and extract the candidate frame images with the highest tilt probability from the preset regions of interest of the spine image to be measured to obtain multiple candidate frame images Vertebrae located in the spine have been located in each candidate frame image, and one candidate frame image corresponds to only one of the vertebrae;

将所述候选框图像输入至预设关键点提取模型中，自所述候选框图像中提取与已定位的所述椎骨关联的多分辨率关键点；一块椎骨关联四个关键点；Input the candidate frame image into a preset key point extraction model, and extract the multi-resolution key points associated with the positioned vertebra from the candidate frame image; one vertebra is associated with four key points;

通过关键点矫正模型对各所述椎骨的所述关键点的相关性进行编码并获取编码结果，根据所述编码结果对所述候选框图像中与所述椎骨关联的所述关键点进行位置矫正，并根据位置矫正结果对所述候选框图像进行修正；Encode the correlation of the key points of each vertebra through the key point correction model and obtain the encoding result, and perform position correction on the key points associated with the vertebra in the candidate frame image according to the encoding result , And correct the candidate frame image according to the position correction result;

根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角。According to the key points in the corrected candidate frame image, the Cobb angle for evaluating the curvature of the spine is determined.
根据权利要求15所述的计算机可读存储介质，其中，所述自脊柱冠状位影像数据集中获取包含预设感兴趣区域的待测量脊柱图像，将获取的所有所述待测量脊柱图像调整为统一的图形尺寸和窗宽窗位时，具体实现以下步骤：The computer-readable storage medium according to claim 15, wherein the spine image to be measured including a preset region of interest is acquired from the coronal image data set of the spine, and all the acquired spine images to be measured are adjusted to be unified When setting the graphic size and window width and level, the following steps are specifically implemented:

通过预设方向投影定位出所述脊柱冠状位影像数据集中包含预设感兴趣区域的所述待测量脊柱图像；Locating the to-be-measured spine image that includes a preset region of interest in the coronal image data set of the spine through a preset direction projection;

对所述待测量脊柱图像按照预设比例表进行裁剪，通过图像插值方法将裁剪后的所述待测量脊柱图像调整为统一图形尺寸，并通过CT成像技术将统一图形尺寸的所述待测量脊柱图像调整为统一的窗宽窗位。The spine image to be measured is cropped according to a preset scale table, the cropped spine image to be measured is adjusted to a uniform graphic size by an image interpolation method, and the spine to be measured with a uniform graphic size is converted by CT imaging technology The image is adjusted to a uniform window width and level.
根据权利要求15所述的计算机可读存储介质，其中，所述预设神经网络模型包括RPN；所述自所述待测量脊柱图像的所述预设感兴趣区域中提取出发生倾斜概率最高的候选框图像时，具体实现以下步骤：The computer-readable storage medium according to claim 15, wherein the preset neural network model comprises RPN; and the one with the highest tilt probability is extracted from the preset interest area of the spine image to be measured When the candidate frame image is selected, the following steps are specifically implemented:

利用所述RPN中的卷积层从包含所述预设感兴趣区域的所述待测量脊柱图像中获取特征图；Using the convolutional layer in the RPN to obtain a feature map from the to-be-measured spine image containing the preset region of interest;

将所述特征图各点的多维特征与所述RPN中的全连接层连接后，获取所述全连接层输出的预测结果；所述预测结果表征出所述全连接层输出的achor box包含目标的目标概率；After connecting the multi-dimensional features of each point of the feature map with the fully connected layer in the RPN, the prediction result output by the fully connected layer is obtained; the prediction result indicates that the achor box output by the fully connected layer contains the target Target probability;

提取出所述目标概率大于预设概率的所述预测结果，将所述预测结果中包含目标的所述achor box进行平移和缩放后，生成预设数量的候选框图像，并利用非极大值筛选出发生倾斜概率最高的所述候选框图像。Extract the prediction result whose target probability is greater than the preset probability, translate and zoom the achor box containing the target in the prediction result, generate a preset number of candidate frame images, and use non-maximum values The candidate frame image with the highest probability of occurrence of tilt is screened out.
根据权利要求15所述的计算机可读存储介质，其中，所述预设关键点提取模型包括CPN；所述自所述候选框图像中提取与已定位的所述脊柱中的所述椎骨关联的多分辨率关键点时，具体实现以下步骤：The computer-readable storage medium according to claim 15, wherein the preset key point extraction model comprises CPN; and the extraction from the candidate frame image is associated with the located vertebrae in the spine For multi-resolution key points, the following steps are specifically implemented:

通过所述CPN中的GlobalNe对各所述候选框图像的第一关键点进行检测；所述第一关键点为属于第一分辨率范围的可见的关键点；Detect the first key point of each candidate frame image through GlobalNe in the CPN; the first key point is a visible key point belonging to a first resolution range;

通过所述CPN中的RefineNe对各所述候选框图像的第二关键点进行检测；所述第二关键点为属于第二分辨率范围的不可见的关键点；Detect the second key points of each candidate frame image through RefineNe in the CPN; the second key points are invisible key points belonging to a second resolution range;

提取检测成功的所述第一关键点和所述第二关键点，并将所述第一关键点和所述第二关键点作为所述脊柱中所有椎骨的多分辨率关键点。The first key point and the second key point that are successfully detected are extracted, and the first key point and the second key point are used as multi-resolution key points of all vertebrae in the spine.
根据权利要求15所述的计算机可读存储介质，其中，所述Cobb角存储于区块链中，所述根据修正后的所述候选框图像中的所述关键点，确定出用于评估脊柱弯曲角度的Cobb角时，具体实现以下步骤：The computer-readable storage medium according to claim 15, wherein the Cobb angle is stored in a blockchain, and the key point in the candidate frame image after the correction is determined to be used for evaluating the spine When bending the Cobb angle of the angle, the following steps are specifically implemented:

以每一张所述候选框图像中的所述椎骨关联的四个所述关键点组装成一个包含第一边线和第二边线的预设形状，并将每一张所述候选框图中的所述第一边线的中心点和第二边线的中心点之间的连线，记录为所述候选框图像对应的所述椎骨的计算线段；所述第一边线和第二边线分别由四个所述关键点中两个不同关键点连接而成；Assemble the four key points associated with the vertebrae in each candidate frame image into a preset shape containing a first edge line and a second edge line, and combine the four key points in each candidate frame image The line between the center point of the first sideline and the center point of the second sideline is recorded as the calculated line segment of the vertebra corresponding to the candidate frame image; the first sideline and the second sideline are respectively determined by Two different key points among the four key points are connected;

计算所有所述计算线段中的任意两条所述计算线段之间所形成的夹角，得到一个包含计算的所有所述夹角的矩阵，从所述矩阵筛选出其中两条所述计算线段之间形成的目标夹角，并将所述目标夹角作为Cobb角。Calculate the included angle formed between any two of the calculated line segments to obtain a matrix containing all the calculated included angles, and filter out one of the two calculated line segments from the matrix The target included angle formed between the two, and the target included angle is taken as the Cobb angle.
根据权利要求15-19任一项所述的计算机可读存储介质，其中，所述计算机程序被处理器执行时还实现以下步骤：The computer-readable storage medium according to any one of claims 15-19, wherein the following steps are further implemented when the computer program is executed by the processor:

根据所述Cobb角评估出所述脊柱弯曲角度。The angle of curvature of the spine is evaluated based on the Cobb angle.