Differentiated CT Image Enhancement Method Based on Noise Level

Abstract

The present invention belongs to the field of medical image processing and provides a differential CT image enhancement method based on noise level, which includes six steps: data preprocessing, constructing and training a denoising network, constructing and training a super-resolution network, constructing and training a noise level classification network, CT image enhancement, and data postprocessing; by determining the noise level of CT images, the present invention uses different deep learning algorithms, namely denoising or super-resolution models, to optimize the quality of CT images, with stronger generalization ability and the ability to handle more complex and variable input data; a variety of traditional filtering methods and super-resolution pre-trained models are used to process noisy data to obtain the gold standard images required for training, so there is no need to spend time and effort in advance to collect high-quality gold standard images; the CT values of the output images are calibrated through postprocessing operations to ensure that the CT value differences before and after enhancement do not affect the accuracy of clinical diagnosis, and it has more practical application value compared with other methods without CT value calibration.

Description

Technical Field

[0001] The present invention belongs to the field of medical image processing, and specifically relates to a differential CT image enhancement method based on noise level. Background Art

[0002] Computed Tomography (CT) is a medical imaging technology that can generate two-dimensional or three-dimensional images representing the internal structure of the human body by using X-rays and computer processing. CT is widely used in medical diagnosis, treatment planning, disease monitoring and other fields. Due to various factors such as CT equipment, scattering of human tissues, reduced scanning dose, and statistical quantum noise, CT scans will introduce a certain degree of noise, which will reduce the image quality, reduce the contrast, and may interfere with doctors' interpretation of the images.

[0003] Improving the quality of low-quality CT images is a current research hotspot. Usually, CT quality improvement includes two parts: denoising and super-resolution reconstruction. First, denoise the CT image, and then perform super-resolution reconstruction on the denoised CT image. In the case of an image pair of high-quality CT images and low-quality CT images, relying on the Convolutional Neural Network (CNN) to learn the non-linear mapping between the source image and the target image end-to-end can achieve better CT quality improvement. However, when there are no high-quality CT images, and only CT images with average quality and noise exist, traditional supervised learning is difficult to train an effective deep learning model for CT quality improvement, and only various unsupervised methods or pre-trained weights can be used to improve its quality. In addition, different from high-level semantic processing such as CT image segmentation, since the film reading preferences of each clinician are different, there is no unified quantitative standard for "good", and the manual intervention required is much greater than the restoration and reconstruction of natural images.

[0004] The reconstruction quality of CT images is affected by various factors, among which the dose size is a key factor. High doses usually lead to high noise, while low doses lead to blurred images and loss of key details. Therefore, developing a CT image enhancement method that can handle various noise levels can not only significantly improve the image quality but also greatly improve the doctor's film reading efficiency.

[0005] Therefore, those skilled in the art have proposed a differential CT image enhancement method based on noise level. Summary of the Invention

[0006] In order to solve the above technical problems, the present invention provides a differential CT image enhancement method based on noise level to solve the problems mentioned in the background art.

[0007] Differentiated CT image enhancement method based on noise level, comprising the following steps:

[0008] S1. Construct a training dataset: Obtain DICOM format CT images with various noise levels; according to the suggestions of clinicians, divide the dataset into a high-noise-level sub-dataset and a low-noise-level sub-dataset;

[0009] S2. Data preprocessing: Normalize the CT images and adjust the data range to between 0 and 1;

[0010] S3. Construct a denoising network and train it: Select SCUNet as the denoising network; use various traditional unsupervised denoising methods to manually tune the parameters of the CT images in the high-noise-level sub-dataset for denoising. After the doctors confirm that it meets the clinical requirements, it is used as the gold standard image, and then a paired training dataset is obtained; train the network, and after convergence, obtain the final denoising model;

[0011] S4. Construct a super-resolution network and train it: Select the LIFF network as the super-resolution network; use the pre-trained model of the SwinIR image super-resolution network to perform super-resolution on the CT images in the low-noise-level sub-dataset. After the doctors confirm that it meets the clinical requirements, it is used as the gold standard image; train the network, and after convergence, obtain the final super-resolution model;

[0012] S5. Construct a noise-level classification network and train it: Design a multi-layer perceptron as the classification network; read the attribute representing the CT dose in the DICOM of the training dataset, and correspond the attribute value with the corresponding noise level determined by the doctors to make a classification network training set; train the network, and after convergence, obtain the final noise-level classification model;

[0013] S6. CT image enhancement: Given a DICOM format CT image, perform normalization using the data preprocessing method described in S1; read the attribute representing the CT dose in the DICOM, and use the noise-level classification network described in S5 to judge its noise level; if it is a high-noise level, use the denoising model for inference, otherwise use the super-resolution model for inference to obtain the output CT image;

[0014] S7. Data post-processing: For the output CT image, first adjust its size to be the same as the input; then crop its data range to between 0 and 1 and denormalize it to the data range of the input image; then use the MeanShift algorithm to adjust the local CT value so that the difference in CT value from the corresponding area on the input image is within the clinically acceptable range to obtain the enhanced CT image; save it as the final result in DICOM format. The calculation method of the MeanShift algorithm is:

[0015] output = MeanShift(input - output') + output'

[0016] Among them, MeanShift is a local average convolution kernel with a kernel size of 51 and a stride of 1. input, output', and output represent the input image, the output CT image after the denormalization step, and the enhanced CT image, respectively.

[0017] Preferably, when constructing the training dataset, the dataset contains a sufficient number of CT images scanned by machines from multiple manufacturers, with various resolutions and doses. Doctors will, based on their clinical film reading experience, divide all CT images into a high-noise-level sub-dataset and a low-noise-level sub-dataset from a diagnostic perspective.

[0018] Preferably, during data preprocessing, all CT images in the dataset are subjected to minimum-maximum normalization with a fixed value, that is

[0019] img = (img' - v min ) / (v max - v min )

[0020] Among them, img' and img represent the images before and after normalization respectively, and v min , v max are 0 and 4095 respectively, representing the maximum data range of CT images in dicom format.

[0021] Preferably, after optimizing each CT image using traditional methods and having it confirmed by doctors to meet clinical requirements as the corresponding gold standard, a deep learning network SCUNet with strong generalization ability and short processing time is trained and set as the denoising network.

[0022] Preferably, a super-resolution network is used to improve the quality of CT images with low noise levels, and the SwinIR natural image super-resolution network is adopted. The pre-trained model is used to perform super-resolution on the CT images in the low-noise-level sub-dataset as the gold standard to train the network.

[0023] Preferably, when training the noise level classification network, the attributes representing CT dose in dicom are read, including but not limited to exposure type, exposure time, tube voltage, and tube current. These attributes are corresponded to the noise level classification results determined by doctors and used as the training set to train the MLP network to obtain the noise level classification model.

[0024] Preferably, the CT value reflects the degree of X-ray absorption by different tissues. Different tissues and lesions, including but not limited to fat, muscle, blood, bone, and tumors, have different CT values. The MeanShift algorithm is used to reduce the local CT value differences on the input and output images through convolution filtering, so that the differences reach the clinically acceptable range standard.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] 1. By determining the noise level of the CT image, the present invention uses different deep learning algorithms, namely denoising or super-resolution models, to optimize the quality of the CT image. Compared with the existing single-model denoising or enhancement methods, it has stronger generalization ability and can handle more complex and variable input data.

[0027] 2. The present invention uses a variety of traditional filtering methods and super-resolution pre-training models to process the noisy data to obtain the gold standard images required for training. Therefore, there is no need to spend time and effort in advance to collect high-quality gold standard images, nor is there a data registration problem.

[0028] 3. The present invention calibrates the CT value of the output image through post-processing operations to ensure that the CT value difference before and after enhancement does not affect the accuracy of clinical diagnosis. Compared with other methods without CT value calibration, it has more practical application value.

[0029] 4. The noise of the CT image processed by the method of the present invention is significantly reduced, the edges are clear, approaching or even exceeding the quality of real high-quality CT images. In practical applications, only low-dose CT images are collected, and the effect of full-dose CT images can be obtained after enhancement, which can reduce the radiation damage suffered by patients during the scanning process and is beneficial to improving the doctor's film reading efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a flowchart of the differential CT image enhancement method based on noise level of the present invention;

[0031] Figure 2 is a diagram of the CT image enhancement module and data post-processing of the present invention;

[0032] Figure 3 is a comparison diagram before and after enhancing the CT image using the present invention, where Figure A.1 is the input original CT image with high noise level, Figure A.2 is the CT image enhanced by the denoising model, Figure B.1 is the input CT image with low noise level, and Figure B.2 is the CT image enhanced by the super-resolution model. DETAILED DESCRIPTION OF THE INVENTION

[0033] The following further describes the implementation manners of the present invention in conjunction with the accompanying drawings and embodiments. The following embodiments are used to illustrate the present invention, but cannot be used to limit the scope of the present invention.

[0034] As shown in the attached Figure 1 to the attached Figure 3 figures:

[0035] Embodiment 1: The present invention provides a differential CT image enhancement method based on noise level, including the following steps: S1. Construct a training data set: Obtain DICOM format CT images with various noise levels; according to the suggestions of clinicians, divide the data set into a high-noise-level sub-data set and a low-noise-level sub-data set. Taking a single-channel CT image with a size of 512×512 as an example, the obtained high- and low-noise image examples are shown in Figure 3 .A.1, 3.B.1;

[0036] S2. Data preprocessing: Normalize the CT images and adjust the data range to between 0 and 1.

[0037] S3. Construct a denoising network and train it: Select SCUNet (Swin-Conv-UNet) as the denoising network; use a variety of traditional unsupervised denoising methods to manually tune the parameters of the CT images in the high-noise-level sub-data set for denoising. After the doctor confirms that it meets the clinical requirements, it is used as the gold standard image, and then a paired training data set is obtained; train the network, and obtain the final denoising model after convergence.

[0038] Among them, the unsupervised denoising methods include one or a combination of more of micro BM3D filtering, discrete cosine transform, wavelet sharpening, Perona-Malik filtering, and TV filtering.

[0039] S4. Construct a super-resolution network and train it: Select the LIFF (Local Implicit Image Function) network as the super-resolution network; use the pre-trained model of the SwinIR (Transformer for Image Restoration) image super-resolution network to perform super-resolution on the CT images in the low-noise-level sub-data set. After the doctor confirms that it meets the clinical requirements, it is used as the gold standard image; train the network, and obtain the final super-resolution model after convergence.

[0040] Among them, the 2-fold or 4-fold pre-trained model of SwinIR is used; for the gold standard image obtained using the 4-fold pre-trained model, in order to reduce the image size, the Lancoz method is used to perform 2-fold image scaling on it.

[0041] S5. Construct and train a noise level classification network: Design a Multi-Layer Perceptron (MLP) as the classification network; Read the attributes representing CT dose in the DICOM of the training dataset, and map the attribute values to the corresponding noise levels determined by doctors one by one to create a training set for the classification network; Train the network, and after convergence, obtain the final noise level classification model.

[0042] Among them, the attributes representing CT dose include exposure type, exposure time, tube voltage, tube current, etc.

[0043] S6. CT image enhancement: Given a CT image in DICOM format, perform normalization using the data preprocessing method described in S1; Read the attributes representing CT dose in the DICOM, and use the noise level classification network described in S5 to judge its noise level; If it is a high noise level, use the denoising model for inference, otherwise use the super-resolution model for inference to obtain the output CT image.

[0044] S7. Data post-processing: For the output CT image, first adjust its size to be the same as the input; Then clip its data range to between 0 and 1 and de-normalize it to the data range of the input image; Then use the MeanShift algorithm to adjust the local CT values so that the difference in CT values from the corresponding regions on the input image is within the clinically acceptable range to obtain the enhanced CT image; Save it as the final result in DICOM format. The calculation method of the MeanShift algorithm is,

[0045] output = MeanShift(input - output') + output' where MeanShift is a local average convolution kernel with a kernel size of 51 and a stride of 1. input, output', and output represent the input image, the output CT image after the de-normalization step, and the enhanced CT image, respectively.

[0046] Embodiment 2: This embodiment is basically the same as the previous embodiment, except that,

[0047] In the data preprocessing step, the image is normalized to other ranges, such as between -1 and 1.

[0048] In the step of constructing and training the super-resolution network, Stable-Diffusion can be selected as the super-resolution network.

[0049] As can be seen from the above, the present invention can improve the quality of CT images without having to collect high-quality gold standard images in advance for use as training data. The noise of the CT images after quality improvement is significantly reduced, the edges are clear, approaching or even exceeding the quality of real high-quality CT images.

[0050] High-quality CT can improve doctors' clinical diagnosis efficiency. However, in practical applications, due to factors such as lack of equipment, high cost, or radiation, it is often impossible to obtain high-quality CT. This invention uses an unsupervised method to optimize low-quality CT images to obtain high-quality CT images, and uses them as the gold standard to train a deep learning network. The quality of the low-dose CT images enhanced by the method of this invention can reach or even exceed that of CT images obtained by standard-dose scanning, greatly reducing the radiation damage suffered by patients during the scanning process.

[0051] As can be seen from the above, this invention optimizes the quality of CT images by determining the noise level of CT images and using different deep learning algorithms, namely denoising or super-resolution models, which has stronger generalization ability and can handle more complex and variable input data; uses a variety of traditional filtering methods and super-resolution pre-training models to process noisy data to obtain the gold standard images required for training, so there is no need to spend time and effort in advance to collect high-quality gold standard images; calibrates the CT values of the output images through post-processing operations to ensure that the difference in CT values before and after enhancement does not affect the accuracy of clinical diagnosis. Compared with other methods without CT value calibration, it has more practical application value; the noise of the CT images processed by the method of this invention is significantly reduced, the edges are clear, approaching or even exceeding the quality of real high-quality CT images. In practical applications, only low-dose CT images are collected, and the effect of full-dose CT images can be obtained after enhancement, which can reduce the radiation damage suffered by patients during the scanning process and is beneficial to improving doctors' film reading efficiency.

[0052] In addition, in order to provide a concise description of exemplary embodiments, not all features of the actual embodiments may be described (i.e., those features that are not relevant to the currently considered best mode of implementing the invention or those features that are not relevant to implementing the invention).

[0053] It should be understood that in the development of any actual implementation, as in any engineering or design project, a large number of specific implementation decisions may be made. Such development efforts may be complex and time-consuming, but for those of ordinary skill in the art who benefit from this disclosure, without excessive experimentation, such development efforts will be a routine task of design, manufacturing, and production.

[0054] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A noise level-based differential CT image enhancement method, characterized in that: The following steps are involved: S1. Construct training dataset: obtain CT images in dicom format with various noise levels; divide the dataset into high noise level sub-dataset and low noise level sub-dataset according to the clinician's suggestion; S2, data preprocessing: normalize the CT image and adjust the data range to between 0 and 1; S3. Build and train the denoising network: Select SCUNet as the denoising network; use a variety of traditional unsupervised denoising methods to manually adjust the parameters of the CT images in the high-noise level sub-dataset for denoising. After the doctor confirms that it meets the clinical requirements, it is used as the gold standard image to obtain a paired training data set; train the network and obtain the final denoising model after convergence; S4. Build and train a super-resolution network: Select the LIIF network as the super-resolution network; use the pre-trained model of the SwinIR image super-resolution network to perform super-resolution on the CT images in the low-noise level sub-dataset, and use them as gold standard images after doctors confirm that they meet clinical requirements; Train the network and obtain the final super-resolution model after convergence; S5. Construct and train a noise level classification network: Design a multi-layer perceptron as a classification network; read the attributes representing CT dose in the DICOM in the training data set, match the attribute values ​​with the corresponding noise levels determined by the doctor, and make a classification network training set; Train the network and obtain the final noise level classification model after convergence; S6, CT image enhancement: given a CT image in dicom format, normalize it using the data preprocessing method described in S1; Read the attribute representing the CT dose in DICOM, and use the noise level classification network described in S5 to determine its noise level; if the noise level is high, use the denoising model for reasoning, otherwise use the super-resolution model for reasoning to obtain the output CT image; S7, data post-processing: For the output CT image, first adjust its size to be consistent with the input; then crop its data range to between 0 and 1, and denormalize it to the data range of the input image; then use the MeanShift algorithm to adjust the local CT value so that the difference between it and the CT value of the corresponding area on the input image is within the clinically acceptable range, and obtain the enhanced CT image; save the final result in dicom format, the calculation method of the MeanShift algorithm is: output=MeanShift(input-output')+output' Where MeanShift is a local mean convolution kernel with a kernel size of 51 and a step size of 1. Input, output', and output represent the input image, the output CT image after the denormalization step, and the enhanced CT image, respectively.

2. The noise level-based differentiated CT image enhancement method according to claim 1, characterized in that: When constructing the training data set, the data set contains a sufficient number of CT images scanned by machines from various manufacturers, with various resolutions and doses. Doctors will divide all CT images into high-noise level sub-datasets and low-noise level sub-datasets from a diagnostic perspective based on their clinical film reading experience.

3. The noise level-based differentiated CT image enhancement method according to claim 1, characterized in that: During data preprocessing, a fixed minimum-maximum normalization operation is used for all CT images in the dataset, that is, img=(img’-v min ) / (v max -v min ) Among them, img', img represent the images before and after normalization, v min， v max They are 0 and 4095 respectively, representing the maximum data range of CT images in dicom format.

4. The noise level-based differentiated CT image enhancement method according to claim 1, characterized in that: After each CT image is optimized using traditional methods and confirmed by doctors to meet clinical requirements, it is used as the corresponding gold standard. A deep learning network SCUNet with strong generalization ability and short processing time is trained and set as a denoising network.

5. The noise level-based differentiated CT image enhancement method according to claim 1, characterized in that: A super-resolution network is used to improve the quality of CT images with low noise levels, and the SwinIR natural image super-resolution network is used to train the network using a pre-trained model to perform super-resolution on CT images in the low-noise level sub-dataset as the gold standard.

6. The noise level-based differentiated CT image enhancement method according to claim 1, characterized in that: When training the noise level classification network, the attributes representing the CT dose in the DICOM are read, including but not limited to the exposure type, exposure time, tube voltage, and tube current. These attributes are matched with the noise level classification results determined by the doctor and used as the training set to train the MLP network to obtain the noise level classification model.

7. The noise level-based differentiated CT image enhancement method according to claim 1, characterized in that: The CT value reflects the degree of X-ray absorption by different tissues. Different tissues and lesions, including but not limited to fat, muscle, blood, bone and tumor, have different CT values. The MeanShit algorithm is used to reduce the local CT value differences between the input and output images through convolution filtering to meet the clinically accepted difference range standards.

Images