CN116630454A - CT-based high-definition imaging method and device - Google Patents
CT-based high-definition imaging method and device Download PDFInfo
- Publication number
- CN116630454A CN116630454A CN202310452785.5A CN202310452785A CN116630454A CN 116630454 A CN116630454 A CN 116630454A CN 202310452785 A CN202310452785 A CN 202310452785A CN 116630454 A CN116630454 A CN 116630454A
- Authority
- CN
- China
- Prior art keywords
- image
- frequency
- module
- noise
- sample set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003384 imaging method Methods 0.000 title claims abstract description 27
- 238000006243 chemical reaction Methods 0.000 claims abstract description 46
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 23
- 230000000694 effects Effects 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims description 38
- 238000000034 method Methods 0.000 claims description 31
- 230000003287 optical effect Effects 0.000 claims description 13
- 230000002159 abnormal effect Effects 0.000 claims description 12
- 230000004580 weight loss Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 8
- 238000001514 detection method Methods 0.000 claims description 3
- 230000002194 synthesizing effect Effects 0.000 claims description 3
- 238000003745 diagnosis Methods 0.000 abstract description 5
- 238000002591 computed tomography Methods 0.000 description 220
- 230000006870 function Effects 0.000 description 62
- 230000008901 benefit Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000003491 array Methods 0.000 description 2
- 230000005251 gamma ray Effects 0.000 description 2
- 238000012827 research and development Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
- G06T11/005—Specific pre-processing for tomographic reconstruction, e.g. calibration, source positioning, rebinning, scatter correction, retrospective gating
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
- G06T11/008—Specific post-processing after tomographic reconstruction, e.g. voxelisation, metal artifact correction
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Abstract
The invention belongs to the technical field of CT high-definition imaging, and discloses a CT-based high-definition imaging method and a CT-based high-definition imaging device, wherein the CT-based high-definition imaging device comprises: the device comprises a scanning module, a main control module, a photoelectric conversion module, a digital-to-analog conversion module, an image generation module, an image denoising module, an image reconstruction module and a display module. According to the invention, the convolutional neural network is iteratively trained by combining the mean square error loss function value and the maximum posterior loss function value through the image denoising module, so that the removal effect of real noise can be greatly improved; meanwhile, the image reconstruction module does not change the existing scanning conditions and reconstruction conditions, and meanwhile, windmill artifacts in CT reconstructed images are effectively eliminated, the quality of the CT reconstructed images is improved, and an accurate basis is provided for subsequent diagnosis based on the CT reconstructed images.
Description
Technical Field
The invention belongs to the technical field of CT high-definition imaging, and particularly relates to a CT-based high-definition imaging method and device.
Background
CT (Computed Tomography) it is an electronic computer tomography, it uses accurate collimated X-ray beam, gamma ray, supersonic wave, etc., and makes one-by-one section scan around a certain part of human body together with the very high-sensitivity detector, it has characteristics of fast scanning time, clear picture, etc., can be used for the inspection of various diseases; the rays used can be classified differently according to the type: x-ray CT (X-CT), gamma-ray CT (gamma-CT), and the like. CT is to scan a certain layer of human body with X-ray beam, to receive X-ray transmitted through the layer by detector, to convert it into visible light, to convert it into electric signal by photoelectric conversion, to convert it into digital by analog/digital converter (analog/digital converter), and to input it into computer for processing. The image formation process includes the steps of dividing a selected layer into a plurality of cuboids with the same volume, which are called voxels (voxels); however, the existing CT-based high-definition imaging device has poor denoising effect and affects CT image quality; meanwhile, in the prior art, windmill artifacts cannot be completely eliminated, so that CT reconstructed images containing the windmill artifacts are obtained, and the quality of the CT reconstructed images is poor.
Through the above analysis, the problems and defects existing in the prior art are as follows:
(1) The existing CT-based high-definition imaging device has poor denoising effect and affects CT image quality.
(2) In the prior art, windmill artifacts cannot be completely eliminated, so that CT reconstructed images containing the windmill artifacts are obtained, and the quality of the CT reconstructed images is poor.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides a CT-based high-definition imaging method and device.
The invention is realized in that a CT-based high-definition imaging device comprises:
the device comprises a scanning module, a main control module, a photoelectric conversion module, a digital-to-analog conversion module, an image generation module, an image denoising module, an image reconstruction module and a display module;
the scanning module is connected with the main control module and used for scanning a patient through an X-ray beam;
the scanning module is used for scanning:
the patient lies on the scanning table; according to the detection requirement of the patient, the parameters of the x-ray scanner, including beam intensity, scanning speed, scanning range and the like, are adjusted to obtain the optimal scanning effect;
starting an x-ray scanner to start scanning the patient; an x-ray scanner will scan a two-dimensional image of the patient surface, showing an image of the interior of the patient;
the main control module is connected with the scanning module, the photoelectric conversion module, the digital-to-analog conversion module, the image generation module, the image denoising module, the image reconstruction module and the display module and used for controlling the normal work of each module;
the photoelectric conversion module is connected with the main control module and used for converting the scanning optical signals into electric signals;
the photoelectric conversion module conversion method comprises the following steps:
the main control module sends the scanning optical signal of the scanning module to the photoelectric conversion module;
the photoelectric conversion module converts the optical signal into an electric signal through a photoelectric conversion circuit; correcting the electric signal;
transmitting the converted electric signal to a digital-to-analog conversion module for processing;
the image generation module is connected with the main control module and used for converting the digital signals into CT images;
the image denoising module is connected with the main control module and is used for denoising the CT image;
the image denoising module denoising method comprises the following steps:
iteratively training the convolutional neural network based on the calculated total loss function value until a training termination condition is met, and outputting a CT image denoising model obtained by training;
denoising the CT image to be denoised according to the CT image denoising model;
the image reconstruction module is connected with the main control module and is used for reconstructing the CT image;
and the display module is connected with the main control module and used for displaying CT images.
A CT-based high-definition imaging method comprises the following steps:
step one, scanning a patient by using an X-ray beam through a scanning module;
step two, the main control module converts the scanning optical signal into an electric signal through the photoelectric conversion module;
step three, converting the electric signal into a digital signal through a digital-to-analog conversion module; converting the digital signal into a CT image through an image generating module; denoising the CT image through an image denoising module;
step four, reconstructing the CT image through an image reconstruction module; and displaying the CT image through a display module.
Further, the denoising method of the image denoising module comprises the following steps:
(1) Acquiring a training sample set, wherein the training sample set comprises a noise CT image sample set and a noise-free CT image sample set corresponding to the noise CT image sample set; inputting the noise CT image sample set into a convolutional neural network for training, and outputting a denoising CT image sample set obtained by training;
(2) Respectively calculating a corresponding mean square error loss function value and a maximum posterior loss function value according to the noiseless CT image sample set and the denoising CT image sample set, and obtaining a total loss function value according to the mean square error loss function value and the maximum posterior loss function value;
the step of calculating the corresponding mean square error loss function value and the maximum posterior loss function value according to the noiseless CT image sample set and the denoising CT image sample set respectively comprises the following steps:
calculating a mean square error loss function value according to the noiseless CT image sample set and the denoising CT image sample set, wherein the mean square error loss function value is specifically as follows:
respectively calculating the square value of each pixel point difference value between each frame of noiseless CT image sample in the noiseless CT image sample set and the corresponding denoising CT image sample in the denoising CT image sample set, and taking the average value of the square values of the calculated pixel point difference values as the mean square error loss function value; and
calculating a maximum posterior loss function value according to the noiseless CT image sample set and the denoising CT image sample set, wherein the maximum posterior loss function value is specifically as follows:
calculating the posterior probability of the poisson noise of each noiseless CT image sample in the noiseless CT image sample set according to each noiseless CT image sample in the noiseless CT image sample set and the poisson noise CT image after the poisson noise is added into each noiseless CT image sample;
calculating the posterior probability of Gaussian noise existing in each Poisson noise CT image according to the Poisson noise CT image after the Poisson noise is added in each noiseless CT image sample and the Gaussian noise CT image after the Gaussian noise is added in each noiseless CT image sample;
calculating to obtain the maximum posterior probability of the poisson Gaussian noise of each noiseless CT image sample according to the posterior probability of the poisson noise of each noiseless CT image sample and the posterior probability of the Gaussian noise of each poisson noise CT image;
calculating a negative logarithmic function value taking the maximum posterior probability of the poisson Gaussian noise of each noiseless CT image sample as an independent variable, and taking the negative logarithmic function value as the maximum posterior loss function value;
(3) And iteratively training the convolutional neural network based on the calculated total loss function value until the training termination condition is met, outputting a CT image denoising model obtained by training, and denoising the CT image to be denoised according to the CT image denoising model.
Further, the step of obtaining a training sample set includes:
acquiring a plurality of frames of noise CT image samples continuously shot by CT image acquisition equipment based on configured CT image acquisition parameters to construct a noise CT image sample set, wherein the CT image acquisition parameters comprise CT image exposure parameters and shooting frequency parameters;
sorting pixel values of each pixel point in the noise CT image sample set according to each frame of noise CT image sample in the noise CT image sample set, and removing abnormal pixel points with the pixel values smaller than a first preset pixel value and the pixel values larger than a second preset pixel value according to sorting results to obtain noise CT image samples with the abnormal pixel points removed;
calculating the average value of each pixel point of the noise CT image sample after eliminating abnormal pixel points of each frame;
and obtaining a noiseless CT image sample set corresponding to the noise CT image sample set according to the relation between the preset noise CT image sample and the corresponding noiseless CT image sample and the average value of each pixel point of the noise CT image sample after eliminating abnormal pixel points of each frame.
Further, the step of inputting the noise CT image sample set into a convolutional neural network for training and outputting a denoised CT image sample set obtained by training includes:
inputting the noise CT image sample set into a convolutional neural network, and sequentially extracting CT image characteristic information of each noise CT image sample in the noise CT image sample set through each layer of convolutional layer in the convolutional neural network;
and aiming at each layer of convolution layer, transmitting the CT image characteristic information of each noise CT image sample extracted by the convolution layer to deconvolution layers symmetrically arranged by the convolution layer through a connection layer, and generating a deconvoluted denoising CT image sample set according to the CT image characteristic information of each noise CT image sample through the deconvolution layers.
Further, the step of obtaining a total loss function value according to the mean square error loss function value and the maximum posterior loss function value includes:
respectively calculating a first weight loss function value of the mean square error loss function value and a second weight loss function value of the maximum posterior loss function value according to a preset weight proportion;
and obtaining the total loss function value according to the first weight loss function value and the second weight loss function value.
Further, the image reconstruction module reconstruction method comprises the following steps:
1) Acquiring an original CT reconstruction image, wherein the original CT reconstruction image comprises windmill artifacts; dividing the frequency of the original CT reconstruction image to obtain a high-frequency image; frequency division combination is carried out on a plurality of thick images reconstructed according to raw data to obtain a low-frequency image, wherein the raw data comprise raw data corresponding to the original CT reconstructed image, and windmill artifacts are not included in each thick image;
2) Synthesizing an intermediate image according to the high-frequency image and the low-frequency image, and outputting the intermediate image as a target CT reconstructed image;
and carrying out frequency division combination on a plurality of thick images reconstructed according to the raw data to obtain the low-frequency image, wherein the method comprises the following steps of:
determining a thick image reconstruction parameter capable of eliminating windmill artifacts;
reconstructing a plurality of thick images by using raw data comprising raw data corresponding to the original CT reconstructed images according to the thick image reconstruction parameters;
and carrying out frequency division on each thick image to obtain a plurality of low-frequency thick images, and carrying out sharpening combination on the plurality of low-frequency thick images to obtain the low-frequency images.
Further, the frequency dividing the original CT reconstructed image to obtain the high frequency image includes:
performing Fourier transform on the original CT reconstruction image, and converting the original CT reconstruction image from a space domain to a frequency domain;
extracting high-frequency components in the frequency domain data of the original CT reconstructed image;
and performing inverse Fourier transform on the extracted high-frequency components to generate the high-frequency image.
Further, the extracting the high-frequency component in the frequency domain data of the original CT reconstructed image includes:
calculating a low-frequency weight coefficient of each frequency position in the frequency domain data;
calculating a low-frequency value of each frequency position according to the value of each frequency position in the frequency domain data and the low-frequency weight coefficient;
the difference between the value of each frequency location and the low frequency value is calculated as the high frequency value of each frequency location,
the high frequency values of all frequency locations constitute high frequency components in the frequency domain data.
Further, the thick image reconstruction parameters comprise reconstruction interval, image thickness and image quantity;
reconstructing a plurality of thick images by using raw data comprising raw data corresponding to the original CT reconstructed images according to the thick image reconstruction parameters, wherein the method comprises the following steps: reconstructing a plurality of thick images along the direction of a scanning bed according to the reconstruction interval by using a group of raw data containing raw data corresponding to the original CT reconstructed images, wherein the thickness of each thick image is the same as that of the image, and the number of the thick images is consistent with that of the images;
frequency division is carried out on each thick image to obtain a plurality of low-frequency thick images, and the method comprises the following steps:
performing Fourier transform on the image to be divided, and converting the image to be divided into a frequency domain from a space domain;
extracting a low-frequency component in the frequency domain data of the thick image to be divided;
performing inverse Fourier transform on the extracted low-frequency component to generate a low-frequency thick image of the image to be divided;
extracting a low-frequency component in the frequency domain data of the thick image to be divided, including:
calculating a low-frequency weight coefficient of each frequency position in the frequency domain data;
calculating a low frequency value for each frequency location based on the value for each frequency location in the frequency domain data and the low frequency weight coefficient,
the low-frequency values of all frequency positions are formed into low-frequency components in the frequency domain data;
sharpening the plurality of low-frequency thick images to obtain the low-frequency image comprises the following steps:
determining weights corresponding to the plurality of low-frequency thick images to be combined;
multiplying the pixel value of each pixel point on each low-frequency thick image to be combined with the corresponding weight to obtain the weighted pixel value of each pixel point on each low-frequency thick image to be combined;
accumulating the weighted pixel values of the same pixel point on the plurality of low-frequency thick images to be combined to obtain accumulated pixel values;
the accumulated pixel values of all the pixel points form the low-frequency image.
In combination with the above technical solution and the technical problems to be solved, please analyze the following aspects to provide the following advantages and positive effects:
first, aiming at the technical problems in the prior art and the difficulty in solving the problems, the technical problems solved by the technical proposal of the invention are analyzed in detail and deeply by tightly combining the technical proposal to be protected, the results and data in the research and development process, and the like, and some technical effects brought after the problems are solved have creative technical effects. The specific description is as follows:
according to the invention, the convolutional neural network is iteratively trained by combining the mean square error loss function value and the maximum posterior loss function value through the image denoising module, so that the removal effect of real noise can be greatly improved; meanwhile, the image reconstruction module obtains a high-frequency image of an original CT reconstruction image and a low-frequency image which does not contain windmill artifacts and corresponds to the original CT reconstruction image according to frequency division processing, and a target CT reconstruction image which basically does not contain windmill artifacts is obtained after the high-frequency image and the low-frequency image are combined, so that the windmill artifacts in the CT reconstruction image are effectively eliminated while the existing scanning conditions and reconstruction conditions are not changed, the quality of the CT reconstruction image is improved, and an accurate basis is provided for the subsequent diagnosis based on the CT reconstruction image.
Secondly, the technical scheme is regarded as a whole or from the perspective of products, and the technical scheme to be protected has the following technical effects and advantages:
according to the invention, the convolutional neural network is iteratively trained by combining the mean square error loss function value and the maximum posterior loss function value through the image denoising module, so that the removal effect of real noise can be greatly improved; meanwhile, the image reconstruction module obtains a high-frequency image of an original CT reconstruction image and a low-frequency image which does not contain windmill artifacts and corresponds to the original CT reconstruction image according to frequency division processing, and a target CT reconstruction image which basically does not contain windmill artifacts is obtained after the high-frequency image and the low-frequency image are combined, so that the windmill artifacts in the CT reconstruction image are effectively eliminated while the existing scanning conditions and reconstruction conditions are not changed, the quality of the CT reconstruction image is improved, and an accurate basis is provided for the subsequent diagnosis based on the CT reconstruction image.
Drawings
Fig. 1 is a flowchart of a CT-based high definition imaging method according to an embodiment of the present invention.
Fig. 2 is a block diagram of a CT-based high definition imaging apparatus according to an embodiment of the present invention.
Fig. 3 is a flowchart of a denoising method of an image denoising module according to an embodiment of the present invention.
Fig. 4 is a flowchart of a reconstruction method of an image reconstruction module according to an embodiment of the present invention.
In fig. 2: 1. a scanning module; 2. a main control module; 3. a photoelectric conversion module; 4. a digital-to-analog conversion module; 5. an image generation module; 6. an image denoising module; 7. an image reconstruction module; 8. and a display module.
Detailed Description
The present invention will be described in further detail with reference to the following examples in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
1. The embodiments are explained. In order to fully understand how the invention may be embodied by those skilled in the art, this section is an illustrative embodiment in which the claims are presented for purposes of illustration.
As shown in fig. 1, the high-definition imaging method based on CT provided by the present invention includes the following steps:
s101, scanning a patient by using an X-ray beam through a scanning module;
s102, a main control module converts scanning optical signals into electric signals through a photoelectric conversion module;
s103, converting the electric signal into a digital signal through a digital-to-analog conversion module; converting the digital signal into a CT image through an image generating module; denoising the CT image through an image denoising module;
s104, reconstructing the CT image through an image reconstruction module; and displaying the CT image through a display module.
As shown in fig. 2, a CT-based high-definition imaging device provided in an embodiment of the present invention includes: the device comprises a scanning module 1, a main control module 2, a photoelectric conversion module 3, a digital-to-analog conversion module 4, an image generation module 5, an image denoising module 6, an image reconstruction module 7 and a display module 8.
The scanning module 1 is connected with the main control module 2 and is used for scanning a patient through an X-ray beam;
the scanning module is used for scanning:
the patient lies on the scanning table; according to the detection requirement of the patient, the parameters of the x-ray scanner, including beam intensity, scanning speed, scanning range and the like, are adjusted to obtain the optimal scanning effect;
starting an x-ray scanner to start scanning the patient; an x-ray scanner will scan a two-dimensional image of the patient surface, showing an image of the interior of the patient;
the main control module 2 is connected with the scanning module 1, the photoelectric conversion module 3, the digital-to-analog conversion module 4, the image generation module 5, the image denoising module 6, the image reconstruction module 7 and the display module 8 and used for controlling the normal work of each module;
the photoelectric conversion module 3 is connected with the main control module 2 and is used for converting the scanning optical signals into electric signals;
the photoelectric conversion module conversion method comprises the following steps:
the main control module sends the scanning optical signal of the scanning module to the photoelectric conversion module;
the photoelectric conversion module converts the optical signal into an electric signal through a photoelectric conversion circuit; correcting the electric signal;
transmitting the converted electric signal to a digital-to-analog conversion module for processing;
the digital-to-analog conversion module 4 is connected with the main control module 2 and is used for converting the electric signals into digital signals;
the image generation module 5 is connected with the main control module 2 and is used for converting the digital signals into CT images;
the image denoising module 6 is connected with the main control module 2 and is used for denoising the CT image;
the image denoising module denoising method comprises the following steps:
iteratively training the convolutional neural network based on the calculated total loss function value until a training termination condition is met, and outputting a CT image denoising model obtained by training;
denoising the CT image to be denoised according to the CT image denoising model;
the image reconstruction module 7 is connected with the main control module 2 and is used for reconstructing CT images;
and the display module 8 is connected with the main control module 2 and is used for displaying CT images.
As shown in fig. 3, the denoising method of the image denoising module provided by the invention is as follows:
s201, acquiring a training sample set, wherein the training sample set comprises a noise CT image sample set and a noise-free CT image sample set corresponding to the noise CT image sample set; inputting the noise CT image sample set into a convolutional neural network for training, and outputting a denoising CT image sample set obtained by training;
s202, respectively calculating a corresponding mean square error loss function value and a maximum posterior loss function value according to the noiseless CT image sample set and the denoising CT image sample set, and obtaining a total loss function value according to the mean square error loss function value and the maximum posterior loss function value;
the step of calculating the corresponding mean square error loss function value and the maximum posterior loss function value according to the noiseless CT image sample set and the denoising CT image sample set respectively comprises the following steps:
calculating a mean square error loss function value according to the noiseless CT image sample set and the denoising CT image sample set, wherein the mean square error loss function value is specifically as follows:
respectively calculating the square value of each pixel point difference value between each frame of noiseless CT image sample in the noiseless CT image sample set and the corresponding denoising CT image sample in the denoising CT image sample set, and taking the average value of the square values of the calculated pixel point difference values as the mean square error loss function value; and
calculating a maximum posterior loss function value according to the noiseless CT image sample set and the denoising CT image sample set, wherein the maximum posterior loss function value is specifically as follows:
calculating the posterior probability of the poisson noise of each noiseless CT image sample in the noiseless CT image sample set according to each noiseless CT image sample in the noiseless CT image sample set and the poisson noise CT image after the poisson noise is added into each noiseless CT image sample;
calculating the posterior probability of Gaussian noise existing in each Poisson noise CT image according to the Poisson noise CT image after the Poisson noise is added in each noiseless CT image sample and the Gaussian noise CT image after the Gaussian noise is added in each noiseless CT image sample;
calculating to obtain the maximum posterior probability of the poisson Gaussian noise of each noiseless CT image sample according to the posterior probability of the poisson noise of each noiseless CT image sample and the posterior probability of the Gaussian noise of each poisson noise CT image;
calculating a negative logarithmic function value taking the maximum posterior probability of the poisson Gaussian noise of each noiseless CT image sample as an independent variable, and taking the negative logarithmic function value as the maximum posterior loss function value;
and S203, iteratively training the convolutional neural network based on the calculated total loss function value until the training termination condition is met, outputting a CT image denoising model obtained by training, and denoising the CT image to be denoised according to the CT image denoising model.
The step of obtaining the training sample set provided by the invention comprises the following steps:
acquiring a plurality of frames of noise CT image samples continuously shot by CT image acquisition equipment based on configured CT image acquisition parameters to construct a noise CT image sample set, wherein the CT image acquisition parameters comprise CT image exposure parameters and shooting frequency parameters;
sorting pixel values of each pixel point in the noise CT image sample set according to each frame of noise CT image sample in the noise CT image sample set, and removing abnormal pixel points with the pixel values smaller than a first preset pixel value and the pixel values larger than a second preset pixel value according to sorting results to obtain noise CT image samples with the abnormal pixel points removed;
calculating the average value of each pixel point of the noise CT image sample after eliminating abnormal pixel points of each frame;
and obtaining a noiseless CT image sample set corresponding to the noise CT image sample set according to the relation between the preset noise CT image sample and the corresponding noiseless CT image sample and the average value of each pixel point of the noise CT image sample after eliminating abnormal pixel points of each frame.
The invention provides a method for inputting the noise CT image sample set into a convolutional neural network for training, and outputting the denoised CT image sample set obtained by training, comprising the following steps:
inputting the noise CT image sample set into a convolutional neural network, and sequentially extracting CT image characteristic information of each noise CT image sample in the noise CT image sample set through each layer of convolutional layer in the convolutional neural network;
and aiming at each layer of convolution layer, transmitting the CT image characteristic information of each noise CT image sample extracted by the convolution layer to deconvolution layers symmetrically arranged by the convolution layer through a connection layer, and generating a deconvoluted denoising CT image sample set according to the CT image characteristic information of each noise CT image sample through the deconvolution layers.
The step of obtaining the total loss function value according to the mean square error loss function value and the maximum posterior loss function value provided by the invention comprises the following steps:
respectively calculating a first weight loss function value of the mean square error loss function value and a second weight loss function value of the maximum posterior loss function value according to a preset weight proportion;
and obtaining the total loss function value according to the first weight loss function value and the second weight loss function value.
As shown in fig. 4, the image reconstruction module reconstruction method provided by the invention is as follows:
s301, acquiring an original CT reconstruction image, wherein the original CT reconstruction image comprises windmill artifacts; dividing the frequency of the original CT reconstruction image to obtain a high-frequency image; frequency division combination is carried out on a plurality of thick images reconstructed according to raw data to obtain a low-frequency image, wherein the raw data comprise raw data corresponding to the original CT reconstructed image, and windmill artifacts are not included in each thick image;
s302, synthesizing an intermediate image according to the high-frequency image and the low-frequency image, and outputting the intermediate image as a target CT reconstructed image;
and carrying out frequency division combination on a plurality of thick images reconstructed according to the raw data to obtain the low-frequency image, wherein the method comprises the following steps of:
determining a thick image reconstruction parameter capable of eliminating windmill artifacts;
reconstructing a plurality of thick images by using raw data comprising raw data corresponding to the original CT reconstructed images according to the thick image reconstruction parameters;
and carrying out frequency division on each thick image to obtain a plurality of low-frequency thick images, and carrying out sharpening combination on the plurality of low-frequency thick images to obtain the low-frequency images.
The method for obtaining the high-frequency image by frequency division of the original CT reconstruction image provided by the invention comprises the following steps:
performing Fourier transform on the original CT reconstruction image, and converting the original CT reconstruction image from a space domain to a frequency domain;
extracting high-frequency components in the frequency domain data of the original CT reconstructed image;
and performing inverse Fourier transform on the extracted high-frequency components to generate the high-frequency image.
The method for extracting the high-frequency component in the frequency domain data of the original CT reconstructed image comprises the following steps:
calculating a low-frequency weight coefficient of each frequency position in the frequency domain data;
calculating a low-frequency value of each frequency position according to the value of each frequency position in the frequency domain data and the low-frequency weight coefficient;
the difference between the value of each frequency location and the low frequency value is calculated as the high frequency value of each frequency location,
the high frequency values of all frequency locations constitute high frequency components in the frequency domain data.
The thick image reconstruction parameters provided by the invention comprise reconstruction interval, image thickness and image quantity;
reconstructing a plurality of thick images by using raw data comprising raw data corresponding to the original CT reconstructed images according to the thick image reconstruction parameters, wherein the method comprises the following steps: reconstructing a plurality of thick images along the direction of a scanning bed according to the reconstruction interval by using a group of raw data containing raw data corresponding to the original CT reconstructed images, wherein the thickness of each thick image is the same as that of the image, and the number of the thick images is consistent with that of the images;
frequency division is carried out on each thick image to obtain a plurality of low-frequency thick images, and the method comprises the following steps:
performing Fourier transform on the image to be divided, and converting the image to be divided into a frequency domain from a space domain;
extracting a low-frequency component in the frequency domain data of the thick image to be divided;
performing inverse Fourier transform on the extracted low-frequency component to generate a low-frequency thick image of the image to be divided;
extracting a low-frequency component in the frequency domain data of the thick image to be divided, including:
calculating a low-frequency weight coefficient of each frequency position in the frequency domain data;
calculating a low frequency value for each frequency location based on the value for each frequency location in the frequency domain data and the low frequency weight coefficient,
the low-frequency values of all frequency positions are formed into low-frequency components in the frequency domain data;
sharpening the plurality of low-frequency thick images to obtain the low-frequency image comprises the following steps:
determining weights corresponding to the plurality of low-frequency thick images to be combined;
multiplying the pixel value of each pixel point on each low-frequency thick image to be combined with the corresponding weight to obtain the weighted pixel value of each pixel point on each low-frequency thick image to be combined;
accumulating the weighted pixel values of the same pixel point on the plurality of low-frequency thick images to be combined to obtain accumulated pixel values;
the accumulated pixel values of all the pixel points form the low-frequency image.
2. Application example. In order to prove the inventive and technical value of the technical solution of the present invention, this section is an application example on specific products or related technologies of the claim technical solution.
According to the invention, the convolutional neural network is iteratively trained by combining the mean square error loss function value and the maximum posterior loss function value through the image denoising module, so that the removal effect of real noise can be greatly improved; meanwhile, the image reconstruction module obtains a high-frequency image of an original CT reconstruction image and a low-frequency image which does not contain windmill artifacts and corresponds to the original CT reconstruction image according to frequency division processing, and a target CT reconstruction image which basically does not contain windmill artifacts is obtained after the high-frequency image and the low-frequency image are combined, so that the windmill artifacts in the CT reconstruction image are effectively eliminated while the existing scanning conditions and reconstruction conditions are not changed, the quality of the CT reconstruction image is improved, and an accurate basis is provided for the subsequent diagnosis based on the CT reconstruction image.
It should be noted that the embodiments of the present invention can be realized in hardware, software, or a combination of software and hardware. The hardware portion may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or special purpose design hardware. Those of ordinary skill in the art will appreciate that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, such as provided on a carrier medium such as a magnetic disk, CD or DVD-ROM, a programmable memory such as read only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The device of the present invention and its modules may be implemented by hardware circuitry, such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., as well as software executed by various types of processors, or by a combination of the above hardware circuitry and software, such as firmware.
3. Evidence of the effect of the examples. The embodiment of the invention has a great advantage in the research and development or use process, and has the following description in combination with data, charts and the like of the test process.
According to the invention, the convolutional neural network is iteratively trained by combining the mean square error loss function value and the maximum posterior loss function value through the image denoising module, so that the removal effect of real noise can be greatly improved; meanwhile, the image reconstruction module obtains a high-frequency image of an original CT reconstruction image and a low-frequency image which does not contain windmill artifacts and corresponds to the original CT reconstruction image according to frequency division processing, and a target CT reconstruction image which basically does not contain windmill artifacts is obtained after the high-frequency image and the low-frequency image are combined, so that the windmill artifacts in the CT reconstruction image are effectively eliminated while the existing scanning conditions and reconstruction conditions are not changed, the quality of the CT reconstruction image is improved, and an accurate basis is provided for the subsequent diagnosis based on the CT reconstruction image.
The foregoing is merely illustrative of specific embodiments of the present invention, and the scope of the invention is not limited thereto, but any modifications, equivalents, improvements and alternatives falling within the spirit and principles of the present invention will be apparent to those skilled in the art within the scope of the present invention.
Claims (10)
1. A CT-based high definition imaging device, comprising:
the device comprises a scanning module, a main control module, a photoelectric conversion module, a digital-to-analog conversion module, an image generation module, an image denoising module, an image reconstruction module and a display module;
the scanning module is connected with the main control module and used for scanning a patient through an X-ray beam;
the scanning module is used for scanning:
the patient lies on the scanning table; according to the detection requirement of the patient, the parameters of the x-ray scanner, including beam intensity, scanning speed, scanning range and the like, are adjusted to obtain the optimal scanning effect;
starting an x-ray scanner to start scanning the patient; an x-ray scanner will scan a two-dimensional image of the patient surface, showing an image of the interior of the patient;
the main control module is connected with the scanning module, the photoelectric conversion module, the digital-to-analog conversion module, the image generation module, the image denoising module, the image reconstruction module and the display module and used for controlling the normal work of each module;
the photoelectric conversion module is connected with the main control module and used for converting the scanning optical signals into electric signals;
the photoelectric conversion module conversion method comprises the following steps:
the main control module sends the scanning optical signal of the scanning module to the photoelectric conversion module;
the photoelectric conversion module converts the optical signal into an electric signal through a photoelectric conversion circuit; correcting the electric signal;
transmitting the converted electric signal to a digital-to-analog conversion module for processing;
the digital-to-analog conversion module is connected with the main control module and is used for converting the electric signals into digital signals;
the image generation module is connected with the main control module and used for converting the digital signals into CT images;
the image denoising module is connected with the main control module and is used for denoising the CT image;
the image denoising module denoising method comprises the following steps:
iteratively training the convolutional neural network based on the calculated total loss function value until a training termination condition is met, and outputting a CT image denoising model obtained by training;
denoising the CT image to be denoised according to the CT image denoising model;
the image reconstruction module is connected with the main control module and is used for reconstructing the CT image;
and the display module is connected with the main control module and used for displaying CT images.
2. The CT-based high definition imaging method of claim 1, comprising the steps of:
step one, scanning a patient by using an X-ray beam through a scanning module;
step two, the main control module converts the scanning optical signal into an electric signal through the photoelectric conversion module;
step three, converting the electric signal into a digital signal through a digital-to-analog conversion module; converting the digital signal into a CT image through an image generating module; denoising the CT image through an image denoising module;
step four, reconstructing the CT image through an image reconstruction module; and displaying the CT image through a display module.
3. The CT-based high definition imaging device of claim 1 wherein the image denoising module denoising method is as follows:
(1) Acquiring a training sample set, wherein the training sample set comprises a noise CT image sample set and a noise-free CT image sample set corresponding to the noise CT image sample set; inputting the noise CT image sample set into a convolutional neural network for training, and outputting a denoising CT image sample set obtained by training;
(2) Respectively calculating a corresponding mean square error loss function value and a maximum posterior loss function value according to the noiseless CT image sample set and the denoising CT image sample set, and obtaining a total loss function value according to the mean square error loss function value and the maximum posterior loss function value;
the step of calculating the corresponding mean square error loss function value and the maximum posterior loss function value according to the noiseless CT image sample set and the denoising CT image sample set respectively comprises the following steps:
calculating a mean square error loss function value according to the noiseless CT image sample set and the denoising CT image sample set, wherein the mean square error loss function value is specifically as follows:
respectively calculating the square value of each pixel point difference value between each frame of noiseless CT image sample in the noiseless CT image sample set and the corresponding denoising CT image sample in the denoising CT image sample set, and taking the average value of the square values of the calculated pixel point difference values as the mean square error loss function value; and
calculating a maximum posterior loss function value according to the noiseless CT image sample set and the denoising CT image sample set, wherein the maximum posterior loss function value is specifically as follows:
calculating the posterior probability of the poisson noise of each noiseless CT image sample in the noiseless CT image sample set according to each noiseless CT image sample in the noiseless CT image sample set and the poisson noise CT image after the poisson noise is added into each noiseless CT image sample;
calculating the posterior probability of Gaussian noise existing in each Poisson noise CT image according to the Poisson noise CT image after the Poisson noise is added in each noiseless CT image sample and the Gaussian noise CT image after the Gaussian noise is added in each noiseless CT image sample;
calculating to obtain the maximum posterior probability of the poisson Gaussian noise of each noiseless CT image sample according to the posterior probability of the poisson noise of each noiseless CT image sample and the posterior probability of the Gaussian noise of each poisson noise CT image;
calculating a negative logarithmic function value taking the maximum posterior probability of the poisson Gaussian noise of each noiseless CT image sample as an independent variable, and taking the negative logarithmic function value as the maximum posterior loss function value;
(3) And iteratively training the convolutional neural network based on the calculated total loss function value until the training termination condition is met, outputting a CT image denoising model obtained by training, and denoising the CT image to be denoised according to the CT image denoising model.
4. The CT-based high definition imaging device of claim 3 wherein the step of acquiring a training sample set comprises:
acquiring a plurality of frames of noise CT image samples continuously shot by CT image acquisition equipment based on configured CT image acquisition parameters to construct a noise CT image sample set, wherein the CT image acquisition parameters comprise CT image exposure parameters and shooting frequency parameters;
sorting pixel values of each pixel point in the noise CT image sample set according to each frame of noise CT image sample in the noise CT image sample set, and removing abnormal pixel points with the pixel values smaller than a first preset pixel value and the pixel values larger than a second preset pixel value according to sorting results to obtain noise CT image samples with the abnormal pixel points removed;
calculating the average value of each pixel point of the noise CT image sample after eliminating abnormal pixel points of each frame;
and obtaining a noiseless CT image sample set corresponding to the noise CT image sample set according to the relation between the preset noise CT image sample and the corresponding noiseless CT image sample and the average value of each pixel point of the noise CT image sample after eliminating abnormal pixel points of each frame.
5. The CT-based high definition imaging apparatus of claim 3 wherein said step of inputting said noisy CT image sample set into a convolutional neural network for training and outputting a trained denoised CT image sample set comprises:
inputting the noise CT image sample set into a convolutional neural network, and sequentially extracting CT image characteristic information of each noise CT image sample in the noise CT image sample set through each layer of convolutional layer in the convolutional neural network;
and aiming at each layer of convolution layer, transmitting the CT image characteristic information of each noise CT image sample extracted by the convolution layer to deconvolution layers symmetrically arranged by the convolution layer through a connection layer, and generating a deconvoluted denoising CT image sample set according to the CT image characteristic information of each noise CT image sample through the deconvolution layers.
6. The CT-based high definition imaging apparatus of claim 3 wherein the step of deriving a total loss function value from the mean square error loss function value and a maximum a posteriori loss function value comprises:
respectively calculating a first weight loss function value of the mean square error loss function value and a second weight loss function value of the maximum posterior loss function value according to a preset weight proportion;
and obtaining the total loss function value according to the first weight loss function value and the second weight loss function value.
7. The CT-based high definition imaging device of claim 1 wherein the image reconstruction module reconstructs the method as follows:
1) Acquiring an original CT reconstruction image, wherein the original CT reconstruction image comprises windmill artifacts; dividing the frequency of the original CT reconstruction image to obtain a high-frequency image; frequency division combination is carried out on a plurality of thick images reconstructed according to raw data to obtain a low-frequency image, wherein the raw data comprise raw data corresponding to the original CT reconstructed image, and windmill artifacts are not included in each thick image;
2) Synthesizing an intermediate image according to the high-frequency image and the low-frequency image, and outputting the intermediate image as a target CT reconstructed image;
and carrying out frequency division combination on a plurality of thick images reconstructed according to the raw data to obtain the low-frequency image, wherein the method comprises the following steps of:
determining a thick image reconstruction parameter capable of eliminating windmill artifacts;
reconstructing a plurality of thick images by using raw data comprising raw data corresponding to the original CT reconstructed images according to the thick image reconstruction parameters;
and carrying out frequency division on each thick image to obtain a plurality of low-frequency thick images, and carrying out sharpening combination on the plurality of low-frequency thick images to obtain the low-frequency images.
8. The CT-based high definition imaging device of claim 7 wherein said dividing said original CT reconstructed image to obtain said high frequency image comprises:
performing Fourier transform on the original CT reconstruction image, and converting the original CT reconstruction image from a space domain to a frequency domain;
extracting high-frequency components in the frequency domain data of the original CT reconstructed image;
and performing inverse Fourier transform on the extracted high-frequency components to generate the high-frequency image.
9. The CT-based high definition imaging device of claim 7 wherein said extracting high frequency components in the frequency domain data of said original CT reconstructed image comprises:
calculating a low-frequency weight coefficient of each frequency position in the frequency domain data;
calculating a low-frequency value of each frequency position according to the value of each frequency position in the frequency domain data and the low-frequency weight coefficient;
the difference between the value of each frequency location and the low frequency value is calculated as the high frequency value of each frequency location,
the high frequency values of all frequency locations constitute high frequency components in the frequency domain data.
10. The CT-based high definition imaging device of claim 7 wherein the thick image reconstruction parameters include reconstruction interval, image thickness, image number;
reconstructing a plurality of thick images by using raw data comprising raw data corresponding to the original CT reconstructed images according to the thick image reconstruction parameters, wherein the method comprises the following steps: reconstructing a plurality of thick images along the direction of a scanning bed according to the reconstruction interval by using a group of raw data containing raw data corresponding to the original CT reconstructed images, wherein the thickness of each thick image is the same as that of the image, and the number of the thick images is consistent with that of the images;
frequency division is carried out on each thick image to obtain a plurality of low-frequency thick images, and the method comprises the following steps:
performing Fourier transform on the image to be divided, and converting the image to be divided into a frequency domain from a space domain;
extracting a low-frequency component in the frequency domain data of the thick image to be divided;
performing inverse Fourier transform on the extracted low-frequency component to generate a low-frequency thick image of the image to be divided;
extracting a low-frequency component in the frequency domain data of the thick image to be divided, including:
calculating a low-frequency weight coefficient of each frequency position in the frequency domain data;
calculating a low frequency value for each frequency location based on the value for each frequency location in the frequency domain data and the low frequency weight coefficient,
the low-frequency values of all frequency positions are formed into low-frequency components in the frequency domain data;
sharpening the plurality of low-frequency thick images to obtain the low-frequency image comprises the following steps:
determining weights corresponding to the plurality of low-frequency thick images to be combined;
multiplying the pixel value of each pixel point on each low-frequency thick image to be combined with the corresponding weight to obtain the weighted pixel value of each pixel point on each low-frequency thick image to be combined;
accumulating the weighted pixel values of the same pixel point on the plurality of low-frequency thick images to be combined to obtain accumulated pixel values;
the accumulated pixel values of all the pixel points form the low-frequency image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310452785.5A CN116630454A (en) | 2023-04-25 | 2023-04-25 | CT-based high-definition imaging method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310452785.5A CN116630454A (en) | 2023-04-25 | 2023-04-25 | CT-based high-definition imaging method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116630454A true CN116630454A (en) | 2023-08-22 |
Family
ID=87635576
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310452785.5A Pending CN116630454A (en) | 2023-04-25 | 2023-04-25 | CT-based high-definition imaging method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116630454A (en) |
-
2023
- 2023-04-25 CN CN202310452785.5A patent/CN116630454A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7444010B2 (en) | Method and apparatus for the reduction of artifacts in computed tomography images | |
| JP6021448B2 (en) | X-ray CT system | |
| US7023951B2 (en) | Method and apparatus for reduction of artifacts in computed tomography images | |
| US20070053477A1 (en) | Method and apparatus of global de-noising for cone beam and fan beam CT imaging | |
| US20110286651A1 (en) | Projection-Space Denoising with Bilateral Filtering in Computed Tomography | |
| US11315220B2 (en) | Distance measuring apparatus, vibration measuring apparatus, and industrial computed tomography apparatus | |
| CN112770838A (en) | System and method for image enhancement using self-attention depth learning | |
| US20070274581A1 (en) | Methods and apparatus for BIS correction | |
| CN106530236B (en) | Medical image processing method and system | |
| US20220130084A1 (en) | Systems and methods for medical image processing using deep neural network | |
| US20140100441A1 (en) | Medical image imaging method, medical diagnostic apparatus using the same, and recording medium therefor | |
| CN111899188A (en) | Neural network learning cone beam CT noise estimation and suppression method | |
| US10019818B2 (en) | Concurrent forward- and back-projection iterative reconstruction system | |
| CN114387359A (en) | Three-dimensional X-ray low-dose imaging method and device | |
| Kobler et al. | Variational deep learning for low-dose computed tomography | |
| CN114359126A (en) | Geometric correction method and device for three-dimensional image, electronic device and storage medium | |
| CN102890817A (en) | Method and system for processing medical images | |
| CN110246199B (en) | Projection domain data noise removing method for energy spectrum CT | |
| CN116630454A (en) | CT-based high-definition imaging method and device | |
| Zhao et al. | Low-dose CT image reconstruction via total variation and dictionary learning | |
| US10600159B2 (en) | Apparatus and method for enhancing spatial resolution of CT image and CT imaging system | |
| Abuazoum | Efficient analysis of medical image de-noising for MRI and ultrasound images | |
| CN111067561B (en) | Energy spectrum CT substance decomposition method and device, CT equipment and CT system | |
| CN113160080B (en) | CR image noise reduction method, device, equipment and medium | |
| US20230363724A1 (en) | X-ray ct apparatus and high-quality image generation device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |