Lung CT intelligent optimization and focus accurate display method based on free breathing

Abstract

The invention provides a lung CT intelligent optimization and focus accurate display method based on free breathing. Relates to the technical field of medical image processing. The method comprises the following steps: acquiring a CT image of a target patient, and performing standardized preprocessing on a low-dose CT image and a standard-dose chest CT image to obtain a first preprocessed image and a second preprocessed image; performing full-automatic lung segmentation on the first preprocessed image by adopting a U-Net convolutional neural network model, and outputting a left lung region and a right lung region; and taking the second pre-processed image as a gold standard, training a deep convolutional neural network CycleGAN to obtain an image enhancement model, and performing image quality enhancement on the left and right lung regions in the first pre-processed image by using the image enhancement model to obtain an enhanced image. The chest image quality can be improved while the radiation dose of the patient is reduced.

Description

Technical Field

[0001] This application relates to the field of medical image processing technology, and in particular to a method for intelligent optimization and accurate display of lesions in lung CT based on free breathing. Background Technology

[0002] PET / CT imaging is an advanced diagnostic tool that combines functional PET imaging with anatomical CT imaging, and is widely used in the diagnosis of oncology, cardiovascular diseases, and neurological disorders. PET provides metabolic information by detecting the distribution of radioactive tracers, while CT provides precise anatomical information. However, CT scans in PET / CT typically use low doses, primarily for attenuation correction and anatomical localization. Their image quality is lower than that of standard-dose CT, making it difficult to meet certain diagnostic needs. Patients often require additional standard-dose CT scans, increasing radiation exposure risk. Furthermore, respiratory motion and scattering can cause image artifacts, further reducing image quality.

[0003] In recent years, deep learning technology has demonstrated tremendous potential in the field of medical image processing. Research shows that deep learning can be used for image denoising, super-resolution reconstruction, and cross-modal image synthesis. However, existing technologies mostly focus on the optimization of a single modality (PET or CT), lacking comprehensive optimization schemes for PET / CT images. Therefore, developing a deep learning-based optimization scheme for PET / CT chest images is a pressing technical problem that needs to be solved. Summary of the Invention

[0004] This application provides a method for intelligent optimization and precise lesion display of lung CT based on free breathing, which aims to reduce the patient's radiation dose while improving the quality of chest images.

[0005] In a first aspect, this application provides a method for intelligent optimization and precise lesion display in lung CT scans based on free breathing, including:

[0006] Obtain the patient's examination information, and filter out target patients based on the set criteria and the patient's examination information;

[0007] CT images of the target patient are acquired, including low-dose CT images and standard-dose chest CT images. The low-dose CT images and standard-dose chest CT images are subjected to standardized preprocessing to obtain a first preprocessed image and a second preprocessed image.

[0008] The U-Net convolutional neural network model is used to perform fully automatic lung segmentation on the first preprocessed image, and the left and right lung regions are output.

[0009] Using the second preprocessed image as the gold standard, a deep convolutional neural network CycleGAN is trained to obtain an image enhancement model. The image enhancement model is then used to enhance the image quality of the left and right lung regions in the first preprocessed image to obtain an enhanced image.

[0010] In one possible design, the target patient is a patient who has undergone routine 18F-FDG PET / CT examinations and has completed image acquisition including a whole-body low-dose CT scan and a standard-dose chest CT scan.

[0011] In one possible design, the normalization preprocessing includes image grayscale normalization, voxel size resampling, spatial registration, and window width / layer thickness adjustment.

[0012] In one possible design, when using the U-Net convolutional neural network model to perform fully automatic lung segmentation on the first preprocessed image, the target region is determined after removing the background and non-lung structures from the first preprocessed image by the U-Net convolutional neural network model, and the target region is output as the left and right lung regions.

[0013] In one possible design, during the training of the deep convolutional neural network CycleGAN, the deep convolutional neural network CycleGAN learns the mapping relationship between a first preset image and a second preset image to achieve detail restoration and noise suppression of the first preprocessed image, thereby achieving image enhancement.

[0014] In one possible design, after obtaining the enhanced image, the method further includes:

[0015] Calculate the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the enhanced image. If both the SNR and CNR of the enhanced image reach the corresponding set thresholds, output the enhanced image.

[0016] In one possible design, the formulas for calculating the signal-to-noise ratio and contrast-to-noise ratio are:

[0017] SNR=HU mean / HU std

[0018] CNR = (HU) 病灶 -HU 背景 ) / HUstd 背景

[0019] In the formula, SNR represents the signal-to-noise ratio, CNR represents the contrast-to-noise ratio, and HU... mean HU represents the average CT value of the left and right lung regions. std HU represents the standard deviation of CT values ​​in the left and right lung regions, reflecting the degree of dispersion of CT values ​​in the left and right lung regions. 病灶HU represents the average CT value of the lesion area, used to measure the density of the lesion. 背景 This represents the average CT value of the background area, used as a comparison reference. (HUstd) 背景 Standard deviation of CT values ​​in the background area

[0020] Secondly, this application provides a lung CT intelligent optimization and precise lesion display device based on free breathing, the device comprising:

[0021] The image acquisition module is configured to acquire CT images of a target patient, the CT images of the target patient including low-dose CT images and standard-dose chest CT images, and to perform standardized preprocessing on the low-dose CT images and standard-dose chest CT images to obtain a first preprocessed image and a second preprocessed image;

[0022] The region output module is configured to use a U-Net convolutional neural network model to perform fully automatic lung segmentation on the first preprocessed image and output the left and right lung regions.

[0023] The image enhancement module is configured to use the second preprocessed image as the gold standard to train a deep convolutional neural network CycleGAN to obtain an image enhancement model. The image enhancement model is then used to enhance the image quality of the left and right lung regions in the first preprocessed image to obtain an enhanced image.

[0024] Thirdly, embodiments of this application provide an electronic device, including: at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, causing the at least one processor to perform the method for intelligent optimization and precise display of lung CT lesions based on free breathing as described in the first aspect and various possible designs of the first aspect.

[0025] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the method for intelligent optimization and precise lesion display of lung CT based on free breathing as described in the first aspect and various possible designs of the first aspect.

[0026] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the method for intelligent optimization and precise lesion display of lung CT based on free breathing as described in the first aspect and various possible designs of the first aspect.

[0027] The method for intelligent optimization and precise lesion display of lung CT based on free breathing provided in this application has at least the following beneficial effects:

[0028] Image quality improvement: Optimized low-dose CT images have lower noise and higher resolution, improving the clarity of lesions and anatomical structures.

[0029] Reduced radiation dose: Optimizing low-dose CT images can reduce the need for standard-dose CT scans and reduce patient radiation exposure.

[0030] Improving diagnostic accuracy: High-quality images help improve the accuracy of lesion detection and disease staging, especially in the field of oncology. Attached Figure Description

[0031] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0032] Figure 1 A flowchart illustrating a method for intelligent optimization and precise lesion display in lung CT based on free breathing, provided in an embodiment of this application;

[0033] Figure 2 A flowchart of CT image acquisition and standardized preprocessing provided in the embodiments of this application;

[0034] Figure 3 A flowchart of lung segmentation for the U-Net convolutional neural network model provided in this application embodiment;

[0035] Figure 4 A flowchart of model training and image quality enhancement provided for embodiments of this application;

[0036] Figure 5 This is a structural diagram of the lung CT intelligent optimization and lesion precise display device based on free breathing provided in the embodiments of this application.

[0037] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0038] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0039] The collection, storage, use, processing, transmission, provision, and disclosure of financial data or user data involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0040] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0041] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0042] This application provides a method for intelligent optimization and precise lesion display in lung CT scans based on free breathing, such as... Figure 1 The diagram shown is a flowchart of the method for intelligent optimization and precise display of lesions in lung CT based on free breathing provided in this application embodiment. The method for intelligent optimization and precise display of lesions in lung CT based on free breathing includes the following steps S100-S300.

[0043] S100: Acquire CT images of the target patient, the CT images of the target patient including low-dose CT images and standard-dose chest CT images, and perform standardized preprocessing on the low-dose CT images and standard-dose chest CT images to obtain a first preprocessed image and a second preprocessed image.

[0044] It should be noted that the target patients were lung disease patients who underwent routine 18F-FDG PET / CT examinations, and all of them completed image acquisition including whole-body low-dose CT scans and standard-dose chest CT scans.

[0045] Low-dose CT images are used for anatomical localization and attenuation correction (AC) in PET scans, while standard-dose CT images are used for clinical diagnosis. However, conventional low-dose CT images suffer from high noise and low resolution, making them unsuitable for diagnostic purposes. By comparing these two types of images, a deep learning optimization model can be constructed, which helps to obtain high-quality images without increasing radiation exposure.

[0046] In this embodiment, the CT images of the target patient were acquired via PET / CT scans, including low-dose CT images (e.g., 120kV, 256mA, 5mm slice thickness) and standard-dose chest CT images (e.g., 120kV, 350mA, 1mm slice thickness). All images were uniformly stored in DICOM format and underwent standardized preprocessing, including image grayscale normalization, voxel size resampling, spatial registration, and window width / slice thickness adjustment. Image standardization effectively eliminates the interference of equipment parameter differences on model training, providing a unified and comparable basis for feature extraction and subsequent model building.

[0047] In one exemplary embodiment, such as Figure 2 The diagram shown is a flowchart of CT image acquisition and standardized preprocessing provided in an embodiment of this application. Step S100 is specifically implemented through the following steps S110-S130.

[0048] S110: Image acquisition preparation, including the following steps S111-S120.

[0049] S111. Screening target patients: Select patients who need to undergo lung CT examination, ensuring that they meet the indications for the examination and have no contraindications for CT examination, such as contrast agent allergy, severe cardiopulmonary insufficiency, etc. If the subsequent enhanced scan is involved, attention should be paid to this. If it is a plain scan, the focus is on basic contraindications.

[0050] S112. Equipment Setup and Scanning: Using CT equipment, perform low-dose CT and standard-dose chest CT scans on the patient sequentially. Set the low-dose CT scan parameters for a full chest scan to obtain low-dose CT images. These parameters include a tube voltage of 120kV, a tube current of 256mA, and a slice thickness of 5mm. Set the standard-dose chest CT scan parameters to focus on the chest area and obtain standard-dose chest CT images. These parameters include a tube voltage of 120kV, a tube current of 350mA, and a slice thickness of 1mm. Both scans must be completed within the same examination cycle (e.g., a single day, short time interval) to ensure consistent patient positioning, respiratory status, and other conditions.

[0051] S120: Image normalization preprocessing, including the following steps S121-S125.

[0052] S121. Format Conversion and Storage: Convert the original scanned images (usually in the original format of the device) into DICOM format and store them in a medical image storage system (such as a PACS system) for easy retrieval and processing later.

[0053] S122. Gray-level normalization: Traverse the pixels of low-dose and standard-dose CT images, calculate the mean and standard deviation of image gray levels, and normalize the pixel gray levels to make the gray-level distribution of different images tend to be consistent.

[0054] S123. Voxel resampling: Due to the different slice thicknesses of low-dose and standard-dose CT images (5mm and 1mm respectively), voxel resampling is required. An interpolation algorithm (such as linear interpolation) is used to uniformly adjust the voxel sizes of both low-dose and standard-dose CT images to the target size (e.g., 1mm × 1mm × 1mm), ensuring consistent spatial resolution during subsequent processing.

[0055] S124. Spatial registration: Based on the patient's anatomical landmarks (such as the thoracic skeleton, hilum, etc.), a rigid registration algorithm is used (since the two scans are completed for the same patient in a short period of time, the deformation is minimal, and rigid registration is sufficient) to spatially align the low-dose CT images with the standard-dose CT images to ensure that the corresponding anatomical positions coincide.

[0056] S125. Window width / slice thickness adjustment: Set an appropriate window width (e.g., lung window: window width 1500-2000HU, window level -500-600HU) according to the lung tissue display requirements, adjust the image display parameters, highlight the lung structure and lesion information, and obtain the first preprocessed image (after low-dose preprocessing) and the second preprocessed image (after standard-dose preprocessing).

[0057] S200: The U-Net convolutional neural network model is used to perform fully automatic lung segmentation on the first preprocessed image, and the left and right lung regions are output.

[0058] In this embodiment, the U-Net convolutional neural network model is used to automatically segment the lungs of the first preprocessed image (a standardized preprocessed low-dose CT image), outputting the left and right lung regions. Lung region extraction is a key preprocessing step for image optimization. After removing the background and non-lung structures, the focus can be placed on the target region, improving the training efficiency of the optimization model, reducing error propagation, and enhancing the specificity of the final image reconstruction.

[0059] In one exemplary embodiment, such as Figure 3 The diagram shown is a flowchart of lung segmentation of the U-Net convolutional neural network model provided in this application embodiment. Step S200 is specifically implemented through the following steps S210-S220.

[0060] S210: Model preparation and data preprocessing, including the following steps S211-S212.

[0061] S211. Model Selection and Loading: Call the pre-trained U-Net convolutional neural network model (which can be based on a public medical image segmentation dataset, such as LUNA16, and pre-trained with weights saved) and load it into the processing platform (such as Python + TensorFlow / PyTorch environment).

[0062] S212 Input Data Adaptation: Extract the first preprocessed image obtained in S100 and convert it into an input format acceptable to the model, such as adjusting it to a two-dimensional image slice sequence of a set size (e.g., 512×512 pixels) (because lung CT is three-dimensional data, it needs to be processed by slice layer by layer), and at the same time normalize the image pixel values ​​(e.g., map to [0,1]) to meet the model input requirements.

[0063] S220: Perform fully automated lung dissection, including the following steps S221-S222.

[0064] S221. Forward inference computation: The processed image slices are input into the U-Net model. The model performs segmentation prediction on the lung region in each slice through the encoding path (downsampling, extracting low-level to high-level features of the image) and the decoding path (upsampling, restoring the image size and refining the segmentation boundary), and outputs a binary segmentation result containing the left and right lung regions (lung tissue is 1, other tissues are 0).

[0065] S222, 3D Result Reconstruction: Stack the binarized segmentation results of all slices according to their spatial positions to reconstruct a 3D lung segmentation image, clearly marking the left and right lung regions, completing fully automatic lung segmentation, and obtaining the segmentation mask of the left and right lung regions for use in subsequent steps.

[0066] S300: Using the second preprocessed image as the gold standard, train a deep convolutional neural network CycleGAN to obtain an image enhancement model. Use the image enhancement model to enhance the image quality of the left and right lung regions in the first preprocessed image to obtain an enhanced image.

[0067] In one exemplary embodiment, such as Figure 4 The diagram shown is a flowchart of model training and image quality enhancement provided in the embodiment of this application. Step S300 is specifically implemented through the following steps S310-S320.

[0068] S310: Data preparation and model initialization, including the following steps S311-S312.

[0069] S311, Data Pairing: Extract the masks of the second preprocessed image (after standard dose chest CT preprocessing) from S100 and the left and right lung regions segmented from S200. Use the masks to extract the lung regions from the second preprocessed image as the "gold standard" samples. Simultaneously, extract the corresponding lung regions from the first preprocessed image based on the masks as low-dose input samples. Pair the two types of samples according to a one-to-one correspondence to construct a training dataset (e.g., 80% for training and 20% for validation).

[0070] S312. CycleGAN Model Construction: In a deep learning framework (such as TensorFlow / PyTorch), construct a deep convolutional neural network (CycleGAN) structure, including a generator G for learning the low-dose to standard-dose image mapping, a generator F for learning the standard-dose to low-dose inverse mapping, and discriminators D_X and D_Y for distinguishing between real and generated low-dose images and standard-dose real / generated images. Initialize model parameters, setting training hyperparameters such as the optimizer (e.g., Adam optimizer) and learning rate (e.g., 0.0002).

[0071] S320: Model training process, including the following steps S321-S324.

[0072] S321. Calculate the adversarial loss: Train the generator G to generate "fake" standard dose images, attempting to deceive the discriminator D_Y; the discriminator D_Y then distinguishes between the real standard dose images and the generated images, calculating the adversarial loss using a binary classification cross-entropy loss function, driving the generator to generate images closer to reality. Similarly, train the generator F and the discriminator D_X.

[0073] S322. Calculate cycle consistency loss: When the "fake" standard dose image generated by generator G is input into generator F, it should be restored back to the low dose input image; conversely, the standard dose image should also be restored after being transformed by generators F and G. By calculating the L1 norm loss between the restored image and the original image, the cycle consistency of image transformation is guaranteed, and the model is constrained to learn a stable mapping relationship.

[0074] S323. Iterative Training: Using paired samples as input, the generator and discriminator are trained alternately. The network parameters are continuously adjusted by the optimizer to minimize the sum of adversarial loss and cycle consistency loss until the loss function converges (e.g., the validation set loss no longer decreases), thus completing the model training and obtaining a stable image enhancement model.

[0075] S330: Perform image quality enhancement, including the following steps S331-S332.

[0076] S331, Enhanced Inference: Extract the left and right lung region masks obtained from the segmentation in S200, extract the lung region from the first preprocessed image, input it into the trained CycleGAN generator G, and the generator performs quality enhancement on the low-dose lung image (such as increasing resolution, reducing noise, and enhancing texture details) based on the learned mapping relationship, and outputs the enhanced lung region image.

[0077] S332, Image Fusion and Output: The enhanced lung region image is fused with the region outside the lung (such as the thoracic cavity, mediastinum, etc.) in the first preprocessed image (by performing a masking inverse operation, the enhanced lung region is replaced back into the original image) to obtain a complete enhanced image, thereby enhancing the image quality of the left and right lung regions in the first preprocessed image, completing the entire method flow, and outputting a high-quality enhanced CT image that can be used for subsequent diagnostic analysis.

[0078] In this embodiment, a second preprocessed image (a standardized preprocessed standard-dose chest CT image) is used as the gold standard to train a deep convolutional neural network, CycleGAN, to enhance the image quality of the lung region in low-dose CT scans. By learning the mapping relationship between low-dose CT and standard-dose CT, image detail restoration and noise suppression are achieved. The optimized CT image retains the radiation advantages of low-dose scanning while possessing near-diagnostic-grade CT image quality, meeting clinical evaluation needs.

[0079] In some embodiments, such as Figure 2 As shown, after obtaining the enhanced image, i.e. after step S300, the method further includes:

[0080] S400: Calculate the signal-to-noise ratio and contrast-to-noise ratio of the enhanced image, and output the enhanced image when both the signal-to-noise ratio and contrast-to-noise ratio of the enhanced image reach the corresponding set thresholds.

[0081] In some embodiments, the formulas for calculating the signal-to-noise ratio and the contrast-to-noise ratio are:

[0082] SNR=HU mean / HU std

[0083] CNR = (HU) 病灶 -HU 背景 ) / HUstd 背景

[0084] In the formula, SNR represents the signal-to-noise ratio, CNR represents the contrast-to-noise ratio, and HU... mean HU represents the average CT value of the left and right lung regions. std HU represents the standard deviation of CT values ​​in the left and right lung regions, reflecting the degree of dispersion of CT values ​​in the left and right lung regions. 病灶 HU represents the average CT value of the lesion area, used to measure the density of the lesion. 背景 This represents the average CT value of the background area, used as a comparison reference. (HUstd) 背景 This represents the standard deviation of CT values ​​in the background area.

[0085] To further illustrate the feasibility and advancement of the method proposed in this application, the image enhancement model will be evaluated below using both objective and subjective evaluation metrics. The objective evaluation metrics are the signal-to-noise ratio and contrast-to-noise ratio proposed in the above embodiments; the specific calculation methods are not detailed here.

[0086] The objective evaluation indicators were determined by having four radiologists use a four-point scale to blindly score the images based on overall image quality, lung texture clarity, and lesion visualization, and to assess whether the optimized CT images were sufficient to replace standard dose chest CT.

[0087] By supplementing objective indicators with subjective evaluations and combining them with actual diagnostic experience, the effectiveness of image optimization in clinical acceptability can be determined.

[0088] The results of evaluating the image enhancement model based on objective and subjective evaluation metrics are as follows:

[0089] Both the SNR and CNR of the CT images before and after optimization were significantly improved (P < 0.01).

[0090] Image quality rating

[0091] Noise control: The optimized CT score was significantly higher than that of the original low-dose CT (P<0.001) and close to that of the standard CT (equivalence test P=0.06).

[0092] Artifacts / Contrast: There was no significant difference between the optimized CT and the standard CT, with the original low-dose CT having the lowest score.

[0093] Lung Cancer Signs Assessment

[0094] Lesion visibility: The optimized CT score was equivalent to that of the standard CT (mean 3.6 vs. 3.8, CI [-0.2, 0.3]), while the original low-dose CT score was 2.1.

[0095] Edge features: The detection rate of spiculation signs by optimized CT was consistent with that of standard CT (P=0.25).

[0096] Diagnostic confidence

[0097] The overall diagnostic confidence score of the optimized CT was equivalent to that of the standard CT (mean 3.7 vs. 3.9, CI [-0.1, 0.3]), and significantly higher than that of the original low-dose CT (2.3).

[0098] Consistency Analysis

[0099] Inter-physician agreement: Fleiss Kappa = 0.65 (moderate agreement), Kendall's W = 0.82 (highly congruent).

[0100] Bland-Altman plots show that the score difference between optimized CT and standard CT is concentrated within ±0.5 points, with no systematic bias.

[0101] This application also provides a lung CT intelligent optimization and precise lesion display device based on free breathing, such as... Figure 5 As shown, this lung CT intelligent optimization and precise lesion display device based on free breathing includes:

[0102] The image acquisition module 501 is configured to acquire CT images of a target patient, the CT images of the target patient including low-dose CT images and standard-dose chest CT images, and to perform standardized preprocessing on the low-dose CT images and standard-dose chest CT images to obtain a first preprocessed image and a second preprocessed image.

[0103] The region output module 502 is configured to use a U-Net convolutional neural network model to perform fully automatic lung segmentation on the first preprocessed image and output the left and right lung regions.

[0104] The image enhancement module 503 is configured to use the second preprocessed image as the gold standard to train a deep convolutional neural network CycleGAN to obtain an image enhancement model, and use the image enhancement model to enhance the image quality of the left and right lung regions in the first preprocessed image to obtain an enhanced image.

[0105] In some embodiments, the target patient is a patient who has undergone routine 18F-FDG PET / CT examination and has completed image acquisition including a whole-body low-dose CT scan and a standard-dose chest CT scan.

[0106] In some embodiments, the normalization preprocessing includes image grayscale normalization, voxel size resampling, spatial registration, and window width / layer thickness adjustment.

[0107] In some embodiments, the region output module is further configured to, when performing fully automatic lung segmentation on the first preprocessed image using the U-Net convolutional neural network model, determine the target region after removing the background and non-lung structures from the first preprocessed image using the U-Net convolutional neural network model, and output the target region as the left and right lung regions.

[0108] In some embodiments, the image enhancement module is further configured to, during the training of the deep convolutional neural network CycleGAN, learn the mapping relationship between the first preset image and the second preset image to achieve detail restoration and noise suppression of the first preprocessed image, thereby achieving image enhancement.

[0109] In some embodiments, the apparatus includes an objective evaluation module, which is configured to:

[0110] Calculate the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the enhanced image. If both the SNR and CNR of the enhanced image reach the corresponding set thresholds, output the enhanced image.

[0111] In some embodiments, the formulas for calculating the signal-to-noise ratio and the contrast-to-noise ratio are:

[0112] SNR=HU mean / HU std

[0113] CNR = (HU) 病灶 -HU 背景 ) / HUstd 背景

[0114] In the formula, SNR represents the signal-to-noise ratio, CNR represents the contrast-to-noise ratio, and HU... mean HU represents the average CT value of the left and right lung regions. std HU represents the standard deviation of CT values ​​in the left and right lung regions, reflecting the degree of dispersion of CT values ​​in the left and right lung regions. 病灶 HU represents the average CT value of the lesion area, used to measure the density of the lesion. 背景 This represents the average CT value of the background area, used as a comparison reference. (HUstd) 背景 This represents the standard deviation of CT values ​​in the background area.

[0115] This application provides an electronic device. The electronic device may include a processor and a memory, wherein the processor and the memory can communicate; exemplarily, the processor and the memory communicate via a communication bus.

[0116] The processor executes computer execution instructions stored in memory, causing the processor to perform the scheme in the above embodiments. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0117] The communication bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. Transceivers are used to enable communication between database access devices and other computers (e.g., clients, read-write libraries, and read-only libraries). Memory may include random access memory (RAM) and may also include non-volatile memory.

[0118] The electronic device provided in this application embodiment can be the terminal device described in the above embodiments.

[0119] This application also provides a computer-readable storage medium storing computer instructions. When the computer instructions are executed on a computer, the computer performs the technical solution of the above-described embodiment of the method for intelligent optimization and precise display of lung CT lesions based on free breathing.

[0120] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When the at least one processor executes the computer program, it can implement the technical solution of the lung CT intelligent optimization and lesion precise display method based on free breathing in the above embodiments.

[0121] 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 merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0122] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0123] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0124] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0125] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0126] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0127] Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Architecture (EISA) buses, etc. Buses can be categorized into address buses, data buses, control buses, etc.

[0128] The aforementioned storage medium can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium accessible to general-purpose or special-purpose computers.

[0129] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic control unit or main control device.

[0130] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for intelligent optimization and precise lesion display in lung CT based on free breathing, characterized in that, The method includes: CT images of the target patient are acquired, including low-dose CT images and standard-dose chest CT images. The low-dose CT images and standard-dose chest CT images are subjected to standardized preprocessing to obtain a first preprocessed image and a second preprocessed image. The U-Net convolutional neural network model is used to perform fully automatic lung segmentation on the first preprocessed image, and the left and right lung regions are output. Using the second preprocessed image as the gold standard, a deep convolutional neural network CycleGAN is trained to obtain an image enhancement model. The image enhancement model is then used to enhance the image quality of the left and right lung regions in the first preprocessed image to obtain an enhanced image.

2. The method for intelligent optimization and precise lesion display of lung CT based on free breathing according to claim 1, characterized in that, The target patients are those who have undergone routine 18F-FDG PET / CT examinations and have completed image acquisition including whole-body low-dose CT scans and standard-dose chest CT scans.

3. The method for intelligent optimization and precise lesion display of lung CT based on free breathing according to claim 1, characterized in that, The standardized preprocessing includes image grayscale normalization, voxel size resampling, spatial registration, and window width / layer thickness adjustment.

4. The method for intelligent optimization and precise lesion display of lung CT based on free breathing according to claim 1, characterized in that, When performing fully automatic lung segmentation on the first preprocessed image using the U-Net convolutional neural network model, the target region is determined after removing the background and non-lung structures from the first preprocessed image using the U-Net convolutional neural network model, and the target region is output as the left and right lung regions.

5. The method for intelligent optimization and precise lesion display of lung CT based on free breathing according to claim 1, characterized in that, During the training of the deep convolutional neural network CycleGAN, CycleGAN learns the mapping relationship between the first preset image and the second preset image to achieve detail restoration and noise suppression of the first preprocessed image, thereby achieving image enhancement.

6. The method for intelligent optimization and precise lesion display of lung CT based on free breathing according to claim 1, characterized in that, After obtaining the enhanced image, the method further includes: Calculate the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the enhanced image. If both the SNR and CNR of the enhanced image reach the corresponding set thresholds, output the enhanced image.

7. The method for intelligent optimization and precise lesion display of lung CT based on free breathing according to claim 6, characterized in that, The formulas for calculating the signal-to-noise ratio and contrast-to-noise ratio are as follows: SNR=HU mean / HU std CNR=(HU 病灶 -HU 背景 ) / HUstd 背景 In the formula, SNR represents the signal-to-noise ratio, CNR represents the contrast-to-noise ratio, and HU... mean HU represents the average CT value of the left and right lung regions. std HU represents the standard deviation of CT values ​​in the left and right lung regions, reflecting the degree of dispersion of CT values ​​in the left and right lung regions. 病灶 HU represents the average CT value of the lesion area, used to measure the density of the lesion. 背景 This represents the average CT value of the background area, used as a comparison reference. (HUstd) 背景 This represents the standard deviation of CT values ​​in the background area.

8. A lung CT intelligent optimization and precise lesion display device based on free breathing, characterized in that, The device includes: The image acquisition module is configured to acquire CT images of a target patient, the CT images of the target patient including low-dose CT images and standard-dose chest CT images, and to perform standardized preprocessing on the low-dose CT images and standard-dose chest CT images to obtain a first preprocessed image and a second preprocessed image; The region output module is configured to use a U-Net convolutional neural network model to perform fully automatic lung segmentation on the first preprocessed image and output the left and right lung regions. The image enhancement module is configured to use the second preprocessed image as the gold standard to train a deep convolutional neural network CycleGAN to obtain an image enhancement model. The image enhancement model is then used to enhance the image quality of the left and right lung regions in the first preprocessed image to obtain an enhanced image.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes the computer execution instructions stored in the memory to implement the method for intelligent optimization and precise display of lung CT lesions based on free breathing as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method for intelligent optimization and precise lesion display of lung CT based on free breathing as described in any one of claims 1-7.

Images