WO2019153289A1 - 运用三维影像套合重建血管结构的方法及其三维模型 - Google Patents

运用三维影像套合重建血管结构的方法及其三维模型 Download PDF

Info

Publication number
WO2019153289A1
WO2019153289A1 PCT/CN2018/076224 CN2018076224W WO2019153289A1 WO 2019153289 A1 WO2019153289 A1 WO 2019153289A1 CN 2018076224 W CN2018076224 W CN 2018076224W WO 2019153289 A1 WO2019153289 A1 WO 2019153289A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
fixed
displacement
blood vessel
dimensional
Prior art date
Application number
PCT/CN2018/076224
Other languages
English (en)
French (fr)
Inventor
陈绍哲
Original Assignee
佛教慈济医疗财团法人大林慈济医院
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 佛教慈济医疗财团法人大林慈济医院 filed Critical 佛教慈济医疗财团法人大林慈济医院
Priority to PCT/CN2018/076224 priority Critical patent/WO2019153289A1/zh
Publication of WO2019153289A1 publication Critical patent/WO2019153289A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis

Definitions

  • the invention relates to a contrast technique, in particular to a method for reconstructing a blood vessel structure by using a three-dimensional image and a three-dimensional model constructed by the method.
  • Blood vessels are an important channel for the human body to supply blood. Modern people suffer from excessive high-fat foods and poor daily life, resulting in high levels of triglycerides in the blood, which are prone to clogging or hardening of blood vessels. Symptoms of blood pressure, along with age, may also increase the risk of cardiovascular disease such as stroke and myocardial infarction.
  • CT computed tomography
  • CTA computed tomography
  • doctors can diagnose and analyze the health of patients through CT and CTA imaging results.
  • CT image is a penetrating image of the neck of the patient by the principle of X-ray.
  • X-ray penetrates the human body, different degrees of penetration images are generated for tissues and organs of different densities, wherein the blood vessels are When the density of other soft tissues is lower, the resulting image of gray-scale color will appear. Conversely, the density of human bones will be higher, and the resulting image will be white.
  • CTA development technology In order to facilitate the analysis and distinguishing images of blood vessels and soft tissues, doctors usually use CTA development technology for diagnosis. CTA images will inject contrast agents into the blood vessels before computed tomography, thus reducing the penetration of X-rays. The penetrating nature of the blood vessels allows the vascular image to approach white, allowing doctors to easily diagnose the location of the patient's blood vessels through CT and CTA imaging results.
  • the current hospital CTA imaging software will remove most of the bones to create a three-dimensional vascular structure, doctors can analyze and diagnose whether the neck vessels are blocked or calcified according to the direction of the blood vessels, and further diagnose the presence of arteries.
  • a vascular disease such as a tumor or an abnormal arteriovenous malformation, and the doctor can clearly evaluate how to perform surgery through a three-dimensional vascular architecture before treatment.
  • the imaging software of the hospital CTA will visualize the processed images, which may cause the doctor to misidentify other soft tissues into blood vessels, and the doctor needs to adjust the window level to facilitate observation. Is the blood vessel position correct?
  • the existing CTA imaging software cannot obtain reconstructed three-dimensional vascular architecture information, such as spatial coordinates or numerical values of blood vessels, and there is no way to perform stereoscopic printing using the reconstructed three-dimensional blood vessels, thereby failing to provide teaching effects.
  • an embodiment of the present invention provides a method for reconstructing a blood vessel structure by using a three-dimensional image, which can automatically and accurately correct and calculate a CT image and accurately fit the CTA image, and the CT image. It is very similar to the characteristic position of the CTA image, and thus accurately reconstructs the vascular structure, which is conducive to the diagnosis and clinical application of the doctor.
  • an embodiment of the present invention provides a method for reconstructing a blood vessel structure by using a three-dimensional image nest, comprising the following steps: initial fitting: using a computed tomography scan and a computed tomography angiography in a target region of the human body respectively Generating a first displacement image of the three-dimensional image and a first fixed image, and converting a plurality of second fixed images of the two-dimensional image by a projection method according to different projection angles of the first displacement image and the first fixed image a plurality of second displacement images, wherein each of the second displacement images respectively corresponds to the feature images of the second fixed images, and fits to form a converted displacement image of the three-dimensional image; and the fit correction: respectively, the converted displacement image and the first fixed image are respectively Setting a variable, the characteristic position of the converted displacement image is close to the first fixed image after measurement and correction; and the blood vessel reconstruction: the converted displacement image and the first fixed image overlap each other, and the converted displacement image and the first fixed image are removed.
  • the projection means employs a maximum intensity projection technique.
  • each of the second fixed images is a cross section (XY), a coronal plane (XZ), and a sagittal plane (YZ) of the first fixed image
  • each of the second displacement images is a first displacement Cross-sectional (XY), coronal (XZ) and sagittal (YZ) images.
  • an optimized fit is included between the calibration fit and the blood vessel reconstruction: the converted displacement image corresponds to the feature position of the first fixed image according to the image type calculation manner, so that the feature of the first displacement image is converted The position produces a very small difference from the feature position of the first fixed image.
  • a noise treatment is further included: the three-dimensional image of the blood vessel of the first fixed image is eliminated by the image erosion method and the image expansion method of the morphology, or the region growing method noise.
  • Another embodiment of the present invention provides a method for reconstructing a blood vessel structure by using a three-dimensional image, comprising the following steps: converting a projection: respectively generating a three-dimensional image by computerized tomography and computed tomography in a target region of the human body.
  • each second displacement image corresponds to the feature image of each second fixed image, and is respectively calculated and assembled by GDB-ICP to form a converted displacement image of the three-dimensional image
  • the fit correction the converted displacement image and the first A fixed image respectively sets a variable, and after the measurement and correction, the characteristic position of the converted displacement image is close to the first fixed image
  • the optimized fit the converted displacement image is calculated according to the image type change, and the characteristic position of the converted displacement image is first Minimize the difference in the feature position of the fixed image
  • revascular reconstruction shift displacement The image and the first fixed image overlap each other, and the portion where the converted displacement image overlaps with the first fixed image is removed, thereby generating a three-dimensional image of the blood vessel.
  • each of the second fixed images is a cross section (XY), a coronal plane (XZ), and a sagittal plane (YZ) of the first fixed image
  • each of the second displacement images is respectively Cross section (XY), coronal plane (XZ) and sagittal plane (YZ) of the first displacement image.
  • a noise processing is further included: the three-dimensional image of the blood vessel of the first fixed image is cancelled by the image erosion method and the image expansion method of the pattern, or the region is grown. The region grows and eliminates noise.
  • An embodiment of the present invention provides a three-dimensional model constructed by using a three-dimensional image to reconstruct a blood vessel structure.
  • the invention can automatically and accurately correct and calculate the first displacement image to form a converted displacement image, and the converted displacement image is accurately fitted with the first fixed image, and the converted displacement image is very similar to the first fixed image feature position. And then accurately reconstruct the vascular structure, which is conducive to the doctor's diagnosis and clinical application.
  • Figure 1 is a flow chart of the steps of the present invention.
  • FIG. 2 is an image diagram of computed tomography and computed tomography angiography of the present invention.
  • 3A is a front view image of a second fixed image and a second displaced image of the present invention.
  • FIG. 3B is a rear view image of the second fixed image and the second displaced image of the present invention.
  • FIG. 4 is a comparison diagram of image fits of a converted displacement image and a first fixed image subjected to nesting correction according to the present invention.
  • FIG. 5 is a comparison diagram of the image fit of the converted displacement image and the first fixed image according to the present invention.
  • Fig. 6 is a view showing a blood vessel image of a first fixed image of the present invention.
  • Computed tomography is a target area that uses X-rays to penetrate the human body.
  • the bones and organ tissues in the human body are of different densities, so that different degrees of penetration are generated, and three-dimensional images of different grayscale values are presented.
  • CTA computed tomography
  • the present invention sets the CT image as a displacement image, and the CTA image is set as a fixed image, so that the feature points and positions of the fixed image mapped fixed image are relatively adjusted and corrected, so that the displacement image is accurately fit on the fixed image. And accurately obtain three-dimensional blood vessel images in fixed images.
  • the present invention provides a method for reconstructing a blood vessel structure by using a three-dimensional image nest, which comprises the following steps:
  • Pre-image processing S1 Computerized Tomography (CT) is used to generate a first displacement image 10 of a three-dimensional image in a target region of the human body in a penetrating contrast mode, and a computerized tomography is used for a period of time ( CTA, Computerized Tomography Angiography) penetrates the same target area of the human body and produces a first fixed image 20 of the three-dimensional image, wherein the target area is a human head.
  • CT Computerized Tomography
  • CTA Computerized Tomography Angiography
  • the time is scanned and monitored, in order to avoid the difference between the position of the first displacement image 10 and the first fixed image 20 caused by the displacement or breathing of the human body, the target area is fixed by the head frame before the monitoring scan, and then respectively CT and CTA image scanning.
  • the first displacement image 10 and the first fixed image 20 generated by the CT and the CTA have obvious headstock signals, and the head frame signal obscures most of the brain information, so the method of using the region growing method is used.
  • the header signal is extracted and excluded to reduce signal interference between the first displacement image 10 and the first fixed image 20.
  • the initial displacement S2 the first displacement image 10 of the three-dimensional image and the first fixed image 20 are easily rotated during the assembly process, thereby causing an error in the image position, thereby the first displacement image 10 and the first fixed image. 20 can accurately align the feature positions with each other.
  • the initial sleeve S2 includes the following steps to achieve the purpose:
  • Converting projection S2A converting the first displacement image 10 into a plurality of second displacement images 11a of the two-dimensional image by a projection means in a cross-sectional (XY), coronal (XZ) and sagittal (YZ) projection angles, 11b, 11c, and the first fixed image 20 is converted into a plurality of second fixed images 21a, 21b, 21c of the two-dimensional image by using the projection angle of view and the projection means.
  • the projection means uses Maximum Intensity Projection (MIP), the principle of which different projection angles of the first fixed image 20 and the first displacement image 10 are respectively projected on a plane by linear coordinates.
  • MIP Maximum Intensity Projection
  • the image results are shown in FIG. 2 and FIG. 3A.
  • the second displacement image 11a is a cross section (XY) of the first displacement image 10, and the second displacement image 11b is a coronal plane (XZ) of the first displacement image 10, and second.
  • the displacement image 11c is a sagittal plane (YZ) of the first displacement image 10
  • the second fixed image 21a is a cross section (XY) of the first fixed image 20
  • the second fixed image 21b is a coronal plane of the first fixed image 20 ( XZ)
  • the second fixed image 21c is the sagittal plane (YZ) of the first fixed image 20.
  • each second displacement image 11a, 11b, 11c and the second fixed image 21a, 21b, 21c of the same projection angle are corresponding to each other by a Generalized Dual Bootstrap-Iterative Closest Point (GDB-ICP) nesting technique
  • GDB-ICP Generalized Dual Bootstrap-Iterative Closest Point
  • the corresponding feature image is found by using the red reference line on each of the second displacement image 11 and each of the second fixed images 21.
  • Feature extraction Calculate the gradient of each of the second displacement images 11 and each of the second fixed images 21.
  • the present invention integrates and converts each of the second displacement images 11 into a converted displacement image 12 of the three-dimensional image.
  • a conversion value is set to 0, and converted into a three-dimensional image by using the following formula.
  • A3D is a matrix parameter of 3*3, and T3D is a displacement parameter.
  • the converted displacement image 12 is firstly fitted to the first fixed image 20 by using a linear interpolation method, so that the converted displacement image 12 is closely combined with the first fixed image 20 in the embodiment of the present invention.
  • the post-conversion error is reduced, and then whether the converted displacement image 12 and the first fixed image 20 are nested in a similar position, and the present invention is published in 2003 by David Mattes, David R Haynor, Hubert Vesselle, Thomas K Lewellen, and William Eubank.
  • the "mutual information method” proposed in the "Pet-ct image registration in the chest using free-form deformations. IEEE transactions on medical imaging" is used to calculate the similarity between the two, and when the shift image 12 is converted to the first fixed image 20 The higher the similarity, the greater the correlation between each other, the smaller the quality of the mutual information, and the calculation formula is as follows:
  • B T Pass the converted image.
  • the nesting optimization is performed.
  • the position correction is performed before the converted displacement image 12 and the first fixed image 20 are assembled, and the gradient is used to find the maximum value in the gradient direction to perform the nesting optimization.
  • the affine transform is used to linearly convert the converted displacement image 12 into the first fixed image 20, and the corresponding stitch correction is performed.
  • the upper left and lower drawings are converted displacement images 12 and
  • the first fixed image 20 corrects the image result before the nesting, in other words, the upper and lower figures correct the assembled image result, and the red line portion is the bone edge of the converted displacement image 12. It is apparent from the result that the displacement image is converted. 12 is very close to the feature position of the first fixed image 20.
  • the present invention further converts the displacement image 12 and the first fixed according to the deformable registration.
  • the image 20 is optimally adjusted, wherein the image-changing image is formed by gradually diffusing the converted displacement image 12 according to the gradient information of the first fixed image 20 toward the first fixed image 20, thereby converting the displacement image 12 and the first fixed image 20.
  • the type is changed, and the difference between the two groups of images is minimal, and the similarity is 0.97.
  • the upper left and lower drawings are the image results before the conversion of the displacement image 12 and the first fixed image 20, in other words, the image results of the upper and lower drawings of the right side are changed.
  • the red line portion is the bone edge of the shift displacement image 12. Obviously, the bone edge of the shift displacement image 12 is closer to the bone position of the first fixed image 20.
  • Vascular reconstruction S4 Finally, the converted displacement image 12 and the first fixed image 20 are superimposed on each other, and the portion of the converted displacement image 12 overlapping with the first fixed image 20 is removed, and then the first fixed image 20 is removed.
  • the three-dimensional image of the blood vessel is relatively generated.
  • the converted displacement image 12 and the first fixed image 20 are nested together, and the high-intensity information is captured, and then the threshold value of the converted displacement image 12 is set to 170, and the first fixed
  • the threshold of the image 20 is 150, thereby removing the information of the subcutaneous tissue and the soft tissue of the first fixed image 20, thereby avoiding the influence of the blood vessel information, and then removing most of the bone information, so that the first fixed image 20 generates blood vessels.
  • 3D imagery is
  • the image results are shown in FIG. 6.
  • the left and right figures respectively correct the first fixed image 20 before and after the removal of the bone and the soft tissue, and the red line is the blood vessel information; thereby, the first accurate image can be accurately
  • the fixed image 20 is reconstructed to obtain a blood vessel of the three-dimensional image, thereby achieving accurate vascular diagnosis.
  • Noise processing S5 This step is used to eliminate the noise of the blood vessels in the first fixed image 20.
  • the noise can be effectively removed by using a mode operation method or a region growing method. .
  • the pattern operation method is calculated by the image erosion method and the image expansion method, wherein the image erosion method enlarges the noise region of the first fixed image 20 to remove the fine noise portion, and then resumes expansion by the image expansion method. .
  • the regional growth rule is to set the initial point, and then grow from the initial point as the starting point, and set the pixel range value to limit the growth, so as to capture the part connected by the blood vessel, and remove most of the surrounding noise, and pass
  • the regional growth method can further remove the calcified area in the blood vessel and improve the accuracy of the vascular tube exchange.
  • I(X) is the image pixel quality, and lower and upper are the lowest pixel quality and the highest pixel quality, respectively.
  • the present invention automatically and accurately corrects and calculates the first displacement image 10 to form a converted displacement image 12, and the converted displacement image 12 is accurately fitted with the first fixed image 20, and the feature positions of each other are very similar.
  • the reconstruction of the vascular structure of the first fixed image 20 can be accurately performed, which is beneficial to the diagnosis and clinical application of the doctor, and the doctor can further construct the printing by reconstructing the blood vessel structure and the skeleton structure.
  • the doctor can further construct the printing by reconstructing the blood vessel structure and the skeleton structure.
  • Into a three-dimensional model in order to have an accurate clinical effect.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Apparatus For Radiation Diagnosis (AREA)
  • Image Processing (AREA)

Abstract

一种运用三维影像套合重建血管结构的方法及其三维模型，所述方法包括初始套合、套合校正及血管重建；其中，初始套合，利用计算机断层扫描及计算机断层血管摄影于人体分别产生三维的一第一位移影像与一第一固定影像，且依据第一位移影像及第一固定影像的不同投影视角，分别投影形成二维的多个第二固定影像及多个第二位移影像，各第二位移图像映射各第二固定影像的特征，而套合形成三维的一转换位移影像；套合校正，将转换位移影像与第一固定影像通过测量及校正，使转换位移影像与第一固定影像非常相似；血管重建，第一固定图像映射重迭转换位移影像，借以精准编辑及重建血管架构。

Description

运用三维影像套合重建血管结构的方法及其三维模型 技术领域
本发明涉及一种造影技术，尤其涉及一种运用三维影像套合重建血管结构的方法及利用该方法建构的三维模型。
背景技术
血管是人体供应血液的重要通道，而现代人因为摄取过多高油脂食物与不良的生活作息，造成血液中的三酸甘油脂过高，而容易形成血管堵塞或硬化，血管硬化通常伴随着高血压的症状，随着年龄的上升，可能也会相对提高发生脑中风和心肌梗塞等心血管疾病的风险。
其中，大多数心血管疾病的病患主要为颈部血管受到堵塞或钙化，而目前医院在心血管疾病的诊断方式大多运用计算机断层扫描(CT,computerized tomography)及计算机断层血管摄影(CTA,computerized tomography angiography)的显影技术，而医生可通过CT和CTA的影像结果来诊断与分析病患的健康情况。进一步说明，CT影像为通过X光原理对病患的颈部进行穿透摄影，当X光穿透人体器官时，对于密度不同的组织和器官会产生不同程度的穿透影像，其中，当血管和其他软组织密度较低，则会呈现灰阶色的结果影像；反之，人体骨骼的密度较高，相对会呈现白色的结果影像。
而医生为了便于分析及区别血管及软组织的影像，通常会共同搭配CTA显影技术进行诊断，而CTA影像会在计算机断层扫描前在血管中注入造影剂(contrast agent)，进而降低X光在穿透血管的穿透性，使血管影像能够趋近于白色，得以让医生通过CT及CTA的影像结果便于诊断病患血管位置。
此外，目前医院CTA的影像软件会将大部分的骨骼去除，以产生三维的血管架构，医生可依据血管的走向进而分析及诊断颈部血管是否堵塞或钙化，还可更进一步诊断是否患有动脉瘤或不正常的动静脉畸形等血管疾病，而且医生在治疗之前，可通过三维的血管架构明确评估如何进行手术治疗。
然而，现阶段的医院CTA的影像软件会把处理过后的影像通过可视化方式呈现，可能会造成医生把其他软组织误辨为血管，而医生还需通过窗值 (window level)的调整，以便于观察血管位置是否正确。
另外，现有的CTA影像软件无法得到重建后的三维血管架构信息，例如：血管的空间坐标或数值等，而且也没办法运用重建后的三维血管进行立体打印，因而无法达到提供教学的效果。
发明内容
为了解决上述课题，本发明的一项实施例提供一种运用三维影像套合重建血管结构的方法，其可自动精准地将CT影像经过校正及计算测量并与CTA影像准确套合，且CT影像与CTA影像的特征位置极为相似，并进而精准地重建血管架构，有利于医生的诊断及临床上的运用。
为达到上述目的，本发明一项实施例提供一种运用三维影像套合重建血管结构的方法，其包括下列步骤：初始套合：利用计算机断层扫描及计算机断层血管摄影于人体的一目标区域分别产生三维影像的一第一位移影像与一第一固定影像，且依据第一位移影像及第一固定影像的不同投影视角，分别以一投影手段转换形成二维影像的多个第二固定影像及多个第二位移影像，而各第二位移影像分别对应各第二固定影像的特征影像，并套合形成三维影像的一转换位移影像；套合校正：将转换位移影像与第一固定影像分别设定变量，通过测量及校正之后转换位移影像的特征位置相靠近于该第一固定影像；以及血管重建：转换位移影像与第一固定影像相互叠合，并且移除转换位移影像与第一固定影像重叠的部分，进而产生血管的三维影像。
本发明的一项实施例中，投影手段运用最大强度投影技术。
本发明的一项实施例中，各第二固定影像分别为第一固定影像的横断面(X-Y)、冠状面(X-Z)及矢状面(Y-Z)，各第二位移影像分别为第一位移影像的横断面(X-Y)、冠状面(X-Z)及矢状面(Y-Z)。
本发明的一项实施例中，于校正套合及血管重建间还包括一优化套合：转换位移影像根据影像型变计算方式对应第一固定影像的特征位置，使转换第一位移影像的特征位置与第一固定影像的特征位置产生极小差异。
本发明的一项实施例中，于血管重建之后还包括一噪声处理：第一固定 影像的血管的三维影像利用型态学的影像侵蚀法及影像膨胀法，或区域成长法(region growing)消除噪声。
本发明另一实施例提供一种运用三维影像套合重建血管结构的方法，其包括下列步骤：转换投影：于人体的一目标区域中分别通过计算机断层扫描及计算机断层血管摄影分别产生三维影像的一第一位移影像与一第一固定影像，第一固定影像及第一位移影像利用最大强度投影技术，于不同投影视角分别投影成二维影像的多个第二固定影像及多个第二位移影像；对应套合：各第二位移影像分别对应各第二固定影像的特征影像，分别通过GDB-ICP计算并套合转换形成三维影像的一转换位移影像；套合校正：转换位移影像与第一固定影像分别设定变量，通过测量及校正之后转换位移影像的特征位置相靠近于第一固定影像；优化套合：转换位移影像根据影像型变计算，将转换位移影像的特征位置与第一固定影像的特征位置产生极小差异；以及血管重建：转换位移影像与第一固定影像相互叠合，并且移除转换位移影像与第一固定影像重叠的部分，进而产生血管的三维影像。
本发明的另一项实施例中，各第二固定影像分别为该第一固定影像的横断面(X-Y)、冠状面(X-Z)及矢状面(Y-Z)，各第二位移影像分别为该第一位移影像的横断面(X-Y)、冠状面(X-Z)及矢状面(Y-Z)。
本发明的另一项实施例中，于该血管重建之后还包括一噪声处理：该第一固定影像的血管的三维影像利用型态学的影像侵蚀法及影像膨胀法消除噪声，或者利用区域成长法(region growing)计算并消除噪声。
本发明一实施例提供一种运用三维影像套合重建血管结构的方法建构的三维模型。
借此，本发明可自动精准地将第一位移影像经过校正及计算测量形成转换位移影像，并转换位移影像与第一固定影像准确套合，且转换位移影像与第一固定影像特征位置极为相似，并进而精准地重建血管架构，有利于医生的诊断及临床上的运用。
附图说明
图1为本发明的步骤流程图。
图2为本发明计算机断层扫描及计算机断层血管摄影的影像图。
图3A为本发明第二固定影像及第二位移影像的套合前影像图。
图3B为本发明第二固定影像及第二位移影像的套合后影像图。
图4为本发明转换位移影像及第一固定影像经过套合校正的影像套合比较图。
图5为本发明转换位移影像及第一固定影像经过优化套合的影像套合比较图。
图6为本发明第一固定影像的血管影像图。
附图标记说明
第一位移影像10
第二位移影像11
第二位移影像11a
第二位移影像11b
第二位移影像11c
转换位移影像12
第一固定影像20
第二固定影像21
第二固定影像21a
第二固定影像21b
第二固定影像21c
前置图像处理S1
初始套合S2
转换投影S2A
对应套合S2B
套合校正S2C
优化套合S3
血管重建S4
噪声处理S5。
具体实施方式
为便于说明本发明于上述发明内容一栏中所表示的中心思想，现以具体实施例表达。实施例中各种不同对象实按适于列举说明的比例，而非按实际组件的比例予以绘制。
计算机断层扫描(CT)是运用X光对应穿透人体的目标区域，而人体中骨头及器官组织分别为不同密度的，因而对应产生不同的穿透程度，进而呈现不同灰阶值的三维影像。
计算机断层血管摄影(CTA)的扫描造影原理与计算机断层扫描(CT)相同，而不同在于计算机断层血管摄影(CTA)在穿透扫描之前会在人体中注射造影剂(contrast agent)，进而在X光在穿透扫描时，降低X光穿透血管的程度，进而在影像中相对产生较高的灰阶值，可较明显观察血管影像。
因而本发明将CT影像设定为位移影像，而CTA影像设定为固定影像，借以操作时，使位移图像映射固定影像的特征点及位置相对调整校正，使位移影像准确套合在固定影像上，并于固定影像精准取得三维的血管影像。
请参阅图1至图6所示，本发明提供一种运用三维影像套合重建血管结构的方法，其包括下列步骤：
前置图像处理S1：利用计算机断层扫描(CT,Computerized Tomography)在人体的一目标区域以穿透造影方式，产生三维影像的一第一位移影像10，并且通过一段时间再运用计算机断层血管摄影(CTA,Computerized Tomography Angiography)在人体的相同目标区域穿透摄影，并产生三维影像的一第一固定影像20，其中，目标区域为人体头部，于本发明实施例中，CT与CTA是在不同时间进行扫描监测，为了避免因此人体位移或呼吸而造成第一位移影像10与第一固定影像20的位置有明显差异，因此在监测扫描之前会利用头架固定所述的目标区域，并随后分别进行CT及CTA影像扫描。
而CT及CTA产生的第一位移影像10与第一固定影像20中会有明显的头 架信号，而头架信号会遮蔽大部分的脑部信息，因此利用区域成长法(region growing)的方式撷取头架信号并将其排除，以减少第一位移影像10与第一固定影像20的信号干扰。
初始套合S2：而三维影像的第一位移影像10与第一固定影像20在套合的过程中，容易受到旋转而导致影像位置有所误差，借以将第一位移影像10与第一固定影像20能够彼此准确对准特征位置，于本发明实施例中，初始套合S2包括有以下步骤以达成目的：
转换投影S2A：将第一位移影像10分别在横断面(X-Y)、冠状面(X-Z)及矢状面(Y-Z)投影视角利用一投影手段转换成二维影像的多个第二位移影像11a、11b、11c，而第一固定影像20运用上述投影视角及投影手段转换成二维影像的多个第二固定影像21a、21b、21c。
于本发明实施例中，投影手段运用最大强度投影技术(MIP,Maximum Intensity Projection)，其原理通过直线坐标将第一固定影像20及第一位移影像10的不同投影视角各别投影在平面上。
影像结果请配合图2及图3A所示，第二位移影像11a为第一位移影像10的横断面(X-Y)，第二位移影像11b为第一位移影像10的冠状面(X-Z)，第二位移影像11c为第一位移影像10的矢状面(Y-Z)，第二固定影像21a为第一固定影像20的横断面(X-Y)，第二固定影像21b为第一固定影像20的冠状面(X-Z)，第二固定影像21c为第一固定影像20的矢状面(Y-Z)。
对应套合S2B：将相同投影视角的各第二位移影像11a、11b、11c与各第二固定影像21a、21b、21c通过Generalized Dual Bootstrap-Iterative Closest Point(GDB-ICP)套合技术将彼此对应套合，而GDB-ICP主要分为初始化(initialization)、撷取特征(Feature extraction)及细部对应(refinement)三个步骤。
初始化(initialization)：是在各第二位移影像11与各第二固定影像21上利用红色参考线找出对应的特征影像。
撷取特征(Feature extraction)：计算各第二位移影像11与各第二固定影像21的梯度。
细部对应(refinement)：则将各第二位移影像11与各第二固定影像21的特征影像的对应关系通过计算并套合形成转换关系，请配合图3A及图3B所示，通过红色参考线观察得知，套合后的各第二位移影像11的特征影像非常接近各第二固定影像21。
接着，本发明将各第二位移影像11整合转换成三维影像的一转换位移影像12。以第二位移影像11a举例来说，将另外一维的参数设定为基准值(identity)，将转换值(translation)设定为0，并运用以下公式转换成三维影像。
Figure PCTCN2018076224-appb-000001
A3D为3*3的矩阵参数，T3D为位移参数。
套合校正S2C：将转换位移影像12与第一固定影像20进行细致套合，于本发明实施例中，首先将运用线性内插方式将转换位移影像12套合在第一固定影像20，以减少转换后的误差，接着比对转换位移影像12与第一固定影像20是否套合在相似位置，并且本发明运用David Mattes,David R Haynor,Hubert Vesselle,Thomas K Lewellen,andWilliam Eubank.于2003发表的“Pet-ct image registration in the chest using free-form deformations.IEEE transactions on medical imaging”文献中提出的mutual information方式来计算彼此的相似度，当转换位移影像12与第一固定影像20的套合相似度越高，则彼此的相关性相对越大，互信息的质也相对越小，而计算公式如下所示：
Figure PCTCN2018076224-appb-000002
其中，
Figure PCTCN2018076224-appb-000003
为最佳转换关系，B T为
Figure PCTCN2018076224-appb-000004
通过转换后的影像。
随后，进行套合优化，于本发明实施例中，在转换位移影像12与第一固定影像20套合之前先进行位置校正，并运用梯度下降法以梯度方向寻找最大值，进行套合优化。
最后，利用affine transform以线性转换方式将转换位移影像12转换套合在第一固定影像20，并且进行对应套合校正，请配合图4所示，左边上、 下附图为转换位移影像12与第一固定影像20校正套合前的影像结果，换言之，右边上、下附图校正套合后的影像结果，而红线部分为转换位移影像12的骨头边缘，由结果显然得知，转换位移影像12与第一固定影像20的特征位置非常接近。
优化套合S3：而套合后的转换位移影像12与第一固定影像20在骨骼边缘有些微误差，因而本发明更进一步根据影像型变计算(Deformable registration)将转换位移影像12与第一固定影像20进行优化调整，其中，影像型变计算是将转换位移影像12依据第一固定影像20的梯度信息进而往第一固定影像20方向逐步扩散，借以转换位移影像12与第一固定影像20形成型变套合，进而两组影像间产生极小差异，而且相似度平均为0.97。
请配合图5所示，左边上、下附图为转换位移影像12与第一固定影像20型变套合前的影像结果，换言之，右边上、下附图型变套合后的影像结果，红线部分为转换位移影像12的骨头边缘，显然地，转换位移影像12的骨骼边缘更贴近于第一固定影像20的骨骼位置。
血管重建S4：最后，将型变套合后的转换位移影像12与第一固定影像20相互叠合，并且移除转换位移影像12与第一固定影像20重叠的部分，进而第一固定影像20会相对产生血管的三维影像，具体来说，将转换位移影像12与第一固定影像20相互套合后，并撷取高亮度信息，接着设定转换位移影像12的阈值为170，第一固定影像20的阈值为150，进而移除第一固定影像20的皮下组织及软组织的信息，可避免血管信息受到影响，随后可移除大部分的骨骼信息，俾使第一固定影像20产生血管的三维影像。
影像结果请配合图6所示，左、右边的附图分别为第一固定影像20校正在移除骨骼及软组织的前后比较图，而红线部分为血管信息；借此，可准确地在第一固定影像20上重建取得三维影像的血管，进而达到准确血管诊断的功效。
噪声处理S5：此步骤则是用来消除第一固定影像20中血管的噪声，于本发明实施例中，可运用型态学操作方式或区域成长法(region growing)两种方式有效移除噪声。
型态学操作方式是利用影像侵蚀法及影像膨胀法方式运算，其中，影像侵蚀法是将第一固定影像20的噪声区域放大，以移除细微的噪声部分，随后再通过影像膨胀法恢复膨胀。
区域成长法则是设定初始点，再由初始点当作起始进行成长，另外设定像素范围值进行成长的限制，借以撷取血管相连的部分，并移除大部分的周围噪声，而且通过区域成长法还可进而移除血管中钙化区域，提升血管管换后的精准性。
其中，像素范围值的公式即为：I(X)∈[lower，uper]
I(X)为影像像素质，lower和upper分别为最低像素质及最高像素质。
综上所述，本发明可自动精准地将第一位移影像10经过校正及计算测量形成转换位移影像12，并转换位移影像12与第一固定影像20准确套合，且彼此的特征位置极为相似，最后，通过两组影像相叠，进而可精准地在第一固定影像20的重建血管架构，有利于医生的诊断及临床上的运用，而且医生还可通过重建血管架构及骨骼架构进一步建构打印成三维模型，借以在临床上具有准确教示的效果。
以上所列举的实施例仅用以说明本发明而已，并非用以限制本发明的保护范围。凡不违本发明精神所从事的种种修改或变化，俱属于本发明意欲保护的范围。

Claims (10)

  1. 一种运用三维影像套合重建血管结构的方法，其特征在于，其包括下列步骤：
    初始套合：利用计算机断层扫描及计算机断层血管摄影于人体的一目标区域分别以穿透造影方式，产生三维影像的一第一位移影像与一第一固定影像，依据该第一位移影像及该第一固定影像的不同投影视角，分别以一投影手段转换形成二维影像的多个第二固定影像及多个第二位移影像，而各第二位移影像分别对应各第二固定影像的特征影像，并套合形成三维影像的一转换位移影像；
    套合校正：将该转换位移影像与该第一固定影像分别设定变量，通过测量及校正之后，该转换位移影像的特征位置相靠近于该第一固定影像；以及
    血管重建：该转换位移影像与该第一固定影像相互叠合，并且移除该转换位移影像与该第一固定影像重叠的部分，进而产生血管的三维影像。
  2. 如权利要求1所述的运用三维影像套合重建血管结构的方法，其特征在于，各第二固定影像分别为该第一固定影像的横断面、冠状面及矢状面，各第二位移影像分别为该第一位移影像的横断面、冠状面及矢状面；该投影手段运用最大强度投影技术。
  3. 如权利要求1所述的运用三维影像套合重建血管结构的方法，其特征在于，于该套合校正及该血管重建步骤之间还包括一优化套合步骤：该转换位移影像根据影像型变计算方式对应该第一固定影像的特征位置，使该转换位移影像的特征位置重叠于该第一固定影像的特征位置。
  4. 如权利要求3所述的运用三维影像套合重建血管结构的方法，其特征在于，于该血管重建步骤之后还包括一噪声处理步骤：该第一固定影像的血管的三维影像利用型态学的影像侵蚀法及影像膨胀法消除噪声。
  5. 如权利要求3所述的运用三维影像套合重建血管结构的方法，其特征在于，于该血管重建步骤之后还包括一噪声处理步骤：该第一固定影像的血管的三维影像运用区域成长法计算并消除噪声。
  6. 一种运用三维影像套合重建血管结构的方法，其特征在于，其包括 下列步骤：
    转换投影：于人体的一目标区域中分别通过计算机断层扫描及计算机断层血管摄影以穿透造影方式，分别产生三维影像的一第一位移影像与一第一固定影像，并利用最大强度投影技术于该第一固定影像及该第一位移影像的不同投影视角，投影成二维影像的多个第二固定影像及多个第二位移影像；
    对应套合：各第二位移影像分别对应各第二固定影像的特征影像，分别通过GDB-ICP计算并套合转换形成三维影像的一转换位移影像；
    套合校正：该转换位移影像与该第一固定影像分别设定变量，通过测量及校正之后，该转换位移影像的特征位置相靠近于该第一固定影像；
    优化套合：该转换位移影像根据影像型变计算，将该转换位移影像的特征位置重叠于该第一固定影像的特征位置；以及
    血管重建：该转换位移影像与该第一固定影像相互叠合，并且移除该转换位移影像与该第一固定影像重叠的部分，进而产生血管的三维影像。
  7. 如权利要求6所述的运用三维影像套合重建血管结构的方法，其特征在于，各第二固定影像分别为该第一固定影像的横断面、冠状面及矢状面，各第二位移影像分别为该第一位移影像的横断面、冠状面及矢状面。
  8. 如权利要求6所述的运用三维影像套合重建血管结构的方法，其特征在于，于该血管重建步骤之后还包括一噪声处理步骤：该第一固定影像的血管的三维影像利用型态学的影像侵蚀法及影像膨胀法消除噪声。
  9. 如权利要求6所述的运用三维影像套合重建血管结构的方法，其特征在于，于该血管重建步骤之后还包括一噪声处理步骤：该第一固定影像的血管的三维影像利用区域成长法计算并消除噪声。
  10. 一种应用如权利要求6所述的运用三维影像套合重建血管结构的方法建构的三维模型。
PCT/CN2018/076224 2018-02-11 2018-02-11 运用三维影像套合重建血管结构的方法及其三维模型 WO2019153289A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/076224 WO2019153289A1 (zh) 2018-02-11 2018-02-11 运用三维影像套合重建血管结构的方法及其三维模型

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/076224 WO2019153289A1 (zh) 2018-02-11 2018-02-11 运用三维影像套合重建血管结构的方法及其三维模型

Publications (1)

Publication Number Publication Date
WO2019153289A1 true WO2019153289A1 (zh) 2019-08-15

Family

ID=67549217

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/076224 WO2019153289A1 (zh) 2018-02-11 2018-02-11 运用三维影像套合重建血管结构的方法及其三维模型

Country Status (1)

Country Link
WO (1) WO2019153289A1 (zh)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150150535A1 (en) * 2013-12-04 2015-06-04 Siemens Medical Solutions Usa, Inc. Motion correction in three-dimensional elasticity ultrasound imaging
CN106327479A (zh) * 2016-09-23 2017-01-11 西安电子科技大学 血管造影中介下先心病术中血管辨识的装置及方法
CN106777582A (zh) * 2016-12-01 2017-05-31 哈尔滨理工大学 一种基于组织分化的长骨骨折愈合仿真系统

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150150535A1 (en) * 2013-12-04 2015-06-04 Siemens Medical Solutions Usa, Inc. Motion correction in three-dimensional elasticity ultrasound imaging
CN106327479A (zh) * 2016-09-23 2017-01-11 西安电子科技大学 血管造影中介下先心病术中血管辨识的装置及方法
CN106777582A (zh) * 2016-12-01 2017-05-31 哈尔滨理工大学 一种基于组织分化的长骨骨折愈合仿真系统

Similar Documents

Publication Publication Date Title
JP7162793B2 (ja) 超音波拓本技術に基づく脊椎画像生成システム及び脊柱手術用のナビゲーション・位置確認システム
US8897514B2 (en) Imaging method for motion analysis
US5889524A (en) Reconstruction of three-dimensional objects using labeled piecewise smooth subdivision surfaces
CN112601487A (zh) 医学图像处理装置、医学图像处理方法和程序
JP5366356B2 (ja) 医用画像処理装置及び医用画像処理方法
US8041088B2 (en) Brain image alignment method and system
JP2013198763A (ja) 医用画像処理装置
TW201219013A (en) Method for generating bone mask
Cheng et al. Airway segmentation and measurement in CT images
US20230130015A1 (en) Methods and systems for computed tomography
KR101118549B1 (ko) 의료용 융합영상 획득장치 및 획득방법
KR20190125592A (ko) 의료 영상을 이용하여 혈관의 3차원 형상을 생성하기 위한 세그멘테이션 방법
WO2019153289A1 (zh) 运用三维影像套合重建血管结构的方法及其三维模型
CN115300809B (zh) 图像处理方法及装置、计算机设备和存储介质
CN111862312B (zh) 一种脑部血管显示装置及方法
CN110858412A (zh) 基于图像配准的心脏冠脉cta模型建立方法
KR20230165022A (ko) 사이클 gan 기반 영상 화질 개선 학습 시스템 및 방법
US20220398720A1 (en) Diagnostic support program
US9235888B2 (en) Image data determination method, image processing workstation, target object determination device, imaging device, and computer program product
JP4540947B2 (ja) Ｘ線ｃｔ装置
JP2020182524A (ja) 画像処理装置、画像処理方法およびプログラム
KR102428579B1 (ko) 전신 ct 스캔 3d 모델링 방법 및 시스템
JP6068177B2 (ja) 医用画像診断装置、医用画像処理装置及び医用画像処理方法
KR102399792B1 (ko) 인공지능 기반 Hounsfield unit (HU) 정규화 및 디노이징 (Denoising)을 통한 전처리 장치 및 방법
Tomaka et al. The dynamics of the stomatognathic system from 4D multimodal data

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18905208

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18905208

Country of ref document: EP

Kind code of ref document: A1