WO2022120983A1 - X射线相位衬度图像提取方法、装置、终端及存储介质 - Google Patents

X射线相位衬度图像提取方法、装置、终端及存储介质 Download PDF

Info

Publication number
WO2022120983A1
WO2022120983A1 PCT/CN2020/139342 CN2020139342W WO2022120983A1 WO 2022120983 A1 WO2022120983 A1 WO 2022120983A1 CN 2020139342 W CN2020139342 W CN 2020139342W WO 2022120983 A1 WO2022120983 A1 WO 2022120983A1
Authority
WO
WIPO (PCT)
Prior art keywords
ray
sample
phase
image
phase contrast
Prior art date
Application number
PCT/CN2020/139342
Other languages
English (en)
French (fr)
Inventor
骆荣辉
李志成
葛永帅
赵源深
Original Assignee
中国科学院深圳先进技术研究院
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 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Publication of WO2022120983A1 publication Critical patent/WO2022120983A1/zh

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
    • A61B6/5205—Devices using data or image processing specially adapted for radiation diagnosis involving processing of raw data to produce diagnostic data

Definitions

  • the invention relates to the technical field of X-ray imaging, in particular to an X-ray phase contrast image extraction method, device, terminal and storage medium.
  • the traditional X-ray absorption contrast imaging technology uses the difference in the internal density distribution of the material to cause the difference in the X-ray absorption characteristics to realize the imaging of the internal structure of the object. It can obtain good imaging contrast for structures or tissues with obvious density distribution differences, such as the interface of different densities of matter distribution, metal knives in luggage, and the distribution of human bones and muscles.
  • materials or biological soft tissues composed of light elements with small differences in density distribution such as articular cartilage, breast, liver and other human soft tissues and polyethylene materials
  • light Z elements are the main components, and the absorption of X-rays is usually very high. weak, so their internal structures are almost impossible to see with traditional absorption contrast-based X-ray imaging techniques.
  • the X-ray phase contrast imaging technology uses the refraction of the imaging material to the X-ray to generate imaging contrast. For light-element substances, when X-rays penetrate the substance, the amount of phase change caused is more than 103 times the amount of absorption change. Therefore, X-ray phase contrast imaging technology has a very broad application research prospect in the future materials science and clinical medicine.
  • X-ray phase imaging techniques including X-ray crystal interferometer phase contrast imaging, X-ray diffraction enhanced phase contrast imaging, X-ray phase propagation phase contrast imaging, X-ray grating phase contrast imaging Contrast imaging, etc.
  • the three imaging methods of crystal interferometer method, diffraction enhancement method and phase propagation method have high requirements on the coherence of X-ray light source, and generally require a synchrotron radiation source as a radiation source, so it is difficult to obtain a wide range of commercial applications.
  • the grating phase contrast imaging method uses the principles of the Talbot effect and the Lau effect, which reduces the requirements for the coherence of the light source, and can use a common X-ray light source to achieve phase contrast imaging.
  • this imaging device also has three micron-scale gratings, two of which must be absorption gratings, and one of the absorption gratings usually needs to be phase-stepped during the imaging process to be successfully extracted. phase image.
  • This not only increases the complexity and cost of the imaging device, but more importantly, the existence of the absorption grating greatly reduces the utilization efficiency of the radiation dose.
  • the phase stepping process also greatly increases the imaging time and makes it difficult to compare with CT scanning technology. fusion.
  • the invention provides an X-ray phase contrast image extraction method, device, terminal and storage medium, so as to solve the problems that the existing X-ray imaging relies on gratings, the process is complex, the cost is high and the radiation dose is large.
  • the present invention provides an X-ray phase contrast image extraction method, which includes: constructing a grating-free Talbot-Lau phase contrast imaging device based on an X-ray light source, a detector, and an object to be measured; adjusting the X-ray The working voltage and working current of the light source, and the projection images of the object to be measured at high energy and low energy are collected, which are recorded as X-ray dual energy absorption contrast images; the X-ray dual energy absorption contrast images are input into the trained deep neural network model , and output the phase contrast image.
  • the deep neural network model is based on the sample phase stepping projection image collected by the Talbot-Lau phase contrast imaging device with grating structure and the dual energy collected by the Talbot-Lau phase contrast imaging device without grating structure. Absorption contrast images are trained.
  • the operating voltage and operating current of the X-ray light source are adjusted, and the projection images of the object to be measured at high energy and low energy are collected, which are recorded as X-ray dual energy absorption contrast images, including: adjusting the low energy of the X-ray light source.
  • Working voltage and working current collect the projected image of the object to be measured, and record it as the X-ray low-energy absorption contrast image; adjust the high-energy working voltage and current of the X-ray light source, collect the projected image of the object to be measured, and record it as X-ray high-energy absorption Contrast image.
  • the present invention also includes: training a deep neural network model, and the step of training the deep neural network model includes: constructing an X-ray light source, a detector, a sample object, a source grating, a phase grating and an absorption grating with a grating structure.
  • Talbot-Lau phase contrast imaging device use a Talbot-Lau phase contrast imaging device with a grating structure to collect multiple sample phase step projection images of different sample objects or sample objects at different angles, and collect a sample phase step at each
  • remove all the gratings in the Talbot-Lau phase contrast imaging device with grating structure and then collect the sample X-ray dual energy absorption contrast image, the corresponding sample phase stepping projection image and the sample X-ray dual energy absorption contrast image.
  • the absorption contrast images form a set of training data; the sample X-ray dual-energy absorption contrast images are input into the initial deep neural network model, and the output results are obtained, and then the output results are compared with the sample phase stepping projection images, and reversed. Update the deep neural network model to the propagation until the training of the deep neural network model is completed.
  • a Talbot-Lau phase contrast imaging device with a grating structure is used to collect multiple sample phase stepping projection images of different sample objects or sample objects at different angles, and each sample phase stepping projection image is collected.
  • the contrast images form a set of training data, including: using a Talbot-Lau phase contrast imaging device with a grating structure to collect the sample phase stepping projection image of the sample object; calculating the average gray value of the sample phase stepping projection image; removing The source grating, the phase grating and the absorption grating in the Talbot-Lau phase contrast imaging device with grating structure, the Talbot-Lau phase contrast imaging device without grating structure is obtained; the low-energy working voltage of the X-ray light
  • the sample X-ray low energy absorption contrast image, the sample X-ray high energy absorption contrast image and the phase contrast image form a set of training data; replace the sample Items or adjust the angle of the sample items, and repeat the above sample collection process to obtain multiple sets of training data.
  • the method further includes: adjusting the Talbot-Lau phase contrast imaging device with a grating structure. The distance between the source grating, phase grating, and absorption grating is performed, and then the sample acquisition process is performed to obtain new sets of training data.
  • the deep neural network model includes 10 convolutional layers, the first layer is the input layer, including the input end of the sample X-ray low energy absorption contrast image and the input end of the sample X-ray high energy absorption contrast image, and the last layer is the input layer.
  • the layer is the output, and the remaining 8 convolutional layers include 4 image mode conversion structures in series.
  • the image mode conversion structure includes three parallel channels.
  • the size of the two concatenated convolution kernels in the first channel is 1*1
  • the size of the convolution kernels of the two convolution layers in the second channel is 1*3 and 3*1
  • the convolution kernel sizes of the two concatenated convolutional layers in the third channel are 1*5 and 5*1, respectively.
  • the present invention also provides an X-ray phase contrast image extraction device, which includes: a building module for building a grating-free Talbot-Lau phase contrast based on an X-ray light source, a detector, and an object to be measured Imaging device; acquisition module, used to adjust the working voltage and current of the X-ray light source, and collect the projection images of the object to be measured at high energy and low energy, recorded as X-ray dual energy absorption contrast image; calculation module, used to convert X-ray The ray dual energy absorption contrast image is input to the trained deep neural network model, and the output phase contrast image is obtained.
  • the deep neural network model is based on the sample phase stepping projection image and It is obtained by training the dual energy absorption contrast images collected by the Talbot-Lau phase contrast imaging device without grating structure.
  • the present invention also provides a terminal, the terminal includes a processor and a memory coupled to the processor, and program instructions are stored in the memory, and when the program instructions are executed by the processor, the processor executes as described above.
  • the present invention also provides a storage medium storing a program file capable of implementing the X-ray phase contrast image extraction method as described above.
  • the X-ray phase contrast image extraction method of the present invention collects the X-ray dual energy absorption contrast image of the object to be measured by adopting a Talbot-Lau phase contrast imaging device without a grating structure, and then inputs it into the
  • the deep neural network model is based on the sample phase stepping projection image collected by the Talbot-Lau phase contrast imaging device with grating structure and the Talbot-Lau phase contrast imaging device without grating structure. Therefore, the phase contrast image can be obtained by inputting the X-ray dual energy absorption contrast image into the deep neural network model, and the phase contrast image can be extracted without relying on the grating.
  • phase contrast image can be completed only with a smaller dose of radiation, which reduces the radiation of the object to be measured.
  • FIG. 1 is a schematic structural diagram of an embodiment of a brushless motor drive system of the present invention
  • FIG. 2 is a schematic flowchart of the first embodiment of the brushless motor driving method of the present invention
  • FIG. 3 is a schematic flowchart of a second embodiment of the brushless motor driving method of the present invention.
  • FIG. 4 is a schematic flowchart of a third embodiment of the brushless motor driving method of the present invention.
  • FIG. 5 is a schematic flowchart of the fourth embodiment of the brushless motor driving method of the present invention.
  • first”, “second” and “third” in the present invention are only used for description purposes, and should not be understood as indicating or implying relative importance or implying the number of indicated technical features. Thus, a feature defined as “first”, “second”, “third” may expressly or implicitly include at least one of that feature.
  • "a plurality of” means at least two, such as two, three, etc., unless otherwise expressly and specifically defined. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship between various components under a certain posture (as shown in the accompanying drawings).
  • FIG. 1 is a schematic flowchart of an X-ray phase contrast image extraction method according to an embodiment of the present application. It should be noted that, if there is substantially the same result, the method of the present application is not limited to the sequence of the processes shown in FIG. 1 . As shown in Figure 1, the method includes the steps:
  • Step S1 constructing a Talbot-Lau phase contrast imaging device without a grating structure based on the X-ray light source, the detector, and the object to be measured.
  • the existing Talbot-Lau phase contrast imaging devices are all composed of an X-ray light source, a detector, an object to be measured, a source grating, a phase grating and an absorption grating.
  • the X-ray light source, the detector, and the object to be measured are used to construct a Talbot-Lau phase contrast imaging device without a grating structure, and the settings of other parameters are the same as those of the existing Talbot-Lau phase contrast imaging device.
  • Step S2 Adjust the working voltage and working current of the X-ray light source, and collect the projection images of the object to be measured at high energy and low energy, which are recorded as X-ray dual-energy absorption contrast images.
  • step S2 the X-ray dual energy absorption contrast image is obtained by using the working voltage and working current of the conditional X imaging, and collecting the projection image of the object to be measured under high energy conditions and the projection image under low energy conditions.
  • this step S2 specifically includes:
  • Adjust the low-energy working voltage and working current of the X-ray light source collect the projection image of the object to be measured, and record it as the X-ray low-energy absorption contrast image.
  • Adjust the high-energy working voltage and working current of the X-ray light source collect the projection image of the object to be measured, and record it as the X-ray high-energy absorption contrast image.
  • Step S3 Input the X-ray dual energy absorption contrast image into the trained deep neural network model, and output a phase contrast image.
  • the deep neural network model is based on the sample phase collected by the Talbot-Lau phase contrast imaging device with grating structure.
  • the step projection image and the dual energy absorption contrast image acquired by the Talbot-Lau phase contrast imaging device without grating structure are obtained by training.
  • step S3 after the X-ray high-energy absorption contrast image is acquired, it is input into the trained deep neural network model for calculation, and the deep neural network model is based on the Talbot-Lau phase contrast imaging device with grating structure
  • the collected sample phase stepping projection image and the dual energy absorption contrast image collected by the Talbot-Lau phase contrast imaging device without grating structure are obtained by training, so that the X-ray high energy absorption contrast image is input into the deep neural network model, namely The corresponding phase contrast image can be obtained.
  • the X-ray dual energy absorption contrast image and the phase contrast image satisfy a certain correspondence, and the correspondence includes that the images of the two modes satisfy the one-to-one correspondence in spatial angle and scale, and the two-mode image.
  • the images of different modes have the same noise level.
  • This correspondence is to minimize the interference of various types of noise on subsequent model training. Therefore, the X-ray dual energy absorption lining can be learned through a deep neural network model.
  • the deep neural network model can obtain the corresponding phase through the new X-ray dual energy absorption contrast image that is not used for network training. Contrast image. Therefore, as shown in Figure 2, before using the deep neural network model, the deep neural network model needs to be trained, which specifically includes the following steps:
  • Step S10 constructing a Talbot-Lau phase contrast imaging device with a grating structure based on the X-ray light source, the detector, the sample object, the source grating, the phase grating and the absorption grating.
  • step S10 the Talbot-Lau phase contrast imaging device with the grating structure is the same as the existing Talbot-Lau phase contrast imaging device, and details are not described herein again.
  • Step S11 Use a Talbot-Lau phase contrast imaging device with a grating structure to collect multiple sample phase-stepping projection images of different sample objects or sample objects at different angles, and each time a sample phase-stepping projection image is collected, Remove all gratings in the Talbot-Lau phase contrast imaging device with grating structure, and then collect the X-ray dual energy absorption contrast image of the sample, which consists of the corresponding sample phase stepping projection image and the sample X-ray dual energy absorption contrast image A set of training data.
  • step S11 first use a Talbot-Lau phase contrast imaging device with a grating structure to collect multiple sample phase stepping projection images of different sample objects or sample objects at different angles, and then use the sample phase stepping projection images as phase information
  • the phase contrast image can be obtained, and the phase contrast image can be used as the real result of training the deep neural network.
  • a Talbot-Lau phase contrast imaging device without grating structure is obtained to collect the sample X-ray dual energy absorption contrast image of the sample object, It is used as the input for training the deep neural network, that is, the sample X-ray dual energy absorption contrast image and the sample phase stepping projection image constitute a set of training data for the deep neural network.
  • step S11 includes:
  • the average gray value is obtained by acquiring the gray value of each pixel in the image, and then calculating the average value of the gray values of all pixels.
  • the preset range is preset, after adjusting the low-energy working voltage of the X-ray light source, when adjusting the working current of the X-ray light source, the average value of the pixel readings of the detector is read in real time, and compared with the average gray value, When the difference between the two is within a preset range, a projection image of the sample object is collected to obtain a sample X-ray low energy absorption contrast image.
  • the average value of the pixel readings of the detector is read in real time, and compared with the average gray value, when the difference between the two is within
  • the projection image of the sample object is collected to obtain the sample X-ray high energy absorption contrast image.
  • the sample X-ray low-energy absorption contrast image, the sample X-ray high-energy absorption contrast image and the phase contrast image form a set of training data.
  • the method further includes:
  • the deep neural network model includes 10 convolutional layers, the first layer is the input layer, including the input end of the sample X-ray low energy absorption contrast image and the input end of the sample X-ray high energy absorption contrast image, and the last layer is At the output, the remaining 8 convolutional layers consist of 4 image mode conversion structures in series.
  • the image mode conversion structure includes three parallel channels.
  • the size of the two concatenated convolution kernels in the first channel is 1*1
  • the size of the convolution kernels of the two convolutional layers in the second channel is 1*3 and 3*1
  • the convolution kernel sizes of the two concatenated convolutional layers in the third channel are 1*5 and 5*1, respectively.
  • the function of the first channel is to ensure that the resolution of the output image of the network is the same as that of the input image
  • the function of the second channel and the third channel is to reflect the operation between the adjacent pixels in the X-ray dual energy absorption contrast image of the sample, which may include phase information.
  • the deep neural network model in this embodiment After the deep neural network model in this embodiment is trained, it can be migrated to other X-ray imaging devices with the same structure type. Therefore, for multiple X-ray imaging devices with the same structure type, only one of the X-ray imaging devices needs to be installed. It is enough to build a Talbot-Lau phase contrast imaging device and complete the experimental acquisition of training image data.
  • the same structure type means that the characteristics of the light source and detector of the imaging device are the same, and the geometric parameters of the device (such as the distance between the gratings) are the same.
  • Step S12 Input the sample X-ray dual energy absorption contrast image into the initial deep neural network model to obtain the output result, then compare the output result with the sample phase stepping projection image, and backpropagate to update the deep neural network model , until the training of the deep neural network model is completed.
  • the X-ray phase contrast image extraction method of this embodiment uses a Talbot-Lau phase contrast imaging device without a grating structure to collect the X-ray dual energy absorption contrast image of the object to be measured, and then input it into the trained deep neural network
  • the deep neural network model is based on the sample phase stepping projection image collected by the Talbot-Lau phase contrast imaging device with grating structure and the dual energy absorption contrast image collected by the Talbot-Lau phase contrast imaging device without grating structure. Therefore, the phase contrast image can be obtained by inputting the X-ray dual energy absorption contrast image into the deep neural network model.
  • the phase contrast image can be obtained by grouping X-ray dual energy absorption contrast images, which makes the whole process of extracting the phase contrast image simpler and shortens the extraction time of the phase contrast image. It will not be lost by the grating, therefore, the phase contrast image can be extracted only with a smaller dose of radiation, which reduces the radiation of the object to be measured.
  • FIG. 3 is a schematic diagram of functional modules of an X-ray phase contrast image extraction apparatus according to an embodiment of the present application.
  • the X-ray phase contrast image extraction device 30 includes: a building module 31 , an acquisition module 32 and a calculation module 33 .
  • the building module 31 is used to build a Talbot-Lau phase contrast imaging device without a grating structure based on the X-ray light source, the detector, and the object to be measured;
  • the acquisition module 32 is used to adjust the working voltage and working current of the X-ray light source, and collect the projection images of the object to be measured at high energy and low energy, which are recorded as X-ray dual energy absorption contrast images;
  • the calculation module 33 is used to input the X-ray dual energy absorption contrast image into the trained deep neural network model, and output the phase contrast image, and the deep neural network model is collected according to the Talbot-Lau phase contrast imaging device with a grating structure.
  • the phase stepping projection image of the sample and the dual energy absorption contrast image collected by the Talbot-Lau phase contrast imaging device without grating structure are obtained by training.
  • the acquisition module 32 adjusts the operating voltage and operating current of the X-ray light source, and collects the projection images of the object to be measured at high energy and low energy, which are recorded as X-ray dual energy absorption contrast images.
  • the operation can also be: adjusting the X-ray The low-energy working voltage and working current of the light source, collect the projected image of the object to be measured, and record it as the X-ray low-energy absorption contrast image; adjust the high-energy operating voltage and operating current of the X-ray light source, and collect the projected image of the object to be measured, marked as X Ray high energy absorption contrast image.
  • the X-ray phase contrast image extraction device 30 further includes a training module 34, the training module 34 is used to train the deep neural network model, and the operation of the training module 34 to train the deep neural network model may be: based on the X-ray Light source, detector, sample object, source grating, phase grating and absorption grating construct a Talbot-Lau phase contrast imaging device with grating structure; use the Talbot-Lau phase contrast imaging device with grating structure to collect different sample objects or sample objects Take multiple sample phase-stepping projection images at different angles, and when collecting a sample phase-stepping projection image, remove all gratings in the Talbot-Lau phase contrast imaging device with grating structure, and then collect sample X
  • the ray dual energy absorption contrast image, the corresponding sample phase step projection image and the sample X-ray dual energy absorption contrast image form a set of training data; the sample X-ray dual energy absorption contrast image is input to the initial deep neural network model , the output results are obtained, and then the output results are
  • the training module 34 uses a Talbot-Lau phase contrast imaging device with a grating structure to collect a plurality of sample phase stepping projection images of different sample objects or sample objects at different angles, and collects a sample phase stepping projection image for each sample.
  • the operation of forming a set of training data from the contrast images can also be: using a Talbot-Lau phase contrast imaging device with a grating structure to collect a sample phase stepping projection image of the sample object; calculating the average gray value of the sample phase stepping projection image ;Remove the source grating, phase grating and absorption grating in the Talbot-Lau phase contrast imaging device with grating structure to obtain a Talbot-Lau phase contrast imaging device without grating structure; adjust the low-energy operating voltage of the X-ray light source, and then adjust The working current of the X-ray light source, until the difference between the average value of the pixel readings of the detector and the average gray value is within the preset range, collect the projection image of the sample object
  • the sample X-ray low energy absorption contrast image, the sample X-ray high energy absorption contrast image and the phase contrast image form a set of training data ; Replace the sample item or adjust the angle of the sample item, and repeat the above sample collection process to obtain multiple sets of training data.
  • the training module 34 replaces the sample item or adjusts the angle of the sample item, and repeats the above-mentioned sample collection process, so that after obtaining multiple sets of training data, it is also used to adjust the Talbot-Lau phase contrast imaging device with the grating structure.
  • the distance between the source grating, phase grating, and absorption grating is performed, and then the sample acquisition process is performed to obtain new sets of training data.
  • the deep neural network model includes 10 convolutional layers, the first layer is the input layer, including the input end of the sample X-ray low energy absorption contrast image and the input end of the sample X-ray high energy absorption contrast image, and the last layer is the output. At the end, the remaining 8 convolutional layers consist of 4 image mode conversion structures in series.
  • the image mode conversion structure includes three parallel channels, the size of the two convolution kernels in the first channel is 1*1, and the size of the convolution kernels of the two convolutional layers in the second channel is 1*3 respectively. and 3*1, the convolution kernel sizes of the two concatenated convolutional layers in the third channel are 1*5 and 5*1, respectively.
  • FIG. 4 is a schematic structural diagram of a terminal according to an embodiment of the present application.
  • the terminal 40 includes a processor 41 and a memory 42 coupled to the processor 41 .
  • the memory 42 stores program instructions, and when the program instructions are executed by the processor 41, the processor 41 executes the steps of the method for extracting an X-ray phase contrast image in the above embodiment.
  • the processor 41 may also be referred to as a CPU (Central Processing Unit, central processing unit).
  • the processor 41 may be an integrated circuit chip with signal processing capability.
  • the processor 41 may also be a general purpose processor, digital signal processor (DSP), application specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components .
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA field programmable gate array
  • a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • FIG. 5 is a schematic structural diagram of a storage medium according to an embodiment of the present application.
  • the storage medium of this embodiment of the present application stores a program file 51 capable of implementing all the above methods, wherein the program file 51 may be stored in the above-mentioned storage medium in the form of a software product, and includes several instructions to enable a computer device (which can be A personal computer, a server, or a network device, etc.) or a processor (processor) executes all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only). Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical discs and other media that can store program codes, or terminal devices such as computers, servers, mobile phones, and tablets.
  • the disclosed terminal, apparatus and method may be implemented in other manners.
  • the apparatus embodiments described above are only illustrative.
  • the division of units is only a logical function division.
  • there may be other division methods for example, multiple units or components may be combined or integrated. to another system, or some features can be ignored, or not implemented.
  • the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

Abstract

本发明公开了一种X射线相位衬度图像提取方法、装置、终端及存储介质，其中方法包括：基于X射线光源、探测器、待测物体构建无光栅结构Talbot-Lau相衬成像装置；调节X射线光源的工作电压和工作电流，采集待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像；将X射线双能吸收衬度图像输入至训练好的深度神经网络模型，输出得到相位衬度图像，深度神经网络模型根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到。通过上述方式，本发明能够在不需要光栅的情况下提取出物体的相衬图像，提取过程更为快捷。

Description

X射线相位衬度图像提取方法、装置、终端及存储介质 技术领域
本发明涉及X射线成像技术领域，特别是涉及一种X射线相位衬度图像提取方法、装置、终端及存储介质。
 背景技术
传统的X射线吸收衬度成像技术是利用物质内部密度分布的差异造成对Ｘ射线吸收特性的不同来实现对物体内部结构的成像。它对存在明显密度分布差异的结构或组织可以获得很好的成像衬度，如不同密度的物质分布的界面，行李箱中的金属刀具，人体骨骼和肌肉的分布等。但对于密度分布差异小的由轻元素组成的材料或生物软组织，如关节软骨、乳腺、肝脏等人体软组织及聚乙烯材料等都是以轻Z元素为主要成分，对X 射线的吸收通常都很弱，所以几乎无法通过传统的基于吸收衬度的X射线成像技术看到它们的内部结构。
而X射线相位衬度成像技术是利用成像物质对X射线的折射作用产生成像衬度的。对于由轻元素物质，X射线穿透物质时，引起的相位变化量是吸收变化量的103倍以上。因此X射线相位衬度成像技术在未来的材料科学及临床医学中具有非常广泛的应用研究前景。
目前，研究人员已提出了多种X射线相位成像技术实现方法，包括X射线晶体干涉仪相衬成像法、X射线衍射增强相衬成像法、X射线相位传播相衬成像法、X射线光栅相衬成像法等。其中晶体干涉仪法、衍射增强法及相位传播法这三种成像方法对X射线光源的相干性要求较高，一般需要同步辐射光源作为射线源，所以难以得到广泛的商业应用。而对于光栅相衬成像方法，其利用Talbot效应和Lau效应的原理，降低了对光源相干性的要求，可以使用普通的X射线光源实现相位衬度成像。但这种成像装置除了光源和平板探测器外，还有三块微米级的光栅，其中有两块光栅须是吸收光栅，且成像过程中通常需要对其中一块吸收光栅作相位步进运动才能成功提取出相位图像。这不仅增加了成像装置的复杂度和造价，更重要是吸收光栅的存在极大降低了辐射剂量的利用效率，同时，相位步进过程也极大地增加了成像时间并且使之难以和CT扫描技术融合。
 技术问题
本发明提供一种X射线相位衬度图像提取方法、装置、终端及存储介质，以解决现有的X成像依赖于光栅来实现，过程复杂、造价高且辐射剂量大的问题。
 技术解决方案
为解决上述技术问题，本发明提供了一种X射线相位衬度图像提取方法，其包括：基于X射线光源、探测器、待测物体构建无光栅结构Talbot-Lau相衬成像装置；调节X射线光源的工作电压和工作电流，采集待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像；将X射线双能吸收衬度图像输入至训练好的深度神经网络模型，输出得到相位衬度图像，深度神经网络模型根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到。
作为本发明的进一步改进，调节X射线光源的工作电压和工作电流，采集待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像，包括：调节X射线光源的低能工作电压和工作电流，采集待测物体的投影图像，记为X射线低能吸收衬度图像；调节X射线光源的高能工作电压和工作电流，采集待测物体的投影图像，记为X射线高能吸收衬度图像。
作为本发明的进一步改进，还包括：训练深度神经网络模型，训练深度神经网络模型的步骤包括：基于X射线光源、探测器、样本物体、源光栅、相位光栅和吸收光栅构建带有光栅结构的Talbot-Lau相衬成像装置；利用带有光栅结构的Talbot-Lau相衬成像装置采集不同样本物体或样本物体在不同角度的多张样本相位步进投影图像，并在每采集一张样本相位步进投影图像时，移除带有光栅结构的Talbot-Lau相衬成像装置中的所有光栅，再采集样本X射线双能吸收衬度图像，对应的样本相位步进投影图像和样本X射线双能吸收衬度图像组成一组训练数据；将样本X射线双能吸收衬度图像输入至初始的深度神经网络模型中，得到输出结果，再将输出结果与样本相位步进投影图像进行比较，并反向传播更新深度神经网络模型，直至深度神经网络模型训练完成。
作为本发明的进一步改进，利用带有光栅结构的Talbot-Lau相衬成像装置采集不同样本物体或样本物体在不同角度的多张样本相位步进投影图像，并在每采集一张样本相位步进投影图像时，移除带有光栅结构的Talbot-Lau相衬成像装置中的所有光栅，再采集样本X射线双能吸收衬度图像，对应的样本相位步进投影图像和样本X射线双能吸收衬度图像组成一组训练数据，包括：利用带有光栅结构的Talbot-Lau相衬成像装置采集样本物体的样本相位步进投影图像；计算样本相位步进投影图像的平均灰度值；移除带有光栅结构的Talbot-Lau相衬成像装置中的源光栅、相位光栅和吸收光栅，得到无光栅结构的Talbot-Lau相衬成像装置；调节X射线光源的低能工作电压，再调节X射线光源的工作电流，直至探测器像素读数的平均值与平均灰度值的差值正在预设范围内时，采集样本物体的投影图像，得到样本X射线低能吸收衬度图像；调节X射线光源的高能工作电压，再调节X射线光源的工作电流，直至探测器像素读数的平均值与平均灰度值的差值正在预设范围内时，采集样本物体的投影图像，得到样本X射线高能吸收衬度图像；对样本相位步进投影图像做相位信息的分离提取，得到相衬图像，样本X射线低能吸收衬度图像、样本X射线高能吸收衬度图像和相衬图像组成一组训练数据；更换样本物品或者调整样本物品的角度，并重复执行上述样本采集过程，从而得到多组训练数据。
作为本发明的进一步改进，更换样本物品或者调整样本物品的角度，并重复执行上述样本采集过程，从而得到多组训练数据之后，还包括：调整带有光栅结构的Talbot-Lau相衬成像装置的源光栅、相位光栅、吸收光栅之间的距离，再执行样本采集过程，得到新的多组训练数据。
作为本发明的进一步改进，深度神经网络模型包括10个卷积层，第一层为输入层，包括样本X射线低能吸收衬度图像输入端和样本X射线高能吸收衬度图像输入端，最后一层为输出端，其余8个卷积层包括串联的4个图像模式转换结构。
作为本发明的进一步改进，图像模式转换结构包括三个并行通道，第一通道中两个串联卷积核大小均为1*1，第二通道中两个卷积层的卷积核大小分别为1*3和3*1，第三通道中两个串联卷积层的卷积核大小分别为1*5和5*1。
为解决上述技术问题，本发明还提供了一种X射线相位衬度图像提取装置，其包括：构建模块，用于基于X射线光源、探测器、待测物体构建无光栅结构Talbot-Lau相衬成像装置；采集模块，用于调节X射线光源的工作电压和工作电流，采集待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像；计算模块，用于将X射线双能吸收衬度图像输入至训练好的深度神经网络模型，输出得到相位衬度图像，深度神经网络模型根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到。
为解决上述技术问题，本发明还提供了一种终端，终端包括处理器、与处理器耦接的存储器，存储器中存储有程序指令，程序指令被处理器执行时，使得处理器执行如上述中任一项权利要求的X射线相位衬度图像提取方法的步骤。
为解决上述技术问题，本发明还提供了一种存储介质，存储有能够实现如上述中任一项的X射线相位衬度图像提取方法的程序文件。
 有益效果
本发明的有益效果是：本发明的X射线相位衬度图像提取方法通过采用无光栅结构的Talbot-Lau相衬成像装置采集待测物体的X射线双能吸收衬度图像，再将其输入至训练好的深度神经网络模型中，而由于深度神经网络模型是根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到，因此，将X射线双能吸收衬度图像输入至深度神经网络模型即可得到相位衬度图像，其不依懒于光栅也可实现相位衬度图像的提取，并且只需要提取一组X射线双能吸收衬度图像即可得到相位衬度图像，使得整个提取相位衬度图像的过程更为简单，缩短了相位衬度图像的提取时间，而且因为去除了光栅，从而使得X射线不会被光栅损耗，因此，只需要更少剂量的辐射也可完成对相位衬度图像的提取，降低了对待测物体的辐射。
 附图说明
图1是本发明无刷电机驱动系统一个实施例的结构示意图；
图2是本发明无刷电机驱动方法第一个实施例的流程示意图；
图3是本发明无刷电机驱动方法第二个实施例的流程示意图；
图4是本发明无刷电机驱动方法第三个实施例的流程示意图；
图5是本发明无刷电机驱动方法第四个实施例的流程示意图。
 本发明的最佳实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本发明的一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
本发明中的术语“第一”、“第二”、“第三”仅用于描述目的，而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”、“第三”的特征可以明示或者隐含地包括至少一个该特征。本发明的描述中，“多个”的含义是至少两个，例如两个，三个等，除非另有明确具体的限定。本发明实施例中所有方向性指示（诸如上、下、左、右、前、后……）仅用于解释在某一特定姿态（如附图所示）下各部件之间的相对位置关系、运动情况等，如果该特定姿态发生改变时，则该方向性指示也相应地随之改变。此外，术语“包括”和“具有”以及它们任何变形，意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、系统、产品或设备没有限定于已列出的步骤或单元，而是可选地还包括没有列出的步骤或单元，或可选地还包括对于这些过程、方法、产品或设备固有的其它步骤或单元。
在本文中提及“实施例”意味着，结合实施例描述的特定特征、结构或特性可以包含在本发明的至少一个实施例中。在说明书中的各个位置出现该短语并不一定均是指相同的实施例，也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是，本文所描述的实施例可以与其它实施例相结合。
图1是本申请实施例的X射线相位衬度图像提取方法的流程示意图。需注意的是，若有实质上相同的结果，本申请的方法并不以图1所示的流程顺序为限。如图1所示，该方法包括步骤：
步骤S1：基于X射线光源、探测器、待测物体构建无光栅结构Talbot-Lau相衬成像装置。
具体地，目前，现有的Talbot-Lau相衬成像装置均是由X射线光源、探测器、待测物体、源光栅、相位光栅和吸收光栅组成。在步骤S1中，则是利用X射线光源、探测器、待测物体构建无光栅结构Talbot-Lau相衬成像装置，其余参数的设置与现有的Talbot-Lau相衬成像装置相同。
步骤S2：调节X射线光源的工作电压和工作电流，采集待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像。
在步骤S2中，通过条件X摄像的工作电压和工作电流，并且采集待测物体在高能条件下的投影图像和在低能条件下的投影图像，从而得到X射线双能吸收衬度图像。
具体地，该步骤S2具体包括：
1、调节X射线光源的低能工作电压和工作电流，采集待测物体的投影图像，记为X射线低能吸收衬度图像。
2、调节X射线光源的高能工作电压和工作电流，采集待测物体的投影图像，记为X射线高能吸收衬度图像。
步骤S3：将X射线双能吸收衬度图像输入至训练好的深度神经网络模型，输出得到相位衬度图像，深度神经网络模型根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到。
在步骤S3中，在获取到X射线高能吸收衬度图像之后，将其输入至训练好的深度神经网络模型中进行计算，而深度神经网络模型是根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到的，从而将X射线高能吸收衬度图像输入至深度神经网络模型即可得到相应的相位衬度图像。
进一步的，需要理解的是，X射线双能吸收衬度图像与相位衬度图像满足一定的对应关系，该对应关系包括两种模式的图像之间满足空间角度及尺度上的一一对应、两种模式的图像之间具有相同的噪声水平，这种对应关系是为了最大限度地降低各种不同类型的噪声对后续模型训练的干扰，因此，可通过深度神经网络模型学习X射线双能吸收衬度图像与相位衬度图像两种模式之间的内在的物理映射关系，经过学习训练之后，深度神经网络模型能够通过新的未用于网络训练的X射线双能吸收衬度图像得到对应的相位衬度图像。因此，如图2所示，在使用该深度神经网络模型之前，需要训练深度神经网络模型，具体包括以下步骤：
步骤S10：基于X射线光源、探测器、样本物体、源光栅、相位光栅和吸收光栅构建带有光栅结构的Talbot-Lau相衬成像装置。
在步骤S10中，该带有光栅结构的Talbot-Lau相衬成像装置与现有的Talbot-Lau相衬成像装置相同，此处不再赘述。
步骤S11：利用带有光栅结构的Talbot-Lau相衬成像装置采集不同样本物体或样本物体在不同角度的多张样本相位步进投影图像，并在每采集一张样本相位步进投影图像时，移除带有光栅结构的Talbot-Lau相衬成像装置中的所有光栅，再采集样本X射线双能吸收衬度图像，对应的样本相位步进投影图像和样本X射线双能吸收衬度图像组成一组训练数据。
在步骤S11中，首先利用带有光栅结构的Talbot-Lau相衬成像装置采集不同样本物体或样本物体在不同角度的多张样本相位步进投影图像，再将样本相位步进投影图像做相位信息的分离提取，即可得到相衬图像，以该相衬图像作为训练深度神经网络的真实结果。然后，再通过移除带有光栅结构的Talbot-Lau相衬成像装置中的所有光栅，得到无光栅结构的Talbot-Lau相衬成像装置来采集样本物体的样本X射线双能吸收衬度图像，将其作为训练深度神经网络的输入，即样本X射线双能吸收衬度图像与样本相位步进投影图像构成一组深度神经网络的训练数据。
具体地，步骤S11包括：
1、利用带有光栅结构的Talbot-Lau相衬成像装置采集样本物体的样本相位步进投影图像。
2、计算样本相位步进投影图像的平均灰度值。
具体地，通过获取图像中每一个像素的灰度值，然后计算所有像素的灰度值的平均值，从而得到平均灰度值。
3、移除带有光栅结构的Talbot-Lau相衬成像装置中的源光栅、相位光栅和吸收光栅，得到无光栅结构的Talbot-Lau相衬成像装置。
4、调节X射线光源的低能工作电压，再调节X射线光源的工作电流，直至探测器像素读数的平均值与平均灰度值的差值正在预设范围内时，采集样本物体的投影图像，得到样本X射线低能吸收衬度图像。
具体地，预设范围预先设定，在调节X射线光源的低能工作电压后，调节X射线光源的工作电流时，实时读取探测器像素读数的平均值，并与平均灰度值进行比较，当两者的差值在预设范围内时，采集样本物体的投影图像，得到样本X射线低能吸收衬度图像。
5、调节X射线光源的高能工作电压，再调节X射线光源的工作电流，直至探测器像素读数的平均值与平均灰度值的差值正在预设范围内时，采集样本物体的投影图像，得到样本X射线高能吸收衬度图像。
具体地，在调节X射线光源的高能工作电压后，调节X射线光源的工作电流时，实时读取探测器像素读数的平均值，并与平均灰度值进行比较，当两者的差值在预设范围内时，采集样本物体的投影图像，得到样本X射线高能吸收衬度图像。
6、对样本相位步进投影图像做相位信息的分离提取，得到相衬图像，样本X射线低能吸收衬度图像、样本X射线高能吸收衬度图像和相衬图像组成一组训练数据。
7、更换样本物品或者调整样本物品的角度，并重复执行上述样本采集过程，从而得到多组训练数据。
具体地，在获取当前样本物体在当前角度的样本数据之后，更换样本物体或调整样本物体的角度，然后再次执行上述1-6的样本采集过程，从而得到更多的训练数据。
进一步的，在更换样本物品或者调整样本物品的角度，并重复执行上述样本采集过程，从而得到多组训练数据之后，还包括：
8、调整带有光栅结构的Talbot-Lau相衬成像装置的源光栅、相位光栅、吸收光栅之间的距离，再执行样本采集过程，得到新的多组训练数据。
具体地，通过调整带有光栅结构的Talbot-Lau相衬成像装置的源光栅、相位光栅、吸收光栅之间的距离后，采集不同情形下的数据进行训练，从而提高深度卷积神经网络模型的迁移效果。
需要说明的是，深度神经网络模型包括10个卷积层，第一层为输入层，包括样本X射线低能吸收衬度图像输入端和样本X射线高能吸收衬度图像输入端，最后一层为输出端，其余8个卷积层包括串联的4个图像模式转换结构。
进一步的，图像模式转换结构包括三个并行通道，第一通道中两个串联卷积核大小均为1*1，第二通道中两个卷积层的卷积核大小分别为1*3和3*1，第三通道中两个串联卷积层的卷积核大小分别为1*5和5*1。其中，第一通道的作用是保证网络输出图像的分辨率与输入图像相同，第二通道和第三通道的作用是反映样本X射线双能吸收衬度图像中邻域像素相互间的运算可能包含的相位信息。
本实施例中的深度神经网络模型在训练完成之后，可以迁移至其他具有相同结构类型的X射线成像设备中，因此对于具有相同结构类型的多台X射线成像设备，只需在其中一台设备中搭建Talbot-Lau相衬成像装置并完成训练图像数据的实验采集即可。其中，相同结构类型是指成像设备的光源及探测器特性相同，装置的几何参数（如光栅之间的距离）相同。
步骤S12：将样本X射线双能吸收衬度图像输入至初始的深度神经网络模型中，得到输出结果，再将输出结果与样本相位步进投影图像进行比较，并反向传播更新深度神经网络模型，直至深度神经网络模型训练完成。
本实施例的X射线相位衬度图像提取方法通过采用无光栅结构的Talbot-Lau相衬成像装置采集待测物体的X射线双能吸收衬度图像，再将其输入至训练好的深度神经网络模型中，而由于深度神经网络模型是根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到，因此，将X射线双能吸收衬度图像输入至深度神经网络模型即可得到相位衬度图像，其不依懒于光栅也可实现相位衬度图像的提取，并且只需要提取一组X射线双能吸收衬度图像即可得到相位衬度图像，使得整个提取相位衬度图像的过程更为简单，缩短了相位衬度图像的提取时间，而且因为去除了光栅，从而使得X射线不会被光栅损耗，因此，只需要更少剂量的辐射也可完成对相位衬度图像的提取，降低了对待测物体的辐射。
图3是本申请实施例的X射线相位衬度图像提取装置的功能模块示意图。如图3所示，该X射线相位衬度图像提取装置30包括：构建模块31、采集模块32和计算模块33。
构建模块31，用于基于X射线光源、探测器、待测物体构建无光栅结构Talbot-Lau相衬成像装置；
采集模块32，用于调节X射线光源的工作电压和工作电流，采集待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像；
计算模块33，用于将X射线双能吸收衬度图像输入至训练好的深度神经网络模型，输出得到相位衬度图像，深度神经网络模型根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到。
可选地，采集模块32调节X射线光源的工作电压和工作电流，采集待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像的操作还可以为：调节X射线光源的低能工作电压和工作电流，采集待测物体的投影图像，记为X射线低能吸收衬度图像；调节X射线光源的高能工作电压和工作电流，采集待测物体的投影图像，记为X射线高能吸收衬度图像。
可选地，该X射线相位衬度图像提取装置30还包括训练模块34，该训练模块34用于训练深度神经网路模型，训练模块34训练深度神经网路模型的操作可以为：基于X射线光源、探测器、样本物体、源光栅、相位光栅和吸收光栅构建带有光栅结构的Talbot-Lau相衬成像装置；利用带有光栅结构的Talbot-Lau相衬成像装置采集不同样本物体或样本物体在不同角度的多张样本相位步进投影图像，并在每采集一张样本相位步进投影图像时，移除带有光栅结构的Talbot-Lau相衬成像装置中的所有光栅，再采集样本X射线双能吸收衬度图像，对应的样本相位步进投影图像和样本X射线双能吸收衬度图像组成一组训练数据；将样本X射线双能吸收衬度图像输入至初始的深度神经网络模型中，得到输出结果，再将输出结果与样本相位步进投影图像进行比较，并反向传播更新深度神经网络模型，直至深度神经网络模型训练完成。
可选地，训练模块34利用带有光栅结构的Talbot-Lau相衬成像装置采集不同样本物体或样本物体在不同角度的多张样本相位步进投影图像，并在每采集一张样本相位步进投影图像时，移除带有光栅结构的Talbot-Lau相衬成像装置中的所有光栅，再采集样本X射线双能吸收衬度图像，对应的样本相位步进投影图像和样本X射线双能吸收衬度图像组成一组训练数据的操作还可以为：利用带有光栅结构的Talbot-Lau相衬成像装置采集样本物体的样本相位步进投影图像；计算样本相位步进投影图像的平均灰度值；移除带有光栅结构的Talbot-Lau相衬成像装置中的源光栅、相位光栅和吸收光栅，得到无光栅结构的Talbot-Lau相衬成像装置；调节X射线光源的低能工作电压，再调节X射线光源的工作电流，直至探测器像素读数的平均值与平均灰度值的差值正在预设范围内时，采集样本物体的投影图像，得到样本X射线低能吸收衬度图像；调节X射线光源的高能工作电压，再调节X射线光源的工作电流，直至探测器像素读数的平均值与平均灰度值的差值正在预设范围内时，采集样本物体的投影图像，得到样本X射线高能吸收衬度图像；对样本相位步进投影图像做相位信息的分离提取，得到相衬图像，样本X射线低能吸收衬度图像、样本X射线高能吸收衬度图像和相衬图像组成一组训练数据；更换样本物品或者调整样本物品的角度，并重复执行上述样本采集过程，从而得到多组训练数据。
可选地，训练模块34更换样本物品或者调整样本物品的角度，并重复执行上述样本采集过程，从而得到多组训练数据之后，还用于调整带有光栅结构的Talbot-Lau相衬成像装置的源光栅、相位光栅、吸收光栅之间的距离，再执行样本采集过程，得到新的多组训练数据。
可选地，深度神经网络模型包括10个卷积层，第一层为输入层，包括样本X射线低能吸收衬度图像输入端和样本X射线高能吸收衬度图像输入端，最后一层为输出端，其余8个卷积层包括串联的4个图像模式转换结构。
可选地，图像模式转换结构包括三个并行通道，第一通道中两个串联卷积核大小均为1*1，第二通道中两个卷积层的卷积核大小分别为1*3和3*1，第三通道中两个串联卷积层的卷积核大小分别为1*5和5*1。
关于上述实施例X射线相位衬度图像提取装置中各模块实现技术方案的其他细节，可参见上述实施例中的X射线相位衬度图像提取方法中的描述，此处不再赘述。
需要说明的是，本说明书中的各个实施例均采用递进的方式描述，每个实施例重点说明的都是与其他实施例的不同之处，各个实施例之间相同相似的部分互相参见即可。对于装置类实施例而言，由于其与方法实施例基本相似，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
请参阅图4，图4为本申请实施例的终端的结构示意图。如图4所示，该终端40包括处理器41及和处理器41耦接的存储器42。
存储器42存储有程序指令，程序指令被处理器41执行时，使得处理器41执行上述实施例中的X射线相位衬度图像提取方法的步骤。
其中，处理器41还可以称为CPU（Central Processing Unit，中央处理单元）。处理器41可能是一种集成电路芯片，具有信号的处理能力。处理器41还可以是通用处理器、数字信号处理器（DSP）、专用集成电路（ASIC）、现场可编程门阵列（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。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。
参阅图5，图5为本申请实施例的存储介质的结构示意图。本申请实施例的存储介质存储有能够实现上述所有方法的程序文件51，其中，该程序文件51可以以软件产品的形式存储在上述存储介质中，包括若干指令用以使得一台计算机设备（可以是个人计算机，服务器，或者网络设备等）或处理器（processor）执行本申请各个实施方式所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器（ROM，Read-Only Memory）、随机存取存储器（RAM，Random Access Memory）、磁碟或者光盘等各种可以存储程序代码的介质，或者是计算机、服务器、手机、平板等终端设备。
在本申请所提供的几个实施例中，应该理解到，所揭露的终端，装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，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信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
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。以上仅为本申请的实施方式，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请的专利保护范围内。

Claims (10)

  1. 一种X射线相位衬度图像提取方法，其特征在于，其包括：
    基于X射线光源、探测器、待测物体构建无光栅结构Talbot-Lau相衬成像装置；
    调节所述X射线光源的工作电压和工作电流，采集所述待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像；
    将所述X射线双能吸收衬度图像输入至训练好的深度神经网络模型，输出得到相位衬度图像，所述深度神经网络模型根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和所述无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到。
  2. 根据权利要求1所述的X射线相位衬度图像提取方法，其特征在于，所述调节所述X射线光源的工作电压和工作电流，采集所述待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像，包括：
    调节所述X射线光源的低能工作电压和工作电流，采集所述待测物体的投影图像，记为X射线低能吸收衬度图像；
    调节所述X射线光源的高能工作电压和工作电流，采集所述待测物体的投影图像，记为X射线高能吸收衬度图像。
  3. 根据权利要求1所述的X射线相位衬度图像提取方法，其特征在于，还包括：训练所述深度神经网络模型，所述训练所述深度神经网络模型的步骤包括：
    基于所述X射线光源、所述探测器、样本物体、源光栅、相位光栅和吸收光栅构建所述带有光栅结构的Talbot-Lau相衬成像装置；
    利用所述带有光栅结构的Talbot-Lau相衬成像装置采集不同所述样本物体或所述样本物体在不同角度的多张样本相位步进投影图像，并在每采集一张所述样本相位步进投影图像时，移除所述带有光栅结构的Talbot-Lau相衬成像装置中的所有光栅，再采集样本X射线双能吸收衬度图像，对应的所述样本相位步进投影图像和所述样本X射线双能吸收衬度图像组成一组训练数据；
    将所述样本X射线双能吸收衬度图像输入至初始的深度神经网络模型中，得到输出结果，再将所述输出结果与所述样本相位步进投影图像进行比较，并反向传播更新所述深度神经网络模型，直至所述深度神经网络模型训练完成。
  4. 根据权利要求3所述的X射线相位衬度图像提取方法，其特征在于，所述利用所述带有光栅结构的Talbot-Lau相衬成像装置采集不同所述样本物体或所述样本物体在不同角度的多张样本相位步进投影图像，并在每采集一张所述样本相位步进投影图像时，移除所述带有光栅结构的Talbot-Lau相衬成像装置中的所有光栅，再采集样本X射线双能吸收衬度图像，对应的所述样本相位步进投影图像和所述样本X射线双能吸收衬度图像组成一组训练数据，包括：
    利用所述带有光栅结构的Talbot-Lau相衬成像装置采集所述样本物体的样本相位步进投影图像；
    计算所述样本相位步进投影图像的平均灰度值；
    移除所述带有光栅结构的Talbot-Lau相衬成像装置中的所述源光栅、所述相位光栅和所述吸收光栅，得到所述无光栅结构的Talbot-Lau相衬成像装置；
    调节所述X射线光源的低能工作电压，再调节所述X射线光源的工作电流，直至所述探测器像素读数的平均值与所述平均灰度值的差值正在预设范围内时，采集所述样本物体的投影图像，得到样本X射线低能吸收衬度图像；
    调节所述X射线光源的高能工作电压，再调节所述X射线光源的工作电流，直至所述探测器像素读数的平均值与所述平均灰度值的差值正在所述预设范围内时，采集所述样本物体的投影图像，得到样本X射线高能吸收衬度图像；
    对所述样本相位步进投影图像做相位信息的分离提取，得到相衬图像，所述样本X射线低能吸收衬度图像、所述样本X射线高能吸收衬度图像和所述相衬图像组成一组训练数据；
    更换所述样本物品或者调整所述样本物品的角度，并重复执行上述样本采集过程，从而得到多组训练数据。
  5. 根据权利要求4所述的X射线相位衬度图像提取方法，其特征在于，所述更换所述样本物品或者调整所述样本物品的角度，并重复执行上述样本采集过程，从而得到多组训练数据之后，还包括：
    调整所述带有光栅结构的Talbot-Lau相衬成像装置的所述源光栅、所述相位光栅、所述吸收光栅之间的距离，再执行样本采集过程，得到新的多组训练数据。
  6. 根据权利要求3所述的X射线相位衬度图像提取方法，其特征在于，所述深度神经网络模型包括10个卷积层，第一层为输入层，包括样本X射线低能吸收衬度图像输入端和样本X射线高能吸收衬度图像输入端，最后一层为输出端，其余8个卷积层包括串联的4个图像模式转换结构。
  7. 根据权利要求6所述的X射线相位衬度图像提取方法，其特征在于，所述图像模式转换结构包括三个并行通道，第一通道中两个串联卷积核大小均为1*1，第二通道中两个卷积层的卷积核大小分别为1*3和3*1，第三通道中两个串联卷积层的卷积核大小分别为1*5和5*1。
  8. 一种X射线相位衬度图像提取装置，其特征在于，其包括：
    构建模块，用于基于X射线光源、探测器、待测物体构建无光栅结构Talbot-Lau相衬成像装置；
    采集模块，用于调节所述X射线光源的工作电压和工作电流，采集所述待测物体在高能和低能时的投影图像，记为X射线双能吸收衬度图像；
    计算模块，用于将所述X射线双能吸收衬度图像输入至训练好的深度神经网络模型，输出得到相位衬度图像，所述深度神经网络模型根据带光栅结构的Talbot-Lau相衬成像装置采集到的样本相位步进投影图像和所述无光栅结构的Talbot-Lau相衬成像装置采集到的双能吸收衬度图像训练得到。
  9. 一种终端，其特征在于，所述终端包括处理器、与所述处理器耦接的存储器，所述存储器中存储有程序指令，所述程序指令被所述处理器执行时，使得所述处理器执行如权利要求1-7中任一项权利要求所述的X射线相位衬度图像提取方法的步骤。
  10. 一种存储介质，其特征在于，存储有能够实现如权利要求1-7中任一项所述的X射线相位衬度图像提取方法的程序文件。
PCT/CN2020/139342 2020-12-10 2020-12-25 X射线相位衬度图像提取方法、装置、终端及存储介质 WO2022120983A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011437945.1 2020-12-10
CN202011437945.1A CN112568923B (zh) 2020-12-10 2020-12-10 X射线相位衬度图像提取方法、装置、终端及存储介质

Publications (1)

Publication Number Publication Date
WO2022120983A1 true WO2022120983A1 (zh) 2022-06-16

Family

ID=75130504

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/139342 WO2022120983A1 (zh) 2020-12-10 2020-12-25 X射线相位衬度图像提取方法、装置、终端及存储介质

Country Status (2)

Country Link
CN (1) CN112568923B (zh)
WO (1) WO2022120983A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116843596A (zh) * 2023-08-28 2023-10-03 浙江大学杭州国际科创中心 X射线光栅多模态图像自适应融合方法、系统及装置

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113367717B (zh) * 2021-05-26 2022-11-22 中国科学院深圳先进技术研究院 一种锥束x射线荧光成像方法、系统、终端以及存储介质
CN114018961B (zh) * 2021-11-03 2023-08-18 北京航空航天大学宁波创新研究院 基于深度学习的单步x射线光栅差分相位衬度成像方法及装置
CN114137002B (zh) * 2021-11-18 2023-07-14 北京航空航天大学 一种基于衬度间增强的低剂量x射线差分相位衬度成像方法
CN114113167A (zh) * 2021-11-26 2022-03-01 中国科学技术大学 一种X射线Talbot-Lau光栅相衬成像方法
CN116309237A (zh) * 2021-12-20 2023-06-23 中国科学院深圳先进技术研究院 深度高分辨相位信息提取方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN209894730U (zh) * 2018-12-29 2020-01-03 深圳大学 X射线双相位光栅相衬成像系统
WO2020039654A1 (ja) * 2018-08-22 2020-02-27 株式会社島津製作所 Ｘ線位相イメージング装置
CN111089871A (zh) * 2019-12-12 2020-05-01 中国科学院苏州生物医学工程技术研究所 X射线光栅相衬图像的相位信息分离方法及系统、储存介质、设备
CN111089869A (zh) * 2019-12-12 2020-05-01 中国科学院苏州生物医学工程技术研究所 多能探测器x射线相衬成像方法及系统、储存介质及设备

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2751442C (en) * 2009-02-05 2018-06-12 Paul Scherrer Institut Low dose single step grating based x-ray phase contrast imaging
CN104132953B (zh) * 2014-08-01 2017-03-29 中国科学技术大学 一种双能x射线相位衬度成像装置及其实现方法
CN107807139B (zh) * 2016-09-05 2020-04-24 天津工业大学 一种无步进装置的双能x射线相衬成像系统及其实现方法
WO2019056309A1 (en) * 2017-09-22 2019-03-28 Shenzhen United Imaging Healthcare Co., Ltd. METHOD AND SYSTEM FOR GENERATING A PHASE CONTRAST IMAGE
JP2019066283A (ja) * 2017-09-29 2019-04-25 株式会社島津製作所 放射線位相差撮影装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020039654A1 (ja) * 2018-08-22 2020-02-27 株式会社島津製作所 Ｘ線位相イメージング装置
CN209894730U (zh) * 2018-12-29 2020-01-03 深圳大学 X射线双相位光栅相衬成像系统
CN111089871A (zh) * 2019-12-12 2020-05-01 中国科学院苏州生物医学工程技术研究所 X射线光栅相衬图像的相位信息分离方法及系统、储存介质、设备
CN111089869A (zh) * 2019-12-12 2020-05-01 中国科学院苏州生物医学工程技术研究所 多能探测器x射线相衬成像方法及系统、储存介质及设备

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116843596A (zh) * 2023-08-28 2023-10-03 浙江大学杭州国际科创中心 X射线光栅多模态图像自适应融合方法、系统及装置
CN116843596B (zh) * 2023-08-28 2023-11-14 浙江大学杭州国际科创中心 X射线光栅多模态图像自适应融合方法、系统及装置

Also Published As

Publication number Publication date
CN112568923B (zh) 2022-08-19
CN112568923A (zh) 2021-03-30

Similar Documents

Publication Publication Date Title
WO2022120983A1 (zh) X射线相位衬度图像提取方法、装置、终端及存储介质
CN109493328B (zh) 医学图像显示方法、查看设备以及计算机设备
Yang et al. Deep learning for automated detection of cyst and tumors of the jaw in panoramic radiographs
Fukuda et al. Comparison of 3 deep learning neural networks for classifying the relationship between the mandibular third molar and the mandibular canal on panoramic radiographs
Young et al. A virtual trial framework for quantifying the detectability of masses in breast tomosynthesis projection data
Ali et al. IMNets: Deep learning using an incremental modular network synthesis approach for medical imaging applications
He et al. Automated model design and benchmarking of deep learning models for covid-19 detection with chest ct scans
US7596204B2 (en) Method and device for the iterative reconstruction of cardiac images
Wu et al. Spatial-temporal total variation regularization (STTVR) for 4D-CT reconstruction
Liu et al. Deep learning-based evaluation of the relationship between mandibular third molar and mandibular canal on CBCT
CN114067013A (zh) 用于经由单应重采样变换进行重投影和反投影的系统和方法
Lee et al. A deep learning-based scatter correction of simulated x-ray images
WO2023142781A1 (zh) 图像三维重建方法、装置、电子设备及存储介质
Arvind et al. Improvised light weight deep CNN based U-Net for the semantic segmentation of lungs from chest X-rays
Ma et al. An encoder-decoder network for direct image reconstruction on sinograms of a long axial field of view PET
US20210048941A1 (en) Method for providing an image base on a reconstructed image group and an apparatus using the same
Liang et al. Bone suppression on chest radiographs with adversarial learning
Hong et al. GAN‐LSTM‐3D: An efficient method for lung tumour 3D reconstruction enhanced by attention‐based LSTM
Li et al. X-ray photon-counting data correction through deep learning
Fan et al. Deep learning-based multi-class COVID-19 classification with x-ray images
Zhang et al. Advancing 3D medical image analysis with variable dimension transform based supervised 3D pre-training
WO2022247234A1 (zh) 一种锥束x射线荧光成像方法、系统、终端以及存储介质
Ji et al. ELCT-YOLO: an efficient one-stage model for automatic lung tumor detection based on CT images
Chen et al. Deep learning-based recognition of periodontitis and dental caries in dental x-ray images
US7260254B2 (en) Comparing images

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: 20964912

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: 20964912

Country of ref document: EP

Kind code of ref document: A1