CT image noise reduction enhancement system and method

Abstract

The invention relates to the technical field of image noise reduction and enhancement, and discloses a CT image noise reduction and enhancement system and method which are used for providing a data interface module of a data transmission channel between the system and imaging equipment. The original data preprocessing module is used for preliminarily correcting the collected original projection data; the noise simulation module is used for establishing a noise model to simulate noise distribution in actual work according to physical characteristics and an imaging principle of the X-ray detector; the noise suppression simulation module is used for reducing noise based on a noise suppression algorithm of a physical model; the deep learning fusion module is used for optimizing the system in combination with a physical model noise reduction method and a deep learning technology; the image protection module is used for particularly paying attention to protection of important anatomical structures in the noise reduction process and avoiding loss of key diagnosis information due to excessive noise reduction; and the post-processing enhancement module is used for improving the image quality after the image processing is completed and ensuring the clinical readability of the image.

Description

Technical Field

[0001] The present invention relates to the technical field of image noise reduction and enhancement, and in particular to a CT image noise reduction and enhancement system and method. Background Art

[0002] Since the commercialization of CT (Computed Tomography) technology in the 1970s, its scanning speed, spatial resolution and density resolution have been significantly improved. However, with the increase in scanning speed and the requirement for dose management, the problem of image noise has become increasingly prominent. With the improvement of patient awareness of radiation protection, low-dose CT scanning has become a trend, but low-dose scanning will lead to a decrease in image quality, especially an increase in noise level, which requires the development of more advanced noise reduction and enhancement technologies to ensure diagnostic quality. In order to accurately identify tiny lesions and fine anatomical structures, doctors need high-definition and low-noise CT images. Therefore, the development of efficient noise reduction algorithms is crucial to improving diagnostic accuracy. In recent years, the significant improvement in computer hardware performance, especially the development of parallel computing devices such as GPUs, has enabled noise reduction methods based on complex mathematical models and deep learning technologies to be realized and applied in real-time processing. It is necessary to deeply understand the physical noise sources in the CT imaging process, such as quantum noise, shot noise, etc., and analyze these noises through modeling, and design corresponding noise reduction algorithms under the guidance of physical models. Therefore, a CT image noise reduction and enhancement system is proposed here.

[0003] At present, various noises are generated in the process of CT scanning, such as quantum noise, shot noise and system noise. The appearance of noise will lead to the decline of image quality, affecting the clarity and readability of the image, and the traditional non-specific noise reduction method will lead to the loss of image edge and detail information. Therefore, a CT image noise reduction and enhancement system is proposed here. By establishing a noise model and suppressing it in a targeted manner, it can better retain the diagnosis-related structural details while avoiding the poor visual effects caused by over-smoothing, better maintain the edge and texture information of the image, avoid the information loss caused by over-smoothing in traditional noise reduction methods, and perform noise reduction more accurately. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a CT image noise reduction and enhancement system and method. In view of the shortcomings of the prior art that easily lead to the loss of image edge and detail information, by establishing a noise model and suppressing it in a targeted manner, it is possible to better retain diagnosis-related structural details while avoiding the poor visual effects caused by over-smoothing, better maintain the edge and texture information of the image, avoid the information loss caused by over-smoothing in traditional noise reduction methods, and perform noise reduction more accurately.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a CT image noise reduction and enhancement system, comprising: a data interface module for providing a data transmission channel between the system and the imaging device;

[0007] A raw data preprocessing module for performing preliminary correction on the collected raw projection data;

[0008] A noise simulation module for building a noise model to simulate the noise distribution in actual work according to the physical characteristics and imaging principles of the X-ray detector;

[0009] Noise suppression simulation module for noise reduction using a physical model-based noise suppression algorithm;

[0010] A deep learning fusion module for optimizing the system by combining physical model denoising methods with deep learning techniques;

[0011] An image protection module that pays special attention to the protection of important anatomical structures during the noise reduction process to avoid the loss of key diagnostic information due to excessive noise reduction;

[0012] A post-processing enhancement module used to improve image quality after image processing is completed to ensure the clinical readability of the image;

[0013] The raw data preprocessing module includes a noise analysis module and a noise modeling module, the noise suppression simulation module is provided with a forward projection model module and a reverse reconstruction algorithm module, the noise suppression simulation module and the deep learning fusion module are both provided with a regularization technology introduction module, and the image protection module is provided with an anatomical structure protection module and an edge detection and retention module.

[0014] As a preferred solution of the CT image noise reduction and enhancement system described in the present invention, the data interface module includes, based on DICOM standard support, generating image files that comply with the DICOM standard through a CT scanning device, reading and parsing data in DICOM format, including metadata patient information, scanning parameters and pixel data. When the system directly receives real-time data from the CT device, it uses the DICOM network service C-STORE to transmit data through an Ethernet connection. The data interface adopts the TCP / IP protocol. For a highly integrated system, the data stream inside the CT device is accessed through an API call.

[0015] As a preferred solution of the CT image noise reduction and enhancement system described in the present invention, wherein: the raw data preprocessing module is used to further preprocess the received data and identify noise;

[0016] The noise analysis module identifies and quantifies the noise source quantum noise in the CT imaging process based on the collected CT image data, and the quantum noise is modeled by Poisson distribution;

[0017] The noise modeling module constructs a Poisson distribution model based on a known average counting rate using a determined average counting rate, and calculates, estimates and quantifies the level of quantum noise for each noise intensity x.

[0018] The noise simulation module performs statistical analysis on the noise samples actually collected, calculates the average gray value activity level of each pixel based on the noise analysis module, and uses it as a parameter of Poisson distribution to compare the simulated noise with the actual noise.

[0019] As a preferred solution of the CT image noise reduction and enhancement system described in the present invention, the noise suppression simulation module uses TV regularization based on the forward projection model module to achieve noise reduction by minimizing a functional that combines data fitting error and total image variation:

[0020]

[0021] Where u is the image function to be optimized, A represents the linear operator from the real image to the observed data, f is the observed data with noise, that is, the data processed by the forward projection model module, TV(u) represents the total variation of the image, which is used to control the edge and detail information of the image to avoid over-smoothing in the denoising process, and λ is the regularization parameter;

[0022] The reverse reconstruction algorithm module performs iterative calculation based on the SplitBregman algorithm, converts the TV regularization term into the form of L1 norm, and introduces an additional variable v:

[0023]

[0024] Among them, u' is the image to be reconstructed, A' is the projection operator, f' is the projection data, is the gradient of the image, μ is the regularization parameter, which is used to balance the weight between the data fitting error and the TV regularization term;

[0025] The inverse reconstruction algorithm module uses the SplitBregman algorithm to perform iterative calculations until the stop condition is met. During the iteration process, the algorithm will continuously adjust the image and v values ​​to minimize the objective function, that is, the combination of the data fitting error and the TV regularization term.

[0026] As a preferred solution of the CT image denoising and enhancement system described in the present invention, the deep learning fusion module uses a physical model denoising method based on TV regularization, takes the image denoised by the physical model and the corresponding noise-free and low-noise image as a training set, and uses a generative adversarial network GANs structure. The adversarial network GANs structure includes a discriminator D and a generator G, and performs joint optimization. The discriminator D guides the update of the generator G by distinguishing the generated image from the real noise-free image. The GANs model is continuously iterated and updated during the adversarial training process. The generator strives to deceive the discriminator, and the discriminator continuously improves its recognition ability, which ultimately prompts the generator to recover a result close to the real noise-free image from the noisy image. At the same time, after the training is completed, the image denoised by the physical model is fused with the image generated by GANs to optimize the image quality.

[0027] As a preferred solution of the CT image noise reduction and enhancement system described in the present invention, the anatomical structure protection module in the image protection module connects pixels belonging to the same tissue structure according to the similarity of pixel intensity, performs regional growth, uses multi-threshold technology to distinguish different tissue levels, uses Otsu's binarization method to find the optimal segmentation threshold, uses morphological operations such as expansion, corrosion, and opening and closing operations to protect the integrity of the fault structure, and learns and identifies the anatomical structure in the image through a training model to perform accurate segmentation.

[0028] As a preferred solution of the CT image noise reduction and enhancement system described in the present invention, the edge detection in the edge detection and retention module performs noise reduction and smoothing on the original image to reduce the influence of noise on the edge detection result, calculates the gradient strength and direction of each pixel in the image by the first-order and second-order derivative operator Sobel, locates the edge position, and sets two high and low thresholds, divides the pixels into strong edges, weak edges and non-edge areas according to the gradient strength, and the edge retention part uses a non-maximum suppression method to exclude points with non-maximum gradient responses after determining the preliminary edge candidate points, ensuring that only true edge points are retained; for the edges of anatomical structures that need special protection, when performing operations that blur the edges, the morphological operation expansion operation is used to increase the weight of the area near the edge to prevent the edge from being over-smoothed and retain the edge details while removing noise;

[0029] The post-processing enhancement module performs contrast adjustment, edge sharpening and detail restoration, color correction and mapping, structure enhancement and protection, and texture analysis and reconstruction on the processed image.

[0030] As a preferred solution of the CT image noise reduction and enhancement method described in the present invention, wherein: image data is obtained from a CT scanning device through the DICOM standard, and noise is identified and quantified on the received raw data to establish a noise model;

[0031] The forward projection model and TV regularization algorithm are used to minimize the data fitting error, and the inverse reconstruction algorithm is applied to iteratively optimize the image to achieve effective noise suppression.

[0032] The denoised image of the physical model is combined with the noise-free image and jointly optimized using a generative adversarial network (GAN) to improve image quality.

[0033] Important anatomical structures are protected by region growing and multi-threshold techniques, combined with edge detection and preservation methods to ensure that edge details are not over-smoothed;

[0034] Contrast adjustment, edge sharpening, and detail restoration are performed on the final image to ensure clinical readability and overall image quality.

[0035] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, steps of a CT image noise reduction and enhancement system are implemented.

[0036] A computer-readable storage medium stores a computer program, wherein the computer program implements the steps of a CT image noise reduction and enhancement system when executed by a processor.

[0037] Beneficial effects of the present invention: The present invention can better preserve the edge and texture information of the image through a noise removal method based on physical mechanisms, taking into account the relationship between the signal and noise in the imaging process, thereby avoiding over-smoothing or loss of important diagnostic information. By establishing a noise model and suppressing it in a targeted manner, it can better retain the structural details related to diagnosis while avoiding the poor visual effects caused by over-smoothing, better preserve the edge and texture information of the image, and avoid the information loss caused by over-smoothing in traditional noise reduction methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0039] Figure 1 A schematic diagram of a CT image noise reduction and enhancement system module provided by an embodiment of the present invention.

[0040] In the figure: 1. Data interface module; 2. Raw data preprocessing module; 3. Noise simulation module; 4. Noise suppression simulation module; 5. Deep learning fusion module; 6. Image protection module; 7. Post-processing enhancement module; 8. Noise analysis module; 9. Noise modeling module; 10. Forward projection model module; 11. Inverse reconstruction algorithm module; 12. Regularization technology introduction module; 13. Anatomical structure protection module; 14. Edge detection and retention module. DETAILED DESCRIPTION

[0041] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0042] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0043] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0044] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0045] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0046] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0047] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides a CT image noise reduction and enhancement system, comprising:

[0048] The data interface module 1 is based on DICOM standard support. It generates image files that comply with DICOM standard through CT scanning equipment, reads and parses data in DICOM format, including metadata patient information, scanning parameters and pixel data. When the system receives real-time data directly from the CT device, it uses DICOM network service C-STORE for data transmission through Ethernet connection. The data interface adopts TCP / IP protocol. For systems with higher integration, the data flow inside the CT device is accessed through API calls.

[0049] The raw data preprocessing module 2 is used to further preprocess the received data and identify noise. The noise analysis module 8 identifies and quantifies the noise source quantum noise in the CT imaging process based on the collected CT image data. The quantum noise can be modeled by Poisson distribution, and its probability density function is:

[0050] P(x)=(λ^x*e^(-λ)) / x!

[0051] Among them, λ is the average counting rate, which reflects the average occurrence rate of noise events. The noise modeling module 9 constructs a Poisson distribution model based on the known value of λ using the determined λ, and performs calculations for each noise intensity x to predict the probability of occurrence of noises of different intensities.

[0052] The noise simulation module 3 performs statistical analysis on the noise samples actually collected, and calculates the average grayscale activity level of each pixel based on the noise analysis module 8. This value is used as the parameter of the Poisson distribution. According to the Poisson distribution formula, a set of simulated noise data is generated using the estimated λ value. For each pixel in the image, a random number is independently drawn from the Poisson distribution whose parameter is the λ value corresponding to the pixel. The simulated noise is compared with the actual noise, and the simulated noise data is compared with the actual collected data. Through histogram comparison, the pixel value distribution histogram of the actual noise and the simulated noise is drawn to determine whether they are consistent, and the statistical indicators are compared to calculate the mean square error MSE evaluation index based on the formula:

[0053] MSE=(1 / n)*Σ(yi-xi)^2

[0054] Among them, n represents the number of samples, yi represents the observed value of actual noise, and xi represents the observed value of simulated noise for comparison, comparing the simulated noise with the actual noise.

[0055] Among them: Calculate the average gray value of each pixel in the image (corresponding to the activity level in the Poisson noise model), which usually involves the following steps: First, load the actual collected noise sample image and convert it into a gray image, and then calculate the average gray value: traverse all pixels in the image and directly use matrix operations to find their average value to obtain the average gray value of the entire image. If you want to estimate the activity level for each pixel separately, you need to further process it, pixel-level estimation: The activity level needs to be estimated independently for each pixel, because each pixel has a different exposure time and other factors that cause the intensity of the Poisson process to change. In this case, the gray value of each pixel is directly used as an estimate of the activity level, and then estimated based on statistical characteristics: In some cases, the activity level is not simply equal to the gray value itself, but a statistical parameter related to the gray value. The gray mean and variance feature quantities in the local window can be used to estimate the activity level of each pixel.

[0056] Furthermore, the noise suppression simulation module 4 uses TV regularization based on the forward projection model module 10 to achieve noise reduction by minimizing a functional that combines the data fitting error and the total image variation.

[0057]

[0058] Among them, u is the image function to be optimized, A is the linear operator from the real image to the observed data, f is the observed data with noise, that is, the data processed by the forward projection model module, TV(u) represents the total variation of the image, which is used to control the edge and detail information of the image to avoid over-smoothing in the denoising process, and λ is the regularization parameter;

[0059] The reverse reconstruction algorithm module performs iterative calculation based on the SplitBregman algorithm, converts the TV regularization term into the form of L1 norm, and introduces an additional variable v:

[0060]

[0061] Among them, u' is the image to be reconstructed, A' is the projection operator, f' is the projection data, is the gradient of the image, μ is the regularization parameter, which is used to balance the weight between the data fitting error and the TV regularization term;

[0062] The inverse reconstruction algorithm module uses the SplitBregman algorithm to perform iterative calculations until the stop condition is met. During the iteration process, the algorithm will continuously adjust the image and v values ​​to minimize the objective function, that is, the combination of the data fitting error and the TV regularization term.

[0063] Furthermore, the deep learning fusion module 5 is based on the TV regularized physical model denoising method, and uses the images denoised by the physical model and the corresponding noise-free and low-noise images as training sets, and uses the generative adversarial network GANs structure. The adversarial network GANs structure includes a discriminator D and a generator G, and performs joint optimization. The discriminator D guides the update of the generator G by distinguishing the generated image from the real noise-free image. The GANs model is continuously iterated and updated during the adversarial training process. The generator strives to deceive the discriminator, while the discriminator continuously improves its recognition ability, which ultimately enables the generator to better recover results close to the real noise-free image from the noisy image. At the same time, after the training is completed, the image denoised by the physical model is appropriately fused with the image generated by GANs to optimize the image quality.

[0064] The anatomical structure protection module 13 in the image protection module 6 connects pixels belonging to the same tissue structure according to the similarity of pixel intensity and performs region growing, that is, selects a set of initial seed points, which usually represent known parts of the tissue structure to be segmented, checks the pixels in the neighborhood around the seed points, and adds these pixels to the growing region if their similarity with the seed points exceeds a certain threshold, and repeats the above process until there are no more pixels that meet the conditions to be added, or the preset growing conditions are reached;

[0065] Then, the multi-threshold technology is used to distinguish different tissue levels. Otsu's binarization method is used to find the best segmentation threshold. According to the image characteristics and tissue structure, multiple thresholds are set, and each pixel in the image is traversed. It is classified into the corresponding category according to its gray value. Then the gray histogram of the image is calculated, and all possible thresholds are traversed. The inter-class variance under each threshold is calculated. The threshold that maximizes the inter-class variance is found, which is the best segmentation threshold. The image is binarized using this threshold.

[0066] Morphological operations such as dilation, erosion, and opening and closing are used to protect the integrity of the fault structure. Finally, the model is trained to learn and identify the anatomical structure in the image for accurate segmentation.

[0067] In the edge detection and preservation module 14, the edge detection first performs noise reduction and smoothing processing on the original image to reduce the influence of noise on the edge detection result, and uses an adaptive filtering method, such as bilateral filtering or guided filtering, which can maintain edge details while reducing noise, and applies an adaptive filtering algorithm to the original image to adjust the filtering parameters to adapt to different areas of the image;

[0068] Then, the gradient intensity and direction of each pixel in the image are calculated by the first-order and second-order derivative operator Sobel to locate the edge position, and then two high and low thresholds are set to divide the pixels into strong edges, weak edges and non-edge areas according to the gradient intensity. After determining the preliminary edge candidate points, the edge preservation part traverses the edge candidate points and applies the non-maximum suppression method to exclude points with non-maximum gradient responses to ensure that only true edge points are retained. For the edges of anatomical structures that require special protection, when blurring the edges, the morphological operation dilation operation is used to increase the weight of the area near the edge to prevent the edge from being over-smoothed while removing noise and retaining the edge details. The post-processing enhancement module 7 performs contrast adjustment, edge sharpening and detail restoration, color correction and mapping, structure enhancement and protection, and texture analysis and reconstruction on the processed image.

[0069] Among them: Contrast adjustment in post-processing enhancement module 7: Use histogram equalization technology to expand the dynamic range of the image and improve the contrast of the overall image, edge sharpening and detail recovery: Apply a high-pass filter (Laplacian operator) to sharpen the image to highlight the boundary information, and use morphological operations (top hat transform) to extract the edge and detail features of the image, color correction and mapping: For color images or multi-channel images, color balance correction can be performed to ensure that the relationship between different colors is correct, and the color space is converted into color mapping as needed, structure enhancement and protection: Use local and global image enhancement algorithms, such as deep learning-based methods, to strengthen important anatomical structures and prevent structural blurring and loss caused by excessive processing, and texture analysis and reconstruction: For images with rich texture information, texture analysis and reconstruction technology can be used to better present and understand the microscopic structure inside the image.

[0070] Embodiment 2, the second embodiment of the present invention, is different from the previous embodiment in that:

[0071] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0072] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0073] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0074] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0075] Embodiment 3 is an embodiment of the present invention, and provides a CT image noise reduction and enhancement method, characterized in that: image data is acquired from a CT scanning device through the DICOM standard, and noise is identified and quantified on the received raw data to establish a noise model;

[0076] The forward projection model and TV regularization algorithm are used to minimize the data fitting error, and the inverse reconstruction algorithm is applied to iteratively optimize the image to achieve effective noise suppression.

[0077] The denoised image of the physical model is combined with the noise-free image and jointly optimized using a generative adversarial network (GAN) to improve image quality.

[0078] Important anatomical structures are protected by region growing and multi-threshold techniques, combined with edge detection and preservation methods to ensure that edge details are not over-smoothed;

[0079] Contrast adjustment, edge sharpening, and detail restoration are performed on the final image to ensure clinical readability and overall image quality.

[0080] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A CT image noise reduction and enhancement system, characterized in that: The noise reduction enhancement system includes, A data interface module for providing a data transmission channel between the system and the imaging device; A raw data preprocessing module for performing preliminary correction on the collected raw projection data; A noise simulation module for building a noise model to simulate the noise distribution in actual work according to the physical characteristics and imaging principles of the X-ray detector; Noise suppression simulation module for noise reduction using a physical model-based noise suppression algorithm; A deep learning fusion module for optimizing the system by combining physical model denoising methods with deep learning techniques; An image protection module that pays special attention to the protection of important anatomical structures during the noise reduction process to avoid the loss of key diagnostic information due to excessive noise reduction; A post-processing enhancement module used to improve image quality after image processing is completed to ensure the clinical readability of the image; The raw data preprocessing module includes a noise analysis module and a noise modeling module, the noise suppression simulation module is provided with a forward projection model module and a reverse reconstruction algorithm module, the noise suppression simulation module and the deep learning fusion module are both provided with a regularization technology introduction module, and the image protection module is provided with an anatomical structure protection module and an edge detection and retention module.

2. A CT image noise reduction and enhancement system as claimed in claim 1, characterized in that: The data interface module includes, based on DICOM standard support, generating image files that comply with DICOM standards through CT scanning equipment, reading and parsing data in DICOM format, including metadata patient information, scanning parameters and pixel data. When the system directly receives real-time data from the CT device, it uses the DICOM network service C-STORE to transmit data through Ethernet connection. The data interface adopts TCP / IP protocol. For highly integrated systems, the data stream inside the CT device is accessed through API calls.

3. A CT image noise reduction and enhancement system as claimed in claim 2, characterized in that: The raw data preprocessing module is used to further preprocess the received data and identify noise; The noise analysis module identifies and quantifies the noise source quantum noise in the CT imaging process based on the collected CT image data, and the quantum noise is modeled by Poisson distribution; The noise modeling module constructs a Poisson distribution model based on a known average counting rate using a determined average counting rate, and calculates, estimates and quantifies the level of quantum noise for each noise intensity x. The noise simulation module performs statistical analysis on the noise samples actually collected, calculates the average gray value activity level of each pixel based on the noise analysis module, and uses it as a parameter of Poisson distribution to compare the simulated noise with the actual noise.

4. A CT image noise reduction and enhancement system as claimed in claim 3, characterized in that: The noise suppression simulation module uses TV regularization based on the forward projection model module to achieve noise reduction by minimizing a functional that combines the data fitting error and the total image variation: Among them, u is the image function to be optimized, A is the linear operator from the real image to the observed data, f is the observed data with noise, that is, the data processed by the forward projection model module, TV(u) represents the total variation of the image, which is used to control the edge and detail information of the image to avoid over-smoothing in the denoising process, and λ is the regularization parameter; The reverse reconstruction algorithm module performs iterative calculation based on the SplitBregman algorithm, converts the TV regularization term into the form of L1 norm, and introduces an additional variable v: Among them, u' is the image to be reconstructed, A' is the projection operator, f' is the projection data, is the gradient of the image, μ is the regularization parameter, which is used to balance the weight between the data fitting error and the TV regularization term; The inverse reconstruction algorithm module uses the Split Bregman algorithm to perform iterative calculations until the stop condition is met. During the iteration process, the algorithm will continuously adjust the image and v values ​​to minimize the objective function, that is, the combination of the data fitting error and the TV regularization term.

5. A CT image noise reduction and enhancement system as claimed in claim 4, characterized in that: The deep learning fusion module is based on the TV regularized physical model denoising method. The images denoised by the physical model and the corresponding noise-free and low-noise images are used as training sets. The generative adversarial network GANs structure is used. The adversarial network GANs structure includes a discriminator D and a generator G. The discriminator D guides the update of the generator G by distinguishing the generated image from the real noise-free image. The GANs model is continuously iterated and updated during the adversarial training process. The generator strives to deceive the discriminator, and the discriminator continuously improves its recognition ability, which ultimately enables the generator to recover results close to the real noise-free image from the noisy image. At the same time, after the training is completed, the image denoised by the physical model is fused with the image generated by GANs to optimize the image quality.

6. A CT image noise reduction and enhancement system as claimed in claim 5, characterized in that: The anatomical structure protection module in the image protection module connects pixels belonging to the same tissue structure according to the similarity of pixel intensity, performs regional growth, uses multi-threshold technology to distinguish different tissue levels, uses Otsu's binarization method to find the optimal segmentation threshold, uses morphological operations such as expansion, erosion, and opening and closing operations to protect the integrity of the fault structure, and learns and identifies the anatomical structure in the image through a training model to perform accurate segmentation.

7. A CT image noise reduction and enhancement system as claimed in claim 6, characterized in that: In the edge detection and preservation module, the original image is subjected to denoising and smoothing to reduce the influence of noise on the edge detection result. The gradient strength and direction of each pixel in the image are calculated by the first-order and second-order derivative operator Sobel, the edge position is located, and two high and low thresholds are set. The pixels are divided into strong edge, weak edge and non-edge area according to the gradient strength. After the preliminary edge candidate points are determined, the edge preservation part uses the non-maximum suppression method to exclude points with non-maximum gradient response to ensure that only the real edge points are retained. For the edges of anatomical structures that need special protection, when performing operations that blur the edges, the morphological operation dilation operation is used to increase the weight of the area near the edge to prevent the edge from being over-smoothed and retain the edge details while removing noise. The post-processing enhancement module performs contrast adjustment, edge sharpening and detail restoration, color correction and mapping, structure enhancement and protection, and texture analysis and reconstruction on the processed image.

8. A method using a CT image noise reduction and enhancement system as claimed in any one of claims 1 to 7, characterized in that: Obtain image data from CT scanning equipment through the DICOM standard, identify and quantify noise in the received raw data, and establish a noise model; The forward projection model and TV regularization algorithm are used to minimize the data fitting error, and the inverse reconstruction algorithm is applied to iteratively optimize the image to achieve effective noise suppression. The denoised image of the physical model is combined with the noise-free image and jointly optimized using a generative adversarial network (GAN) to improve image quality. Important anatomical structures are protected by region growing and multi-threshold techniques, combined with edge detection and preservation methods to ensure that edge details are not over-smoothed; Contrast adjustment, edge sharpening, and detail restoration are performed on the final image to ensure clinical readability and overall image quality.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images