CN110246216B - Spine model generation method, spine model generation system and terminal - Google Patents
Spine model generation method, spine model generation system and terminalDownload PDFInfo
- Publication number
- CN110246216B CN110246216BCN201910450859.5ACN201910450859ACN110246216BCN 110246216 BCN110246216 BCN 110246216BCN 201910450859 ACN201910450859 ACN 201910450859ACN 110246216 BCN110246216 BCN 110246216B
- Authority
- CN
- China
- Prior art keywords
- image
- spine
- model
- lesion
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30008—Bone
- G06T2207/30012—Spine; Backbone
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
Description
Translated fromChinese技术领域technical field
本申请属于医学工程技术领域，尤其涉及一种脊柱模型生成方法、脊柱模型生成系统及终端。The application belongs to the technical field of medical engineering, and in particular relates to a spine model generation method, a spine model generation system and a terminal.
背景技术Background technique
全脊柱切除术(TES)是指完整切除一个或多个脊柱节段，包括椎体、椎间盘、椎弓根、椎板、棘突、韧带以及其他脊柱附属软组织一并完全切除的手术。该手术是目前国际上公认的治疗脊柱肿瘤疾病复发率最低的手术方式。Total spondylectomy (TES) refers to the complete resection of one or more spinal segments, including vertebral bodies, intervertebral discs, pedicles, laminae, spinous processes, ligaments, and other soft tissues attached to the spine. This operation is currently internationally recognized as the operation method with the lowest recurrence rate in the treatment of spinal tumors.
但随着手术的完成，人体的脊柱受手术影响连续性出现中断。为实现脊柱稳定性的重建，当前主要采用的方法是通过3D打印技术制作人工椎体，并将制作出的假体植入患者体内，以替代原有脊柱部位实现人体脊椎功能。因此假体的设计和制作对于患者而言是极其重要的。But with the completion of the operation, the continuity of the human spine affected by the operation is interrupted. In order to achieve the reconstruction of spinal stability, the main method currently used is to manufacture artificial vertebral bodies through 3D printing technology, and implant the prosthesis into the patient's body to replace the original spinal column to realize the function of the human spine. Therefore, the design and manufacture of prostheses is extremely important for patients.
而在现有技术中，往往需要制作数十个3D打印所需的植入物的术前模型，该些术前模型的建立与确定需要极富经验的临床医生把关才能实现。且受限于现有技术的影响，制作出的假体与理想假体存在较大的误差，造成患者在植入假体后运动功能受限，术后脊椎在轴向旋转的稳定性不够，与上下椎体的结合强度低等问题出现。However, in the existing technology, it is often necessary to make dozens of preoperative models of the implants required for 3D printing. The establishment and determination of these preoperative models requires the checks of highly experienced clinicians to realize. And limited by the influence of the existing technology, there is a large error between the manufactured prosthesis and the ideal prosthesis, resulting in limited movement function of the patient after implantation of the prosthesis, and insufficient stability of the axial rotation of the spine after surgery. Problems such as low bonding strength with the upper and lower vertebral bodies appear.
发明内容Contents of the invention
有鉴于此，本申请实施例提供了一种脊柱模型生成方法、脊柱模型生成系统及终端，以解决现有技术中需要制作数十个3D打印所需的植入物的术前模型，且制作出的脊椎假体与理想假体存在较大的误差的问题。In view of this, the embodiment of the present application provides a spine model generation method, a spine model generation system and a terminal to solve the need to make dozens of preoperative models of implants required for 3D printing in the prior art, and to make There is a large error between the proposed spinal prosthesis and the ideal prosthesis.
本申请实施例的第一方面提供了一种脊柱模型生成方法，包括：The first aspect of the embodiment of the present application provides a method for generating a spine model, including:
根据非病变脊柱图像，训练生成对抗网络模型；Training a generative adversarial network model based on non-lesional spine images;
通过训练后的所述生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，获得目标脊柱图像，其中，所述目标脊柱图像中所述病变发生部位被修复为非病变脊柱形态；Using the trained GAN model, perform image repair on the lesion occurrence site in the lesion spine image to obtain a target spine image, wherein the lesion occurrence site in the target spine image is repaired to a non-lesion spine shape;
对所述目标脊柱图像进行三维重建，获取对应的目标脊柱模型。Three-dimensional reconstruction is performed on the image of the target spine to obtain a corresponding target spine model.
本申请实施例的第二方面提供了一种脊柱模型生成系统，包括：The second aspect of the embodiment of the present application provides a spine model generation system, including:
模型训练模块，用于根据非病变脊柱图像，训练生成对抗网络模型；The model training module is used to train and generate an adversarial network model according to the non-pathological spine image;
图像修复模块，用于通过训练后的所述生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，获得目标脊柱图像，其中，所述目标图像中所述病变发生部位被修复为非病变脊柱形态；The image repair module is used to perform image repair on the lesion occurrence part in the lesion spine image through the trained generation confrontation network model, and obtain the target spine image, wherein the lesion occurrence part in the target image is repaired as a non- Diseased spine morphology;
模型建立模块，用于对所述目标脊柱图像进行三维重建，获取对应的目标脊柱模型。The model building module is used to perform three-dimensional reconstruction on the image of the target spine, and obtain a corresponding target spine model.
本申请实施例的第三方面提供了一种终端，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如第一方面所述方法的步骤。The third aspect of the embodiments of the present application provides a terminal, including a memory, a processor, and a computer program stored in the memory and operable on the processor. When the processor executes the computer program, the The steps of the method as described in the first aspect.
本申请实施例的第四方面提供了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如第一方面所述方法的步骤。A fourth aspect of the embodiments of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.
本申请的第五方面提供了一种计算机程序产品，所述计算机程序产品包括计算机程序，所述计算机程序被一个或多个处理器执行时实现如上述第一方面所述方法的步骤。A fifth aspect of the present application provides a computer program product, where the computer program product includes a computer program, and when the computer program is executed by one or more processors, the steps of the method described in the first aspect above are realized.
由上可见，本申请实施例中，通过基于非病变脊柱图像，训练生成对抗网络模型，利用训练后的生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，将病变发生部位修复为非病变脊柱形态，得到目标脊柱图像，对目标脊柱图像进行三维重建，获取最终的目标脊柱模型，该过程通过利用生成对抗网络，对病变脊柱图像进行复原，以尽可能得到一个理想状态下的椎体模型，实现一个尽可能正常、完整的椎体的重建，避免目前需要制作数十个植入物的三维术前模型的问题，提升脊柱假体制作的精准度，减少假体植入患者体内后的活动限制。It can be seen from the above that in the embodiment of the present application, based on the non-pathological spine image, the GAN model is trained, and the trained GAN model is used to perform image repair on the lesion occurrence site in the lesion spine image, and the lesion occurrence site is repaired as The shape of the non-pathological spine is obtained, and the target spine image is obtained, and the target spine image is 3D reconstructed to obtain the final target spine model. This process uses the generative confrontation network to restore the lesioned spine image to obtain an ideal spine as much as possible. body model to achieve a normal and complete reconstruction of the vertebral body as much as possible, avoid the current problem of making dozens of three-dimensional preoperative models of implants, improve the accuracy of spinal prosthesis production, and reduce the number of implants implanted in patients subsequent activity restrictions.
附图说明Description of drawings
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。In order to more clearly illustrate the technical solutions in the embodiments of the present application, the accompanying drawings that need to be used in the descriptions of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings in the following description are only for the present application For some embodiments, those of ordinary skill in the art can also obtain other drawings based on these drawings without paying creative efforts.
图1是本申请实施例提供的一种脊柱模型生成方法的流程图一；Fig. 1 is a flowchart one of a method for generating a spine model provided by an embodiment of the present application;
图2是本申请实施例提供的一种脊柱模型生成方法的流程图二；Fig. 2 is a flow chart 2 of a method for generating a spine model provided by an embodiment of the present application;
图3是本申请实施例中的脊柱掩码图像；Fig. 3 is the spine mask image in the embodiment of the present application;
图4是本申请实施例中的修复后的目标脊柱图像；Fig. 4 is the target spine image after repairing in the embodiment of the present application;
图5是本申请实施例中的基于V-Net神经网络进行语义分割得到的脊柱区域的二值图像；Fig. 5 is the binary image of the spinal column area that carries out semantic segmentation based on V-Net neural network in the embodiment of the present application;
图6是本申请实施例提供的一种脊柱模型生成系统的结构图；FIG. 6 is a structural diagram of a spine model generation system provided by an embodiment of the present application;
图7是本申请实施例提供的一种终端的结构图。FIG. 7 is a structural diagram of a terminal provided by an embodiment of the present application.
具体实施方式Detailed ways
以下描述中，为了说明而不是为了限定，提出了诸如特定系统结构、技术之类的具体细节，以便透彻理解本申请实施例。然而，本领域的技术人员应当清楚，在没有这些具体细节的其它实施例中也可以实现本申请。在其它情况中，省略对众所周知的系统、装置、电路以及方法的详细说明，以免不必要的细节妨碍本申请的描述。In the following description, specific details such as specific system structures and technologies are presented for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.
应当理解，当在本说明书和所附权利要求书中使用时，术语“包括”指示所描述特征、整体、步骤、操作、元素和/或组件的存在，但并不排除一个或多个其它特征、整体、步骤、操作、元素、组件和/或其集合的存在或添加。It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and/or components, but does not exclude one or more other features. , whole, step, operation, element, component and/or the presence or addition of a collection thereof.
还应当理解，在此本申请说明书中所使用的术语仅仅是出于描述特定实施例的目的而并不意在限制本申请。如在本申请说明书和所附权利要求书中所使用的那样，除非上下文清楚地指明其它情况，否则单数形式的“一”、“一个”及“该”意在包括复数形式。It should also be understood that the terminology used in the specification of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural referents unless the context clearly dictates otherwise.
还应当进一步理解，在本申请说明书和所附权利要求书中使用的术语“和/或”是指相关联列出的项中的一个或多个的任何组合以及所有可能组合，并且包括这些组合。It should also be further understood that the term "and/or" used in the description of the present application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations .
如在本说明书和所附权利要求书中所使用的那样，术语“如果”可以依据上下文被解释为“当...时”或“一旦”或“响应于确定”或“响应于检测到”。类似地，短语“如果确定”或“如果检测到[所描述条件或事件]”可以依据上下文被解释为意指“一旦确定”或“响应于确定”或“一旦检测到[所描述条件或事件]”或“响应于检测到[所描述条件或事件]”。As used in this specification and the appended claims, the term "if" may be construed as "when" or "once" or "in response to determining" or "in response to detecting" depending on the context . Similarly, the phrase "if determined" or "if [the described condition or event] is detected" may be construed, depending on the context, to mean "once determined" or "in response to the determination" or "once detected [the described condition or event] ]” or “in response to detection of [described condition or event]”.
具体实现中，本申请实施例中描述的终端包括但不限于诸如具有触摸敏感表面(例如，触摸屏显示器和/或触摸板)的移动电话、膝上型计算机或平板计算机之类的其它便携式设备。还应当理解的是，在某些实施例中，所述设备并非便携式通信设备，而是具有触摸敏感表面(例如，触摸屏显示器和/或触摸板)的台式计算机。In a specific implementation, the terminals described in the embodiments of the present application include but are not limited to other portable devices such as mobile phones, laptop computers or tablet computers with touch-sensitive surfaces (eg, touch screen displays and/or touch pads). It should also be appreciated that in some embodiments, the device is not a portable communication device, but a desktop computer with a touch-sensitive surface (eg, a touchscreen display and/or a touchpad).
在接下来的讨论中，描述了包括显示器和触摸敏感表面的终端。然而，应当理解的是，终端可以包括诸如物理键盘、鼠标和/或控制杆的一个或多个其它物理用户接口设备。In the ensuing discussion, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that a terminal may include one or more other physical user interface devices such as a physical keyboard, mouse and/or joystick.
终端支持各种应用程序，例如以下中的一个或多个：绘图应用程序、演示应用程序、文字处理应用程序、网站创建应用程序、盘刻录应用程序、电子表格应用程序、游戏应用程序、电话应用程序、视频会议应用程序、电子邮件应用程序、即时消息收发应用程序、锻炼支持应用程序、照片管理应用程序、数码相机应用程序、数字摄影机应用程序、web浏览应用程序、数字音乐播放器应用程序和/或数字视频播放器应用程序。The terminal supports various applications such as one or more of the following: drawing application, presentation application, word processing application, website creation application, disk burning application, spreadsheet application, gaming application, telephony application programs, video conferencing applications, email applications, instant messaging applications, exercise support applications, photo management applications, digital camera applications, digital video camera applications, web browsing applications, digital music player applications, and and/or digital video player applications.
可以在终端上执行的各种应用程序可以使用诸如触摸敏感表面的至少一个公共物理用户接口设备。可以在应用程序之间和/或相应应用程序内调整和/或改变触摸敏感表面的一个或多个功能以及终端上显示的相应信息。这样，终端的公共物理架构(例如，触摸敏感表面)可以支持具有对用户而言直观且透明的用户界面的各种应用程序。Various applications that can be executed on the terminal can use at least one common physical user interface device, such as a touch-sensitive surface. One or more functions of the touch-sensitive surface and corresponding information displayed on the terminal may be adjusted and/or changed between applications and/or within the respective applications. In this way, the common physical architecture (eg, touch-sensitive surface) of the terminal can support various applications with a user interface that is intuitive and transparent to the user.
应理解，本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 sequence numbers of the steps in this embodiment do not mean the order of execution, and the execution order 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.
为了说明本申请所述的技术方案，下面通过具体实施例来进行说明。In order to illustrate the technical solutions described in this application, specific examples are used below to illustrate.
参见图1，图1是本申请实施例提供的一种脊柱模型生成方法的流程图一。如图1所示，一种脊柱模型生成方法，该方法包括以下步骤：Referring to FIG. 1 , FIG. 1 is a flowchart 1 of a method for generating a spine model provided by an embodiment of the present application. As shown in Figure 1, a spine model generation method, the method includes the following steps:
步骤101，根据非病变脊柱图像，训练生成对抗网络模型。
该非病变脊柱图像为基于未发生病变的脊柱而得到的图像。该非病变脊柱图像具体为一脊柱区域的二值图像，或者可以基于该非病变脊柱图像，辅助实现生成对抗网络模型训练的其他类型图像，例如为灰度图像。The image of the non-lesioned spine is an image obtained based on an undisturbed spine. The non-disordered spine image is specifically a binary image of a spinal region, or other types of images, such as grayscale images, can be assisted in generating an adversarial network model training based on the non-disordered spine image.
具体地，该生成对抗网络模型包含两个网络模型：生成网络模型和判别网络模型。其中，生成网络是一个生成式的网络，它接受一个随机的噪声，通过这个噪声生成图像，输出图像G；判别网络用来判断一张图片是不是真实的，它的输入x是一张图片，输出D(x)表示x为真实图片的概率，如果为1，就表示x是一张真实的图片，而输出为0，就表示不可能是真实的图片。模型训练的过程中，生成网络的目标就是尽量生成真实的图片去欺骗判别网络，而判别网络的目标就是尽量判别出生成网络生成的图片和真实的图像，这样G和D构成了一个动态的“博弈过程”，最终的平衡点即纳什平衡点。Specifically, the generative confrontation network model includes two network models: a generative network model and a discriminative network model. Among them, the generation network is a generative network, which accepts a random noise, generates an image through this noise, and outputs an image G; the discriminant network is used to judge whether a picture is real, and its input x is a picture, The output D(x) indicates the probability that x is a real picture. If it is 1, it means that x is a real picture, and if the output is 0, it means that it cannot be a real picture. In the process of model training, the goal of the generative network is to generate real pictures as much as possible to deceive the discriminant network, and the goal of the discriminant network is to distinguish the pictures generated by the generative network from real images as much as possible, so G and D constitute a dynamic " Game process", the final equilibrium point is the Nash equilibrium point.
该生成对抗网络模型应用到深度学习神经网络上来说，就是通过生成网络和判别网络的不断博弈，进而使得生成网络学习到数据的分布，如果用到图片生成上，则训练完成后，生成网络可以从一段随机数中生成逼真的图像。The application of the generative confrontation network model to the deep learning neural network is through the continuous game between the generative network and the discriminant network, so that the generative network can learn the distribution of data. If it is used for image generation, after the training is completed, the generative network can Generate realistic images from a segment of random numbers.
步骤102，通过训练后的所述生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，获得目标脊柱图像。
其中，所述目标图像中所述病变发生部位被修复为非病变脊柱形态。Wherein, the lesion occurrence site in the target image is restored to a non-lesion spine shape.
具体地，通常情况下，病变脊柱图像中，病变发生部位会受肿瘤等病变细胞影响使得脊柱图像中的原本正常脊柱区域信息丢失，例如脊柱轮廓不完整、脊柱区域异常变形等。该些图像区域需要进行图像修复，以得到正常脊柱状态下的目标脊柱图像。Specifically, usually, in the image of the lesioned spine, the location of the lesion is affected by diseased cells such as tumors, so that the information of the original normal spine area in the image of the spine is lost, such as incomplete outline of the spine, abnormal deformation of the spine area, and the like. These image regions need to be inpainted to obtain the target spine image in a normal spine state.
本步骤中，通过经非病变脊柱图像训练后的生成对抗网络模型，将病变脊柱图像进行修复，以获取病变发生部位被修复为非病变脊柱形态的目标脊柱图像，以能够对病变脊柱图像进行复原，将复原后的图像应用于全脊柱切除术后对应脊柱部位的三维模型建立过程中，以实现理想脊柱假体的生产制作，减少对人为因素的依赖，提升脊柱假体制作的精准度，减少假体植入患者体内后的不适症状。In this step, the lesioned spine image is repaired by using the generative adversarial network model trained by the non-lesioned spine image, so as to obtain the target spine image where the lesion has been repaired into a non-lesioned spine shape, so as to be able to restore the lesioned spine image , the restored image is applied to the establishment of the 3D model of the corresponding spinal part after total spondylectomy, in order to realize the production of ideal spinal prosthesis, reduce the dependence on human factors, improve the accuracy of spinal prosthesis production, reduce the Symptoms of discomfort after implantation of the prosthesis in the patient.
具体实现时，图像修复的实现过程可以理解为：在一幅图像上挖一个洞，如何利用其他的信息将这个洞补全，并且让人眼无法辨别出补全的部分。这个问题对于人类很容易，但是对于计算机却显得格外困难，首先这个问题有没有唯一确定的解，其次如何利用其他的信息，还有如何判断补全的结果是否足够真实。In specific implementation, the implementation process of image restoration can be understood as: dig a hole in an image, how to use other information to complete the hole, and make it impossible for the human eye to distinguish the completed part. This problem is easy for humans, but it is extremely difficult for computers. First, is there a unique solution to this problem? Second, how to use other information, and how to judge whether the completion result is true enough.
在本申请实施例中，作为一可选的实施方式，其中，通过训练后的所述生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，获得目标脊柱图像，包括：In the embodiment of the present application, as an optional implementation manner, wherein, through the trained generated confrontation network model, image repair is performed on the lesion occurrence site in the lesioned spine image to obtain the target spine image, including:
基于所述病变脊柱图像及预设的二值掩码矩阵，生成在所述病变发生部位添加掩码的脊柱掩码图像。Based on the lesion spine image and a preset binary mask matrix, a spine mask image with a mask added to the lesion occurrence site is generated.
基于所述脊柱掩码图像及脊柱修复参照图像，通过训练后的所述生成对抗网络模型，生成修复后的所述目标脊柱图像F′＝M⊙F+(1-M)⊙G(z)*；Based on the spine mask image and the spine repair reference image, the trained target spine image F'=M⊙F+(1-M)⊙G(z)* is generated by the trained GAN model. ;
其中，F为所述病变脊柱图像；M为所述预设的二值掩码矩阵，所述二值掩码矩阵中的元素与所述病变脊柱图像的像素相对应；M⊙F为所述脊柱掩码图像；G(z)*为所述脊柱修复参照图像，所述脊柱修复参照图像为所述生成对抗网络模型在模型训练过程中生成的符合设定参照条件的图像；⊙为矩阵元素与图像中像素点值相乘的运算。Wherein, F is the lesion spine image; M is the preset binary mask matrix, and the elements in the binary mask matrix correspond to the pixels of the lesion spine image; M⊙F is the Spine mask image; G(z)* is the reference image for the repair of the spine, and the reference image for the repair of the spine is an image that meets the set reference conditions generated by the generation confrontation network model during the model training process; ⊙ is a matrix element The operation of multiplying the pixel value in the image.
该过程，首先，定义了几个符号用于图像修复。M表示一个二值的掩码矩阵，其中的数值只有0或者1；0表示一个图片中需要修复的区域，1表示同样的图片中要保留的部分。The process, first, defines several symbols for image inpainting. M represents a binary mask matrix, the value of which is only 0 or 1; 0 represents the area to be repaired in a picture, and 1 represents the part to be kept in the same picture.
然后，我们定义如何通过给定的掩码修复病变脊柱图像F，具体做法就是让M矩阵中的元素和F中的对应的像素点值相乘，这种矩阵与对应像素点值相乘的方法叫做哈大马乘积，用M⊙F表示，进而在病变脊柱图像基础上，得到在其中的病变发生部位添加了掩码的脊柱掩码图像，如图3所示。接下来，我们从生成对抗网络模型在模型训练过程中，生成器生成的结果图像里找到一张符合设定参照条件的图片G(z)*，作为在脊柱掩码图像基础上进行图像修复时的脊柱修复参照图像。Then, we define how to repair the lesioned spine image F through a given mask. The specific method is to multiply the elements in the M matrix with the corresponding pixel values in F. The method of multiplying the matrix with the corresponding pixel values It is called the Hadama product, represented by M⊙F, and then based on the lesioned spine image, a spine mask image with a mask added to the lesion occurred is obtained, as shown in Figure 3. Next, we find a picture G(z)* that meets the set reference conditions from the result image generated by the generator during the model training process of the generative confrontation network model, as the image inpainting on the basis of the spine mask image Reference image for spine repair.
我们就可以定义修复后的目标脊柱图像F′＝M⊙F+(1-M)⊙G(z)*。We can define the repaired target spine image F'=M⊙F+(1-M)⊙G(z)* .
由于之前过程中，我们依据获得训练好的生成对抗网络模型(具体为WGAN-GP网络模型)，将添加掩码后的脊柱掩码图像M⊙F作为其输入，得到修复后的结果即目标脊柱图像F′，如图4所示。Because in the previous process, based on the trained generative confrontation network model (specifically the WGAN-GP network model), we took the spine mask image M⊙F after adding the mask as its input, and obtained the repaired result, that is, the target spine Image F', as shown in Figure 4.
其中，该符合设定参照条件的图片G(z)*为前述生成对抗网络模型在模型训练过程中生成的图像中与图像F间对应像素点在值的分布上的尽可能的相似的一个最优图，即相似度超出阈值的一个最优图。Among them, the picture G(z)* that meets the set reference conditions is the most similar as possible in the value distribution of the corresponding pixel points between the image F in the image generated by the aforementioned generative confrontation network model during the model training process. An optimal graph is an optimal graph whose similarity exceeds the threshold.
图像G(z)*中包含了与病变脊柱图像中病变发生部位对应的脊柱区域，及除该脊柱区域之外的其他区域；(1-M)⊙G(z)*则求出了将图像G(z)*中其他区域添加掩码，保留与病变脊柱图像中病变发生部位对应的脊柱区域部分的脊柱修复参照掩码图像。将M⊙F代表的脊柱掩码图像与(1-M)⊙G(z)*代表的脊柱修复参照掩码图像像素叠加，得到该修复后的目标脊柱图像F′。The image G(z)* contains the spine region corresponding to the lesion in the lesion spine image, and other regions except the spine region; (1-M)⊙G(z)* calculates the image Masks are added to other regions in G(z)* , and the spine restoration reference mask image of the spine region corresponding to the location of the lesion in the lesioned spine image is reserved. The spine mask image represented by M⊙F is superimposed with the spine restoration reference mask image pixels represented by (1-M)⊙G(z)* to obtain the repaired target spine image F'.
进一步地，为实现对最优图G(z)*的确定，作为一可选的实施方式，所述基于所述脊柱掩码图像及脊柱修复参照图像，通过训练后的所述生成对抗网络模型，生成修复后的所述目标脊柱图像之前，还包括：Further, in order to realize the determination of the optimal graph G(z)* , as an optional implementation, the generated adversarial network model after training is based on the spine mask image and the spine repair reference image. , before generating the inpainted target spine image, also include:
获取所述生成对抗网络模型在模型训练过程中生成的输出图像；Obtain the output image generated by the generated confrontation network model during the model training process;
根据所述输出图像，依据损失函数L＝Lcontextual+λLperceptual，从所述输出图像中获取所述脊柱修复参照图像；According to the output image, according to the loss function L=Lcontextual + λLperceptual , obtain the reference image for spine restoration from the output image;
其中，λ为超参数；Among them, λ is a hyperparameter;
Lcontextual＝||M⊙G(z)-M⊙F||1，G(z)为所述输出图像；Lcontextual =||M⊙G(z)-M⊙F||1 , G(z) is the output image;
Lperceptual＝log(1-D(G(z)))，D(G(z))为训练后的所述生成对抗网络模型中判别器对所述输出图像的判别结果值；Lperceptual =log(1-D(G(z))), D(G(z)) is the discrimination result value of the discriminator to the output image in the trained confrontational network model;
所述脊柱修复参照图像为在所述损失函数的值符合设定要求时的一个所述输出图像。The reference image for spine restoration is one of the output images when the value of the loss function meets the set requirements.
具体地，该符合设定要求具体可以是：该损失函数的值的变化走向符合设定要求，例如为该损失函数的值的变化走向为朝向一个目标值进行趋近变化的走向，具体为处于朝向0进行趋近变化的走向，即该损失函数的值的变化走向为下降趋势。或者该符合设定要求具体可以是：该损失函数的值处于设定范围内，例如为接近于0的一个数值范围中。Specifically, meeting the setting requirements may specifically mean that the change trend of the value of the loss function meets the set requirements, for example, the change trend of the value of the loss function is a trend towards a target value, specifically in The trend of approaching change towards 0, that is, the change trend of the value of the loss function is a downward trend. Alternatively, meeting the set requirement may specifically mean that: the value of the loss function is within a set range, for example, in a value range close to 0.
从公式F′＝M⊙F+(1-M)⊙G(z)，并结合生成对抗网络模型在模型训练过程中生成的输出图像G(z)有多张的事实情况，我们可以看出，对修复后的目标脊柱图像最重要的是第二项生成部分，也就是需要实现良好的图像修复损失区域算法。找出G(z)中的最优图像G(z)*。From the formula F′=M⊙F+(1-M)⊙G(z), combined with the fact that there are multiple output images G(z) generated by the GAN model during model training, we can see that, The most important part of the inpainted target spine image is the second generation part, which is the need to implement a good image inpainting loss region algorithm. Find the optimal image G(z)* in G(z).
为了实现这一目的，我们需要两个重要信息：上下文信息和感知信息，这两个信息的获取主要是通过损失函数实现，损失函数越小，图像修复的结果越好。In order to achieve this goal, we need two important information: context information and perceptual information. The acquisition of these two information is mainly achieved through loss function. The smaller the loss function, the better the image restoration result.
为保证输入图片相同的上下文信息，图片F与G(z)中对应的像素尽可能的相似，因此需要对产生不相似像素的G(z)作出惩罚，损失函数为：In order to ensure the same context information of the input picture, the corresponding pixels in the picture F and G(z) are as similar as possible, so it is necessary to punish G(z) that produces dissimilar pixels. The loss function is:
Lcontextual＝||M⊙G(z)-M⊙F||1；Lcontextual =||M⊙G(z)-M⊙F||1 ;
最理想的情况是F和G(z)的所有像素值是相同的，损失函数值为0。The ideal situation is that all pixel values of F and G(z) are the same, and the loss function value is 0.
为了获得高质量的修复图片，我们需要让判别器具有正确分辨真实图片的能力，损失函数为：In order to obtain high-quality repaired pictures, we need to allow the discriminator to have the ability to correctly distinguish real pictures. The loss function is:
Lperceptual＝log(1-D(G(z)))；Lperceptual = log(1-D(G(z)));
因此最终的损失函数为：So the final loss function is:
L＝Lcontextual+λLperceptual；L=Lcontextual +λLperceptual ;
其中，λ为事先设置好的，是一个超参数；D(G(z))∈[0,1]。Among them, λ is set in advance and is a hyperparameter; D(G(z))∈[0,1].
考虑到图像处理中的实际和图像信息可控范围内的损失，具体实现时，脊柱修复参照图像为在所述损失函数的值处于设定范围内时的一个输出图像；该设定范围为接近于0的一个范围。Considering the actual loss in image processing and the loss within the controllable range of image information, during specific implementation, the spine restoration reference image is an output image when the value of the loss function is within a set range; the set range is close to in a range of 0.
步骤103，对所述目标脊柱图像进行三维重建，获取对应的目标脊柱模型。
该目标脊柱图像为一个二值图像。最后，将修复后的目标脊柱图像进行三维重建，获得正常的脊柱模型，该模型可以通过3D打印技术用于TES手术。The target spine image is a binary image. Finally, 3D reconstruction is performed on the repaired target spine image to obtain a normal spine model, which can be used in TES surgery through 3D printing technology.
本申请实施例中，通过基于非病变脊柱图像，训练生成对抗网络模型，利用训练后的生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，将病变发生部位修复为非病变脊柱形态，得到目标脊柱图像，对目标脊柱图像进行三维重建，获取最终的目标脊柱模型，该过程通过利用生成对抗网络，对病变脊柱图像进行复原，以尽可能得到一个理想状态下的椎体模型，实现一个尽可能正常、完整的椎体的重建，避免目前需要制作数十个植入物的三维术前模型的问题，提升脊柱假体制作的精准度，减少假体植入患者体内后的活动限制。In the embodiment of the present application, by training the generated confrontational network model based on the non-diseased spine image, the trained generated confrontational network model is used to repair the lesion in the lesioned spine image, and restore the lesioned spine to the non-diseased spine shape , get the target spine image, perform three-dimensional reconstruction on the target spine image, and obtain the final target spine model. This process restores the diseased spine image by using the generative confrontation network, so as to obtain an ideal vertebral body model as much as possible, and achieve A reconstruction of the vertebral body that is as normal and complete as possible, avoiding the current need to make dozens of implants in three-dimensional preoperative models, improving the accuracy of spinal prosthesis production, and reducing the movement restrictions after the prosthesis is implanted in the patient's body .
本申请实施例中还提供了脊柱模型生成方法的不同实施方式。The embodiments of the present application also provide different implementations of the spine model generation method.
参见图2，图2是本申请实施例提供的一种脊柱模型生成方法的流程图二。如图2所示，一种脊柱模型生成方法，该方法包括以下步骤：Referring to FIG. 2 , FIG. 2 is a flowchart 2 of a spine model generation method provided by an embodiment of the present application. As shown in Figure 2, a spine model generation method, the method includes the following steps:
步骤201，对非病变脊柱的计算机断层扫描CT图像进行语义分割，获得脊柱区域的第一二值图像。
首先，通过神经网络对非病变脊柱CT图像进行语义分割，获得能明确区分脊柱区域和非脊柱区域的二值图。具体地，该第一二值图像中脊柱区域灰度值为255，非脊柱区域灰度值为0。First, a neural network is used to semantically segment non-lesional spinal CT images to obtain a binary image that can clearly distinguish spinal regions from non-spine regions. Specifically, the gray value of the spine region in the first binary image is 255, and the gray value of the non-spine region is 0.
具体地,对非病变脊柱的计算机断层扫描CT图像进行语义分割，可以是基于V-Net神经网络进行的脊柱CT图像分割。Specifically, the semantic segmentation of computed tomography CT images of non-pathological spines may be performed based on the V-Net neural network for spine CT image segmentation.
V-Net是一个典型的端到端的网络结构，针对的问题是三维图像分割。这里，我们将其用来进行二维图像分割。在利用V-Net神经网络进行非病变脊柱CT图像分割时，将非病变脊柱CT图像的二维图像信息进行输入，忽略深度信息。V-Net is a typical end-to-end network structure, which is aimed at the problem of three-dimensional image segmentation. Here, we use it for 2D image segmentation. When using the V-Net neural network to segment non-lesional spine CT images, the two-dimensional image information of non-lesional spine CT images is input, and the depth information is ignored.
具体地，这里的V-Net神经网络结构中，左侧为一个下采样的过程，用来捕获上下文信息，分5组卷积操作进行，每组卷积操作后进行一次类似pooling操作，具体来说就是步幅为2的2*2的卷积核进行反卷积，将图像缩小为原来的1/2。通过四组操作将大小为512*512的图像变成大小为32*32的图像。结构中的右侧为上采样过程，结合下采样各层信息来还原细节信息，使用4组反卷积，每次上采样将图片扩展为原来的2倍，然后将相应的特征图进行复制，和左侧下采样的结果进行连结，上采样过程结束后得到512*512*16大小的图，最后用一个1*1的卷积核将通道数降为1。在下采样过程中作卷积时没有改变图像的大小，这样做的好处是省去了裁剪的工作，因为对应层的大小是一样的，这样能够保留卷积操作提取到的全部信息。分割后得到的脊柱区域的第一二值图像如图5所示。Specifically, in the V-Net neural network structure here, the left side is a downsampling process, which is used to capture context information, and it is divided into 5 groups of convolution operations. After each group of convolution operations, a similar pooling operation is performed. Specifically, That is to say, a 2*2 convolution kernel with a stride of 2 performs deconvolution to reduce the image to 1/2 of its original size. The image with the size of 512*512 is changed into the image with the size of 32*32 through four groups of operations. The right side of the structure is the upsampling process, combining the information of each layer of downsampling to restore the detailed information, using 4 sets of deconvolution, each upsampling expands the image to twice the original size, and then copies the corresponding feature map, Connect with the result of downsampling on the left. After the upsampling process, a 512*512*16 size image is obtained. Finally, a 1*1 convolution kernel is used to reduce the number of channels to 1. The size of the image is not changed during the convolution during the downsampling process. The advantage of this is that the work of cropping is omitted, because the size of the corresponding layer is the same, so that all the information extracted by the convolution operation can be preserved. The first binary image of the spine region obtained after segmentation is shown in FIG. 5 .
步骤202，基于所述第一二值图像，训练所述生成对抗网络模型。
将获得的正常脊柱CT图像的二值图像放入生成对抗网络进行训练，用训练好的生成对抗网络模型对脊柱肿瘤发生部位进行图像修复，获得该部位的正常脊柱二值图像。Put the obtained binary image of the normal spine CT image into the generative confrontation network for training, and use the trained generative confrontation network model to repair the image of the spinal tumor site, and obtain the normal spine binary image of this part.
具体地，在得到非病变脊柱CT图像的二值图像时，可能会出现图像中脊柱区域相比于背景区域太小的情况，因此还可以对该第一二值图像进一步优化，将第一二值图像进行裁剪，将裁剪后的图像进行放大，得到优化后的该第一二值图像，进而训练生成对抗网络模型。具体可以是，将初始大小为512*512的图像进行裁剪和缩放，变为64*64；将处理后的图像放入WGAN-GP进行训练，最后得到训练模型，训练模型会输出模型训练结果，即前述实施方式中生成对抗网络模型在模型训练过程中生成的图像。Specifically, when obtaining a binary image of a CT image of a non-pathological spine, the spine area in the image may be too small compared to the background area, so the first binary image can be further optimized, and the first two The value image is cropped, and the cropped image is enlarged to obtain the optimized first binary image, and then the generative confrontation network model is trained. Specifically, the image with an initial size of 512*512 is cropped and scaled to 64*64; the processed image is put into WGAN-GP for training, and finally the training model is obtained, and the training model will output the model training results. That is, the images generated by the adversarial network model during the model training process in the foregoing embodiments are generated.
步骤201及步骤202实现根据非病变脊柱图像，训练生成对抗网络模型的实施过程。
步骤203，对病变脊柱的CT图像进行语义分割，获得脊柱区域的第二二值图像。
该步骤中，实现的是对病变脊椎的CT图像的处理，生成脊柱区域的第二二值图像。该基于CT图像进行语义分割获得脊柱区域的二值图像的处理过程与可参见步骤201中对非病变脊柱的CT图像进行语义分割，获得脊柱区域的二值图像的处理过程，此处不再赘述。In this step, what is realized is to process the CT image of the lesioned spine to generate a second binary image of the spinal column region. The processing process of performing semantic segmentation on the basis of CT images to obtain binary images of the spine region can refer to the processing process of performing semantic segmentation on CT images of non-pathological spines in
步骤204，确定所述第二二值图像为所述病变脊柱图像。
在通过训练后的生成对抗网络模型进行图像修复时，修复的图像需为一个二值图像，因此，在本步骤中，将语义分割并处理得到的脊柱区域的第二二值图像确定为病变脊柱图像，为后续图像修复做准备。When image repair is performed through the trained GAN model, the repaired image needs to be a binary image. Therefore, in this step, the second binary image of the spine region obtained by semantic segmentation and processing is determined as the lesioned spine image to prepare for subsequent image restoration.
步骤205，通过训练后的所述生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，获得目标脊柱图像。
该步骤的实现过程与前述实施方式中的步骤102的实现过程相同，此处不再赘述。The implementation process of this step is the same as the implementation process of
步骤206，通过移动立方体算法对所述目标脊柱图像进行三维重建，得到对应的所述目标脊柱模型。Step 206: Perform three-dimensional reconstruction on the image of the target spine through a moving cube algorithm to obtain a corresponding model of the target spine.
这里，利用移动立方体算法对修复后的脊柱肿瘤二值图像进行三维重建，得到最终的目标脊柱模型，技术容易实现。Here, the moving cube algorithm is used to perform three-dimensional reconstruction on the repaired binary image of the spinal tumor to obtain the final target spine model, which is easy to implement.
本申请实施例中，通过基于非病变脊柱图像，训练生成对抗网络模型，利用训练后的生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，将病变发生部位修复为非病变脊柱形态，得到目标脊柱图像，对目标脊柱图像进行三维重建，获取最终的目标脊柱模型，该过程通过利用生成对抗网络，对病变脊柱图像进行复原，以尽可能得到一个理想状态下的椎体模型，实现一个尽可能正常、完整的椎体的重建，避免目前需要制作数十个植入物的三维术前模型的问题，提升脊柱假体制作的精准度，减少假体植入患者体内后的活动限制。In the embodiment of the present application, by training the generated confrontational network model based on the non-diseased spine image, the trained generated confrontational network model is used to repair the lesion in the lesioned spine image, and restore the lesioned spine to the non-diseased spine shape , get the target spine image, perform three-dimensional reconstruction on the target spine image, and obtain the final target spine model. This process restores the diseased spine image by using the generative confrontation network, so as to obtain an ideal vertebral body model as much as possible, and achieve A reconstruction of the vertebral body that is as normal and complete as possible, avoiding the current need to make dozens of implants in three-dimensional preoperative models, improving the accuracy of spinal prosthesis production, and reducing the movement restrictions after the prosthesis is implanted in the patient's body .
参见图6，是图6是本申请实施例提供的一种脊柱模型生成系统的结构图，为了便于说明，仅示出了与本申请实施例相关的部分。Referring to FIG. 6, FIG. 6 is a structural diagram of a spine model generation system provided by the embodiment of the present application. For convenience of description, only the parts related to the embodiment of the present application are shown.
所述脊柱模型生成系统300包括：模型训练模块301、图像修复模块302及模型建立模块303。The spine
模型训练模块301，用于根据非病变脊柱图像，训练生成对抗网络模型；The
图像修复模块302，用于通过训练后的所述生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，获得目标脊柱图像，其中，所述目标图像中所述病变发生部位被修复为非病变脊柱形态；The
模型建立模块303，用于对所述目标脊柱图像进行三维重建，获取对应的目标脊柱模型。The
其中，所述模型训练模块301具体用于：Wherein, the
对非病变脊柱的计算机断层扫描CT图像进行语义分割，获得脊柱区域的第一二值图像；Semantic segmentation of computed tomography CT images of the non-diseased spine to obtain a first binary image of the spine region;
基于所述第一二值图像，训练所述生成对抗网络模型。Based on the first binary image, train the generative adversarial network model.
所述图像修复模块具体用于：The image restoration module is specifically used for:
基于所述病变脊柱图像及预设的二值掩码矩阵，生成在所述病变发生部位添加掩码的脊柱掩码图像；Based on the lesion spine image and a preset binary mask matrix, generate a spine mask image in which a mask is added to the lesion occurrence site;
基于所述脊柱掩码图像及脊柱修复参照图像，通过训练后的所述生成对抗网络模型，生成修复后的所述目标脊柱图像F′＝M⊙F+(1-M)⊙G(z)*；Based on the spine mask image and the spine repair reference image, the trained target spine image F'=M⊙F+(1-M)⊙G(z)* is generated by the trained GAN model. ;
其中，F为所述病变脊柱图像；M为所述预设的二值掩码矩阵，所述二值掩码矩阵中的元素与所述病变脊柱图像的像素相对应；M⊙F为所述脊柱掩码图像；G(z)*为所述脊柱修复参照图像，所述脊柱修复参照图像为所述生成对抗网络模型在模型训练过程中生成的符合设定参照条件的图像；⊙为矩阵元素与图像中像素点值相乘的运算。Wherein, F is the lesion spine image; M is the preset binary mask matrix, and the elements in the binary mask matrix correspond to the pixels of the lesion spine image; M⊙F is the Spine mask image; G(z)* is the reference image for the repair of the spine, and the reference image for the repair of the spine is an image that meets the set reference conditions generated by the generation confrontation network model during the model training process; ⊙ is a matrix element The operation of multiplying the pixel value in the image.
所述图像修复模块还用于：The image restoration module is also used for:
获取所述生成对抗网络模型在模型训练过程中生成的输出图像；Obtain the output image generated by the generated confrontation network model during the model training process;
根据所述输出图像，依据损失函数L＝Lcontextual+λLperceptual，从所述输出图像中获取所述脊柱修复参照图像；According to the output image, according to the loss function L=Lcontextual + λLperceptual , obtain the reference image for spine restoration from the output image;
其中，λ为超参数；Among them, λ is a hyperparameter;
Lcontextual＝||M⊙G(z)-M⊙F||1，G(z)为所述输出图像；Lcontextual =||M⊙G(z)-M⊙F||1 , G(z) is the output image;
Lperceptual＝log(1-D(G(z)))，D(G(z))为训练后的所述生成对抗网络模型中判别器对所述输出图像的判别结果值；Lperceptual =log(1-D(G(z))), D(G(z)) is the discrimination result value of the discriminator to the output image in the trained confrontational network model;
所述脊柱修复参照图像为在所述损失函数的值符合设定要求时的一个所述输出图像。The reference image for spine restoration is one of the output images when the value of the loss function meets the set requirements.
脊柱模型生成系统还包括：The spine model generation system also includes:
获得模块，用于对病变脊柱的CT图像进行语义分割，获得脊柱区域的第二二值图像；An obtaining module, configured to perform semantic segmentation on the CT image of the lesion spine, and obtain a second binary image of the spine region;
确定模块，用于确定所述第二二值图像为所述病变脊柱图像。A determining module, configured to determine that the second binary image is the lesioned spine image.
所述模型建立模块303具体用于：The
通过移动立方体算法对所述目标脊柱图像进行三维重建，得到对应的所述目标脊柱模型。Performing three-dimensional reconstruction on the image of the target spine by using a moving cube algorithm to obtain a corresponding model of the target spine.
本申请实施例提供的脊柱模型生成系统能够实现上述脊柱模型生成方法的实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。The spine model generation system provided by the embodiment of the present application can realize the various processes of the above-mentioned spine model generation method embodiment, and can achieve the same technical effect. To avoid repetition, details are not repeated here.
图7是本申请实施例提供的一种终端的结构图。如该图7所示，该实施例的终端4包括：处理器40、存储器41以及存储在所述存储器41中并可在所述处理器40上运行的计算机程序42。FIG. 7 is a structural diagram of a terminal provided by an embodiment of the present application. As shown in FIG. 7 , the
示例性的，所述计算机程序42可以被分割成一个或多个模块/单元，所述一个或者多个模块/单元被存储在所述存储器41中，并由所述处理器40执行，以完成本申请。所述一个或多个模块/单元可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述所述计算机程序42在所述终端4中的执行过程。例如，所述计算机程序42可以被分割成模型训练模块、图像修复模块、模型建立模块、获得模块及确定模块，各模块具体功能如下：Exemplarily, the
模型训练模块，用于根据非病变脊柱图像，训练生成对抗网络模型；The model training module is used to train and generate an adversarial network model according to the non-pathological spine image;
图像修复模块，用于通过训练后的所述生成对抗网络模型，对病变脊柱图像中病变发生部位进行图像修复，获得目标脊柱图像，其中，所述目标图像中所述病变发生部位被修复为非病变脊柱形态；The image repair module is used to perform image repair on the lesion occurrence part in the lesion spine image through the trained generation confrontation network model, and obtain the target spine image, wherein the lesion occurrence part in the target image is repaired as a non- Diseased spine morphology;
模型建立模块，用于对所述目标脊柱图像进行三维重建，获取对应的目标脊柱模型。The model building module is used to perform three-dimensional reconstruction on the image of the target spine, and obtain a corresponding target spine model.
其中，所述模型训练模块具体用于：Wherein, the model training module is specifically used for:
对非病变脊柱的计算机断层扫描CT图像进行语义分割，获得脊柱区域的第一二值图像；Semantic segmentation of computed tomography CT images of the non-diseased spine to obtain a first binary image of the spine region;
基于所述第一二值图像，训练所述生成对抗网络模型。Based on the first binary image, train the generative adversarial network model.
所述图像修复模块具体用于：The image restoration module is specifically used for:
基于所述病变脊柱图像及预设的二值掩码矩阵，生成在所述病变发生部位添加掩码的脊柱掩码图像；Based on the lesion spine image and a preset binary mask matrix, generate a spine mask image in which a mask is added to the lesion occurrence site;
基于所述脊柱掩码图像及脊柱修复参照图像，通过训练后的所述生成对抗网络模型，生成修复后的所述目标脊柱图像F′＝M⊙F+(1-M)⊙G(z)*；Based on the spine mask image and the spine repair reference image, the trained target spine image F'=M⊙F+(1-M)⊙G(z)* is generated by the trained GAN model. ;
其中，F为所述病变脊柱图像；M为所述预设的二值掩码矩阵，所述二值掩码矩阵中的元素与所述病变脊柱图像的像素相对应；M⊙F为所述脊柱掩码图像；G(z)*为所述脊柱修复参照图像，所述脊柱修复参照图像为所述生成对抗网络模型在模型训练过程中生成的符合设定参照条件的图像；⊙为矩阵元素与图像中像素点值相乘的运算。Wherein, F is the lesion spine image; M is the preset binary mask matrix, and the elements in the binary mask matrix correspond to the pixels of the lesion spine image; M⊙F is the Spine mask image; G(z)* is the reference image for the repair of the spine, and the reference image for the repair of the spine is an image that meets the set reference conditions generated by the generation confrontation network model during the model training process; ⊙ is a matrix element The operation of multiplying the pixel value in the image.
所述图像修复模块还用于：The image restoration module is also used for:
获取所述生成对抗网络模型在模型训练过程中生成的输出图像；Obtain the output image generated by the generated confrontation network model during the model training process;
根据所述输出图像，依据损失函数L＝Lcontextual+λLperceptual，从所述输出图像中获取所述脊柱修复参照图像；According to the output image, according to the loss function L=Lcontextual + λLperceptual , obtain the reference image for spine restoration from the output image;
其中，λ为超参数；Among them, λ is a hyperparameter;
Lcontextual＝||M⊙G(z)-M⊙F||1，G(z)为所述输出图像；Lcontextual =||M⊙G(z)-M⊙F||1 , G(z) is the output image;
Lperceptual＝log(1-D(G(z)))，D(G(z))为训练后的所述生成对抗网络模型中判别器对所述输出图像的判别结果值；Lperceptual =log(1-D(G(z))), D(G(z)) is the discrimination result value of the discriminator to the output image in the trained confrontational network model;
所述脊柱修复参照图像为在所述损失函数的值符合设定要求时的一个所述输出图像。The reference image for spine restoration is one of the output images when the value of the loss function meets the set requirements.
获得模块，用于对病变脊柱的CT图像进行语义分割，获得脊柱区域的第二二值图像；An obtaining module, configured to perform semantic segmentation on the CT image of the lesion spine, and obtain a second binary image of the spine region;
确定模块，用于确定所述第二二值图像为所述病变脊柱图像。A determining module, configured to determine that the second binary image is the lesioned spine image.
所述模型建立模块具体用于：The model building module is specifically used for:
通过移动立方体算法对所述目标脊柱图像进行三维重建，得到对应的所述目标脊柱模型。Performing three-dimensional reconstruction on the image of the target spine by using a moving cube algorithm to obtain a corresponding model of the target spine.
所述终端4可以是桌上型计算机、笔记本、掌上电脑及云端服务器等计算设备。所述终端4可包括，但不仅限于，处理器40、存储器41。本领域技术人员可以理解，图7仅仅是终端4的示例，并不构成对终端4的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如所述终端还可以包括输入输出设备、网络接入设备、总线等。The
所称处理器40可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。The so-called
所述存储器41可以是所述终端4的内部存储单元，例如终端4的硬盘或内存。所述存储器41也可以是所述终端4的外部存储设备，例如所述终端4上配备的插接式硬盘，智能存储卡(Smart Media Card，SMC)，安全数字(Secure Digital，SD)卡，闪存卡(Flash Card)等。进一步地，所述存储器41还可以既包括所述终端4的内部存储单元也包括外部存储设备。所述存储器41用于存储所述计算机程序以及所述终端所需的其他程序和数据。所述存储器41还可以用于暂时地存储已经输出或者将要输出的数据。The
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。实施例中的各功能单元、模块可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中，上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。另外，各功能单元、模块的具体名称也只是为了便于相互区分，并不用于限制本申请的保护范围。上述系统中单元、模块的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。Those skilled in the art can clearly understand that for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used for illustration. In practical applications, the above-mentioned functions can be assigned to different functional units, Completion of modules means that the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment may be integrated into one processing unit, or each unit may exist separately physically, or two or more units may be integrated into one unit, and the above-mentioned integrated units may adopt hardware It can also be implemented in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the protection scope of the present application. For the specific working processes of the units and modules in the above system, reference may be made to the corresponding processes in the aforementioned method embodiments, and details will not be repeated here.
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述或记载的部分，可以参见其它实施例的相关描述。In the above-mentioned embodiments, the descriptions of each embodiment have their own emphases, and for parts that are not detailed or recorded in a certain embodiment, refer to the relevant descriptions of other embodiments.
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。Those skilled in the art can appreciate that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be regarded as exceeding the scope of the present application.
在本申请所提供的实施例中，应该理解到，所揭露的终端和方法，可以通过其它的方式实现。例如，以上所描述的终端实施例仅仅是示意性的，例如，所述模块或单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通讯连接可以是通过一些接口，装置或单元的间接耦合或通讯连接，可以是电性，机械或其它的形式。In the embodiments provided in this application, it should be understood that the disclosed terminal and method may be implemented in other ways. For example, the terminal embodiments described above are only illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be Incorporation may either be integrated into another system, or some features may be omitted, or not implemented. In another point, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.
所述集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请实现上述实施例方法中的全部或部分流程，也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计算机可读存储介质中，该计算机程序在被处理器执行时，可实现上述各个方法实施例的步骤。其中，所述计算机程序包括计算机程序代码，所述计算机程序代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。所述计算机可读介质可以包括：能够携带所述计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、电载波信号、电信信号以及软件分发介质等。需要说明的是，所述计算机可读介质包含的内容可以根据司法管辖区内立法和专利实践的要求进行适当的增减，例如在某些司法管辖区，根据立法和专利实践，计算机可读介质不包括电载波信号和电信信号。If the integrated module/unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments in the present application can also be completed by instructing related hardware through computer programs. The computer programs can be stored in a computer-readable storage medium, and the computer When the program is executed by the processor, the steps in the above-mentioned various method embodiments can be realized. Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM, Read-Only Memory) , random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, computer-readable media Excludes electrical carrier signals and telecommunication signals.
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-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still implement the foregoing embodiments Modifications to the technical solutions described in the examples, or equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the application, and should be included in the Within the protection scope of this application.
Claims (10)
Translated fromChinesePriority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910450859.5ACN110246216B (en) | 2019-05-28 | 2019-05-28 | Spine model generation method, spine model generation system and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910450859.5ACN110246216B (en) | 2019-05-28 | 2019-05-28 | Spine model generation method, spine model generation system and terminal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110246216A CN110246216A (en) | 2019-09-17 |
| CN110246216Btrue CN110246216B (en) | 2023-02-28 |
Family
ID=67885234
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910450859.5AActiveCN110246216B (en) | 2019-05-28 | 2019-05-28 | Spine model generation method, spine model generation system and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110246216B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111127412B (en)* | 2019-12-18 | 2021-05-28 | 广东工业大学 | A Pathological Image Recognition Device Based on Generative Adversarial Network |
| CN111274990B (en)* | 2020-02-11 | 2024-01-12 | 广东同天科技产业发展有限公司 | Computer device and storage medium for classifying spinal morphology |
| CN111353995B (en)* | 2020-03-31 | 2023-03-28 | 成都信息工程大学 | Cervical single cell image data generation method based on generation countermeasure network |
| CN112734740B (en)* | 2021-01-18 | 2024-02-23 | 推想医疗科技股份有限公司 | Method for training target detection model, target detection method and device |
| CN115456947A (en)* | 2022-08-11 | 2022-12-09 | 中国人民解放军总医院第四医学中心 | Repair method of acetabular injury based on deep learning |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105723423A (en)* | 2013-10-30 | 2016-06-29 | 皇家飞利浦有限公司 | Volumetric image data visualization |
| CN107945140A (en)* | 2017-12-20 | 2018-04-20 | 中国科学院深圳先进技术研究院 | A kind of image repair method, device and equipment |
| CN108460830A (en)* | 2018-05-09 | 2018-08-28 | 厦门美图之家科技有限公司 | Image repair method, device and image processing equipment |
| CN108765417A (en)* | 2018-06-15 | 2018-11-06 | 西安邮电大学 | It is a kind of that system and method is generated based on the femur X-ray film of deep learning and digital reconstruction irradiation image |
| CN108765340A (en)* | 2018-05-29 | 2018-11-06 | Oppo（重庆）智能科技有限公司 | Fuzzy image processing method, apparatus and terminal device |
| CN109242865A (en)* | 2018-09-26 | 2019-01-18 | 上海联影智能医疗科技有限公司 | Medical image auto-partition system, method, apparatus and storage medium based on multichannel chromatogram |
| WO2019079778A1 (en)* | 2017-10-20 | 2019-04-25 | Nuvasive, Inc. | Intervertebral disc modeling |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3470006B1 (en)* | 2017-10-10 | 2020-06-10 | Holo Surgical Inc. | Automated segmentation of three dimensional bony structure images |
| US10892058B2 (en)* | 2017-09-29 | 2021-01-12 | K2M, Inc. | Systems and methods for simulating spine and skeletal system pathologies |
| CN107977971A (en)* | 2017-11-09 | 2018-05-01 | 哈尔滨理工大学 | The method of vertebra positioning based on convolutional neural networks |
| CN108038860A (en)* | 2017-11-30 | 2018-05-15 | 杭州电子科技大学 | Spine segmentation method based on the full convolutional neural networks of 3D |
- 2019
- 2019-05-28CNCN201910450859.5Apatent/CN110246216B/enactiveActive
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105723423A (en)* | 2013-10-30 | 2016-06-29 | 皇家飞利浦有限公司 | Volumetric image data visualization |
| WO2019079778A1 (en)* | 2017-10-20 | 2019-04-25 | Nuvasive, Inc. | Intervertebral disc modeling |
| CN107945140A (en)* | 2017-12-20 | 2018-04-20 | 中国科学院深圳先进技术研究院 | A kind of image repair method, device and equipment |
| CN108460830A (en)* | 2018-05-09 | 2018-08-28 | 厦门美图之家科技有限公司 | Image repair method, device and image processing equipment |
| CN108765340A (en)* | 2018-05-29 | 2018-11-06 | Oppo（重庆）智能科技有限公司 | Fuzzy image processing method, apparatus and terminal device |
| CN108765417A (en)* | 2018-06-15 | 2018-11-06 | 西安邮电大学 | It is a kind of that system and method is generated based on the femur X-ray film of deep learning and digital reconstruction irradiation image |
| CN109242865A (en)* | 2018-09-26 | 2019-01-18 | 上海联影智能医疗科技有限公司 | Medical image auto-partition system, method, apparatus and storage medium based on multichannel chromatogram |
Non-Patent Citations (2)
| Title |
|---|
| 基于深度学习的脊柱CT图像分割;刘忠利等;《计算机应用与软件》;20181012;第35卷(第10期);第200-204页，第273页* |
| 生成对抗网络在医学图像处理中的应用;陈锟等;《生命科学仪器》;20181025;第16卷;第71-80页，第91页* |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110246216A (en) | 2019-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110246216B (en) | Spine model generation method, spine model generation system and terminal | |
| CN109754389B (en) | An image processing method, device and equipment | |
| Jiao et al. | Burn image segmentation based on mask regions with convolutional neural network deep learning framework: more accurate and more convenient | |
| CN110400328B (en) | Calculation method, calculation system and terminal for motion speed of surgical assistant robot | |
| CN110276408B (en) | 3D image classification method, device, equipment and storage medium | |
| JP2013521844A (en) | Increased probability of model-based segmentation | |
| WO2021081841A1 (en) | Splenic tumor recognition method based on vrds 4d medical image, and related apparatus | |
| CN111918611A (en) | Abnormal display control method for chest X-ray image, abnormal display control program, abnormal display control device, and server device | |
| WO2020168694A1 (en) | Vrds 4d medical image-based ai processing method and product for tumors | |
| KR20230125412A (en) | Classification method and analysis apparatus for uric acid stone using radiomics parameters of computed tomography | |
| Zedan et al. | RMHA-Net: Robust Optic Disc and Optic Cup Segmentation Based on Residual Multiscale Feature Extraction With Hybrid Attention Networks | |
| CN111613300B (en) | Tumor and blood vessel Ai processing methods and products based on VRDS 4D medical imaging | |
| CN113205566A (en) | Abdomen three-dimensional medical image conversion generation method based on deep learning | |
| CN117671052A (en) | Mammary gland image processing and reconstructing method and system based on data driving | |
| CN117830328A (en) | Kidney CTA image segmentation and reconstruction method and system for intraoperative augmented reality guidance | |
| CN116739951A (en) | An image generator, image style conversion device and method | |
| WO2021081839A1 (en) | Vrds 4d-based method for analysis of condition of patient, and related products | |
| CN116052263A (en) | A control method and electronic device | |
| Jerg et al. | Diffuse domain method for needle insertion simulations | |
| WO2021081850A1 (en) | Vrds 4d medical image-based spine disease recognition method, and related devices | |
| Wu et al. | 3D residual attention hierarchical fusion for real-time detection of the prostate capsule | |
| JP2021111076A (en) | Diagnostic devices, diagnostic systems, and programs that use AI | |
| Kim et al. | Multi-Stage Hierarchical Feature Extraction for Efficient 3D Medical Image Segmentation. | |
| CN117058405A (en) | Image-based emotion recognition method, system, storage medium and terminal | |
| KR20240081583A (en) | Apparatus and method for creating 3d liver objects using artificial intelligence based volumne segmenmentation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |