DR image analysis method and DR imaging system

Abstract

The invention discloses a DR image analysis method and a DR imaging system. The method comprises the following steps: generating a digital X-ray radiography DR image; extracting image features from the DR image, wherein the image features comprise at least one of a gray scale entropy feature, a texture feature, a noise feature, a gradient feature and a divergence feature; converting the DR image into an energy space to obtain an energy distribution diagram of the DR image; extracting energy characteristics from the energy distribution diagram, wherein the energy characteristics comprise at least one of an energy density characteristic, an energy gravity center characteristic and an energy effective value characteristic; obtaining a DR feature index according to the at least one image feature and the at least one energy feature, wherein the DR feature index is used for representing the comprehensive level between the radiation dose of the X-ray and the image quality of the DR image; and outputting the DR characteristic index. The DR image index is obtained according to the image features and the energy features of the DR image, and the comprehensive level between the X-ray radiation dose and the image quality of the DR image can be reflected.

Description

Technical Field

[0001] The present application relates to the field of medical image technology, and more specifically to a DR image analysis method and a DR imaging system. Background Art

[0002] Digital Radiography (DR) images are a common type of medical digital images and are widely used in physical examinations and routine medical imaging diagnosis. With the development of digital X-ray detectors and digital image processing systems, DR images can often present diagnostic image effects within a wide range of exposure doses. However, this process often relies on the experience of the operator to grasp the final image effect, and the accuracy of the judgment is easily affected by various factors such as experience differences, subjective differences, and post-processing.

[0003] In order to determine the exposure dose of DR images, one way is to use the exposure index (EI) to indicate the size of a single exposure. However, the exposure index is usually simply calculated based on the grayscale information of the DR image and cannot adapt well to complex and changing clinical scenarios. Summary of the invention

[0004] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further described in detail in the Detailed Description of the Invention section. The Summary of the Invention section of this application does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the scope of protection of the claimed technical solution.

[0005] An embodiment of the present invention provides a method for analyzing a DR image, the method comprising:

[0006] emitting X-rays to a target part of a target object so that the X-rays pass through the target part;

[0007] receiving X-rays after passing through the target part, and generating a digital X-ray photography DR image according to the X-rays after passing through the target part, wherein the DR image includes at least one of a DR original image and an image after the DR original image is processed;

[0008] Extracting image features from the DR image, the image features comprising at least one of grayscale entropy features, texture features, noise features, gradient features, and divergence features;

[0009] transforming the DR image into an energy space to obtain an energy distribution map of the DR image;

[0010] Extracting energy features from the energy distribution diagram, the energy features comprising at least one of energy density features, energy center of gravity features, and energy effective value features;

[0011] Obtaining a DR feature index according to at least one of the image features and at least one of the energy features, wherein the DR feature index is used to characterize a comprehensive level between the radiation dose of the X-ray and the image quality of the DR image;

[0012] The DR characteristic index is output.

[0013] In one embodiment, transforming the DR image into an energy space to obtain an energy distribution map of the DR image includes:

[0014] An orthogonal transformation is performed on the DR image to obtain the energy distribution map.

[0015] In one embodiment, the energy feature includes the energy density feature, and extracting the energy density feature from the energy distribution diagram includes:

[0016] The energy density feature is obtained according to the energy distribution concentration and dispersion trend in the energy distribution diagram.

[0017] In one embodiment, the energy feature includes the energy center of gravity feature, and extracting the energy center of gravity feature from the energy distribution diagram includes:

[0018] The energy center of gravity feature is obtained according to the energy concentration degree in the energy distribution diagram.

[0019] In one embodiment, the energy feature includes the energy effective value feature, and extracting the energy effective value feature from the energy distribution diagram includes:

[0020] The energy effective value feature is obtained according to the overall energy level of the energy distribution diagram.

[0021] In one embodiment, the image feature includes the grayscale entropy feature, and extracting the grayscale entropy feature from the DR image includes:

[0022] Obtaining a first probability statistical distribution of each grayscale source in the DR image;

[0023] Determining a first probability statistical distribution of a non-redundant grayscale information source from the first probability statistical distribution of the grayscale information source according to the size of the first probability statistical distribution;

[0024] Obtaining a second probability statistical distribution of each non-redundant grayscale information source according to a ratio of the first probability statistical distribution of each non-redundant grayscale information source to the sum of the first probability statistical distributions of all non-redundant grayscale information sources;

[0025] The entropy of the second probability statistical distribution of the non-redundant grayscale information source is calculated to obtain the grayscale entropy feature.

[0026] In one embodiment, the image feature includes the texture feature, and extracting the texture feature from the DR image includes:

[0027] A texture feature description matrix is ​​obtained according to the gray value of each pixel in the DR image, wherein the texture feature description matrix describes the distance between different pixel points and the change of the gray value in the direction between different pixel points in the DR image;

[0028] At least one two-dimensional component of the texture feature description matrix is ​​extracted as the texture feature.

[0029] In one embodiment, the image feature includes the noise feature, and extracting the noise feature from the DR image includes:

[0030] Performing filtering on the DR image to obtain a high-frequency image;

[0031] Calculating the root mean square of the neighborhood of each pixel in the high-frequency image to obtain a noise distribution image;

[0032] The noise value distribution in the noise distribution image is statistically analyzed to obtain the noise feature.

[0033] In one embodiment, the image feature includes the gradient feature, and extracting the gradient feature from the DR image includes:

[0034] Determining the regional distribution of different tissues in the DR image;

[0035] The clarity of the boundaries of different tissue region distributions is obtained as the gradient feature.

[0036] In one embodiment, the image feature includes the divergence feature, and extracting the divergence feature from the DR image includes:

[0037] Determining the regional distribution of different tissues in the DR image;

[0038] The transition strength and / or trend consistency of the boundaries of different tissue region distributions are obtained as the divergence feature.

[0039] In one embodiment, obtaining the DR feature index according to at least one of the image features and at least one of the energy features includes:

[0040] At least one of the image features and at least one of the energy features are input into a target model to obtain the DR feature index output by the target model.

[0041] In one embodiment, the target model includes a pre-trained network model, and the training process of the network model includes:

[0042] Acquire a DR image set, where the DR image set includes a plurality of DR images;

[0043] Obtaining an energy feature set and an image feature set according to the DR image set, wherein the energy feature set includes a plurality of energy features extracted from the plurality of DR images, and the image feature set includes a plurality of image features extracted from the plurality of DR images;

[0044] Obtaining a score set according to the DR image set, wherein the score set includes scores obtained by evaluating the plurality of DR images;

[0045] The network model is trained using the energy feature set, the image feature set and the score set as a training sample set to obtain a trained network model.

[0046] Another aspect of the present invention provides a method for analyzing a DR image, the method comprising:

[0047] Acquire a DR image, wherein the DR image includes at least one of an original DR image and an image obtained by processing the original DR image;

[0048] Extracting image features from the DR image, the image features comprising at least one of grayscale entropy features, texture features, noise features, gradient features, and divergence features;

[0049] Extracting energy features from the DR image, the energy features comprising at least one of energy density features, energy center of gravity features, and energy effective value features;

[0050] Obtaining a DR feature index according to at least one of the image features and at least one of the energy features, wherein the DR feature index is used to characterize a comprehensive level between the radiation dose of the X-ray and the image quality of the DR image;

[0051] The DR characteristic index is output.

[0052] In one embodiment, extracting energy features from the DR image includes:

[0053] The DR image is input into a pre-trained feature extraction model to obtain the energy feature,

[0054] Alternatively, the DR image is transformed into an energy space to obtain an energy distribution map of the DR image, and the energy feature is extracted from the energy distribution map of the DR image.

[0055] Another aspect of an embodiment of the present invention provides a DR imaging system, the DR imaging system comprising: a transmitting device, a receiving device, a processor and an output device;

[0056] The transmitting device is used to transmit X-rays to a target part of a target object so that the X-rays pass through the target part;

[0057] The receiving device is used to receive the X-rays after passing through the target part;

[0058] The processor is used to execute the DR image analysis method as described above to obtain a DR feature index;

[0059] The output device is used to output the DR characteristic index.

[0060] The DR image analysis method and DR imaging system of the embodiment of the present invention extract the image features and energy features of the DR image, and then obtain the DR image index, which can objectively reflect the comprehensive level between the X-ray radiation dose and the image quality of the DR image, thereby providing users with an objective judgment basis, and can be used for in-hospital quality control and guidance of dose reduction imaging and other directions. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0062] Figure 1 A schematic flow chart showing a method for analyzing a DR image according to an embodiment of the present invention;

[0063] Figure 2 A schematic diagram showing model training and DR imaging analysis using the trained model according to an embodiment of the present invention;

[0064] Figure 3 A schematic flow chart showing a method for analyzing a DR image according to another embodiment of the present invention;

[0065] Figure 4 A schematic block diagram of a DR imaging system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions and advantages of the present application more obvious, the example embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application described in the present application, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present application.

[0067] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present application. However, it is apparent to those skilled in the art that the present application can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present application, some technical features well known in the art are not described.

[0068] It should be understood that the present application can be implemented in different forms and should not be interpreted as being limited to the embodiments set forth herein. On the contrary, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0069] The purpose of the terms used herein is only to describe specific embodiments and is not intended to be limiting of the present application. When used herein, the singular forms "one", "an" and "said / the" are also intended to include plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "consisting of" and / or "comprising", when used in this specification, determine the presence of the features, integers, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts and / or groups. When used herein, the term "and / or" includes any and all combinations of the relevant listed items.

[0070] In order to thoroughly understand the present application, a detailed structure will be proposed in the following description to illustrate the technical solution proposed by the present application. The optional embodiments of the present application are described in detail as follows, but in addition to these detailed descriptions, the present application may also have other implementation methods.

[0071] Next, first refer to Figure 1 A method 100 for analyzing a DR image according to an embodiment of the present application is described. Figure 1 As shown, the DR image analysis method 100 may include the following steps:

[0072] In step S110, X-rays are emitted to a target part of a target object so that the X-rays pass through the target part;

[0073] In step S120, receiving the X-ray after passing through the target part, and generating a digital X-ray photography DR image according to the X-ray after passing through the target part, wherein the DR image includes at least one of a DR original image and an image after the DR original image is processed;

[0074] In step S130, image features are extracted from the DR image, wherein the image features include at least one of grayscale entropy features, texture features, noise features, gradient features, and divergence features;

[0075] In step S140, the DR image is transformed into an energy space to obtain an energy distribution map of the DR image;

[0076] In step S150, energy features are extracted from the energy distribution diagram, wherein the energy features include at least one of energy density features, energy center of gravity features, and energy effective value features;

[0077] In step S160, a DR feature index is obtained according to at least one of the image features and at least one of the energy features, wherein the DR feature index is used to characterize a comprehensive level between the radiation dose of the X-ray and the image quality of the DR image;

[0078] In step S170, the DR feature index is output.

[0079] The DR image analysis method 100 of the embodiment of the present invention extracts the image features and energy features of the DR image, and then obtains the DR image index, which can objectively reflect the comprehensive level between the X-ray radiation dose and the image quality of the DR image, thereby providing users with an objective judgment basis, and can be used for in-hospital quality control and guidance of dose reduction imaging and other directions.

[0080] In step S110, the DR imaging system is controlled to emit X-rays to the target tissue part of the target object, and the X-rays are controlled to pass through the target tissue part; the X-rays that pass through the target tissue part are received, and the X-rays that pass through the target tissue part are processed to obtain a digital X-ray photography DR image. The target tissue part can be a tissue part to be examined in a human body or other animal body, such as the head, abdomen, etc.

[0081] In step S120, the X-rays after passing through the target part are received, and a DR image is generated according to the X-rays after passing through the target part. The DR image of the embodiment of the present invention may be at least one of a DR original image and an image after the DR original image is processed. In the DR imaging process, the X-rays emitted by the X-ray generator are attenuated to varying degrees after passing through the human body and are received by the detector. Exemplarily, the upper layer of the detector is an X-ray conversion medium, which has photoconductivity and can convert X-rays into electronic signals, generating positive and negative charges, which move along the electric field in the form of current under the action of the bias voltage and are collected by the detector array of the detector, each detection unit corresponding to a pixel of the DR image. A digital signal, i.e., a DR original image, can be obtained by amplifying and performing analog-to-digital conversion on the electrical signal. The DR original image is a grayscale image that has not been post-processed by contrast adjustment, brightness adjustment, etc. The processed image of the DR original image may be an image of the DR original image after a certain transformation relationship, or an image of image contrast, image brightness adjustment, etc., i.e., the processed image of the DR original image may be an image of the DR original image after any post-processing, which is not specifically limited here.

[0082] In step S120, image features are extracted from the DR image, and the image features include at least one of grayscale entropy features, texture features, noise features, gradient features, and divergence features. That is, the image features extracted in step S120 may be all of the grayscale entropy features, texture features, noise features, gradient features, and divergence features, or any one or more of them. Among them, the grayscale entropy features, texture features, noise features, gradient features, and divergence features reflect the grayscale information, texture information, noise information, gradient information, and divergence information of the DR image, respectively, and multi-dimensional image features are conducive to obtaining a more accurate DR feature index.

[0083] In one embodiment, the grayscale entropy feature is the information content obtained by statistically screening out redundant grayscale information sources in the DR image. According to the transmission characteristics of X-ray quanta converted into digital signals by the detector of the DR system, each independent grayscale in the DR image can be regarded as a source, and the DR imaging process can be regarded as a process of transmitting information through grayscale information sources. Among all the grayscales, there are some grayscales that do not have grayscale values, so they are regarded as redundant grayscale information sources. Since the extraction process of the grayscale entropy feature of the embodiment of the present invention screens out redundant grayscale information sources, more effective image information can be extracted.

[0084] For example, the grayscale entropy feature can be extracted from the DR image in the following manner:

[0085] First, the first probability statistical distribution pH(i) of each grayscale source in the DR image is obtained, where i is the image grayscale value of the DR image. Then, according to the size of the first probability statistical distribution, the first probability statistical distribution pNH(i) of the non-redundant grayscale source is determined from the first probability statistical distribution pH(i) of the grayscale source. Exemplarily, if the first probability statistical distribution pH(i) is greater than 0, it is determined to be the first probability statistical distribution pNH(i) of the non-redundant grayscale source. Conversely, if the first probability statistical distribution pH(i) is greater than 0, it is determined to be the first probability statistical distribution pNH(i) of the non-redundant grayscale source, that is:

[0086]

[0087] After obtaining the first probability statistical distribution pNH(i) of the non-redundant grayscale information source, the second probability statistical distribution pH'(i) of each non-redundant grayscale information source is obtained according to the ratio of the first probability statistical distribution pNH(i) of each non-redundant grayscale information source to the sum of the first probability statistical distributions ∑pNH(i) of all non-redundant grayscale information sources, that is:

[0088] pH'(i)=pNH(i) / ∑pNH(i) Formula (2)

[0089] Finally, entropy is calculated based on the second probability statistical distribution pH'(i) of the non-redundant grayscale information source to obtain the grayscale entropy feature H(Image). The method of calculating entropy includes but is not limited to taking the natural logarithm ln with e as the base or taking the logarithm with other values ​​as the base, which is expressed as follows:

[0090]

[0091] Thus, the grayscale entropy feature can be obtained after filtering out the redundant grayscale signal sources.

[0092] Texture features represent the spatial variation relationship of different grayscale values ​​on the DR image, which can reflect the abstract features of the law of changes in human tissue. For example, a statistical method can be used to extract the texture features of the DR image, which obtains the statistical characteristics of the texture area based on the grayscale attributes of the pixel and its neighborhood. The following mainly describes the specific details of the statistical method for extracting texture features, but the method for extracting texture features can also include geometric methods, model methods or other suitable texture feature extraction methods.

[0093] Exemplarily, when using statistical methods to extract texture features from a DR image, a texture feature description matrix is ​​first obtained based on the grayscale value of each pixel in the DR image; then, at least one two-dimensional component of the texture feature description matrix is ​​extracted as at least one texture feature.

[0094] In one embodiment, the texture feature description matrix describes the distance between different pixels and the change of grayscale values ​​in the direction between different pixels in the DR image. Assuming that the DR image Image = f(x, y), its texture feature description matrix P(i, j) is:

[0095] P(i,j)=#{(x1,y1),(x2,y2)∈M*N|f(x1,y1)=i,f(x2,y2)=j} Formula (4)

[0096] Among them, #(x) represents the number of elements in the set x. Assuming that the distance between two points (x1, y1) and (x2, y2) in the image is k, the texture feature description matrix in different directions l can be expanded to P(i, j, k, l). The texture feature description matrix can be expanded by counting the changes in grayscale values ​​at different distances and directions to obtain information in more dimensions. For example, in order to calculate the actual eigenvalues ​​at different angles, super-resolution interpolation can be performed on the DR image in each direction to obtain the sub-pixel grayscale values ​​corresponding to the new target position, and the texture feature description matrix in different directions can be obtained according to the sub-pixel grayscale values.

[0097] After obtaining the texture feature description matrix, extract at least one two-dimensional component of the texture feature description matrix to obtain a texture feature, where the texture feature reflects the key characteristics of the texture feature description matrix. Exemplarily, the texture feature includes at least one of the following:

[0098] The first texture feature P1 is used to count the value distribution of the texture feature description matrix and the overall distribution of texture changes in the DR image; the value of P1 can reflect the clarity of the DR image and the depth of the texture grooves. The deeper the texture grooves, the larger the P1 value, and vice versa, the shallower the grooves, the smaller the P1 value;

[0099] The second texture feature P2 is used to count the similarity between the value distribution in the texture feature description matrix and the parallel and normal directions in the DR image. The larger the P2 value, the greater the similarity of the grayscale of the image in different directions.

[0100] The third texture feature P3 is used to count the value distribution in the texture feature description matrix and the uniformity of the grayscale change distribution in the DR image. The size of the P3 value can reflect the uniformity of the image grayscale distribution and the coarseness of the texture. The larger the P3 value, the more stable the texture change of the DR image.

[0101] The fourth texture feature P4 is used to measure the distribution of values ​​in the texture feature description matrix and the local texture variation of the DR image. The larger the P4 value is, the stronger the regularity of the texture is.

[0102] The texture features extracted from the texture feature description matrix are not limited to the above four types. In other embodiments, other two-dimensional components of the texture feature description matrix may also be extracted as texture features.

[0103] Image noise mainly includes X-ray quantum noise, and the distribution of X-ray quanta obeys Poisson distribution, and its variance is proportional to the average number of quantum detections. Based on this characteristic of X-ray quantum noise, the fluctuation degree of X-ray quanta can be statistically analyzed from the DR image as the noise feature, that is, the noise feature reflects the fluctuation degree of X-ray quanta in the DR image.

[0104] In one embodiment, extracting noise features from a DR image includes the following steps: filtering the DR image to obtain a high-frequency image, wherein the high-frequency image obtained from the DR image mainly contains noise information; calculating the root mean square of the neighborhood of each pixel point in the high-frequency image to obtain a noise distribution image; and statistically analyzing the distribution of noise values ​​in the noise distribution image to obtain the noise features.

[0105] As an implementation, filtering the DR image to obtain a high-frequency image includes: performing low-frequency filtering on the DR image I to filter out the low-frequency components therein to obtain a low-frequency image I1; and obtaining the high-frequency image I2 according to the difference between the low-frequency image and the DR image. In other implementations, the DR image I may also be directly subjected to high-frequency filtering to obtain the high-frequency image I2.

[0106] Exemplarily, the DR image I can be subjected to Gaussian low-pass filtering to obtain a low-frequency image I1. Since the image is a two-dimensional signal, a two-dimensional Gaussian function is used for Gaussian low-pass filtering, and the two-dimensional Gaussian filter kernel is:

[0107]

[0108] The high-frequency image I2 can be obtained by calculating the difference between the low-frequency image I1 and the image I, that is, I2=I1-I. The noise distribution image I3 can be obtained by extracting the effective information in the high-frequency image I2. In one embodiment, the noise distribution image I3 is an image formed by calculating the root mean square of the neighborhood of each pixel point in the high-frequency image. The method for calculating the root mean square includes using the L1 norm or other approximate or equivalent measures, for example:

[0109]

[0110] Among them, I3(i,j) is the value of the noise distribution image I3 at the pixel point (i,j), and I2(l,k) is the value of the high frequency I2 at the pixel point (l,k).

[0111] After acquiring the noise distribution image I3, the noise value distribution thereof is statistically analyzed to obtain the acoustic feature. For example, the noise value interval with the highest noise value distribution probability in the noise distribution image can be determined, and the noise value of the noise value interval with the highest noise value distribution probability is used as the noise feature.

[0112] Specifically, first, the pixel values ​​of the noise distribution image I3 are divided into M intervals, the pixel value interval of each interval is d, the initialization histogram vector is h, the length is M, and each component h(i) of the initialization histogram represents the number of pixel values ​​in the i-th interval. For the pixel point (i, j) in the noise distribution image I3, the interval corresponding to the pixel value of this point is R[I3(i, j)]. After traversing each pixel point of the noise distribution image I3 and counting h(R(I3(i, j))), after obtaining h, the maximum value of its main peak is max(h), corresponding to the interval R0=arg max(h), then the noise distribution probability R0*d is calculated as the noise feature.

[0113] In some embodiments, the image features also include gradient features. Due to the different absorption coefficients of X-rays after penetrating different tissues of the human body, regional distributions of different tissues will be formed in the DR image, and there will be different degrees of boundaries between different tissue regions. The clarity of these boundaries can be quantified by extracting gradient features, that is, gradient features are used to characterize the clarity of boundaries between different tissues in the DR image. In one embodiment, extracting gradient features from a DR image includes: determining the regional distribution of different tissues in the DR image; obtaining the clarity of the boundaries of different tissue regional distributions as gradient features.

[0114] Exemplarily, the gradient feature can be calculated as follows:

[0115]

[0116] Among them, Grad(x,y) is used to represent the gradient size at the coordinates (x,y), and Image(x,y) is used to represent the pixel value at the coordinates (x,y).

[0117] In some embodiments, the image features also include divergence features. In DR images, the boundaries of different tissues are not strict boundaries, but there is a certain degree of transition. The strength and / or trend consistency of this transition can be quantified by extracting the divergence features, that is, the divergence features are used to characterize the strength and / or trend consistency of the transition of the boundaries of different tissues in the DR image. In one embodiment, extracting the divergence features from the DR image includes: determining the regional distribution of different tissues in the DR image; obtaining the strength and / or trend consistency of the transition of the boundaries of the regional distribution of different tissues as the divergence features.

[0118] For example, the divergence feature can be calculated as follows:

[0119]

[0120] Among them, Diver(x,y) is used to represent the divergence size at the coordinates (x,y), Image(x,y) is used to represent the pixel value at the coordinates (x,y), and actan represents the inverse tangent function.

[0121] In some embodiments, in addition to at least one of the above image features, other image features can also be extracted from the DR image for obtaining the DR feature index, such as grayscale gradient, grayscale mean, variance, pixel value information, signal-to-noise ratio, contrast-to-noise ratio, etc. The specific features can be determined according to actual conditions and are not specifically limited here.

[0122] In step S140, the DR image is transformed into the energy space to obtain an energy distribution map of the DR image. By transforming the DR image from the spatial domain to the energy domain, features that are not prominent in the spatial domain can be highlighted in the energy domain, which is helpful for extracting features of the DR image.

[0123] Among them, the energy distribution map can be obtained by performing an orthogonal transformation on the DR image. The purpose of orthogonal transformation is to concentrate the energy of the image on a small number of coefficients as much as possible, thereby removing the correlation in the original data to the maximum extent, so as to facilitate image feature extraction. Orthogonal transformation includes Fourier transform, discrete cosine transform, wavelet transform, etc.

[0124] Optionally, the following function can be used to perform orthogonal transformation on the DR image:

[0125]

[0126] Wherein, I(x, y) is the pixel matrix of the DR image, M and N are the two-dimensional dimensions of the DR image, and F(u, v) is the energy distribution matrix. The transformation function used in the orthogonal transformation can be in other function forms, which are not specifically limited here.

[0127] After the energy distribution map is obtained, in step S150, energy features are extracted from the energy distribution map, and the energy features reflect the energy distribution features of the DR image. The energy features include at least one of energy density features, energy center of gravity features, and energy effective value features.

[0128] Among them, the energy density feature (Den) is used to reflect the energy distribution trend of the DR image in the energy space, which can correspond to the distribution difference of different energy segments in the DR image. Therefore, the energy density feature can be obtained according to the energy distribution trend in the energy distribution diagram. It can be calculated according to the following formula:

[0129]

[0130] Among them, the higher the Den value, the more concentrated the energy distribution of the DR image is in the high energy segment.

[0131] The energy center of gravity feature is used to reflect the energy concentration level of the DR image in the energy space, and can correspond to the characteristics of the energy concentration level of the DR image. Therefore, the energy center of gravity feature can be obtained according to the energy concentration level in the energy distribution diagram. It can be calculated according to the following formula:

[0132]

[0133] Among them, the higher the Gra value, the higher the energy concentration of the DR image.

[0134] The energy effective value feature is used to reflect the overall energy level of the DR image in the energy space, and can correspond to the overall energy level measurement of the DR image. Therefore, the energy effective value feature can be obtained according to the overall energy level of the energy distribution diagram. Specifically, it can be calculated according to the following formula:

[0135]

[0136] Among them, the higher the value of Val is, the higher the overall energy level of the DR image is.

[0137] Thereafter, in step S160, a DR feature index is obtained according to at least one image feature and at least one energy feature, and the DR feature index is used to characterize the comprehensive level between the radiation dose of the X-ray and the image quality of the DR image.

[0138] Specifically, at least one image feature among grayscale entropy feature, texture feature, noise feature, gradient feature and divergence feature and at least one energy feature among energy density feature, energy center of gravity feature and energy effective value feature can be input into the target model, and DR image index can be output. The DR image index can objectively reflect the comprehensive level between the radiation dose of X-rays received by the target object and the image quality of the DR image, thereby providing an objective judgment basis for the operator of the DR imaging device, and can guide the direction of in-hospital quality control and dose reduction imaging.

[0139] Among them, the target model can be trained in an offline manner, such as Figure 2 As shown, the training method mainly includes:

[0140] First, a DR image set is obtained, and the DR image set includes multiple DR images, that is, DR images used as training samples. According to the DR image set, an energy feature set and an image feature set can be obtained. The energy feature set includes energy features extracted from multiple DR images, and the image feature set includes image features extracted from multiple DR images. Exemplarily, for each DR image in the DR image set, at least one image feature among grayscale entropy features, texture features, noise features, gradient features, and divergence features is extracted respectively, and multiple image features extracted from multiple DR images in the DR image set together constitute an image feature set, and each DR image in the DR image set corresponds to at least one sample image feature in the image feature set. The method for extracting image features from DR images can refer to the method for extracting image features from DR images in step S130. Furthermore, for each DR image in the DR image set, at least one energy feature among energy density feature, energy center of gravity feature and energy effective value feature is extracted respectively, and multiple energy features extracted from multiple DR images in the DR image set together constitute an energy feature set, and each DR image in the DR image set corresponds to at least one energy feature in the energy feature set. The method for extracting energy features from DR images can refer to the method for extracting energy features from DR images in step S150.

[0141] A scoring set can also be obtained according to the DR image set, and the scoring set includes the scores obtained by evaluating multiple DR images. Exemplarily, if the DR images in the DR image set are DR original images, the multiple DR images in the DR image set can be image processed separately to obtain multiple DR diagnostic images, that is, visual images used by doctors for diagnosis, and the multiple DR diagnostic images constitute a DR diagnostic image set, and the above-mentioned scoring set is obtained according to the DR diagnostic image set, and the scoring set includes the scores obtained by evaluating multiple DR diagnostic images. In some embodiments, clinical experts can evaluate the DR diagnostic images in the DR diagnostic image set according to the quality of the DR diagnostic images and give expert scores. The higher the score of the DR diagnostic image, the higher the quality of the corresponding DR original image; conversely, the lower the score of the DR diagnostic image, the lower the quality of the corresponding DR original image. The expert scores of the DR diagnostic images constitute an expert scoring set. Optionally, if the DR images in the DR image set are visual images used by doctors for diagnosis, the scoring set can be directly obtained according to the DR image set.

[0142] Afterwards, the network model is trained using the acquired image feature set, energy feature set and score set as training sample sets to obtain a trained network model. The network model may be a traditional machine learning model or a deep learning model, including but not limited to neural networks, support vector machines, linear discriminant analysis, etc.; the training methods include but are not limited to linear regression, gradient descent and other model training methods. Exemplarily, an optimal mapping function from image features to scores can be learned so that the error between the DR feature index obtained by the image feature mapping and the actual calibrated expert score is minimized. By executing this optimal mapping function for the image features acquired in step S130 and the energy features obtained in step S150, a prediction result of the DR feature index closest to the expert score can be obtained.

[0143] In one embodiment, the network model includes three fully connected layers, the fully connected layer includes a fully connected network and an activation function, and the ReLU function is used for activation. The last fully connected layer has no activation function. The network model uses the mean squared error (MSE) loss function, which is the square of the difference between the predicted value and the target value. The calculation formula of MSE is as follows:

[0144]

[0145] Where n is the number of data points, and MSE is the mean of the sum of squares of the errors between the predicted data and the original data. The image feature set, energy feature set, and score set are imported into the network model for training, and the model training parameters are adjusted until the network converges. In addition, the network model can also use an open source neural network model, some variants of a neural network model, or a nested set of multiple neural network models.

[0146] After obtaining the trained target model in the model training stage, in the actual application process, the image features obtained in S130 and the energy features obtained in step S150 are input into the trained network model, and a DR feature index for characterizing the comprehensive level between the radiation dose of X-rays received by the target object and the image quality of the DR image can be obtained, and the DR feature index can be displayed through a display device or output in other ways. It can be understood that the comprehensive level in this article refers to the quality of the DR image and the comprehensive level of the radiation dose of the X-rays received. If the user thinks that the image quality is good and meets the requirements, the DR feature index can be lowered to achieve the effect of reducing the radiation dose of the X-rays received; similarly, if the image quality is considered to be poor, the DR feature index can be increased to achieve the effect of improving the image quality.

[0147] Based on the above description, the DR image analysis method 100 according to the embodiment of the present application obtains a DR feature index that can objectively reflect the comprehensive level between the radiation dose of X-rays received by the target object and the image quality of the DR image based on the image features and energy features of the DR image, thereby providing an objective judgment basis for the operator.

[0148] Below, reference Figure 3 A method 300 for analyzing a DR image according to another embodiment of the present application is described. Figure 3 As shown, the DR image analysis method 300 may include the following steps:

[0149] In step S310, a DR image is acquired, wherein the DR image includes at least one of an original DR image and an image obtained by processing the original DR image;

[0150] In step S320, image features are extracted from the DR image, wherein the image features include at least one of grayscale entropy features, texture features, noise features, gradient features, and divergence features;

[0151] In step S330, energy features are extracted from the DR image, where the energy features include at least one of energy density features, energy center of gravity features, and energy effective value features;

[0152] In step S340, a DR feature index is obtained according to at least one of the image features and at least one of the energy features, wherein the DR feature index is used to characterize a comprehensive level between the radiation dose of the X-ray and the image quality of the DR image;

[0153] In step S350, the DR feature index is output.

[0154] Exemplarily, acquiring a DR image may include acquiring a DR image locally from a DR imaging system, or may include acquiring a DR image from other external devices, which is not specifically limited herein. Acquiring from the local may be acquiring a DR image in real time by the DR imaging system, or may be acquiring a DR image that is acquired non-real time and stored locally in the DR imaging system.

[0155] In one embodiment, extracting energy features from a DR image includes transforming the DR image into an energy space to obtain an energy distribution map of the DR image, and extracting energy features from the energy distribution map of the DR image. For details, reference may be made to the DR image analysis method 100. In addition, the DR image may also be input into a pre-trained feature extraction model to obtain energy features.

[0156] The DR image analysis method 300 of the embodiment of the present application obtains a DR feature index that can objectively reflect the comprehensive level between the radiation dose of X-rays received by the target object and the image quality of the DR image based on the image features and energy features of the DR image, thereby providing an objective basis for judgment for the operator. For more specific details of the DR image analysis method 300, please refer to the relevant description above, which will not be repeated here.

[0157] Another embodiment of the present invention provides a DR imaging system, see Figure 4 The DR imaging system 400 includes a transmitting device 410, a receiving device 420, a processor 430 and an output device 440. The transmitting device 410 is used to transmit X-rays to a target part of a target object, so that the X-rays pass through the target part; the receiving device 420 is used to receive the X-rays after passing through the target part; the processor 430 is used to perform the above-mentioned DR image analysis method to obtain a DR feature index; the output device 440 is used to output the DR feature index

[0158] Exemplarily, the transmitting device 410 includes an X-ray generator, which includes a tube for transmitting X-rays; the receiving device 420 includes a flat panel detector, specifically including an X-ray conversion medium located in the upper layer and a detection unit array located in the lower layer, the X-ray conversion medium converts X-rays into electronic signals, generates positive and negative charges, these charges move along the electric field in the form of current under the action of the bias voltage, are collected by the detection unit array, and converted into electrical signals. Flat panel detectors include but are not limited to amorphous silicon, amorphous selenium, CCD, etc. When using the DR imaging system 400 for image acquisition, it is necessary to fix the position of the transmitting device 410 and the position of the receiving device 420 in a specified manner, so that the X-rays emitted by the transmitting device 410 can irradiate the receiving device 420 after passing through the target tissue, so that the receiving device 420 can collect the X-rays after passing through the target tissue site to obtain a DR image.

[0159] The processor 430 is used to process the X-rays after passing through the target tissue site to obtain a digital X-ray photography DR image. The processor 430 is also used to execute the above-mentioned DR image analysis method 100 or DR image analysis method 300 to obtain a DR characteristic index reflecting the comprehensive level between the radiation dose of the X-rays received by the target object and the image quality of the DR image; the display 440 is used to display the DR characteristic index. The specific details of the DR image analysis method can be referred to above. According to the DR imaging system 400 of the embodiment of the present application, based on the image features and energy features of the DR image, obtains a DR characteristic index that can objectively reflect the comprehensive level between the radiation dose of the X-rays received by the target object and the image quality of the DR image, thereby providing an objective judgment basis for the operator.

[0160] In addition, according to an embodiment of the present invention, a computer storage medium is also provided, on which program instructions are stored, and when the program instructions are executed by a computer or a processor, the corresponding steps of the DR image analysis method of any embodiment of the present invention are executed. In some embodiments, the computer storage medium is a non-volatile computer-readable storage medium, which may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required by at least one function; the data storage area may store data created during the execution of program instructions, etc. In addition, the non-volatile computer-readable storage medium may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the non-volatile computer-readable storage medium may optionally include a memory remotely arranged relative to the processor.

[0161] Exemplarily, the computer storage medium may include, for example, a hard disk of a personal computer, a storage component of a tablet computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, a memory card of a smart phone, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

[0162] In addition, according to an embodiment of the present invention, a computer program is also provided, which can be stored in a cloud or local storage medium. When the computer program is executed by a computer or a processor, it is used to perform the corresponding steps of the DR image analysis method of the embodiment of the present invention.

[0163] Although example embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above example embodiments are merely exemplary and are not intended to limit the scope of the present application to this. Those of ordinary skill in the art may make various changes and modifications therein without departing from the scope and spirit of the present application. All these changes and modifications are intended to be included within the scope of the present application as required by the appended claims.

[0164] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0165] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0166] In the description provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present application can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0167] Similarly, it should be understood that in order to streamline the present application and help understand one or more of the various inventive aspects, in the description of the exemplary embodiments of the present application, the various features of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the method of the present application should not be interpreted as reflecting the following intention: the claimed application requires more features than the features clearly stated in each claim. More specifically, as reflected in the corresponding claims, the inventive point is that the corresponding technical problem can be solved with features less than all the features of a single disclosed embodiment. Therefore, the claims following the specific embodiment are hereby explicitly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the present application.

[0168] It will be understood by those skilled in the art that, except for mutually exclusive features, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this specification may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0169] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present application and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0170] The various component embodiments of the present application can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all of the functions of some modules in the article analysis device according to the embodiment of the present application. The present application can also be implemented as a device program (e.g., computer program and computer program product) for executing part or all of the methods described herein. Such a program implementing the present application can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0171] It should be noted that the above embodiments illustrate the present application rather than limit the present application, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets should not be constructed as a limitation to the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of multiple such elements. The present application may be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In a unit claim that lists several devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names.

[0172] The above is only a specific implementation or description of a specific implementation of the present application, and the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. The protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A DR image analysis method, characterized in that: The method comprises: emitting X-rays to a target part of a target object so that the X-rays pass through the target part; receiving X-rays after passing through the target part, and generating a digital X-ray photography DR image according to the X-rays after passing through the target part, wherein the DR image includes at least one of a DR original image and an image after the DR original image is processed; Extracting image features from the DR image, the image features comprising at least one of grayscale entropy features, texture features, noise features, gradient features, and divergence features; transforming the DR image into an energy space to obtain an energy distribution map of the DR image; Extracting energy features from the energy distribution diagram, the energy features comprising at least one of energy density features, energy center of gravity features, and energy effective value features; Obtaining a DR feature index according to at least one of the image features and at least one of the energy features, wherein the DR feature index is used to characterize a comprehensive level between the radiation dose of the X-ray and the image quality of the DR image; The DR characteristic index is output.

2. The method according to claim 1, characterized in that The step of transforming the DR image into an energy space to obtain an energy distribution map of the DR image includes: An orthogonal transformation is performed on the DR image to obtain the energy distribution map.

3. The method according to claim 1, characterized in that The energy feature includes the energy density feature, and extracting the energy density feature from the energy distribution diagram includes: The energy density feature is obtained according to the energy distribution concentration and dispersion trend in the energy distribution diagram.

4. The method according to claim 1, characterized in that The energy feature includes the energy center of gravity feature, and extracting the energy center of gravity feature from the energy distribution diagram includes: The energy center of gravity feature is obtained according to the energy concentration degree in the energy distribution diagram.

5. The method according to claim 1, characterized in that The energy feature includes the energy effective value feature, and extracting the energy effective value feature from the energy distribution diagram includes: The energy effective value feature is obtained according to the overall energy level of the energy distribution diagram.

6. The method according to claim 1, characterized in that The image feature includes the grayscale entropy feature, and extracting the grayscale entropy feature from the DR image includes: Obtaining a first probability statistical distribution of each grayscale source in the DR image; Determining a first probability statistical distribution of a non-redundant grayscale information source from the first probability statistical distribution of the grayscale information source according to the size of the first probability statistical distribution; Obtaining a second probability statistical distribution of each non-redundant grayscale information source according to a ratio of the first probability statistical distribution of each non-redundant grayscale information source to the sum of the first probability statistical distributions of all non-redundant grayscale information sources; The entropy of the second probability statistical distribution of the non-redundant grayscale information source is calculated to obtain the grayscale entropy feature.

7. The method according to claim 1, characterized in that The image feature includes the texture feature, and extracting the texture feature from the DR image includes: A texture feature description matrix is ​​obtained according to the gray value of each pixel in the DR image, wherein the texture feature description matrix describes the distance between different pixel points and the change of the gray value in the direction between different pixel points in the DR image; At least one two-dimensional component of the texture feature description matrix is ​​extracted as the texture feature.

8. The method according to claim 1, characterized in that The image feature includes the noise feature, and extracting the noise feature from the DR image includes: Performing filtering on the DR image to obtain a high-frequency image; Calculating the root mean square of the neighborhood of each pixel in the high-frequency image to obtain a noise distribution image; The noise value distribution in the noise distribution image is statistically analyzed to obtain the noise feature.

9. The method according to claim 1, characterized in that: The image feature includes the gradient feature, and extracting the gradient feature from the DR image includes: Determining the regional distribution of different tissues in the DR image; The clarity of the boundaries of different tissue region distributions is obtained as the gradient feature.

10. The method according to claim 1, characterized in that The image feature includes the divergence feature, and extracting the divergence feature from the DR image includes: Determining the regional distribution of different tissues in the DR image; The transition strength and / or trend consistency of the boundaries of different tissue region distributions are obtained as the divergence feature.

11. The method according to claim 1, characterized in that The obtaining of the DR feature index according to at least one of the image features and at least one of the energy features comprises: At least one of the image features and at least one of the energy features are input into a target model to obtain the DR feature index output by the target model.

12. The method according to claim 11, characterized in that The target model includes a pre-trained network model, and the training process of the network model includes: Acquire a DR image set, where the DR image set includes a plurality of DR images; Obtaining an energy feature set and an image feature set according to the DR image set, wherein the energy feature set includes a plurality of energy features extracted from the plurality of DR images, and the image feature set includes a plurality of image features extracted from the plurality of DR images; Obtaining a score set according to the DR image set, wherein the score set includes scores obtained by evaluating the plurality of DR images; The network model is trained using the energy feature set, the image feature set and the score set as a training sample set to obtain a trained network model.

13. A DR image analysis method, characterized in that: The method comprises: Acquire a DR image, wherein the DR image includes at least one of an original DR image and an image obtained by processing the original DR image; Extracting image features from the DR image, the image features comprising at least one of grayscale entropy features, texture features, noise features, gradient features, and divergence features; Extracting energy features from the DR image, the energy features comprising at least one of energy density features, energy center of gravity features, and energy effective value features; Obtaining a DR feature index according to at least one of the image features and at least one of the energy features, wherein the DR feature index is used to characterize a comprehensive level between the radiation dose of the X-ray and the image quality of the DR image; The DR characteristic index is output.

14. The method according to claim 13, characterized in that The extracting energy features from the DR image comprises: The DR image is input into a pre-trained feature extraction model to obtain the energy feature, Alternatively, the DR image is transformed into an energy space to obtain an energy distribution map of the DR image, and the energy feature is extracted from the energy distribution map of the DR image.

15. A DR imaging system, characterized in that: The DR imaging system comprises: a transmitting device, a receiving device, a processor and an output device; The transmitting device is used to transmit X-rays to a target part of a target object so that the X-rays pass through the target part; The receiving device is used to receive the X-rays after passing through the target part; The processor is used to execute the DR image analysis method according to any one of claims 1 to 14 to obtain a DR feature index; The output device is used to output the DR characteristic index.

Images