Multi-modal medical image noise reduction optimization system and optimization method

Abstract

The invention discloses a multi-modal medical image noise reduction optimization system and optimization method, and the system comprises a data preprocessing module which is used for receiving and preprocessing multi-modal medical image data, and obtaining the preprocessed multi-modal medical image data; the CNN noise reduction module is used for performing noise reduction processing on the preprocessed multi-modal medical image data by using the trained CNN model to obtain the multi-modal medical image data after noise reduction processing; and the post-processing module is used for optimizing the multi-modal medical image data subjected to noise reduction processing to obtain optimized multi-modal medical image data. It can be seen that the neural network model in the application is obtained by training a large amount of sample data, the noise in the multi-modal medical image data can be effectively removed, and the quality and definition of the medical image are improved. And the multi-modal medical image data after noise reduction processing is further optimized, so that the noise reduction effect is improved, and the medical image data after noise reduction is more accurate.

Description

Technical Field

[0001] The present application relates to the field of medical image processing technology, and in particular to a multimodal medical image noise reduction optimization system and optimization method. Background Art

[0002] Multimodal medical imaging data refers to the imaging data obtained by different medical imaging technologies. Medical image denoising technology refers to removing noise from medical images by applying digital signal processing technology to improve the quality and readability of images.

[0003] At present, medical images can be de-noised by spatial domain filtering, for example, by using Gaussian filter, convolving the image with Gaussian function to achieve noise smoothing. It is suitable for Gaussian noise, but may cause image blur, resulting in poor noise reduction effect of medical images. Summary of the invention

[0004] In view of this, the present application provides a multimodal medical image denoising optimization system and optimization method to improve the denoising effect of medical images.

[0005] In a first aspect, the present application provides a multimodal medical image noise reduction optimization system, the system comprising:

[0006] A data preprocessing module, used for receiving and preprocessing multimodal medical image data to obtain preprocessed multimodal medical image data;

[0007] A CNN denoising module, used to perform denoising on the pre-processed multimodal medical image data using a trained CNN model to obtain denoised multimodal medical image data; the CNN model is trained using multimodal medical image sample data with noise labels;

[0008] The post-processing module is used to optimize the multimodal medical image data after the noise reduction process to obtain optimized multimodal medical image data.

[0009] Optionally, the data preprocessing module includes:

[0010] A data collection submodule, used to collect multimodal medical imaging data;

[0011] A data cleaning submodule, used to remove irrelevant information and noise in the multimodal medical image data to obtain cleaned multimodal medical image data;

[0012] The data standardization submodule is used to standardize the cleaned multimodal medical image data to obtain standardized multimodal medical image data.

[0013] Optionally, the CNN denoising module includes:

[0014] CNN model building submodule, used to build a CNN model for multimodal medical imaging data denoising;

[0015] The training data preparation submodule is used to prepare the training data required for training the CNN model;

[0016] The model training submodule is used to train the CNN model using the training data to obtain a trained CNN denoising model.

[0017] Optionally, the post-processing module includes:

[0018] A post-processing algorithm determination submodule is used to determine the post-processing algorithm corresponding to the current multimodal medical image data;

[0019] A post-processing execution submodule, used to optimize the multi-modal medical image data after the noise reduction processing by using the post-processing algorithm to obtain optimized multi-modal medical image data;

[0020] The result output submodule is used to output the optimized multimodal medical imaging data.

[0021] Optionally, the CNN model is a deep convolutional neural network model, including multiple convolutional layers, pooling layers and fully connected layers, which is used to extract features in multimodal medical imaging data and perform noise reduction processing.

[0022] Optionally, the training data includes multimodal medical imaging data with noise labels;

[0023] During the training process of the CNN model, the loss function is calculated by comparing the denoised medical image data with the original medical image data to adjust the parameters of the CNN model.

[0024] Optionally, the post-processing algorithm includes one or more of a histogram equalization algorithm, a median filtering algorithm, and an edge-preserving filtering algorithm.

[0025] In a second aspect, the present application provides a multimodal medical image noise reduction optimization method, which is applied to the system as described in the first aspect, and the method includes:

[0026] Acquire multimodal medical image data to be denoised;

[0027] Inputting the multimodal medical image data to be denoised into the CNN model to obtain the multimodal medical image data after denoising;

[0028] The multimodal medical image data after the noise reduction process is optimized to obtain optimized multimodal medical image data.

[0029] Optionally, the training process of the CNN model includes:

[0030] Acquire a training sample set, wherein the training sample set includes multimodal medical image sample data with noise labels;

[0031] The CNN model is trained using the training data set, and the model parameters of the CNN model are optimized based on a back propagation algorithm and a gradient descent algorithm until the model convergence condition is reached, thereby obtaining the CNN model.

[0032] Optionally, after obtaining the CNN model, the method further includes:

[0033] The CNN model is tested using a test dataset.

[0034] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device includes a controller, the controller includes a memory and a processor;

[0035] The memory is used to store one or more programs;

[0036] The processor; when the one or more programs are executed by the processor, the multimodal medical image denoising optimization method as described in the second aspect is implemented.

[0037] In a fourth aspect, an embodiment of the present application provides a computer storage medium, in which a program is stored. When the program is executed by a processor, the multimodal medical image denoising optimization method described in the second aspect is implemented.

[0038] The above technical solution has the following beneficial effects:

[0039] The embodiment of the present application provides a multimodal medical image denoising optimization system, which includes: a data preprocessing module for receiving and preprocessing multimodal medical image data to obtain preprocessed multimodal medical image data; a CNN denoising module for performing denoising on the preprocessed multimodal medical image data using a trained CNN model to obtain denoised multimodal medical image data; a post-processing module for optimizing the denoised multimodal medical image data to obtain optimized multimodal medical image data. It can be seen that the neural network model in the present application is obtained by training a large amount of sample data, which can effectively remove noise from multimodal medical image data, improve the quality and clarity of medical images, and facilitate further analysis and processing of medical images. And the denoised multimodal medical image data is further optimized by the post-processing module, thereby improving the denoising effect, so that the denoised medical image data is more accurate and clear. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in this embodiment or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A schematic diagram of the structure of a multimodal medical image noise reduction optimization system provided in an embodiment of the present application;

[0042] Figure 2 A flowchart of a multimodal medical image denoising optimization method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0043] The technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0044] In order to facilitate further understanding of the technical solution provided by this application, the background technology involved in this application will be explained below.

[0045] At present, the methods for medical image denoising include:

[0046] 1. Spatial domain filtering method

[0047] 1-1) Mean filter

[0048] By calculating the average value of the pixel neighborhood to replace the original pixel value, noise smoothing is achieved, but it may cause image edges to be blurred.

[0049] 1-2) Median filter

[0050] Sort the pixel values ​​in the pixel neighborhood and take the median as the new pixel value. It works better for salt and pepper noise, but may change image details.

[0051] 1-3) Gaussian filter

[0052] Use Gaussian function to convolve the image to smooth the noise. It is suitable for Gaussian noise, but may cause image blur.

[0053] 2. Frequency domain filtering method

[0054] 2-1) Low-pass filter

[0055] Reduce noise by filtering out high-frequency components of the image. Suitable for periodic noise, but may lose image details and edge information.

[0056] 2-2) Bandpass filter

[0057] Allows signals within a specific frequency range to pass through and filters out other frequency components. Applicable to certain types of noise, but requires selecting an appropriate passband range.

[0058] 2-3) Homomorphic filter

[0059] The illumination and reflectance components of the image are processed in the frequency domain to achieve noise reduction and image enhancement. However, the calculation complexity is high and the effect is greatly affected by the parameters.

[0060] In order to overcome the above technical problems, the embodiment of the present application provides a multimodal medical image denoising optimization system, which includes: a data preprocessing module for receiving and preprocessing multimodal medical image data to obtain preprocessed multimodal medical image data; a CNN denoising module for using a trained CNN model to perform denoising on the preprocessed multimodal medical image data to obtain denoised multimodal medical image data; a post-processing module for optimizing the denoised multimodal medical image data to obtain optimized multimodal medical image data. It can be seen that the neural network model in the present application is obtained by training a large amount of sample data, which can effectively remove noise from multimodal medical image data, improve the quality and clarity of medical images, and facilitate further analysis and processing of medical images. And the denoised multimodal medical image data is further optimized by the post-processing module, thereby improving the denoising effect, so that the denoised medical image data is more accurate and clear.

[0061] See also Figure 1 , is a structural diagram of a multimodal medical image denoising and optimization system provided in an embodiment of the present application. The multimodal medical image denoising and optimization system 100 includes: a data preprocessing module 101, a CNN denoising module 102 and a post-processing module 103.

[0062] The data preprocessing module 101 is used to receive and preprocess the multimodal medical image data to obtain preprocessed multimodal medical image data.

[0063] The CNN denoising module 102 is used to perform denoising on the pre-processed multimodal medical image data using a trained CNN model to obtain denoised multimodal medical image data.

[0064] The post-processing module 103 is used to optimize the multimodal medical image data after the noise reduction process to obtain optimized multimodal medical image data.

[0065] In a possible implementation, the data preprocessing module includes:

[0066] A data collection submodule, used to collect multimodal medical imaging data;

[0067] A data cleaning submodule, used to remove irrelevant information and noise in the multimodal medical image data to obtain cleaned multimodal medical image data;

[0068] The data standardization submodule is used to standardize the cleaned multimodal medical image data to obtain standardized multimodal medical image data.

[0069] In this embodiment, after acquiring the multimodal medical image data, the data cleaning submodule removes irrelevant information and noise in the multimodal medical image data to obtain cleaned multimodal medical image data. The purpose of this step is to deal with problems such as outliers, missing values, and noise in the multimodal medical image data. Outliers may significantly affect the performance of the model, so they need to be detected and processed.

[0070] Furthermore, the data standardization submodule performs standardization processing on the cleaned multimodal medical image data to obtain standardized multimodal medical image data. Data standardization can make the features in different dimensions have the same weight, thereby improving the optimization efficiency.

[0071] It can be understood that the embodiment of the present application removes irrelevant information and noise in the multimodal medical imaging data through the data cleaning submodule, and standardizes the cleaned multimodal medical imaging data through the data standardization submodule, thereby improving the accuracy of subsequent noise reduction and optimization processes.

[0072] In one possible implementation, the CNN denoising module includes:

[0073] CNN model building submodule, used to build a CNN model for multimodal medical imaging data denoising;

[0074] The training data preparation submodule is used to prepare the training data required for training the CNN model;

[0075] The model training submodule is used to train the CNN model using the training data to obtain a trained CNN denoising model.

[0076] In this embodiment, the CNN model construction submodule constructs a CNN model for denoising multimodal medical imaging data.

[0077] The multimodal medical image denoising algorithm optimization model based on CNN (convolutional neural network) is a complex deep learning framework specially designed to process multimodal image data from different medical imaging devices. The main goal of this model is to remove or reduce noise in the image so that the anatomical structure and lesion details can be displayed more clearly, thereby improving the accuracy of medical diagnosis.

[0078] In one possible implementation, the CNN model is a deep convolutional neural network model, including multiple convolutional layers, pooling layers, and fully connected layers, which is used to extract features from multimodal medical imaging data and perform noise reduction processing.

[0079] Next, the optimization model of the multimodal medical image denoising algorithm based on CNN (convolutional neural network) is introduced in detail.

[0080] Specifically, in the embodiment of the present application, the CNN model may include:

[0081] Input Layer: Responsible for receiving raw multimodal medical imaging data. The input data can come from different devices such as MRI, CT, ultrasound or X-ray, usually in the form of two-dimensional or three-dimensional images.

[0082] Convolutional Layer: The convolutional layer is the core of the CNN model and is used to extract local features of the input image. In the denoising task, the convolutional layer can learn filters or kernels that reflect the image content (such as edges, textures, etc.).

[0083] Pooling Layer: The pooling layer is located after the convolutional layer and is used to reduce the dimension and computational complexity of the feature map while retaining important spatial information. Common pooling operations include maximum pooling and average pooling.

[0084] Batch Normalization Layer: The batch normalization layer is used to improve the training speed and stability of the model. It normalizes the input data of each batch to reduce internal covariate shift.

[0085] Residual Connection Layer: For deeper network structures, introducing residual connections can help solve the problem of gradient disappearance or gradient explosion. Through skip connections, the model can better learn the identity mapping, thereby improving the noise reduction performance.

[0086] Feature Fusion Layer: In multimodal medical image denoising, the feature fusion layer is used to integrate image features from different modalities. This helps the model learn richer information and improves denoising effects and diagnostic accuracy.

[0087] Output Layer: The output layer is responsible for generating the denoised medical image. Usually, this layer uses the same data type and dimension as the input layer to facilitate comparison and evaluation with the actual medical image.

[0088] In addition to the above key layers, the model can also have other types of layers, such as activation function layers (such as ReLU, sigmoid or tanh, etc.), fully connected layers (for classification or regression tasks), and regularization layers (such as Dropout, used to prevent overfitting).

[0089] During the training process, the training data preparation submodule first prepares the training data required for training the CNN model. The model training submodule uses the training data to train the CNN model to obtain a trained CNN denoising model.

[0090] It should be noted that in order to improve the noise reduction accuracy of the CNN model, the embodiment of the present application provides a large number of training sample sets, and the training sample sets are introduced in detail below.

[0091] In this embodiment, the training sample set may include:

[0092] 1. CT and MRI paired data: This dataset usually contains CT and MRI images of the same patient. CT images provide structural information, while MRI provides more soft tissue contrast. By training with images from these two modalities, the ability of the denoising algorithm to preserve structure and contrast can be improved.

[0093] 2. MRI multiple sequence data: MRI can usually collect multiple different sequences, such as T1, T2, PD (proton density), etc. Each sequence reflects different tissue characteristics. Including multiple sequence images in the training data set at the same time can improve the algorithm's adaptability to different tissue characteristics.

[0094] 3. Image data of different resolutions: In actual applications, the resolution of medical images may vary due to limitations of equipment, acquisition parameters, etc. Including images of different resolutions in the training data set can help the algorithm learn how to adapt to images of different resolutions, thereby improving its performance in actual applications.

[0095] 4. Image pairs with and without noise: In order to train an effective denoising algorithm, it is necessary to prepare some image pairs with and without noise as training samples. These image pairs can be generated by adding artificial noise to the original images, or they can be obtained from actual clinical data. By incorporating these multimodal medical images into the training dataset, a more robust and general CNN-based medical image denoising algorithm can be trained to better meet the actual clinical application needs.

[0096] 5. Synthetic Data: To increase the number and diversity of training samples, additional training data can be generated using image synthesis techniques. This can be achieved by simulating different types of noise and adding it to the original image.

[0097] 6. Labeled data: In some cases, images in the training dataset may have labels or annotations associated with them, such as annotated lesion areas. These labels can be used for supervised learning, helping the algorithm learn denoising and feature extraction more accurately.

[0098] It can be understood that the embodiment of the present application utilizes a large amount of data for training, so that a more accurate CNN model for multimodal medical image data denoising can be trained, and the trained CNN model has higher denoising accuracy.

[0099] In a possible implementation, the training data includes multimodal medical imaging data with noise labels;

[0100] During the training process of the CNN model, the loss function is calculated by comparing the denoised medical image data with the original medical image data to adjust the parameters of the CNN model.

[0101] In this embodiment, during the training and optimization process, the model will compare the denoised medical image data with the original medical image data, calculate the loss function (such as mean square error, structural similarity index, etc.) to measure the denoising effect, and use gradient descent or other optimization algorithms to adjust the network parameters, thereby gradually improving the denoising performance. At the same time, in order to cope with the differences between images of different modalities, the model can also adopt modality-specific preprocessing and parameter adjustment strategies.

[0102] After training the CNN model for multimodal medical image data denoising, the CNN model is further tested using test data.

[0103] Test data can include:

[0104] 1. Unseen multimodal images: The test dataset should contain multimodal image data that has not been seen during the training phase to ensure that the algorithm has generalization ability. This may include images from different scanners, different parameter settings, or different patient groups.

[0105] 2. Challenging images: To evaluate the performance of the algorithm in complex and difficult situations, the test dataset should contain challenging images, such as those with high noise levels, complex anatomical structures, or artifacts.

[0106] 3. Labeled images: Similar to the training dataset, the images in the test dataset can also contain labels or annotations. These labels can be used to quantitatively evaluate the performance of the algorithm, such as by calculating the similarity between the denoised image and the original image or by evaluating the detection accuracy of the lesion area.

[0107] 4. External datasets: In addition to imaging data from the same hospital or research team, you can also consider using external datasets for testing. This helps evaluate the versatility of the algorithm under different data sources and conditions.

[0108] After training the CNN model, the performance of the multimodal medical image denoising algorithm based on CNN (convolutional neural network) can be evaluated in the following ways:

[0109] 1. Noise suppression performance: By comparing the images before and after noise reduction, we can observe whether the noise is effectively suppressed. The Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) can be used to measure the effect of noise suppression. The higher the PSNR, the closer the SSIM is to 1, indicating that the noise reduction effect is better.

[0110] 2. Image quality: Evaluate whether the denoised image retains important information of the original image, such as edges, textures, and contrast. Image quality can be evaluated through subjective evaluation or by calculating image features such as gradients and textures.

[0111] 3. Model complexity: Considering the actual needs of medical image processing, model complexity is also an important evaluation indicator. The model complexity can be measured by the number of model parameters, computational effort (FLOPs), or inference time. Under the premise of ensuring the denoising effect, choose a model with low complexity as much as possible.

[0112] 4. Multimodal compatibility: Multimodal medical image denoising algorithms need to process different types of medical images (such as CT, MRI, ultrasound, etc.). Therefore, it is crucial to evaluate the performance of the model on images of different modalities. The multimodal compatibility of the model can be evaluated by comparing the denoising effect of the model on images of each modality and the processing effect of cross-modal images.

[0113] 5. Robustness: In practical applications, medical images may be interfered by various noises and artifacts. Therefore, it is very important to evaluate the performance of the model under different noise levels and different noise types. The robustness of the model can be evaluated by adding noises of different intensities and different types to the test dataset.

[0114] 6. Clinical applicability: Ultimately, it is crucial to evaluate the performance of the model in clinical applications. Doctors can be invited to make subjective evaluations of the denoised images, or clinical indicators (such as diagnostic accuracy) can be used to evaluate the value of the model in clinical applications.

[0115] It should be noted that in order to comprehensively evaluate the performance of the CNN-based multimodal medical image denoising algorithm, the above methods and indicators can be combined for evaluation when evaluating the CNN model. At the same time, it is also necessary to pay attention to maintaining the fairness and repeatability of the evaluation process to ensure the credibility and reliability of the evaluation results.

[0116] In summary, the multimodal medical image denoising algorithm based on CNN (convolutional neural network) trained in this application has the ability to process different types of medical image data, including:

[0117] 1. Feature extraction capability: CNN can extract local features in images through convolutional layers, and gradually abstract and integrate these features through layer-by-layer convolution and pooling operations to form a deep understanding of the image content. This feature extraction capability enables CNN to process medical images of different modalities, such as CT, MRI, X-ray, etc., and extract information useful for noise reduction.

[0118] 2. Generalization ability: The deep learning structure of CNN gives it strong generalization ability, that is, it can maintain good performance on different types of medical imaging data. By training a large amount of medical imaging data, CNN can learn the inherent laws and commonalities between various imaging data, thereby achieving rapid adaptation to new types of imaging data.

[0119] 3. Parameter sharing: In CNN, the parameters of the convolution kernel are shared, which means that no matter how the size and modality of the input image change, the convolution kernel can extract the corresponding features through sliding operations. This parameter sharing feature makes it unnecessary for CNN to design algorithms for each modality of image when processing different types of medical imaging data, thereby improving the versatility and efficiency of the algorithm.

[0120] 4. Nonlinear mapping capability: CNN introduces nonlinear mapping through activation functions, which enables the network to fit complex nonlinear relationships. In the task of medical image denoising, the relationship between noise and image content is often nonlinear. The nonlinear mapping capability of CNN enables it to effectively suppress noise while retaining useful information in the image.

[0121] In summary, the CNN-based multimodal medical image denoising algorithm can process different types of medical image data and achieve effective denoising through its powerful feature extraction, generalization, parameter sharing and nonlinear mapping capabilities.

[0122] In a possible implementation, the post-processing module includes:

[0123] A post-processing algorithm determination submodule is used to determine the post-processing algorithm corresponding to the current multimodal medical image data;

[0124] A post-processing execution submodule, used to optimize the multi-modal medical image data after the noise reduction processing by using the post-processing algorithm to obtain optimized multi-modal medical image data;

[0125] The result output submodule is used to output the optimized multimodal medical imaging data.

[0126] In this embodiment, after obtaining the multimodal medical image data after denoising through the CNN denoising module, the post-processing algorithm determination submodule first determines the post-processing algorithm corresponding to the current multimodal medical image data, and then the post-processing execution submodule uses the post-processing algorithm to optimize the multimodal medical image data after denoising, thereby obtaining the optimized multimodal medical image data, and finally the result output submodule outputs the optimized multimodal medical image data.

[0127] Among them, the post-processing algorithm may include one or more of a histogram equalization algorithm, a median filtering algorithm, and an edge-preserving filtering algorithm. The present application does not limit the post-processing algorithm.

[0128] It can be understood that the embodiment of the present application further processes and optimizes the denoised data through a post-processing module to obtain denoised medical image data, which can improve the denoising effect and make the denoised medical image data more accurate and clear.

[0129] In summary, this application mainly realizes the ability to process different types of medical imaging data through improvements and expansions in the following aspects:

[0130] First, the versatility of the network structure. In order to process different types of medical imaging data, we designed the network structure to be versatile. This includes designing network layers that can process data of multiple resolutions and modalities, such as using multi-scale feature extraction technology and customizing convolutional layers for the characteristics of specific medical imaging data.

[0131] Secondly, improve the adaptability and generalization ability of the model. One of the goals of the optimization algorithm is to enhance the adaptability and generalization ability of the model to different image data. This application achieves this by introducing regularization techniques (such as Dropout, L1 / L2 regularization, etc.) to reduce overfitting and enhance the adaptability of the model to new data through data enhancement.

[0132] Third, cross-modal learning. For multimodal medical imaging data, the optimization algorithm should also consider the ability of cross-modal learning. This is usually achieved by building a model that can jointly learn the features of multiple modal data, such as using a multi-way convolutional neural network structure that shares underlying features, or creating cross-modal representations by fusing features between modalities.

[0133] Then, parameter optimization and transfer learning are performed. During the optimization process, this patent uses appropriate optimization algorithms (such as Adam, RMSProp, etc.) and loss functions to enable the model to better learn the specific features of different data sets. In addition, through transfer learning, the knowledge learned from one data set can be used to accelerate the training process on another data set, thereby improving the efficiency and performance of the model.

[0134] In addition, a multimodal data integration strategy is adopted. When processing multimodal data, this patent designs an effective data integration strategy. This includes fusion at the data level, such as early fusion, late fusion or hybrid fusion methods; and fusion at the decision level, which helps to improve the algorithm's comprehensive understanding and processing capabilities of different modal data.

[0135] In addition, data preprocessing and post-processing techniques need to be improved. Because this patent targets different types of medical imaging data, data preprocessing and post-processing techniques also need to be improved accordingly. This may involve image standardization, normalization, denoising, and other processing, as well as specific post-processing steps for different medical imaging data characteristics.

[0136] In summary, by comprehensively applying the above strategies and methods, the CNN-based multimodal medical image denoising algorithm can optimize its structure and enhance its ability to process different types of medical image data. These improvements not only help improve the denoising effect of the algorithm, but also enhance its applicability to different data sets and disease types.

[0137] See also Figure 2, is a flow chart of a multimodal medical image denoising optimization method provided in an embodiment of the present application, and the method is applied to the multimodal medical image denoising optimization system described in the above embodiment.

[0138] The method may include:

[0139] Step S201: Acquire multimodal medical image data to be denoised.

[0140] Among them, multimodal medical imaging data refers to the imaging data obtained by different medical imaging technologies.

[0141] Step S202: inputting the multimodal medical image data to be denoised into a CNN model to obtain the multimodal medical image data after denoising.

[0142] After the multimodal medical image data to be denoised is obtained in step S201, the multimodal medical image data to be denoised is further input into the CNN model to obtain the multimodal medical image data after denoising.

[0143] The CNN model is trained using multimodal medical image sample data with noise labels.

[0144] In one possible implementation, the training process of the CNN model includes:

[0145] Acquire a training sample set, wherein the training sample set includes multimodal medical image sample data with noise labels;

[0146] The CNN model is trained using the training data set, and the model parameters of the CNN model are optimized based on a back propagation algorithm and a gradient descent algorithm until the model convergence condition is reached, thereby obtaining the CNN model.

[0147] Specifically, when training the CNN model, a training sample set is first obtained, where the training sample set includes multimodal medical image sample data with noise labels.

[0148] The CNN model is trained using the training data set, and the model parameters of the CNN model are optimized based on a back propagation algorithm and a gradient descent algorithm until the model convergence condition is reached, thereby obtaining the CNN model.

[0149] It should be noted that the model convergence condition can be set according to actual conditions. For example, when the change in model parameters between two iterations is less than or equal to a change threshold, it is considered that the model convergence condition is met.

[0150] Specifically, the training sample set has been introduced in detail above and will not be repeated here.

[0151] In order to test the noise reduction performance of the CNN model, in a possible implementation, after obtaining the CNN model, the method further includes:

[0152] The CNN model is tested using a test dataset.

[0153] Specifically, after the CNN model is trained, the CNN model is tested using a test data set to determine the noise reduction performance of the CNN model.

[0154] For example: Use the test data set to evaluate the trained model, and use the validation data to evaluate the model's performance, including accuracy, speed, and stability. The model needs to be able to show the same accuracy and generalization ability on the test data as on the training data set.

[0155] Specifically, the test data set has been introduced in detail above and will not be repeated here.

[0156] Step S203: Optimizing the multimodal medical image data after the noise reduction process to obtain optimized multimodal medical image data.

[0157] In this embodiment, after obtaining the multimodal medical image data after denoising using the trained CNN model, the multimodal medical image data after denoising is further optimized to obtain optimized multimodal medical image data.

[0158] It can be seen from the above embodiments that the neural network model in the present application is obtained by training with a large amount of sample data, which can effectively remove noise from multimodal medical image data, improve the quality and clarity of medical images, and facilitate further analysis and processing of medical images. In addition, the multimodal medical image data after noise reduction is further optimized to improve the noise reduction effect, making the noise-reduced medical image data more accurate and clear.

[0159] An embodiment of the present application provides an electronic device, characterized in that the electronic device includes a controller, and the controller includes a memory and a processor;

[0160] The memory is used to store one or more programs;

[0161] The processor; when the one or more programs are executed by the processor, the multimodal medical image denoising optimization method as described in the above embodiment is implemented.

[0162] An embodiment of the present application provides a computer storage medium, in which a program is stored. When the program is executed by a processor, the multimodal medical image denoising optimization method described in the above embodiment is implemented.

[0163] The "first" and "second" in the names such as "first" and "second" (if any) mentioned in the embodiments of the present application are only used as name identifiers and do not represent the first or second in order.

[0164] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0165] Those skilled in the art can understand that the flowchart shown in the figure is only an example in which the embodiments of the present application can be implemented, and the scope of application of the embodiments of the present application is not limited by any aspect of the flowchart.

[0166] In the several embodiments provided in the present application, it should be understood that the disclosed methods, devices and equipment can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the device or unit can be electrical, mechanical or other forms.

[0167] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, the functional units in the various embodiments of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0168] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multimodal medical image denoising optimization system, characterized in that: The system comprises: A data preprocessing module, used for receiving and preprocessing multimodal medical image data to obtain preprocessed multimodal medical image data; A CNN denoising module, used to perform denoising on the pre-processed multimodal medical image data using a trained CNN model to obtain denoised multimodal medical image data; wherein the CNN model is trained using multimodal medical image sample data with noise labels; The post-processing module is used to optimize the multimodal medical image data after the noise reduction process to obtain optimized multimodal medical image data.

2. The system according to claim 1, characterized in that The data preprocessing module comprises: A data collection submodule, used to collect multimodal medical imaging data; A data cleaning submodule, used to remove irrelevant information and noise in the multimodal medical image data to obtain cleaned multimodal medical image data; The data standardization submodule is used to standardize the cleaned multimodal medical image data to obtain standardized multimodal medical image data.

3. The system according to claim 1, characterized in that The CNN denoising module includes: CNN model building submodule, used to build a CNN model for multimodal medical imaging data denoising; The training data preparation submodule is used to prepare the training data required for training the CNN model; The model training submodule is used to train the CNN model using the training data to obtain a trained CNN denoising model.

4. The system according to claim 1, characterized in that The post-processing module comprises: A post-processing algorithm determination submodule is used to determine the post-processing algorithm corresponding to the current multimodal medical image data; A post-processing execution submodule, used to optimize the multi-modal medical image data after the noise reduction processing by using the post-processing algorithm to obtain optimized multi-modal medical image data; The result output submodule is used to output the optimized multimodal medical imaging data.

5. The system according to claim 1, characterized in that The CNN model is a deep convolutional neural network model, including multiple convolutional layers, pooling layers and fully connected layers, which is used to extract features from multimodal medical imaging data and perform noise reduction processing.

6. The system according to claim 1, characterized in that The training data includes multimodal medical imaging data with noise labels; During the training process of the CNN model, the loss function is calculated by comparing the denoised medical image data with the original medical image data to adjust the parameters of the CNN model.

7. The system according to claim 4, characterized in that The post-processing algorithm includes one or more of a histogram equalization algorithm, a median filtering algorithm, and an edge-preserving filtering algorithm.

8. A multimodal medical image denoising optimization method, characterized in that: The method is applied to the system according to any one of claims 1 to 7, and the method comprises: Acquire multimodal medical image data to be denoised; Inputting the multimodal medical image data to be denoised into the CNN model to obtain the multimodal medical image data after denoising; The multimodal medical image data after the noise reduction process is optimized to obtain optimized multimodal medical image data.

9. The method according to claim 8, characterized in that The training process of the CNN model includes: Acquire a training sample set, wherein the training sample set includes multimodal medical image sample data with noise labels; The CNN model is trained using the training data set, and the model parameters of the CNN model are optimized based on a back propagation algorithm and a gradient descent algorithm until the model convergence condition is reached, thereby obtaining the CNN model.

10. The method according to claim 9, characterized in that After obtaining the CNN model, the method further includes: The CNN model is tested using a test dataset.

Images