Low-dose X-ray multi-contrast high-precision signal analysis method

Abstract

The invention provides a low-dose X-ray multi-contrast high-precision signal analysis method, and belongs to the technical field of deep learning and X-ray grating contrast imaging. Comprising the following steps: acquiring a grating stepping projection sequence under a low-dose condition; a convolutional neural network model is constructed, the model is composed of a contrast analysis module and a U-net, the contrast analysis module is used for extracting a contrast signal, and the U-net is used for further image enhancement and noise removal; training a convolutional neural network model; and obtaining high-quality absorption, phase and dark field projection by using a trained convolutional neural network model and utilizing a low-dose grating stepping projection sequence. According to the method, the physical information in the grating stepping projection sequence is extracted by using the convolutional neural network, so that the influence of noise introduced by low-dose X-rays on contrast signals can be reduced, the radiation dose borne by a sample is effectively reduced, the imaging quality is remarkably improved, and the application potential of X-ray grating differential phase contrast imaging is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of deep learning and X-ray grating contrast imaging, and in particular to a low-dose X-ray multi-contrast high-precision signal analysis method. Background Art

[0002] Compared with traditional absorption contrast imaging, grating interferometric imaging technology based on the Talbot-Lau effect can obtain absorption, phase and dark field contrast images of objects. Absorption images have good imaging effects for heavy elements such as metals, bones and teeth. Differential phase contrast imaging can obtain higher imaging contrast for light elements such as biological tissues. Dark field contrast works best for low-density tissue imaging, such as human lungs. Grating interferometric imaging technology based on the Talbot-Lau effect can use an ordinary X-ray source to generate coherent X-rays through the source grating G0, which pass through the phase grating G1 and propagate freely for a certain distance, then pass through the absorption grating G2 and are finally received by the detector. However, since the grating absorbs most of the X-rays and the absorption grating G2 needs to move multiple times (usually 4-8 times) to complete the extraction of absorption, phase and dark field contrast signals, the radiation dose in the imaging process is greatly increased.

[0003] Reducing the tube current is the simplest and most easily implemented measure to reduce the radiation dose. However, reducing the radiation dose will introduce quantum noise into the projection data. In addition, since the fast Fourier transform is required to perform phase extraction on the step projection data, as long as there is noise in a certain pixel point of the projection sequence image, the noise will be introduced into the phase contrast projection image of the corresponding pixel point, causing significant image degradation in the differential phase contrast projection. Not only will the local fine structure be damaged by the noise, but the overall structure will also be affected by the noise and disappear. Summary of the invention

[0004] In order to overcome the above technical problems, the present invention provides a low-dose X-ray multi-contrast high-precision signal analysis method, which reduces the radiation dose by reducing the tube current, and uses deep learning technology to realize contrast analysis and image enhancement of low-dose projection. The convolutional neural network can learn the mapping between low-dose projection images and high-quality projection images. While completing contrast extraction, it restores image details destroyed by noise, ensures the quality of contrast analysis under low-dose conditions, and promotes the application of grating differential phase contrast imaging technology in the clinical field.

[0005] In order to achieve the above object, the present invention adopts the following technical scheme:

[0006] A low-dose X-ray multi-contrast high-precision signal analysis method comprises the following steps:

[0007] Step S101, using an X-ray grating differential phase contrast imaging device based on the Talbot-Lau effect, obtaining at least three step projection sequences including samples and at least three step projection sequences including backgrounds under normal tube current conditions, and obtaining at least three step projection sequences including samples and at least three step projection sequences including backgrounds under low tube current conditions;

[0008] Step S102, constructing a convolutional neural network model, wherein the network model includes a contrast analysis module and a U-net module connected in series, wherein the contrast analysis module is used to extract contrast signals, and the U-net module is used to enhance and remove noise from the extracted contrast signals;

[0009] Step S103, using the at least three step projection sequences including the sample and the at least three step projection sequences including the background obtained under the low tube current condition as inputs of the network model, performing Fourier analysis on the at least three step projection sequences including the sample and the at least three step projection sequences including the background under the normal tube current condition, using the obtained absorption, phase and dark field images as labels, and training the convolutional neural network model;

[0010] Step S104: input at least three step projection sequences including samples and at least three step projection sequences including backgrounds under the low tube current condition into the trained network model to obtain high-precision absorption, phase and dark field contrast images.

[0011] The beneficial effects of the present invention are:

[0012] Compared with the existing X-ray grating contrast signal analysis method, the present invention can effectively reduce the radiation dose received by the sample and enhance the application potential of imaging technology based on the Talbot-Lau effect; at the same time, it fully exploits the continuous physical information in the step-by-step projection sequence, thereby effectively enhancing the contrast image affected by low-dose noise. It avoids the loss of details and local structure damage of the phase contrast image caused by the use of fast Fourier transform to analyze the low-dose step-by-step projection sequence, and improves the final imaging quality.

[0013] The low-dose X-ray multi-contrast high-precision signal analysis method of the present invention does not require modification of existing imaging devices, has strong scalability, and can effectively reduce radiation dose. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of a low-dose X-ray multi-contrast high-precision signal analysis method of the present invention;

[0015] Figure 2 This is the imaging principle diagram based on Talbot-Lau grating interferometer;

[0016] Figure 3 This is a 3D reconstruction of the mouse used in the experiment;

[0017] Figure 4 This is a structural diagram of a convolutional neural network used in an embodiment of the present invention;

[0018] Figure 5 The figures are comparative diagrams of the implementation effects of the present invention and the prior art, wherein (a) is the absorption, phase and dark field contrast diagram of the standard dose Fourier analysis, (b) is the absorption, phase and dark field contrast diagram of the one-eighth dose Fourier analysis, and (c) is the absorption, phase and dark field contrast diagram obtained by the present invention at one-eighth dose. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in this embodiment to clearly and completely describe the technical solution in this embodiment. Obviously, the described embodiment is only a part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] like Figure 1 As shown, this embodiment discloses a low-dose X-ray multi-contrast high-precision signal analysis method, comprising the following steps:

[0021] Step S101, obtaining a step projection sequence: using an X-ray grating differential phase contrast imaging device based on the Talbot-Lau effect, obtaining at least three step projection sequences including samples and at least three step projection sequences including backgrounds under normal tube current conditions, and obtaining at least three step projection sequences including samples and at least three step projection sequences including backgrounds under low tube current conditions, as follows:

[0022] like Figure 2 As shown, when X-rays penetrate an object, absorption attenuation, phase shift and scattering will occur. The X-ray grating differential phase contrast imaging device includes six parts: an X-ray source, a source grating G0, a sample, a phase grating G1, an absorption grating G2 and a detector;

[0023] Among them, the source grating G0 is used to generate coherent X-rays; the duty cycle of the phase grating G1 is 50%, and the X-rays are used to generate a phase shift of π; the duty cycle of the absorption grating G2 is 50%, a part of which can completely absorb X-rays, and the other part can transmit X-rays, and the detector is used to convert X-ray energy into digital images;

[0024] The corresponding relationship between the imaging experimental parameters of the X-ray grating differential phase contrast imaging device is as follows:

[0025] (1)

[0026] (2)

[0027] (3)

[0028] (4)

[0029] Wherein, d is the distance between the phase grating G1 and the absorption grating G2, m is an integer, representing m times the fractional Talbot distance, k = (L+d) / L is the magnification ratio, g1 is the period of the phase grating G1, λ is the wavelength of the X-ray, g2 is the period of the absorption grating G2, g0 is the period of the source grating G0, L is the distance between the source grating G0 and the phase grating G1, and s is the width of the source grating G0 that allows X-rays to pass through in each period;

[0030] During the imaging process, at least three projection sequences are obtained by stepping the absorption grating, and then the curve of light intensity changing with position is obtained by fast Fourier transform, which is called step curve. The grating step curve can be approximated as:

[0031] (5)

[0032] Among them, x is the independent variable, representing the position of the absorption grating, f(x) represents the X-ray intensity corresponding to the absorption grating at position x, a0 is the constant term, a1 is the amplitude, and φ1 is the initial phase. Then use the formula:

[0033] (6)

[0034] (7)

[0035] (8)

[0036] Where A represents the absorption contrast image, and represent the constant terms of the sample and background in formula (5), φ represents the phase contrast image, and represent the initial phases of the sample and background in formula (5), V represents the dark field contrast image, and v s and v r They are the dark field contrast images of the sample and the background, and Represent the amplitudes of the sample and background in formula (5) respectively.

[0037] The data were acquired by stepping the absorption grating three times under low tube current and normal tube current conditions, respectively, to obtain four projection images of the sample and background and the corresponding high-quality absorption, phase and dark-field contrast projections. Figure 3 This is a 3D reconstruction of the mouse used in the experiment.

[0038] Step S102, constructing a convolutional neural network model: the network model includes a contrast analysis module and a U-net module connected in series, the contrast analysis module is used to extract contrast signals, and the U-net module is used to enhance and remove noise from the extracted contrast signals.

[0039] like Figure 4 As shown in the figure, the contrast parsing network consists of the following three parts: 1) Inception block 1, which consists of a series of convolutional layers to form the first Inception block, most of which use 1×1Conv, which can reduce the channel dimension while introducing nonlinear transformations to improve the network's expressiveness. It can be observed that the Inception block receives features from five branches, namely, the 8-channel feature layer of the first layer, the 16-channel feature layer of the second layer, the 32-channel feature layer of the third layer, the 32-channel 1×1Conv feature layer and the 32-channel 1×3Conv feature layer of the fourth layer, and the 64-channel 1×1 convolution layer and the 64-channel 3×1 convolution layer of the fifth layer, and merges them. 2) ResNet, the role of introducing ResNet into the two Inception blocks is to introduce residual connections, allowing information to directly skip some levels in the network, thereby better propagating gradients and information. This can solve the problem of gradient disappearance, allowing the network to be stacked deeper and better capture complex features in the image. The residual block consists of a 128-channel 3×3 convolution layer and a 64-channel 1×1 convolution layer. 3) Inception block 2, further expanding the representation capability based on the first Inception block. Similar to the first Inception block, it contains multiple parallel convolutional layers. Specifically, the Inception block receives features from three branches, namely, three 64-channel 1×1Conv layers of the 11th layer, two 32-channel 1×1Conv layers of the 12th layer and a 16-channel 1Conv layer, and finally outputs a preliminary differential phase contrast projection image. The contrast resolution module is connected in series with the U-net to construct a convolutional neural network, where the contrast resolution module consists of an Inception module and a residual module, which is used to extract phase contrast.

[0040] Furthermore, the U-net described in step 2 mainly includes: 1) Encoder. The main function of the encoder is to convert the input image into a low-resolution, high-level feature representation to capture the global context information in the image. As shown in the purple rectangle in the figure, the input image is reduced in resolution through a series of convolutional layers, and increasingly abstract feature representations are extracted. It is cascaded through 5 consecutive Conv layers with 16, 32, 64, 128 and 256 channels, 3×3 convolution kernels, and stride 2 convolution layers to obtain five feature representations of different scales. 2) Decoder. The main function of the decoder is to use the multi-scale features extracted by the encoder and retain more detailed information while restoring the resolution to obtain more accurate image processing results. As shown in the figure, the decoder accepts the output feature map of the encoder part and gradually restores the resolution of the projected image through a series of upsampling and convolution operations. It is cascaded through 4 consecutive Conv layers with 128, 64, 32 and 16 channels, 3×3 convolution kernels, and stride 2 upsampling convolution layers. 3) Skip connection: Skip connection helps solve the problem of local structure loss and resolution loss in image enhancement tasks, and improves the context perception and accuracy of the projection enhancement network. Skip connection acts between the encoder and the decoder, connecting the features of the encoder with the corresponding layers of the decoder, and can pass low-level image feature information to the decoder, helping it to further restore the local structure of the image and improve image enhancement capabilities.

[0041] It is worth noting that the contrast resolution network and the projection enhancement network form an overall deep convolutional network and are trained in a unified manner, rather than a cascade of network models trained separately. Therefore, there is no logical order between contrast resolution and projection enhancement, but they work together to achieve the goal of outputting a differential phase contrast projection image with correct structure and complete information.

[0042] Step S103, training a convolutional neural network model based on a step projection sequence: using at least three step projection sequences including samples and at least three step projection sequences including backgrounds acquired under the low tube current condition as inputs of the network model, performing Fourier analysis on at least three step projection sequences including samples and at least three step projection sequences including backgrounds acquired under the normal tube current condition, using the acquired absorption, phase and dark field images as labels, and training the convolutional neural network model.

[0043] Step S104, using the trained network model to obtain a high-precision contrast image: input at least three step projection sequences including samples and at least three step projection sequences including backgrounds under the low tube current condition into the trained network model to obtain high-precision absorption, phase and dark field contrast images.

[0044] Figure 5(a) is the absorption, phase and dark field contrast map of the standard dose Fourier analysis, (b) is the absorption, phase and dark field contrast map of the one-eighth dose Fourier analysis, and (c) is the absorption, phase and dark field contrast map obtained by the one-eighth dose of the present invention. As can be seen from the figure, the example of the present invention has absolute advantages in noise reduction, artifact suppression and structure recovery. This is due to the existence of the contrast analysis module, which enables the network to effectively extract continuous physical information in the step projection sequence, forming a mutual enhancement of effective information between projection sequences, thereby restoring the detail information destroyed due to low tube current.

[0045] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A low-dose X-ray multi-contrast high-precision signal analysis method, characterized in that: The steps include: Step S101, using an X-ray grating differential phase contrast imaging device based on the Talbot-Lau effect, obtaining at least three step projection sequences including samples and at least three step projection sequences including backgrounds under normal tube current conditions, and obtaining at least three step projection sequences including samples and at least three step projection sequences including backgrounds under low tube current conditions; Step S102, constructing a convolutional neural network model, wherein the network model includes a contrast analysis module and a U-net module connected in series, wherein the contrast analysis module is used to extract contrast signals, and the U-net module is used to enhance and remove noise from the extracted contrast signals; Step S103, using the at least three step projection sequences including the sample and the at least three step projection sequences including the background obtained under the low tube current condition as inputs of the network model, performing Fourier analysis on the at least three step projection sequences including the sample and the at least three step projection sequences including the background under the normal tube current condition, using the obtained absorption, phase and dark field images as labels, and training the convolutional neural network model; Step S104: input at least three step projection sequences including samples and at least three step projection sequences including backgrounds under the low tube current condition into the trained network model to obtain high-precision absorption, phase and dark field contrast images.

2. A low-dose X-ray multi-contrast high-precision signal analysis method according to claim 1, characterized in that: In step S101, the X-ray grating differential phase contrast imaging device includes an X-ray source, a source grating, a sample, a phase grating, an absorption grating and a detector arranged in sequence along an optical path; The corresponding relationship between the imaging experimental parameters of the X-ray grating differential contrast imaging device is as follows: (1) (2) (3) (4) Wherein, d is the distance between the phase grating and the absorption grating, m is an integer, representing m times the fractional Talbot distance, k = (L + d) / L is the magnification ratio, g1 is the period of the phase grating, λ is the wavelength of the X-ray, g2 is the period of the absorption grating, g0 is the period of the source grating, L is the distance between the source grating and the phase grating, and s is the width of the source grating that allows X-rays to pass through in each period; Under the conditions of low tube current and normal tube current respectively, stepping is performed through the absorption grating to obtain at least three stepping projection sequences including samples and at least three stepping projection sequences including backgrounds.

3. The low-dose X-ray multi-contrast high-precision signal analysis method according to claim 1, characterized in that: In step S102, the contrast analysis module includes a first Inception block, a ResNet block, and a second Inception block connected in series.

4. A low-dose X-ray multi-contrast high-precision signal analysis method according to claim 3, characterized in that: The first Inception block includes several parallel convolutional layers; the ResNet block is used to introduce residual connections; the second Inception block includes several parallel convolutional layers.

5. The low-dose X-ray multi-contrast high-precision signal analysis method according to claim 3, characterized in that: The U-net module includes an encoder and a decoder, and the encoder and the decoder connect their corresponding layers through a jump connection.

6. The low-dose X-ray multi-contrast high-precision signal analysis method according to claim 1, characterized in that: In step S103, during the imaging process, at least three step projection sequences including samples and at least three step projection sequences including backgrounds under normal tube current conditions are obtained by stepping the absorption grating, and a step curve of light intensity varying with position is obtained by fast Fourier transform: (5) Among them, x is the independent variable, representing the position of the absorption grating, f(x) represents the X-ray intensity corresponding to the absorption grating at position x, a0 is a constant term, a1 is the amplitude, φ1 is the initial phase, and the absorption, phase and dark field contrast images of the sample and the background are obtained using the following formulas: (6) (7) (8) Where A represents the absorption contrast image, and represent the constant terms of the sample and background in formula (5), φ represents the phase contrast image, and represent the initial phases of the sample and background in formula (5), V represents the dark field contrast image, and v s and v r They are the dark field contrast images of the sample and the background, and Represent the amplitudes of the sample and background in formula (5) respectively; The acquired absorption contrast map A, phase contrast map φ, and dark field contrast map V are used as labels to train the convolutional neural network model.

Images