A chest CT image processing system

Abstract

This invention discloses a chest CT image processing system, including a data acquisition module, an image enhancement module, a multi-resolution fusion module, and a high-fidelity export module. This invention relates to the field of intelligent CT image processing technology, specifically a chest CT image processing system. This system performs intensity cropping and normalization on CT images, and combines logarithmic domain bias field regularization iterative correction to restore tissue contrast. First, it locates the lung field, refines the boundaries using variational segmentation, and uses mask-constrained nonlocal mean denoising within the lung to preserve small lesions. It then combines a Hessian matrix with an improved Frangi metric for structural enhancement, fusing the original image, denoised image, and enhanced image. Based on structural response and gradient, it constructs a voxel-level importance mapping, performs high- and low-resolution resampling, and seamlessly synthesizes the images according to weights, preserving high resolution and detail in key areas, improving the accuracy of lung field and lesion boundaries, while reducing computational and storage overhead and effectively avoiding artifacts.

Description

Technical Field

[0001] This invention relates to the field of intelligent CT image processing technology, specifically to a chest CT image processing system. Background Technology

[0002] With the rapid growth in demand for chest disease screening and precise diagnosis and treatment, chest CT has become a core imaging tool for clinical diagnosis and follow-up. The widespread adoption of low-dose CT screening, radiomics, and computer-aided diagnosis has driven an urgent need for high-fidelity, highly reproducible, and quantifiable image processing technologies. At the same time, the improvement in computing power and the development of multi-scale image processing methods have provided a feasible technical foundation for restoring tissue contrast, preserving minute lesions, and achieving automated analysis. Clinically-oriented CT image processing systems must not only improve visualization quality but also ensure that quantitative results can be reliably used for downstream diagnostic, segmentation, and radiomics research, thereby contributing to early diagnosis and personalized treatment decisions.

[0003] Existing technologies still have several key shortcomings in practical applications. Inconsistencies in bias fields and HUs caused by different scanning protocols and equipment lead to tissue contrast distortion. Simple denoising and full-image thresholding often destroy nodule boundaries or misclassify the pleura and cardiac silhouette, reducing the sensitivity and segmentation accuracy of lesion detection. Single-scale structural enhancement methods have difficulty distinguishing between blood vessels and nodules simultaneously. Most resampling strategies use full-volume high resolution, resulting in large computational and storage costs and potential artifacts. Fixed resolution, on the other hand, loses details of small nodules, limiting the reliability of clinical application and quantitative analysis. Summary of the Invention

[0004] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a chest CT image processing system that addresses the limitations of traditional methods. Image preprocessing and enhancement methods exist Addressing issues such as lung field deviation, coarse lung field segmentation leading to missed or misclassified cases, and insufficient ability to differentiate between different morphological structures like blood vessels and nodules, this approach utilizes specific... Intensity clipping and linear normalization are performed on intervals; a smooth bias field is established in the logarithmic domain and iterative optimization with regularization is used to recover the true tissue contrast; thresholding and connectivity are used for fast localization before... Variational refinement of the lung mask, nonlocal mean denoising within the lung using mask constraints to suppress random noise while preserving small lesions, and improvement by combining multi-scale Hessian matrix Scale-selective structure enhancement is performed using metrics, and the original image, denoised image, and structure-enhanced image are ultimately fused using multi-scale Laplacian and Gaussian pyramid methods; this addresses the traditional... Commonly used resampling and display methods, such as fixed voxel spacing or full-volume oversampling, suffer from resolution loss in key nodules and small bronchi, high computational and storage overhead due to full-volume high resolution, and artifacts at resolution switching points. This solution addresses these issues by constructing a voxel-level content importance mapping based on multi-scale structural response and fused image gradients, and then applying this mapping to the fused image. The images are resampled to high resolution and low resolution respectively, and then weighted point by point according to voxel-level weights to achieve seamless fusion.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides a chest CT image processing system, including a data acquisition module, an image enhancement module, a multi-resolution fusion module and a high-fidelity export module;

[0006] The data acquisition module collects chest data. image;

[0007] The image enhancement module, for the original Intensity cropping, logarithmic domain bias field estimation and correction, fine segmentation of lung fields based on threshold and variational segmentation, nonlocal mean denoising, and multi-scale structural enhancement are performed. Then, the original image, the denoised image, and the structural enhancement image are fused together by multi-scale weighted fusion to output an enhanced fused image.

[0008] The multi-resolution fusion module calculates the importance weight of the content based on the structural response and gradient intensity, allocates resolution resources at the voxel level according to the importance of the image, resamples the fused image into high resolution and low resolution respectively, and then uses the importance weight to perform voxel-by-voxel weighted synthesis.

[0009] The high-fidelity export module saves the normalized version of the final output image. Restore the formula and write the set of key parameters into the metadata.

[0010] Furthermore, the data acquisition module receives clinical chest data. The output of the scanning device Formatted image data, extract the chest area to be processed. Images and Metadata.

[0011] Furthermore, the image enhancement module enhances the acquired original image. Image preprocessing, bias field correction, lung field segmentation, denoising, and structural enhancement specifically include the following units:

[0012] Intensity trimming unit, will remove the original Voxel values ​​are clipped to a fixed value. The range of values ​​is determined, and then linear normalization is performed.

[0013] The bias field modeling unit estimates the smooth bias field in the logarithmic domain and removes... Low-frequency multiplicative noise exists, restoring the true tissue contrast;

[0014] The lung field segmentation unit first uses thresholding and connectivity to quickly obtain candidate lung field regions, removes external thoracic structures, and then uses variational segmentation. The algorithm refines the boundaries within the candidate region to obtain a smooth, closed lung mask;

[0015] The nonlocal mean denoising unit uses repeated blocks of recurring features in the image for weighted averaging to remove random noise while preserving fine structures as much as possible.

[0016] The structural enhancement unit utilizes multi-scale Hessian matrix features to detect and enhance tubular and spherical structures, specifically including the following sub-units:

[0017] The eigenvalue calculation subunit uses Gaussian kernel convolution to calculate the scale space, and calculates the Hessian matrix based on the scale space to obtain the three eigenvalues ​​of the Hessian matrix.

[0018] Enhanced output subunit, introducing improved The type metric calculates the maximum response across all scales as a candidate augmentation map;

[0019] Laplacian fusion unit, for the original image, nonlocal means denoised image and The structure enhancement map results are subjected to multi-scale Gaussian pyramid decomposition, and weighted fusion is performed at each scale using saliency weights to preserve edge, detail and structural consistency. Finally, the fused image is reconstructed.

[0020] Furthermore, the multi-resolution fusion module allocates resolution resources based on the importance of image content, performs multi-scale resampling and weighted fusion on the enhanced image, retains high resolution in key structural regions, and reduces resolution in non-critical regions. Specifically, it includes the following units:

[0021] The content importance mapping unit calculates the importance weight of each voxel based on structural response and edge strength, and assigns key structures to high resolution and strong detail.

[0022] The multi-resolution resampling unit first resamples the fused image into high-resolution and low-resolution images respectively, and then performs voxel-by-voxel weighted synthesis according to the importance of the content. High resolution is retained in important areas and low resolution is used in unimportant areas, resulting in seamless fusion.

[0023] Furthermore, the fidelity export module saves the normalized version of the final output image and... The formula is restored, and the key parameter set is written into the metadata, specifically including the following units:

[0024] Normalized storage unit;

[0025] Metadata record unit.

[0026] The beneficial effects achieved by adopting the above solution are as follows:

[0027] (1) Targeting traditional Image preprocessing and enhancement methods exist Addressing issues such as lung field deviation, coarse lung field segmentation leading to missed or misclassified cases, and insufficient ability to differentiate between different morphological structures like blood vessels and nodules, this approach utilizes specific... Intensity clipping and linear normalization are performed on intervals; a smooth bias field is established in the logarithmic domain and iterative optimization with regularization is used to recover the true tissue contrast; thresholding and connectivity are used for fast localization before... Variational refinement of the lung mask, nonlocal mean denoising within the lung using mask constraints to suppress random noise while preserving small lesions, and improvement by combining multi-scale Hessian matrix The algorithm performs scale-selective structure enhancement, and finally merges the original image, denoised image, and structure-enhanced image using multi-scale Laplacian and Gaussian pyramid methods, thereby achieving... Correct image restoration Compare and improve the accuracy of lung field and lesion boundaries, and preserve and amplify diagnostic details.

[0028] (2) Targeting traditional Commonly used resampling and display methods, such as fixed voxel spacing or full-volume oversampling, suffer from resolution loss in key nodules and small bronchi, high computational and storage overhead due to full-volume high resolution, and artifacts at resolution switching points. This solution addresses these issues by constructing a voxel-level content importance mapping based on multi-scale structural response and fused image gradients, and then applying this mapping to the fused image. Images are resampled to high and low resolution respectively, and then weighted point-by-point according to voxel-level weights to achieve seamless fusion. This preserves high spatial detail and measurement accuracy in key lesions, blood vessels, and bronchial regions, while reducing sampling in background areas to save resources. Furthermore, smooth weight transitions effectively avoid resolution boundary artifacts, achieving a balance between... Diagnostic visibility and clinical treatment efficiency. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of a chest CT image processing system provided by the present invention;

[0030] Figure 2 This is a schematic diagram of the image enhancement module;

[0031] Figure 3 This is a schematic diagram of a multi-resolution fusion module;

[0032] Figure 4This is a schematic diagram of the export module for high fidelity.

[0033] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0034] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0035] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0036] Example 1, see Figure 1 The present invention provides a chest CT image processing system, including a data acquisition module, an image enhancement module, a multi-resolution fusion module and a high-fidelity export module;

[0037] The data acquisition module receives clinical chest data. The output of the scanning device Formatted image data, extract the chest area to be processed. image as well as Metadata, and send the data to the image enhancement module;

[0038] The image enhancement module receives data sent by the data acquisition module and processes the original image. The system performs intensity cropping, logarithmic domain bias field estimation and correction, fine segmentation of lung fields based on threshold and variational segmentation, nonlocal mean denoising, and multi-scale structural enhancement. Then, the original image, the denoised image, and the structural enhancement image are fused through multi-scale weighted fusion to output an enhanced fused image, and the data is sent to the multi-resolution fusion module.

[0039] The multi-resolution fusion module receives data sent by the image enhancement module, calculates the importance weight of the content based on the structural response and gradient intensity, allocates resolution resources to the image at the voxel level according to its importance, resamples the fused image into high resolution and low resolution respectively, performs voxel-by-voxel weighted synthesis using the importance weight, and sends the data to the fidelity export module.

[0040] The high-fidelity export module receives data sent by the multi-resolution fusion module and saves the normalized version of the final output image. Restore the formula and write the set of key parameters into the metadata.

[0041] Example 2, see Figure 1 This embodiment is based on the above embodiment, wherein the data acquisition module receives clinical chest data. The output of the scanning device Formatted image data, extract the chest area to be processed. image as well as Metadata, preserved Floating-point precision, recording the original voxel spacing and The range of values, where, Represents the three-dimensional spatial coordinates of a single voxel.

[0042] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. The image enhancement module enhances the acquired original image. Image preprocessing, bias field correction, lung field segmentation, denoising, and structural enhancement specifically include the following units:

[0043] Intensity trimming unit, will remove the original Voxel values ​​cropped to of The value range is then linearly normalized to... , means as follows:

[0044] ;

[0045] in, Represents the original Image in coordinates voxel values ​​at the location, and Lower and upper limits for cropping. Indicates the original voxel value Limited to the scope between, This represents the voxel value after clipping. This represents the normalized voxel values, ranging from... ;

[0046] The bias field modeling unit estimates the smooth bias field in the logarithmic domain and removes... The presence of low-frequency multiplicative noise, when compared to the restored true tissue, is shown below:

[0047] ;

[0048] in, The translated image, Indicates the bias field. Indicates observation noise. Represents a true, unbiased image. Represents the logarithmic bias field. This indicates taking the minimum value. Represents the three-dimensional Laplacian operator. The regularization weights are listed, and their values ​​range from [value range missing]. , This indicates that the minimum value is being sought, and the iteration termination condition of the objective function is set as follows: or When the iteration stops, Indicates the number of iterations. and They represent the first The second iteration and the first Log-bias field of the next iteration Represents the natural logarithm function;

[0049] The lung field segmentation unit first uses thresholding and connectivity to quickly obtain candidate lung field regions, removes external thoracic structures, and then uses variational segmentation. The algorithm refines the boundaries within the candidate region to obtain a smooth, closed lung mask, as shown below:

[0050] ;

[0051] in, express Energy function Represents a segmented contour surface. The area of ​​the contour surface is represented by the number of contour voxels multiplied by the voxel scale. and Let the inner and outer domains of the contour surface be represented respectively, and define the level set function. , Indicates the outer domain. Indicates the inner domain. Represents a contour surface. and These represent the average gray values ​​of the inner and outer regions of the contour surface, respectively. This represents the contour smoothing parameter, with an initial value set to 0.1; and This represents the data fitting parameters, initially set to 1. Indicates in the region Integrating within, Dirac function, The value is 0.5; Indicates curvature. Represents the gradient operator. The time derivative is used for iterative updates: ;in, Indicates the iteration step as The outline of time, Indicates the iteration step as The outline of time, Indicates the time step; the implementation process is as follows: First, initialization: use a fixed threshold. Threshold segmentation is performed, followed by binary morphological opening and closing operations to obtain the initial mask. Then run the above within the mask. Refine 200 times;

[0052] The nonlocal mean denoising unit uses a weighted average of repeating blocks of recurring features in the image to remove random noise while preserving as much fine structure as possible, as shown below:

[0053] ;

[0054] in, This represents the estimated value after denoising. This refers to the search window, specifically... Centered Square neighborhood This represents the three-dimensional coordinates of another single voxel within the search window. express right Contribution weight, The mask constraint image obtained after lung field region segmentation retains only the voxels of the lung field region and sets the voxels of non-lung field regions to 0. This represents the patch vector, specifically centered at x. The small window flattens into a vector. This represents taking the square of the Euclidean norm. Indicates the smoothing parameter. , This represents the standard deviation of noise, where, Indicates the background region outside the lung field. This represents the voxel mean of the background region. Indicates the total number of voxels in the background region;

[0055] The structural enhancement unit utilizes multi-scale Hessian matrix features to detect and enhance tubular and spherical structures, specifically including the following sub-units:

[0056] The eigenvalue calculation subunit uses Gaussian kernel convolution to calculate the scale space and scale. The Hessian matrix below ,get The three eigenvalues ​​are represented as follows:

[0057] ;

[0058] in, This represents a standard Gaussian kernel with a scale parameter of . ,set up As a set of scale values ​​for multi-scale analysis, the specific values ​​are... The unit is millimeters; defined as , Represents an exponential function with the natural constant as its base; Represents the convolution operator. The scale is represented as Scale-space images, Representing scale The Hessian matrix below, Represents coordinates The second-order partial derivative operator is obtained. The three eigenvalues ​​sorted by absolute value from smallest to largest , , Specifically ;

[0059] Enhanced output subunit, introducing improved The type metric calculates the maximum response across all scales, which serves as a candidate augmentation map, as shown below:

[0060] ;

[0061] in, Representing scale Enhanced response under the following conditions The maximum response across all scales is used as a candidate augmentation map. This indicates taking the maximum value. Indicates the measurement of tubular structure. Represents a measure of spherical structure. Indicates structural strength. , and This indicates the adjustment parameter, with the initial value set to... , , , Indicates or, Indicates other situations;

[0062] Laplacian fusion unit, for the original image, nonlocal means denoised image and The structure enhancement map results are subjected to multi-scale Gaussian pyramid decomposition. Weighted fusion is performed at each scale using saliency weights to preserve edge, detail, and structural consistency. Finally, the fused image is reconstructed, as shown below:

[0063] ;

[0064] in, Indicates the first The input image is at scale The Gaussian pyramid has three layers. Indicates the first The input image is at scale Laplace layer, Indicates layered output. Indicates the fusion weight. Indicates the first The input image is at scale The gradient of the Gaussian pyramid layer satisfies 'k' represents the index of the fused input source, which includes the original image, the non-local means denoised image, and... Structural reinforcement diagram, This represents an operator that performs upsampling and smooth interpolation. This represents the reconstructed fused image.

[0065] By performing the above operations, the traditional Image preprocessing and enhancement methods exist Addressing issues such as lung field deviation, coarse lung field segmentation leading to missed or misclassified cases, and insufficient ability to differentiate between different morphological structures like blood vessels and nodules, this approach utilizes specific... Intensity clipping and linear normalization are performed on intervals; a smooth bias field is established in the logarithmic domain and iterative optimization with regularization is used to recover the true tissue contrast; thresholding and connectivity are used for fast localization before... Variational refinement of the lung mask, nonlocal mean denoising within the lung using mask constraints to suppress random noise while preserving small lesions, and improvement by combining multi-scale Hessian matrix The algorithm performs scale-selective structure enhancement, and finally merges the original image, denoised image, and structure-enhanced image using multi-scale Laplacian and Gaussian pyramid methods, thereby achieving... Correct image restoration Compare and improve the accuracy of lung field and lesion boundaries, and preserve and amplify diagnostic details.

[0066] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. The multi-resolution fusion module allocates resolution resources based on the importance of image content, performs multi-scale resampling and weighted fusion on the enhanced image, retains high resolution in key structural regions, and reduces resolution in non-critical regions. Specifically, it includes the following units:

[0067] The content importance mapping unit calculates the importance weight of each voxel based on structural response and edge strength, assigning key structures to high resolution and strong detail, as shown below:

[0068] ;

[0069] in, Represents the original importance measure, and represents the enhancement term resulting from the product of structural strength and edge strength. This represents the importance weight of the content after full graph normalization. This indicates that the maximum value of y is obtained for all prime numbers.

[0070] The multi-resolution resampling unit first resamples the fused image into high-resolution and low-resolution versions, then weights them according to content importance. Voxel-weighted synthesis is performed, retaining high resolution in important regions and using low resolution in non-important regions, ultimately achieving seamless fusion, as shown below:

[0071] ;

[0072] in, This represents the resampling operator, which resamples the input to the voxel spacing. The grid, and These represent the high-resolution and low-resolution voxel spacing of the target, respectively, in millimeters; This indicates a high-resolution resampled image. This indicates a low-resolution resampled image. This represents the image after inductive resampling and fusion.

[0073] By performing the above operations, the traditional Commonly used resampling and display methods, such as fixed voxel spacing or full-volume oversampling, suffer from resolution loss in key nodules and small bronchi, high computational and storage overhead due to full-volume high resolution, and artifacts at resolution switching points. This solution addresses these issues by constructing a voxel-level content importance mapping based on multi-scale structural response and fused image gradients, and then applying this mapping to the fused image. Images are resampled to high and low resolution respectively, and then weighted point-by-point according to voxel-level weights to achieve seamless fusion. This preserves high spatial detail and measurement accuracy in key lesions, blood vessels, and bronchial regions, while reducing sampling in background areas to save resources. Furthermore, smooth weight transitions effectively avoid resolution boundary artifacts, achieving a balance between... Diagnostic visibility and clinical treatment efficiency.

[0074] Example 5, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. The fidelity export module saves the normalized version of the final output image and... The formula is restored, and the key parameter set is written into the metadata, specifically including the following units:

[0075] Normalized storage units are represented as follows:

[0076] ;

[0077] in, Indicates that it can be written back of Unit image;

[0078] Metadata record unit, which includes the following parameter set Write Private tags are represented as follows:

[0079] ;

[0080] in, Represents the parameter set.

[0081] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0082] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0083] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A chest CT image processing system, characterized in that: It includes a data acquisition module, an image enhancement module, a multi-resolution fusion module, and a high-fidelity export module; The data acquisition module collects chest data. image; The image enhancement module, for the original Intensity cropping, logarithmic domain bias field estimation and correction, fine segmentation of lung fields based on threshold and variational segmentation, nonlocal mean denoising, and multi-scale structural enhancement are performed. Then, the original image, the denoised image, and the structural enhancement image are fused together by multi-scale weighted fusion to output an enhanced fused image. The multi-resolution fusion module calculates the importance weight of the content based on the structural response and gradient intensity, allocates resolution resources at the voxel level according to the importance of the image, resamples the fused image into high resolution and low resolution respectively, and then uses the importance weight to perform voxel-by-voxel weighted synthesis. The fidelity export module saves the normalized version of the final output image. Restore the formula and write the set of key parameters into the metadata.

2. The chest CT image processing system according to claim 1, characterized in that: The image enhancement module enhances the original image data. Image preprocessing, bias field correction, lung field segmentation, denoising, and structural enhancement specifically include the following units: Intensity trimming unit, will remove the original Voxel values ​​are clipped to a fixed value. The range of values ​​is determined, and then linear normalization is performed. The bias field modeling unit estimates the smooth bias field in the logarithmic domain and removes... Low-frequency multiplicative noise exists, restoring the true tissue contrast; The lung field segmentation unit first uses thresholding and connectivity to quickly obtain candidate lung field regions, removes external thoracic structures, and then uses variational segmentation. The algorithm refines the boundaries within the candidate region to obtain a smooth, closed lung mask; The nonlocal mean denoising unit uses repeated blocks of recurring features in the image for weighted averaging to remove random noise while preserving fine structures as much as possible. The structural enhancement unit utilizes multi-scale Hessian matrix features to detect and enhance tubular and spherical structures.

3. The chest CT image processing system according to claim 2, characterized in that: The structural reinforcement unit specifically includes the following sub-units: The eigenvalue calculation subunit uses Gaussian kernel convolution to calculate the scale space, and calculates the Hessian matrix based on the scale space to obtain the three eigenvalues ​​of the Hessian matrix. Enhanced output subunit, introducing improved The type metric calculates the maximum response across all scales as a candidate augmentation map.

4. The chest CT image processing system according to claim 2, characterized in that: The image enhancement module further includes the following units: Laplacian fusion unit, for the original image, nonlocal means denoised image and The structure enhancement map results are subjected to multi-scale Gaussian pyramid decomposition, and weighted fusion is performed at each scale using saliency weights to preserve edge, detail and structural consistency. Finally, the fused image is reconstructed.

5. A chest CT image processing system according to claim 1, characterized in that: The multi-resolution fusion module allocates resolution resources based on the importance of image content, performs multi-scale resampling and weighted fusion on the enhanced image, retains high resolution in key structural regions, and reduces resolution in non-critical regions. Specifically, it includes the following units: The content importance mapping unit calculates the importance weight of each voxel based on structural response and edge strength, and assigns key structures to high resolution and strong detail. The multi-resolution resampling unit first resamples the fused image into high-resolution and low-resolution images respectively, and then performs voxel-by-voxel weighted synthesis according to the importance of the content. High resolution is retained in important areas and low resolution is used in unimportant areas, resulting in seamless fusion.

6. The chest CT image processing system according to claim 1, characterized in that: The fidelity export module saves the normalized version of the final output image. The formula is restored, and the key parameter set is written into the metadata, specifically including the following units: Normalized storage unit; Metadata record unit.

7. A chest CT image processing system according to claim 1, characterized in that: The data acquisition module receives clinical chest data. The output of the scanning device Formatted image data, extract the chest area to be processed. Images and Metadata.

Images