WO2021068643A1 - 一种复合能谱ct成像方法 - Google Patents

一种复合能谱ct成像方法 Download PDF

Info

Publication number
WO2021068643A1
WO2021068643A1 PCT/CN2020/108820 CN2020108820W WO2021068643A1 WO 2021068643 A1 WO2021068643 A1 WO 2021068643A1 CN 2020108820 W CN2020108820 W CN 2020108820W WO 2021068643 A1 WO2021068643 A1 WO 2021068643A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
voltage
scanning
low
scan
Prior art date
Application number
PCT/CN2020/108820
Other languages
English (en)
French (fr)
Inventor
曾凯
徐丹
Original Assignee
南京安科医疗科技有限公司
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 南京安科医疗科技有限公司 filed Critical 南京安科医疗科技有限公司
Publication of WO2021068643A1 publication Critical patent/WO2021068643A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/02—Arrangements for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
    • A61B6/03—Computed tomography [CT]
    • A61B6/032—Transmission computed tomography [CT]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/40—Arrangements for generating radiation specially adapted for radiation diagnosis
    • A61B6/4064—Arrangements for generating radiation specially adapted for radiation diagnosis specially adapted for producing a particular type of beam
    • A61B6/4085—Cone-beams
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/42—Arrangements for detecting radiation specially adapted for radiation diagnosis
    • A61B6/4208—Arrangements for detecting radiation specially adapted for radiation diagnosis characterised by using a particular type of detector
    • A61B6/4241—Arrangements for detecting radiation specially adapted for radiation diagnosis characterised by using a particular type of detector using energy resolving detectors, e.g. photon counting
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/42—Arrangements for detecting radiation specially adapted for radiation diagnosis
    • A61B6/4266—Arrangements for detecting radiation specially adapted for radiation diagnosis characterised by using a plurality of detector units
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
    • A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
    • A61B6/5229—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data combining image data of a patient, e.g. combining a functional image with an anatomical image
    • A61B6/5235—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data combining image data of a patient, e.g. combining a functional image with an anatomical image combining images from the same or different ionising radiation imaging techniques, e.g. PET and CT

Definitions

  • the invention relates to the technical field of energy spectrum imaging, in particular to a composite energy spectrum CT imaging method.
  • Energy spectrum imaging technology is of great significance to medical imaging diagnosis. It can separate information of different energies of substances, significantly suppress ray hardening artifacts, and bring more evidence for clinical diagnosis. But to achieve energy spectrum imaging, it is necessary to use more advanced hardware systems than traditional CT, such as Siemens' dual-ray source dual-detector CT system, General Electric's high-speed switching CT scanning system, and Philips's dual-layer detector CT system. These systems rely on their own proprietary hardware technology to scan and obtain material information of different energies. Therefore, it is difficult for traditional CT systems to perform energy spectrum imaging in actual clinical applications.
  • Energy spectrum CT has been widely used in scientific research and clinical practice since Siemens launched the dual-source dual-detector solution in 2006.
  • Traditional CT imaging provides an image of the effective X-ray absorption coefficient of the scanned object under a certain scan kV, which is related to the size of the object, the ray filter used, and so on. Even under the same scanning conditions, the CT values on different CT scanning systems are different. Therefore, quantitative analysis is relatively difficult on traditional CT.
  • Energy spectrum CT scans the object under different energies, and can sense the attenuation of the object to different energy X-rays, so it can distinguish the composition of the material in the scanned object, and the obtained image is less affected by the scanning conditions and can be more Accurately provide quantitative analysis.
  • Energy spectrum CT measures the absorption of x-rays of different energies by the scanned object (patient) to achieve the purpose of material decomposition. In order to obtain better image quality, generally speaking, this measurement under two or more energies requires good simultaneity and large energy separation.
  • Table 1 there are several implementation methods shown in Table 1 in the existing products on the market: one is the fast switching of the energy spectrum adopted by General Electric, the second is the dual-source dual-detector of Siemens, and the third is the dual-sensor of Philips. Layer detector.
  • photon counting CT is under development, and the market has no products that can be used in clinical practice.
  • the simultaneity and energy separation of photon counting CT are the best, but its cost is also quite expensive.
  • the stability problem of photon counting detectors at high doses has not been well solved, so it is difficult to walk on the road to commercialization.
  • the fast energy spectrum switching has good simultaneity and low cost, its switching speed also has a limit. When the switching speed becomes faster and faster, due to the rise and fall times of the tube voltage, the difference between the switched energy spectra will become smaller and smaller, resulting in the deterioration of the effect of reconstructing the image.
  • the cost of dual-source dual-detector is also high, so energy spectrum scanning technology has been only available in Siemens' highest-end CT products.
  • the double-layer detector has good simultaneity, its energy resolution capability is the worst among all technologies, and the cost of the detector itself is also relatively high.
  • the present invention proposes a composite energy spectrum CT imaging method, which is based on the traditional CT system, mainly through scan control, suitable image reconstruction algorithms and The energy spectrum decomposition technology to obtain the energy spectrum image does not require any changes on the hardware.
  • a composite energy spectrum CT imaging method including the steps:
  • the high voltage generator switches the low voltage and the high voltage according to the specified frequency to obtain the composite scanning data, which specifically includes any one of the switching methods described in a, b, and c in a scanning period:
  • the high and low voltage is switched every time the ray source rotates
  • the scan obtained by scanning includes axial scanning and spiral scanning.
  • the scanning pitch p satisfies:
  • the ray off time between the previous high-pressure/low-pressure scan and the next low-pressure/high-pressure scan is
  • the method of image reconstruction includes: filtered back projection method, iterative reconstruction method and compressed sensing method.
  • the image reconstruction method for removing artifacts is used in the step (2) to perform image reconstruction on the low-voltage scan data and the high-voltage scan data respectively, and the specific steps include:
  • Construct training data Obtain a clear and artifact-free CT image as the target image; obtain the spiral scanning cone beam projection data by performing spiral CT scanning on the phantom corresponding to the target image, and pass the spiral scanning cone beam projection data through
  • the existing image reconstruction method performs image reconstruction to obtain the initial image; the initial image is subjected to a numerical simulation process to generate the corresponding simulated cone-beam projection data; the simulated cone-beam projection data is reconstructed by the same image reconstruction method as the initial image.
  • Img represents the output image of the convolutional neural network
  • Img ture represents the target image corresponding to Img
  • Img k represents the pixel value of the k-th pixel in the image Img
  • Img ture,k represents the k-th pixel in the image Img ture Pixels represent the pixel value
  • step (54) For the low-voltage helical scanning cone beam projection data and the high-voltage helical scanning cone beam projection data obtained in step (1), the method described in step (51) is used to obtain the initial image and the secondary image; The obtained initial image and secondary image are sent to the trained convolutional neural network, and the corresponding clear and artifact-free high/voltage reconstructed image is obtained.
  • the convolutional neural network includes: CNN, ResNet, and Unet.
  • the present invention has the following advantages:
  • the present invention implements dual-energy scanning through high-voltage switching like GE
  • the present invention does not need to switch every sampling like the GE system.
  • the switching speed can be adjusted according to the existing hardware conditions without any changes to the hardware, so the implementation cost is extremely low. Due to the reduced switching speed, the voltage can maintain the same stability as the single-energy scan, so no special calibration is required.
  • the exposure current is also easy to modify, and it has the ability to realize automatic current modulation.
  • the present invention mainly has the advantages of low cost, easy implementation, large energy separation, and low dose.
  • Figure 1 is a flow chart of the present invention
  • Figure 2 is a schematic diagram of two complete data scans at low voltage and high voltage respectively when the axis scan mode is adopted;
  • Figure 3 is a schematic diagram of scanning data of two adjacent circles obtained by switching high and low voltages in the same circle several times when the axis scanning mode is adopted;
  • Figure 4 is a schematic diagram of scanning two half-turn data at high and low voltages respectively when the axis scanning mode is adopted;
  • Figure 5 is a schematic diagram of scanning data for switching between low voltage and high voltage every other circle in the spiral mode
  • Figure 6 is a schematic diagram of scanning data for switching between low voltage and high voltage every 0.75 turns in spiral mode
  • Fig. 7 is a schematic diagram of scanning data of reducing the current in a specific direction or completely turning off the ray in the spiral mode
  • Figure 8 is a structural diagram of a convolutional neural network.
  • the energy spectrum CT imaging method proposed by the present invention is based on the traditional CT system.
  • the conventional CT system realizes a more flexible voltage switching scanning mode to collect data, and uses image reconstruction and energy spectrum decomposition techniques to obtain energy spectrum images. .
  • This method does not require any changes to the hardware.
  • the present invention proposes a method of using existing CT hardware to switch between low voltage and high voltage on the basis of only changing the scan control mode to obtain a set of scan data under composite energy, and then use advanced reconstruction technology to achieve Dual-energy scanning method.
  • the high-voltage controller In a scan of the CT system, the high-voltage controller repeatedly switches between low voltage and high voltage at the speed supported by the existing hardware to achieve the purpose of dual-energy scanning.
  • the scanning mode can be axial scanning or spiral scanning.
  • the high voltage generator switches back and forth between low voltage and high voltage. The switching frequency can be adjusted to ensure that under each energy, there is enough data to independently reconstruct low-voltage and high-voltage images.
  • Scanning usually has an axis scan mode and a spiral scan mode.
  • the bed does not move.
  • at least half a circle of complete data is required under each voltage.
  • multiple switching methods can be designed.
  • Figures 2 to 4 show schematic diagrams of scan data in three axis scan modes when the scan cycle is 2 revolutions.
  • Fig. 2 is the complete data of scanning one circle at low voltage and high voltage respectively when the axis scanning mode is adopted.
  • Figure 3 shows the complete data of two circles scanned when the axis scan mode is adopted, where each circle is switched several times, but the switching mode of the first circle and the second circle is opposite, that is, the high voltage area in the first circle and the second circle The low-voltage regions of the first circle overlap with each other, and the high-voltage regions of the second circle overlap each other.
  • Figure 4 is a schematic diagram of scanning two half-turn data at high and low voltages respectively when the axis scanning method is adopted. In these switching modes, both low voltage and high voltage have enough data to reconstruct the image.
  • the pitch is p (movement distance per revolution/scan width).
  • Figures 5 to 7 show the schematic diagrams of the scan data in the three helical scan modes. Among them, Fig. 5 is the ray source switching voltage once per revolution in spiral mode; Fig. 6 is the voltage switching once per 0.75 revolution of the ray source in spiral mode; Fig.
  • the high and low voltages are switched every N turns of the ray source, 0.5 ⁇ N ⁇ 1. It is only necessary to ensure that at least half of the high-voltage scan data and low-voltage scan data are obtained in one scan cycle. Turn off or not turn off the ray between the high pressure/low pressure scan and the next low pressure/high pressure scan.
  • low-voltage and high-voltage exposure data can independently reconstruct the image in the entire scanning range. After obtaining low-voltage and high-voltage images, the material can be decomposed directly in the image domain.
  • Step1 Based on the voltage switching mode illustrated in Figures 2-7, scan using the existing CT system.
  • the voltage switching interval can adopt the speed that the system hardware can reach. For example, the scanning speed is 0.5 seconds per revolution, and 0.75 revolutions are used to switch once, the low voltage is 80kVp, and the high voltage is 140kVp.
  • Step2 After the data is collected by scanning, the low-voltage and high-voltage data are corrected and preprocessed respectively, and the images of the two voltages are reconstructed separately using CT image reconstruction technology.
  • its reconstruction algorithms include but are not limited to: filtered back projection, iterative reconstruction, compressed sensing, etc.
  • an improved spiral reconstruction method is preferably provided here to reduce the impact of cone beam artifacts on image quality, which specifically includes the following steps:
  • Construct training data Obtain a clear and artifact-free CT image as the target image; perform spiral CT scanning on the phantom corresponding to the target image to obtain the spiral scan cone beam projection data, and pass the current pattern to the spiral scan cone beam projection data.
  • Some image reconstruction methods perform image reconstruction to obtain the initial image; the initial image is subjected to a numerical simulation process to generate the corresponding simulated cone-beam projection data; the simulated cone-beam projection data is reconstructed by the same image reconstruction method as the initial image to obtain a secondary image;
  • Img represents the output image of the convolutional neural network
  • Img ture represents the target image corresponding to Img
  • Img k represents the pixel value of the k-th pixel in the image Img
  • Img ture,k represents the k-th pixel in the image Img ture Pixels represent the pixel value
  • step 1) For the low-voltage helical scanning cone beam projection data and the high-voltage helical scanning cone beam projection data obtained in step (1), the method described in step 1) is used to obtain the initial image and the secondary image; The initial image and the secondary image are sent to the trained convolutional neural network, and the corresponding clear and artifact-free high/voltage reconstructed image is obtained.
  • the existing image reconstruction methods include, but are not limited to, filtered back projection method, iterative reconstruction method, and compressed sensing method
  • the convolutional neural network includes: CNN, ResNet, and Unet.
  • Step3 The obtained reconstructed images of the two voltages can be directly decomposed in the image domain.
  • the decomposition function F of the polynomial in the image domain can be solved by the following method:
  • F wat and F iod respectively represent the decomposition function of the image threshold material to be solved, usually expressed by a polynomial; j represents the index value of each pixel of the input high-voltage image/low-voltage image; Img low and Img high respectively represent The low-voltage image and the high-voltage image obtained by image reconstruction, Img wat and Img iod respectively represent the real water-based image and the iodine-based image, which can usually be set in advance through phantom measurement or numerical simulation.
  • the current commercial CT system that can perform dual-energy scanning through high-voltage switching is GE's high-speed switching spectrum CT.
  • the product needs to change the design of the high-voltage generator in hardware, while reducing the requirements for voltage stabilization to achieve the purpose of high-speed switching.
  • This implementation has a cost investment in hardware.
  • the energy spectrum scan of the system requires special calibration, and the exposure current cannot be arbitrarily modified in the scan protocol, resulting in the inability to perform current modulation to reduce radiation dose.
  • the present invention implements dual-energy scanning through high-voltage switching like GE
  • the present invention does not need to switch every sampling like the GE system.
  • the switching speed can be adjusted according to the existing hardware conditions without any changes to the hardware, so the implementation cost is extremely low. Due to the reduced switching speed, the voltage can maintain the same stability as the single-energy scan, so no special calibration is required.
  • the exposure current is also easy to modify, and it has the ability to realize automatic current modulation.
  • the present invention mainly has the advantages of low cost, easy implementation, large energy separation, and low dose.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Medical Informatics (AREA)
  • Radiology & Medical Imaging (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Optics & Photonics (AREA)
  • Pathology (AREA)
  • Physics & Mathematics (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Pulmonology (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

一种复合能谱CT成像方法，在扫描过程中通过高压发生器按照指定频率切换低电压和高电压，获得复合扫描数据；对低电压扫描数据和高电压扫描数据分别进行图像重建，得到低电压图像和高电压图像；对重建后的低电压图像和高电压图像进行图像域的物质分解得到基物质图像。该方法基于传统CT系统实现，电压切换的速度可以根据现有硬件的条件来调整，不需要对硬件进行任何改动，因此实现成本极低。由于切换速度的降低，电压能保持和单能扫描一样的稳定性，因此不需要进行特殊的校准。具有成本低，易实现，能量分离大，剂量低的优点。

Description

一种复合能谱CT成像方法 技术领域
本发明涉及能谱成像技术领域，尤其是一种复合能谱CT成像方法。
背景技术
能谱成像技术对医学影像诊断有着重要的意义，他能够分离物质不同能量的信息，显著抑制射线硬化伪影，给临床诊断带来更多的依据。但是要达到能谱成像必须要采用比传统CT更加先进的硬件系统，比如西门子的双射线源双探测器CT系统，通用电气的高速切换CT扫描系统，飞利浦的双层探测器CT系统。这些系统都是依赖他们自己专有的硬件技术来扫描获得不同能量的物质信息。因此传统的CT系统难以在实际临床应用中进行能谱成像。
能谱CT自从西门子在2006年推出双源双探测器的解决方案开始，已经在科研和临床上得到了广泛应用。传统CT成像提供的是被扫描物体在某个扫描kV下对X射线的有效吸收系数图像，与物体的大小、所使用的射线过滤器等等都相关。即使在同一扫描条件下，在不同CT扫描系统上的CT值也有差异。因此在传统CT上，定量分析相对比较困难。能谱CT通过在不同能量下对物体进行扫描，能感知物体对不同能量X射线的衰减情况，因此能分辨出被扫描物体中物质的组成，而且得到的图像受扫描条件的影响小，能更准确地提供定量的分析。
能谱CT是通过测量被扫描物体(病人)对不同能量的x射线的吸收情况来达到物质分解的目的。为了获得更好的图像质量，一般来说，这在两个或多个能量下的测量要求有好的同时性，以及大的能量分离。目前来说市面上已有产品中有表1所示的几种实现方式：一是通用电气公司采用的能谱快速切换，二是西门子公司的双源双探测器，第三是飞利浦公司的双层探测器。另外还有光子计数CT正在研发中，市场还没有能用于临床的产品。
表1 现有能谱CT技术
能谱CT技术 同时性 能量分离 造价
通用电气快速能谱切换 良 中 低
西门子双源双探测器 差 好 高
飞利浦双层探测器 优 差 中
光子计数 优 优 高
目前的技术水平来说，光子计数CT的同时性和能量分离是最好的，但是其造价也是相当昂贵。而且光子计数探测器在高剂量下的稳定性问题一直没有得到很好的解决，因此在商业化的道路上走得比较艰难。快速能谱切换虽然有很好的同时性以及低廉的成本，但是其切换速度也是有极限的。当切换速度越来越快时，由于球管电压存在一定的上升和下降时间，切换的能谱之间的差异将越来越小，导致重建图像的效果变差。双源双探测器的成本也居高不下，因此一直只在西门子的最高端CT产品中才有能谱扫描技术。双层探测器虽然有很好的同时性，但是它的能量分辨能力是所有技术中最差的，而且探测器本身的造价也较高。
发明内容
发明目的：为了降低CT成像对于能谱CT系统的硬件要求，本发明提出一种复合能谱CT成像方法，该方法基于传统CT系统而提出的，主要是通过扫描控制、适合的图像重建算法和能谱分解技术来得到能谱图像，在硬件上不需要进行任何更改。
技术方案：为实现上述目的，本发明提出的技术方案为：
一种复合能谱CT成像方法，包括步骤：
(1)在扫描过程中通过高压发生器按照指定频率切换低电压和高电压，获得复合扫描数据，具体包括在一个扫描周期内进行a、b、c所述的任意一种切换方式：
a.射线源每旋转一圈切换一次高低电压；
b.射线源每旋转N圈切换一次高低电压，0.5≤N＜1，且保证一个扫描周期内至少分别获得半圈高电压扫描数据和低电压扫描数据，前一次高压/低压扫描和后一次低压/高压扫描之间关闭或不关闭射线；
c.在每一圈扫描过程中切换多次，相邻两圈扫描过程中电压切换方式互补，使获得的每一圈扫描数据中高电压扫描数据区域和低电压扫描数据区域交替分布，而相邻两圈扫描数据中，前一圈的高电压扫描数据区域与后一圈的低电压扫描数据区域在位置上一一对应；
(2)对低电压扫描数据和高电压扫描数据分别进行图像重建，得到低电压图像和高电压图像；
(3)对重建后的低电压图像和高电压图像进行图像域的物质分解得到基物质图像。
进一步的，所述扫描获得扫描包括轴扫描和螺旋扫描。
进一步的，当采用螺旋扫描时，对于b所述的切换方式，扫描的螺距p满足：
Figure PCTCN2020108820-appb-000001
前一次高压/低压扫描和后一次低压/高压扫描之间的射线关闭时间为
Figure PCTCN2020108820-appb-000002
进一步的，所述图像重建的方法包括：滤波反投影法、迭代重建法和压缩感知法。
进一步的，当采用螺旋扫描方式进行扫描时，所述步骤(2)中采用去伪影的图像重建方法对低电压扫描数据和高电压扫描数据分别进行图像重建，具体步骤包括：
(51)构建训练数据：获得清晰的无伪影CT图像作为目标图像；通过对与目标图像相应的体模进行螺旋CT扫描得到螺旋扫描锥形束投影数据，对螺旋扫描锥形束投影数据通过现有的图像重建方法进行图像重建，得到初始图像；将初始图像经过数值模拟过程产生对应的模拟锥形束投影数据；对模拟锥形束投影数据采用与初始图像相同的图像重建方法重建得到二次图像；
(52)搭建卷积神经网络，在卷积神经网络的输入层增加一个输入通道，即构建好的卷积神经网络具有两个输入通道；将初始图像和二次图像分别送入两个输入通道，通过卷积神经网络提取初始图像和二次图像的灰度信息、结构信息，并根据提取出的灰度信息、结构信息进行自主学习后估计真实图像；
(53)构建关于目标图像和估计真实图像的损失函数，利用梯度下降法训练所述卷积神经网络；所述损失函数为：
Figure PCTCN2020108820-appb-000003
其中，Img表示卷积神经网络的输出图像，Img ture表示与Img对应的目标图像，Img k表示图像Img中第k个像素点的表示像素值，Img ture,k表示图像Img ture中第k个像素点的表示像素值；
(54)对于步骤(1)中得到的低电压螺旋扫描锥形束投影数据和高电压螺旋扫描锥形束投影数据，分别采用步骤(51)所述的方法得到初始图像和二次图像；将得到的初始图像和二次图像送入训练好的卷积神经网络，得到相应的清晰无伪影的高/电压重建图像。
具体的，所述卷积神经网络包括：CNN、ResNet和Unet。
有益效果：与现有技术相比，本发明具有以下优势：
本发明在实现方式上虽然与GE一样是通过高压切换来实现双能扫描，但是本发明不需要像GE的系统一样在每次采样都进行切换。切换的速度可以根据现有硬件的条件来调整，不需要对硬件进行任何改动，因此实现成本极低。由于切换速度的降低，电压能保持和单能扫描一样的稳定性，因此不需要进行特殊的校准。曝光电流也容易修改，而且有能力实现自动电流调制。总体来说，相对于GE的高速切换方式，本发明主要有成本低，易实现，能量分离大，剂量低的优点。
附图说明
图1为本发明的流程图；
图2为采用轴扫描方式时，分别在低电压和高电压扫描完整的两圈数据示意图；
图3为采用轴扫描方式时，在同一圈内切换多次高低电压，得到的相邻两圈扫描数据的示意图；
图4为采用轴扫描方式时，在高、低电压分别扫描两个半圈数据的示意图；
图5为螺旋模式下每隔一圈切换低电压和高电压的扫描数据示意图；
图6为螺旋模式下每隔0.75圈切换低电压和高电压的扫描数据示意图；
图7为螺旋模式下在特定的方向上降低电流或完全关闭射线的扫描数据示意图；
图8为卷积神经网络的结构图。
具体实施方式
下面结合附图对本发明作更进一步的说明。
本发明所提出的能谱CT成像方法是基于传统CT系统而提出的，通过常规CT系统实现更加灵活的电压切换的扫描方式来采集数据，并利用图像重建和能谱分解技术来得到能谱图像。该方法在硬件上不需要进行任何更改。
对于能谱成像来说，最关键的就是要获得不同能谱的物质信息。本发明提出了一种利用现有CT硬件，在只改变扫描控制方式的基础上，在低电压和高电压之间进行切换，得到一组复合能量下的扫描数据，然后利用高级重建技术，实现双能扫描的方法。CT系统的一次扫描中，高压控制器在低电压和高电压中以现有硬件支持的速度反复切换，达到双能扫描的目的，扫描方式可以是轴扫，也可以是螺旋扫描。扫描过程中，高压发生器在低电压和高电压之间来回切换。切换的频率可以调整，保证在每个能量下，能有足够的数据分别独立重建低电压和高电压的图像。
扫描通常有轴扫描模式和螺旋扫描模式。
在轴扫模式下，病床不动，为了让曝光后在低电压和高电压都能有足够的数据重建图像，一般来说每个电压下面都需要有至少半圈完整的数据。这种模式下，可以设计多种切换方式。
图2至4给出了在扫描周期为2圈的情况下的3种轴扫模式下的扫描数据示意图。其中，图2为采用轴扫描方式时，分别在低电压和高电压扫描一圈完整的数据。图3为采用轴扫描方式时，扫描两圈完整的数据，其中每圈内切换数次，但是第一圈和第二圈的切换方式相反，即第一圈内的高电压区域与第二圈的低电压区域互相重合，第一圈的低电压区域与第二圈的高电压区域互相重合。图4为采用轴扫描方式时，在高、低电压分别扫描两个半圈数据的示意图。在这些切换模式下，低电压和高电压都分别有足够的数据重建出图像。
在螺旋扫描模式下，我们可以通过控制螺距和切换间隔，来保证低电压和高电压下能分别重建出完整的图像。在螺旋扫描模式下，螺距是p(每圈的移动距离/扫描宽度)。每个扫描切换的周期可以按照1/p圈来规定，比如p＝0.5，每两圈就是一个扫描周期。对于每一个扫描周期，图5至7给出了3种螺旋扫描模式下的扫描数据示意图。其中，图5是螺旋模式下射线源每旋转一圈切换一次电压；图6是螺旋模式下射线源每旋转0.75圈切换一次电压；图7是螺旋模式下射线源以高电压扫描0.75圈后切换低电压再扫描0.75圈，但是前一次扫描和后一次扫描之间引入时间间隔(约为0.25圈)，在这个时间间隔内降低电流或者完全关闭射线。
如图7这种情况，射线源每旋转N圈切换一次高低电压，0.5≤N＜1，只需要保证一个扫描周期内至少分别获得半圈高电压扫描数据和低电压扫描数据即可，前一次高压/低压扫描和后一次低压/高压扫描之间关闭或不关闭射线。这个N也可以取其他满足条件的值，比如如p＝0.5,N＝0.75圈，这样情况下如果在前后两次扫描之间关闭射线的话，关闭时间
Figure PCTCN2020108820-appb-000004
此处就是vof＝0.5圈；或者可以取p＝0.75，N＝0.75圈，这样vof＝0圈。
上述的轴扫描和螺旋双能扫描方式，低电压和高电压的曝光数据都能独立重建出整个扫描范围内的图像。在获得低电压和高电压图像后，可以直接在图像域进行物质分解。
下面通过具体实施方式来进一步说明本发明的技术方案，实施例提出的复合能谱CT成像方法流程如图1所示，包括步骤：
Step1:基于图2至7中举例说明的电压切换形式，利用现有CT系统进行扫描。电压的切换间隔可以采用系统硬件能够达到的速度。比如扫描转速是0.5秒一圈，采用0.75圈切换一次，低电压是80kVp，高电压是140kVp。
Step2:通过扫描采集数据后，对低电压和高电压数据分别进行校正和预处理，利用CT图像重建技术对两种电压的图像分别进行重建。
对于轴扫描数据，其重建算法包括但不限于：滤波反投影，迭代重建，压缩感知等。
对于螺旋扫描数据，在分别重建低电压和高电压的扫描数据的时候，由于切换的频率比较低，再加上重建误差和锥角(正比于探测器的排数)的平方成正比，所以当探测器排数增大到128甚至256排的时候就会带来很大的误差，具体体现在重建后的图像中存在较为严重的伪影。为了消除图像伪影，这里优选提供了一种改进的螺旋重建方法来减少锥形束伪影对图像质量带来的影响，具体包括以下步骤：
1)构建训练数据：获得清晰的无伪影CT图像作为目标图像；通过对与目标图像相应的体模进行螺旋CT扫描得到螺旋扫描锥形束投影数据，对螺旋扫描锥形束投影数据通过现有的图像重建方法进行图像重建，得到初始图像；将初始图像经过数值模拟过程产生对应的模拟锥形束投影数据；对模拟锥形束投影数据采用与初始图像相同的图像重建方法重建得到二次图像；
2)搭建如图8所示的卷积神经网络，在卷积神经网络的输入层增加一个输入通道，即构建好的卷积神经网络具有两个输入通道；将初始图像和二次图像分别送入两个输入通道，通过卷积神经网络提取初始图像和二次图像的灰度信息、结构信息，并根据提取出的灰度信息、结构信息进行自主学习后估计真实图像；
3)构建关于目标图像和估计真实图像的损失函数，利用梯度下降法训练所述卷积神经网络；所述损失函数为：
Figure PCTCN2020108820-appb-000005
其中，Img表示卷积神经网络的输出图像，Img ture表示与Img对应的目标图像，Img k表示图像Img中第k个像素点的表示像素值，Img ture,k表示图像Img ture中第k个 像素点的表示像素值；
4)对于步骤(1)中得到的低电压螺旋扫描锥形束投影数据和高电压螺旋扫描锥形束投影数据，分别采用步骤1)所述的方法得到初始图像和二次图像；将得到的初始图像和二次图像送入训练好的卷积神经网络，得到相应的清晰无伪影的高/电压重建图像。
上述改进的螺旋重建方法中，所述的现有的图像重建方法包括但不限于滤波反投影法、迭代重建法和压缩感知法，所述卷积神经网络包括：CNN、ResNet和Unet。
Step3：得到的两个电压的重建图像可以直接在图像域进行物质分解。物质分解时，图像域的多项式的分解函数F可以通过下面的方法来求解得到：
Figure PCTCN2020108820-appb-000006
其中，F wat、F iod分别表示待求解的图像阈物质分解的函数，通常用多项式表示；j表示输入的高电压图像/低电压图像的每个像素的索引值；Img low、Img high分别表示图像重建得到的低电压图像和高电压图像，Img wat、Img iod分别表示真实的水基图像和碘基图像，通常可以通过体模测量或数值仿真提前设定得到。
目前能通过高压切换进行双能扫描的商业CT系统是GE的高速切换能谱CT。该产品在硬件上需要改变高压发生器的设计，同时降低对稳压的要求来达到高速切换的目的。这种实现方式在硬件上有成本投入，系统的能谱扫描需要进行特殊的校准，而且扫描协议中不能随意修改曝光电流，导致无法进行电流调制来降低辐射剂量。
本发明在实现方式上虽然与GE一样是通过高压切换来实现双能扫描，但是本发明不需要像GE的系统一样在每次采样都进行切换。切换的速度可以根据现有硬件的条件来调整，不需要对硬件进行任何改动，因此实现成本极低。由于切换速度的降低，电压能保持和单能扫描一样的稳定性，因此不需要进行特殊的校准。曝光电流也容易修改，而且有能力实现自动电流调制。总体来说，相对于GE的高速切换方式，本发明主要有成本低，易实现，能量分离大，剂量低的优点。
以上所述仅是本发明的优选实施方式，应当指出：对于本技术领域的普通技术人员来说，在不脱离本发明原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本发明的保护范围。

Claims (6)

  1. 一种复合能谱CT成像方法，其特征在于，包括步骤：
    (1)在扫描过程中通过高压发生器按照指定频率切换低电压和高电压，获得复合扫描数据，具体包括在一个扫描周期内进行a、b、c所述的任意一种切换方式：
    a.射线源每旋转一圈切换一次高低电压；
    b.射线源每旋转N圈切换一次高低电压，0.5≤N＜1，且保证一个扫描周期内至少分别获得半圈高电压扫描数据和低电压扫描数据，前一次高压/低压扫描和后一次低压/高压扫描之间关闭或不关闭射线；
    c.在每一圈扫描过程中切换多次，相邻两圈扫描过程中电压切换方式互补，使获得的每一圈扫描数据中高电压扫描数据区域和低电压扫描数据区域交替分布，而相邻两圈扫描数据中，前一圈的高电压扫描数据区域与后一圈的低电压扫描数据区域在位置上一一对应；
    (2)对低电压扫描数据和高电压扫描数据分别进行图像重建，得到低电压图像和高电压图像；
    (3)对重建后的低电压图像和高电压图像进行图像域的物质分解得到基物质图像。
  2. 根据权利要求1所述的一种复合能谱CT成像方法，其特征在于，所述扫描获得扫描包括轴扫描和螺旋扫描。
  3. 根据权利要求2所述的一种复合能谱CT成像方法，其特征在于，当采用螺旋扫描时，对于b所述的切换方式，扫描的螺距p满足：
    Figure PCTCN2020108820-appb-100001
    前一次高压/低压扫描和后一次低压/高压扫描之间的射线关闭时间为
    Figure PCTCN2020108820-appb-100002
  4. 根据权利要求1所述的一种复合能谱CT成像方法，其特征在于，所述图像重建的方法包括：滤波反投影法、迭代重建法和压缩感知法。
  5. 根据权利要求2所述的一种复合能谱CT成像方法，其特征在于，当采用螺旋扫描方式进行扫描时，所述步骤(2)中采用去伪影的图像重建方法对低电压扫描数据和高电压扫描数据分别进行图像重建，具体步骤包括：
    (51)构建训练数据：获得清晰的无伪影CT图像作为目标图像；通过对与目标图像相应的体模进行螺旋CT扫描得到螺旋扫描锥形束投影数据，对螺旋扫描锥形束投影数据通过现有的图像重建方法进行图像重建，得到初始图像；将初始图像经过数值模拟 过程产生对应的模拟锥形束投影数据；对模拟锥形束投影数据采用与初始图像相同的图像重建方法重建得到二次图像；
    (52)搭建卷积神经网络，在卷积神经网络的输入层增加一个输入通道，即构建好的卷积神经网络具有两个输入通道；将初始图像和二次图像分别送入两个输入通道，通过卷积神经网络提取初始图像和二次图像的灰度信息、结构信息，并根据提取出的灰度信息、结构信息进行自主学习后估计真实图像；
    (53)构建关于目标图像和估计真实图像的损失函数，利用梯度下降法训练所述卷积神经网络；所述损失函数为：
    Figure PCTCN2020108820-appb-100003
    其中，Im g表示卷积神经网络的输出图像，Im g ture表示与Img对应的目标图像，Im g k表示图像Im g中第k个像素点的表示像素值，Im g ture,k表示图像Im g ture中第k个像素点的表示像素值；
    (54)对于步骤(1)中得到的低电压螺旋扫描锥形束投影数据和高电压螺旋扫描锥形束投影数据，分别采用步骤(51)所述的方法得到初始图像和二次图像；将得到的初始图像和二次图像送入训练好的卷积神经网络，得到相应的清晰无伪影的高/电压重建图像。
  6. 根据权利要求5所述的一种复合能谱CT成像方法，其特征在于，所述卷积神经网络包括：CNN、ResNet和Unet。
PCT/CN2020/108820 2019-10-08 2020-08-13 一种复合能谱ct成像方法 WO2021068643A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910950111.1 2019-10-08
CN201910950111.1A CN110742635B (zh) 2019-10-08 2019-10-08 一种复合能谱ct成像方法

Publications (1)

Publication Number Publication Date
WO2021068643A1 true WO2021068643A1 (zh) 2021-04-15

Family

ID=69277715

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/108820 WO2021068643A1 (zh) 2019-10-08 2020-08-13 一种复合能谱ct成像方法

Country Status (2)

Country Link
CN (1) CN110742635B (zh)
WO (1) WO2021068643A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114403912A (zh) * 2022-01-17 2022-04-29 明峰医疗系统股份有限公司 一种基于深度学习的能谱ct有效原子序数估计方法
CN115227275A (zh) * 2022-09-23 2022-10-25 中国科学技术大学 一种旋转阻挡器和基于该旋转阻挡器的散射伪影校正方法

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110742635B (zh) * 2019-10-08 2021-10-08 南京安科医疗科技有限公司 一种复合能谱ct成像方法
CN112656436A (zh) * 2020-12-21 2021-04-16 明峰医疗系统股份有限公司 一种ct变速扫描调节剂量方法
CN113509651A (zh) * 2021-07-21 2021-10-19 苏州雷泰医疗科技有限公司 一种具备能谱ct功能的直线加速器、控制方法及存储介质
CN118159197A (zh) * 2021-11-04 2024-06-07 深圳帧观德芯科技有限公司 在计算机断层扫描中使用辐射检测器的成像方法
CN116458906B (zh) * 2023-06-19 2023-09-22 有方（合肥）医疗科技有限公司 侧位图生成方法、装置、电子设备及存储介质

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103202707A (zh) * 2012-01-16 2013-07-17 株式会社东芝 X射线ct装置
CN103913779A (zh) * 2012-12-31 2014-07-09 清华大学 多能ct成像系统以及成像方法
US20150071400A1 (en) * 2013-09-06 2015-03-12 Siemens Aktiengesellschaft Method and x-ray system for dual-energy spectra ct scanning and image reconstruction
CN104882350A (zh) * 2015-06-11 2015-09-02 杭州与盟医疗技术有限公司 一种提供多能量和更大覆盖范围x射线球管系统
CN106859677A (zh) * 2017-01-16 2017-06-20 沈阳东软医疗系统有限公司 双能ct扫描图像重建方法及装置
CN109448071A (zh) * 2018-11-06 2019-03-08 深圳安科高技术股份有限公司 一种能谱图像重建方法及系统
WO2019048294A1 (en) * 2017-09-08 2019-03-14 Koninklijke Philips N.V. METHOD AND DEVICE FOR DETERMINING A SPECTRAL SCANNING PROTOCOL OF TOMODENSITOMETRY
CN110060315A (zh) * 2019-04-22 2019-07-26 深圳安科高技术股份有限公司 一种基于人工智能的图像运动伪影消除方法及系统
CN110742635A (zh) * 2019-10-08 2020-02-04 南京安科医疗科技有限公司 一种复合能谱ct成像方法

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5389324B2 (ja) * 2006-12-18 2014-01-15 ジーイー・メディカル・システムズ・グローバル・テクノロジー・カンパニー・エルエルシー Ｘ線断層撮影装置
US8165264B2 (en) * 2009-01-28 2012-04-24 Kabushiki Kaisha Toshiba Method of pre-reconstruction decomposition for fast kV-switching acquisition in dual energy computed tomography (CT)
CN203149136U (zh) * 2012-12-31 2013-08-21 清华大学 多能ct成像系统
CN107356615B (zh) * 2016-05-10 2020-01-21 清华大学 一种用于双能x射线ct的方法和系统
US10475214B2 (en) * 2017-04-05 2019-11-12 General Electric Company Tomographic reconstruction based on deep learning
AU2017420781B2 (en) * 2017-06-26 2020-12-17 Elekta, Inc. Method for improving cone-beam CT image quality using a deep convolutional neural network
CN109242920B (zh) * 2017-07-11 2020-06-02 清华大学 物质分解方法、装置和系统
CN108010098B (zh) * 2017-12-04 2020-12-25 首都师范大学 一种双能谱ct基材料图像迭代重建方法
WO2019128660A1 (zh) * 2017-12-29 2019-07-04 清华大学 训练神经网络的方法和设备、图像处理方法和设备以及存储介质
CN108564553A (zh) * 2018-05-07 2018-09-21 南方医科大学 基于卷积神经网络的低剂量ct图像噪声抑制方法
CN109064521A (zh) * 2018-07-25 2018-12-21 南京邮电大学 一种使用深度学习的cbct去伪影方法
CN109272472B (zh) * 2018-10-15 2022-07-15 天津大学 面向医用能谱ct图像的噪声及伪影消除方法

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103202707A (zh) * 2012-01-16 2013-07-17 株式会社东芝 X射线ct装置
CN103913779A (zh) * 2012-12-31 2014-07-09 清华大学 多能ct成像系统以及成像方法
US20150071400A1 (en) * 2013-09-06 2015-03-12 Siemens Aktiengesellschaft Method and x-ray system for dual-energy spectra ct scanning and image reconstruction
CN104882350A (zh) * 2015-06-11 2015-09-02 杭州与盟医疗技术有限公司 一种提供多能量和更大覆盖范围x射线球管系统
CN106859677A (zh) * 2017-01-16 2017-06-20 沈阳东软医疗系统有限公司 双能ct扫描图像重建方法及装置
WO2019048294A1 (en) * 2017-09-08 2019-03-14 Koninklijke Philips N.V. METHOD AND DEVICE FOR DETERMINING A SPECTRAL SCANNING PROTOCOL OF TOMODENSITOMETRY
CN109448071A (zh) * 2018-11-06 2019-03-08 深圳安科高技术股份有限公司 一种能谱图像重建方法及系统
CN110060315A (zh) * 2019-04-22 2019-07-26 深圳安科高技术股份有限公司 一种基于人工智能的图像运动伪影消除方法及系统
CN110742635A (zh) * 2019-10-08 2020-02-04 南京安科医疗科技有限公司 一种复合能谱ct成像方法

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114403912A (zh) * 2022-01-17 2022-04-29 明峰医疗系统股份有限公司 一种基于深度学习的能谱ct有效原子序数估计方法
CN115227275A (zh) * 2022-09-23 2022-10-25 中国科学技术大学 一种旋转阻挡器和基于该旋转阻挡器的散射伪影校正方法
CN115227275B (zh) * 2022-09-23 2022-12-30 中国科学技术大学 一种旋转阻挡器和基于该旋转阻挡器的散射伪影校正方法

Also Published As

Publication number Publication date
CN110742635B (zh) 2021-10-08
CN110742635A (zh) 2020-02-04

Similar Documents

Publication Publication Date Title
WO2021068643A1 (zh) 一种复合能谱ct成像方法
CN105608721B (zh) 计算机断层成像伪影校正方法及装置
Shikhaliev et al. Photon counting spectral CT versus conventional CT: comparative evaluation for breast imaging application
JP5469293B2 (ja) ハイブリッド型コンピュータ断層撮影検出器を使用して画像再構成を行うための方法
JP6492005B2 (ja) Ｘ線ｃｔ装置、再構成演算装置、及び再構成演算方法
CN109272472B (zh) 面向医用能谱ct图像的噪声及伪影消除方法
US10827992B2 (en) Energy-discriminating photon-counting detector and the use thereof
Lee et al. A feasibility study of low-dose single-scan dual-energy cone-beam CT in many-view under-sampling framework
CN110811660B (zh) 一种校正ct射线束硬化伪影的方法
CN102254310A (zh) 用于减少ct图像数据中的图像伪影、尤其是金属伪影的方法
WO2014034618A1 (ja) 医用画像処理装置及びｘ線コンピュータ断層撮影装置
US20170090039A1 (en) Layered radiation detector
DE102009053664A1 (de) Verfahren zur empirischen Bestimmung einer Korrekturfunktion zur Korrektur von Strahlungsaufhärtungs- und Streustrahleneffekten in der Projektionsradiografie und in der Computertomografie
CN103622717A (zh) 由单源单次扫描x光ct图像生成双能x光ct图像方法
DE102012222714A1 (de) Ermittlung eines Mehrfachenergie-Bildes
KR102348139B1 (ko) 이중 해상도의 관심 영역 내외 투영 데이터를 이용한 체내 단층 촬영 방법 및 시스템
Feng et al. An experimental method to correct low-frequency concentric artifacts in photon counting CT
Sheth et al. Technical assessment of 2D and 3D imaging performance of an IGZO‐based flat‐panel X‐ray detector
CN110047114B (zh) 用于改善计算机断层扫描中的空间分辨率的系统和方法
CN103654818A (zh) 一种实现多能量ct扫描重建的方法及装置
DE102012105678A1 (de) Verfahren und System zur dosisreduzierten Röntgenbildgebung
Yang et al. An object-specific and dose-sparing scatter correction approach for a dedicated cone-beam breast CT system using a parallel-hole collimator
CN109448071A (zh) 一种能谱图像重建方法及系统
Tesic et al. Single-slit digital radiography: some practical considerations
CN103136772A (zh) 基于加权阿尔法散度约束的x射线低剂量ct图像重建方法

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: 20874808

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: 20874808

Country of ref document: EP

Kind code of ref document: A1